diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index a62c8e72836c518d904dcaa72e2551ac5cf70331..d46bf68c88781de6cdcf080c6ba26035cb3376e0 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": "20260609T060025Z", - "generated_at": "2026-06-09T06:09:53Z", + "snapshot_id": "20260609T120115Z", + "generated_at": "2026-06-09T12:12:37Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1838,104 +1838,6 @@ } ] }, - { - "cluster_id": "cluster-46159-3", - "summary": "Cluster of 1 issues and 2 PRs centered on issue #46159.", - "status": "open", - "confidence": 0.75, - "canonical_issue_number": 46159, - "canonical_pr_number": 46174, - "issue_numbers": [ - 46159 - ], - "pr_numbers": [ - 46174, - 46513 - ], - "evidence_types": [ - "closing_reference", - "shared_issue_target" - ], - "pr_comparisons": [ - { - "left_pr_number": 46174, - "right_pr_number": 46513, - "code_similarity": 0.505, - "size_similarity": 0.203, - "file_overlap": 0.222, - "area_overlap": 0.862, - "patch_similarity": 0.639, - "shared_filenames": [ - "src/transformers/models/rt_detr/modeling_rt_detr.py", - "src/transformers/models/rt_detr_v2/modeling_rt_detr_v2.py" - ], - "shared_file_areas": [ - { - "filename": "src/transformers/models/rt_detr/modeling_rt_detr.py", - "left_ranges": [ - [ - 847, - 854 - ], - [ - 857, - 867 - ], - [ - 870, - 878 - ] - ], - "right_ranges": [ - [ - 847, - 853 - ], - [ - 856, - 867 - ], - [ - 870, - 876 - ] - ] - }, - { - "filename": "src/transformers/models/rt_detr_v2/modeling_rt_detr_v2.py", - "left_ranges": [ - [ - 976, - 983 - ], - [ - 986, - 996 - ], - [ - 999, - 1007 - ] - ], - "right_ranges": [ - [ - 976, - 982 - ], - [ - 985, - 996 - ], - [ - 999, - 1005 - ] - ] - } - ] - } - ] - }, { "cluster_id": "cluster-45561-3", "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", @@ -2194,277 +2096,31 @@ ] }, { - "cluster_id": "cluster-44018-2", + "cluster_id": "cluster-44024-2", "summary": "Cluster of 2 related pull requests linked by soft_similarity.", "status": "open", "confidence": 0.5, "canonical_issue_number": null, - "canonical_pr_number": 44068, + "canonical_pr_number": 44076, "issue_numbers": [], "pr_numbers": [ - 44018, - 44068 + 44024, + 44076 ], "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 - ] - ] - } - ] + "left_pr_number": 44024, + "right_pr_number": 44076, + "code_similarity": 0.269, + "size_similarity": 0.976, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.492, + "shared_filenames": [], + "shared_file_areas": [] } ] } @@ -2525,13 +2181,13 @@ "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, + "cluster_id": "cluster-44024-2", + "canonical_pr_number": 44076, "duplicate_pr_numbers": [ - 44018 + 44024 ], "target_issue_number": null, - "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + "reason": "PRs in cluster-44024-2 are treated as duplicates because they share soft_similarity evidence." }, { "cluster_id": "cluster-45081-3", @@ -2579,15 +2235,6 @@ "target_issue_number": 45987, "reason": "PRs in cluster-45987-3 are treated as duplicates because they converge on issue #45987 with closing_reference, shared_issue_target evidence." }, - { - "cluster_id": "cluster-46159-3", - "canonical_pr_number": 46174, - "duplicate_pr_numbers": [ - 46513 - ], - "target_issue_number": 46159, - "reason": "PRs in cluster-46159-3 are treated as duplicates because they converge on issue #46159 with closing_reference, shared_issue_target evidence." - }, { "cluster_id": "cluster-46291-3", "canonical_pr_number": 46293, diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 0026370e22c5f114837ca6488e8d9b2b1912bf09..53c758c274bdf38a0552507b34c1c81bcbadd73d 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": "20260609T060025Z", - "generated_at": "2026-06-09T06:09:53Z", + "snapshot_id": "20260609T120115Z", + "generated_at": "2026-06-09T12:12:37Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -10,7 +10,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 1, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -127,333 +127,345 @@ "issue:43039|issue:43828", "issue:30277|issue:36331", "issue:34928|issue:36010", - "issue:45237|issue:46015", "issue:43039|issue:44805", + "issue:45237|issue:46015", "issue:43866|issue:46402", "issue:45800|issue:46304" ], - "prepared_review_unit_hash": "f5bff87f55adf1c60230b8445dce577027e5fbd37a5c51be76c0d4cc7b86cc03", + "prepared_review_unit_hash": "822e6c0633dcf95296382bf06616acc0b17ecf8120e53d24c72a1195659d4086", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 63, - "serialized_chars": 46515, - "estimated_input_tokens": 11629, - "estimated_eval_tokens": 23514 + "serialized_chars": 46529, + "estimated_input_tokens": 11633, + "estimated_eval_tokens": 23522 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mixed cluster with several unrelated issue families: distributed-training shape/index errors, model-specific incompatibilities, docs cleanup, and torch/version regressions. Only a small number of soft pairs look like true duplicates; most are just thematically similar.", - "confidence": 0.63, - "canonical_issue_reason": "No single canonical issue cleanly represents this heterogeneous set; the items split into multiple unrelated bugs and docs regressions. If one anchor must be chosen, issue 34928 is the strongest standalone report because it is the most detailed and heavily discussed.", + "summary": "Mostly distinct bug reports, with a few tight duplicate clusters: DeepSpeed ZeRO-3/Bert index-out-of-bounds, Qwen3.5 MoE rotary-embedding/device-materialization issues, and torch 2.4.1 compatibility reports. No pull requests are present.", + "confidence": 0.68, + "canonical_issue_reason": "issue:43596 is the clearest representative of the largest high-similarity cluster: the DeepSpeed ZeRO-3 + BertModel index-out-of-bounds failure, with a near-duplicate in 43638 and no broader, more generic framing than needed.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 34928 is the best single issue candidate overall: it has the richest reproduction details, the most discussion, and a concrete distributed-training failure mode. It is still not a representative canonical for the whole mixed cluster.", + "best_issue_reason": "issue:43596 is the strongest global issue representative because it names the concrete failure mode and model path cleanly, and it anchors the most convincing duplicate cluster in the set.", "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 runtime error; same model family but not the same bug." + "reason": "Related model family, but different failures: missing config attribute vs streaming chat/completions bug." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Different models, backends, and error paths (AMD GPU inference vs MPS histogram op on Int)." + "reason": "Different hardware/runtime problems on different backends (AMD GPU vs MPS Int histogram)." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Both are hardware/model execution problems, but the affected model, modality, and failure mode are unrelated." + "reason": "Different model-specific dtype mismatch vs AMD GPU execution failure." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Different code paths and symptoms; autocast dtype mismatch is not the same issue as AMD GPU failure." + "reason": "Autocast dtype mismatch in Phi MoE is not the same bug as GPT-OSS failing on AMD GPUs." }, { "left": "issue:45237", "right": "issue:46468", "accept": false, - "reason": "Different model and different runtime failure; only broad GPU execution context overlaps." + "reason": "Different models and symptoms; not the same concrete bug." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Both mention MoE-ish model execution, but the concrete bugs and failing methods are different." + "reason": "Both are MoE-related, but the failure modes and code paths are different." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "Distributed ZeRO3 collective mismatch and ONNX export error are unrelated." + "reason": "DeepSpeed collective mismatch and ONNX export error are unrelated issues." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "Export-to-ONNX failure vs Qwen3Moe FP8 failure; same broad area only by coincidence." + "reason": "Different model/export problems with no shared concrete failure path." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "Different models and failure modes; ONNX export is unrelated to auto device mapping cuda assert." + "reason": "ONNX export failure and device-mapping CUDA assert are distinct bugs." }, { "left": "issue:45901", "right": "issue:46015", "accept": false, - "reason": "Task crash vs outdated RAG examples; not the same underlying bug or change." + "reason": "Task crash and outdated RAG examples are not the same underlying issue." }, { "left": "issue:44509", "right": "issue:44512", - "accept": true, - "reason": "Both are v5 docs cleanup issues about stale removed functionality references; one docs fix could address both." + "accept": false, + "reason": "Both are v5 docs regressions, but they concern different removed references and are not one bug/change." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export error and GenerationMixin import error are separate API breakages." + "reason": "Export-to-ONNX and GenerationMixin import failure are unrelated." }, { "left": "issue:44918", "right": "issue:46165", "accept": false, - "reason": "Trainer embedding unpacking bug vs resume/trust_remote_code issue; different code paths." + "reason": "Different training/trainer problems; no shared code-path bug is evident." }, { "left": "issue:44805", "right": "issue:45823", "accept": false, - "reason": "Both are device/mask mismatch style failures, but they affect different models and different forwarding logic." + "reason": "Mask-shape IndexError and Gemma4 device-mismatch are different forward-path failures." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Both are tensor-shape related, but one is ZeRO3/BERT empty-indexing and the other is a mask/index shape mismatch; not clearly the same bug." + "reason": "Both involve errors during loading/training, but the concrete failures are different and not clearly the same bug." }, { "left": "issue:43638", "right": "issue:46165", "accept": false, - "reason": "Different training/resume behaviors and different failure signatures." + "reason": "Zero-3 Bert loading error vs resume-training issue are unrelated." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Both involve shape/index errors, but the BERT ZeRO3 empty tensor path is distinct from the mask/tensor indexing failure." + "reason": "Different concrete failures: Bert zero-size index error vs mask/tensor shape mismatch." }, { "left": "issue:44805", "right": "issue:46294", "accept": false, - "reason": "Different modalities and error causes; mask/index mismatch is not the same as video feature/token mismatch." + "reason": "Unrelated issues: tensor masking vs video chat-template feature/token mismatch." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Different subsystems: generic tensor indexing error vs tensor-parallel MoE routing problem." + "reason": "Different subsystems and failure modes." }, { "left": "issue:46165", "right": "issue:46468", "accept": false, - "reason": "Resume-training trust_remote_code problem is unrelated to the Nemotron example failure." + "reason": "Resume-training with local model is unrelated to the Nemotron example breakage." }, { "left": "issue:44560", "right": "issue:46165", "accept": false, - "reason": "Video chat-template mismatch and resume-training issue are unrelated." + "reason": "Video StopIteration and training-resume problems are unrelated." }, { "left": "issue:43596", "right": "issue:46165", "accept": false, - "reason": "Zero3/BERT initialization failure is unrelated to model resume/trust_remote_code behavior." + "reason": "Different bugs: Zero-3/Bert loading vs resume-training with trust_remote_code." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "Processor voice_preset bug and ONNX export failure are unrelated." + "reason": "BarkProcessor voice preset and ModernBERT ONNX export are unrelated." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Different distributed-training bugs with different trigger conditions and error signatures." + "reason": "Both are DeepSpeed-related, but the concrete failures and code paths differ." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Both report the same DeepSpeed ZeRO-3 + BERT zero-sized parameter/index-0 failure, with only minor wording differences." + "reason": "Same DeepSpeed ZeRO-3/BertModel index-out-of-bounds failure; 43638 is just a non-pretrained Bert variant of the same bug." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Liger cross-entropy interception and GPT-OSS tensor-parallel failure are unrelated." + "reason": "Liger kernel cross-entropy call and GPT-OSS TP failure are unrelated." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Detokenization regression and LoRA merge/reload inconsistency are different bugs." + "reason": "Wrong detokenization and LoRA merge/save divergence are different bugs." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "NVFP4 runtime error and LoRA save/reload mismatch are unrelated." + "reason": "NVFP4 runtime failure and LoRA save/load mismatch are unrelated." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Both are training-time tensor shape issues, but they involve different mechanisms (activation checkpointing/FSDP vs indexing/mask mismatch)." + "reason": "FSDP activation-checkpointing shape recomputation and mask index mismatch are not the same bug." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "DeepSpeed weight-shape error and collective mismatch are different distributed-training failures." + "reason": "Both mention DeepSpeed, but one is a weight-dimension error and the other is a collective mismatch." }, { "left": "issue:45859", "right": "issue:45902", "accept": true, - "reason": "Both describe the same Qwen3_5MoeTextRotaryEmbedding offload/materialization bug: buffers are not correctly handled when moving from meta/offload to CUDA." + "reason": "Same Qwen3_5MoeTextRotaryEmbedding buffer/materialization bug; both failures come from mishandled inv_freq/device state." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "Different model families and failure modes; ZeRO3/BERT indexing is unrelated to GPT-OSS tensor-parallel behavior." + "reason": "Zero-3 Bert loading error vs GPT-OSS tensor-parallel issue are unrelated." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "Activation-checkpointing/FSDP recompute mismatch and custom compute_loss signature breakage are separate trainer issues." - }, - { - "left": "issue:43596", - "right": "issue:45161", - "accept": false, - "reason": "Different runtime and model-parallel problems; not the same bug." + "reason": "Activation-checkpointing/FSDP mismatch and CustomTrainer signature breakage are unrelated." }, { "left": "issue:43716", "right": "issue:45507", "accept": false, - "reason": "Different model-specific failures: dtype mismatch in preprocessing vs invalid method call in GraniteMoEHybrid." + "reason": "Different model families and different failure points." }, { "left": "issue:45874", "right": "issue:46294", "accept": false, - "reason": "Both involve multimodal Gemma/Qwen inputs, but one is reverse embedding lookup and the other is video token/feature mismatch." + "reason": "Inputs_embeds reverse lookup issue and video chat-template mismatch are unrelated." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "Different model/runtime failures and hardware contexts." + "reason": "Different GPU/runtime issues with no shared concrete code path." }, { "left": "issue:39290", "right": "issue:45081", "accept": false, - "reason": "Gemma3/vLLM attribute error and Mistral tokenizer regex patch crash are unrelated." + "reason": "Gemma3/vLLM config regression and Mistral tokenizer patch crash are unrelated." }, { "left": "issue:45800", "right": "issue:46515", "accept": true, - "reason": "Both are the same torch 2.4.1 compatibility regression; one report is just more specific about the transformers version." + "reason": "Both describe the same torch 2.4.1 compatibility regression; one is simply a more specific transformers version report." }, { "left": "issue:45823", "right": "issue:45874", "accept": false, - "reason": "Different Gemma4 bugs: device placement mismatch vs expensive reverse embedding lookup." + "reason": "Device-map forward mismatch and expensive reverse embedding lookup are different Gemma4 issues." }, { "left": "issue:43425", "right": "issue:44292", "accept": false, - "reason": "Generic torch incompatibility and Qwen-3 NVFP4 runtime error are not the same issue." + "reason": "Torch 2.4.1 incompatibility and Qwen-3 NVFP4 runtime failure are not the same bug." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP checkpointing mismatch and Qwen3 auto device-map CUDA assert are unrelated." }, { "left": "issue:43828", - "right": "issue:45161", + "right": "issue:45685", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to GPT-OSS tensor-parallel failure." + "reason": "Different dtype/hardware bugs on different models and backends." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "TRL SFT embedding unpacking and MPS histogram error are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger-kernel cross-entropy routing and Phi autocast dtype mismatch are different bugs." }, { "left": "issue:30277", "right": "issue:36331", "accept": false, - "reason": "Distributed collectives mismatch and trainer compute_loss signature error are unrelated." + "reason": "DeepSpeed collective mismatch and trainer loss-signature incompatibility are unrelated." }, { "left": "issue:34928", "right": "issue:36010", "accept": false, - "reason": "FSDP activation checkpointing mismatch and GenerationMixin import failure are unrelated." + "reason": "FSDP activation-checkpointing issue and GenerationMixin import failure are unrelated." }, { - "left": "issue:45237", - "right": "issue:46015", + "left": "issue:43039", + "right": "issue:44805", "accept": false, - "reason": "GPU inference failure and outdated RAG docs are unrelated." + "reason": "Cross-entropy dispatch issue and tensor mask shape mismatch are different failures." }, { - "left": "issue:43039", - "right": "issue:44805", + "left": "issue:45237", + "right": "issue:46015", "accept": false, - "reason": "Liger kernel cross-entropy routing and mask/index shape mismatch are unrelated." + "reason": "GPU inference failure and RAG docs cleanup are unrelated." }, { "left": "issue:43866", "right": "issue:46402", "accept": false, - "reason": "Checkpoint corruption and missing text encoder weights are different model-loading problems." + "reason": "Checkpoint corruption and missing text encoder weight loading are different bugs." }, { "left": "issue:45800", "right": "issue:46304", "accept": false, - "reason": "Both mention torch/version support, but the concrete breakage and affected runtime paths are different." + "reason": "Both are compatibility regressions, but one is torch-version specific while the other is a USE_DISTRIBUTED=0 breakage; not the same concrete bug." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative about the cluster being mixed. The canonical-anchor choice for issue 34928 is supported by its high discussion activity and detailed report. The accepted soft pairs are also reasonably conservative: 43596/43638 and 45859/45902 look like true duplicates, and 44509/44512 are both docs-v5 cleanup issues in the same general removal family, so treating them as mergeable is defensible." + "feedback": "The summary is grounded in the packet: it sticks to issue titles, correctly notes that all items are issues, and only elevates clusters that are supported by the stronger similarity pairs. The accepted soft-edge pairs are plausible duplicate clusters, and the rejected pairs are conservative rather than overstated. " }, "evaluator_used": true, "retried": false, @@ -466,7 +478,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 2, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -535,8 +547,8 @@ "issue:43039|issue:44560", "issue:43975|issue:46015", "issue:44863|issue:46015", - "issue:45812|issue:45920", "issue:45020|issue:46304", + "issue:45812|issue:45920", "issue:43866|issue:44863", "issue:44292|issue:46015", "issue:43872|issue:46304", @@ -566,251 +578,284 @@ "issue:43541|issue:45161", "issue:43828|issue:45464" ], - "prepared_review_unit_hash": "fd0d2e3b9304a3492e7d9176bfcc5dd2b8b250fe6fa3c20e1b26e14ef690f4f4", + "prepared_review_unit_hash": "9d41620646e18ede194c2c5a2491f9efc7ad563dc74eebbab6077f6ea7945423", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 41, - "serialized_chars": 39729, - "estimated_input_tokens": 9933, - "estimated_eval_tokens": 20122 + "serialized_chars": 39736, + "estimated_input_tokens": 9934, + "estimated_eval_tokens": 20124 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Loose similarity cluster around transformers regressions, model-loading failures, and compatibility issues, but the soft pairs are not true duplicates; they describe different models, APIs, or failure modes.", - "confidence": 0.94, - "canonical_issue_reason": "issue:45812 is the broadest representative in this set: it\u2019s a concrete v4\u2192v5 tokenizer regression affecting multiple model families, whereas most other issues are narrower or on different code paths.", + "summary": "Mostly heterogeneous. The only supported duplicate subgroups are the ZeRO-3 loading IndexErrors (issue:41762 / issue:43596) and the v5 AutoTokenizer wrong-token-ID regression (issue:45812 / issue:45920). The mask/tensor-shape IndexError reports are not strong enough to merge and should remain separate.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45920 is the best canonical anchor for the accepted tokenizer-regression subgroup because it is broader and more current, and it clearly covers the same v5 AutoTokenizer wrong-ID bug seen in issue:45812.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the best overall anchor for the cluster because it has the widest scope and clearest cross-model impact among the issues listed.", + "best_issue_reason": "issue:45920 is the strongest overall issue in the set because it captures a cross-model v5 tokenizer regression rather than a narrow model-specific instance.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43296", "right": "issue:46304", "accept": false, - "reason": "Both are version-related load/runtime problems, but one is a vLLM PaddleOCR-VL load failure and the other is a PyTorch USE_DISTRIBUTED support regression." + "reason": "Different failures: PaddleOCR-VL loading in vLLM versus a USE_DISTRIBUTED=0 support breakage." }, { "left": "issue:41762", "right": "issue:43596", - "accept": false, - "reason": "Both involve DeepSpeed ZeRO-3 index errors, but one is Gemma3 loading and the other is BertModel; different model paths and triggers." + "accept": true, + "reason": "Same ZeRO-3 loading path and same out-of-bounds index-0 failure; these look like one shared bug." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Gemma3 sliding-window config breakage and Qwen3 auto device map CUDA asserts are different failures in different model families." + "reason": "Both involve runtime/model-loading errors, but the concrete bugs differ: Gemma3 sliding_window_pattern vs Qwen3 auto device mapping cuda assert." }, { "left": "issue:44589", "right": "issue:46304", "accept": false, - "reason": "One is a float8 storage/type lookup error; the other is a distributed-build compatibility issue. Different underlying bugs." + "reason": "Float8 storage lookup failure and USE_DISTRIBUTED=0 incompatibility are unrelated runtime problems." }, { "left": "issue:43901", "right": "issue:44512", "accept": false, - "reason": "This is docs drift versus a removed CLI command; they are unrelated documentation issues." + "reason": "Both are docs-related, but they cover different topics and do not share the same underlying change or bug." }, { "left": "issue:45823", "right": "issue:46294", "accept": false, - "reason": "Both mention device/input handling, but one is Gemma4 PLE device mismatch and the other is Qwen3.5 video token/template mismatch." + "reason": "Gemma4 device-map mismatch and Qwen3.5 video/chat-template mismatch are different code paths." }, { "left": "issue:45237", "right": "issue:45901", "accept": false, - "reason": "GPT-OSS on AMD GPUs and a table-question-answering crash are unrelated failures." + "reason": "AMD GPU support for GPT-OSS-20B and table-question-answering crashes are unrelated." }, { "left": "issue:44841", "right": "issue:45823", "accept": false, - "reason": "Voxtral processor failure and Gemma4 device mismatch are different code paths and symptoms." + "reason": "Processor failure for Voxtral and Gemma4 device mismatch are distinct model-specific issues." }, { "left": "issue:43039", "right": "issue:44560", "accept": false, - "reason": "Liger Kernel cross_entropy dispatch and Qwen3-vl video StopIteration are not the same bug." + "reason": "Liger Kernel cross_entropy dispatch and Qwen3-vl video StopIteration are different bugs." }, { "left": "issue:43975", "right": "issue:46015", "accept": false, - "reason": "DeepSeek detokenization regression and outdated RAG examples are unrelated." + "reason": "Incorrect detokenization in DeepSeek-Coder v5 is unrelated to outdated RAG docs." }, { "left": "issue:44863", "right": "issue:46015", "accept": false, - "reason": "NemotronH checkpoint loading and RAG docs examples are different issues." - }, - { - "left": "issue:45812", - "right": "issue:45920", - "accept": false, - "reason": "Both are tokenizer regressions, but 45812 is Granite-specific while 45920 spans a different set of model families and wording." + "reason": "NemotronH checkpoint loading failure and RAG documentation are unrelated." }, { "left": "issue:45020", "right": "issue:46304", "accept": false, - "reason": "Remote_code breakage and USE_DISTRIBUTED build support are different compatibility regressions." + "reason": "Remote_code regressions and distributed-build support are different failure modes." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same v5 AutoTokenizer regression producing wrong token IDs across multiple model families; 45920 is the broader tracker." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Ovis2 checkpoint corruption and NemotronH loading failure are distinct model-specific problems." + "reason": "Corrupted Ovis2 checkpoint and NemotronH checkpoint loading failure are unrelated." }, { "left": "issue:44292", "right": "issue:46015", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime failure and RAG documentation updates are unrelated." + "reason": "Qwen-3-8B-NVFP4 runtime failure and RAG docs are unrelated." }, { "left": "issue:43872", "right": "issue:46304", "accept": false, - "reason": "bitsandbytes Int8Params constructor incompatibility is a quantization-library API issue, not the distributed build problem in 46304." + "reason": "bitsandbytes Int8Params incompatibility is unrelated to the distributed-build support regression." }, { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Both mention sliding_window, but Gemma3 config missing an attribute and Qwen3-MoE sliding_window behavior are different defects." + "reason": "Both mention sliding-window behavior, but the reported bugs are different Gemma3 and Qwen3-MoE issues." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Both are IndexErrors during loading, but one is a mask/tensor shape mismatch and the other is ZeRO-3 init with non-pretrained BertModel." + "reason": "Both are IndexErrors, but the mask/tensor-shape mismatch and the ZeRO-3 zero-sized-tensor load failure do not clearly share one root cause." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "GLM-4.7-Flash test loading failure and Ovis2 checkpoint corruption are unrelated model issues." + "reason": "Failed model loading in unit tests versus a corrupted checkpoint are different problems." }, { "left": "issue:43792", "right": "issue:45901", "accept": false, - "reason": "Whisper loading failure and table QA crash are different tasks and failure modes." - }, - { - "left": "issue:43425", - "right": "issue:45070", - "accept": false, - "reason": "Torch 2.10 incompatibility and the PretrainedConfig pydantic field regression are separate versioning problems." + "reason": "Whisper-large-v2 failing to run and table-question-answering crashing are unrelated." }, { - "left": "issue:43596", - "right": "issue:45800", + "left": "issue:43975", + "right": "issue:45901", "accept": false, - "reason": "Both are compatibility regressions, but one is a ZeRO-3 Bert init failure and the other is a torch/transformers version mismatch." + "reason": "Incorrect detokenization and table-question-answering crash are unrelated." }, { - "left": "issue:43975", - "right": "issue:45901", + "left": "issue:44863", + "right": "issue:46015", "accept": false, - "reason": "DeepSeek detokenization and table QA crashes are unrelated." + "reason": "Checkpoint loading failure and docs issue are unrelated." }, { - "left": "issue:43792", - "right": "issue:45070", + "left": "issue:45020", + "right": "issue:46304", "accept": false, - "reason": "Whisper cannot run and the pydantic PretrainedConfig regression are different issues." + "reason": "Remote_code breakage and distributed-build support are different issues." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Both involve Qwen3.5, but one is a chat crash on a large model and the other is a streaming inference API failure on a smaller model." + "reason": "Both involve Qwen3.5, but the symptoms and code paths differ: chat crash versus streaming API failure." }, { "left": "issue:44292", "right": "issue:45901", "accept": false, - "reason": "A Qwen-3-8B-NVFP4 runtime error and a table QA crash do not share the same underlying bug." + "reason": "Qwen-3-8B-NVFP4 runtime failure and table-question-answering crash are unrelated." }, { "left": "issue:39290", "right": "issue:45823", "accept": false, - "reason": "Gemma3 sliding-window attribute regression and Gemma4 device-map mismatch are different model-specific bugs." + "reason": "Gemma3/vLLM sliding-window config and Gemma4 device-mapping mismatch are different bugs." }, { "left": "issue:43404", "right": "issue:43866", "accept": false, - "reason": "A tied lm_head weight bug in Mistral3 and a corrupted Ovis2 checkpoint are unrelated." + "reason": "Tied lm_head weights in Mistral3 and a corrupted Ovis2 checkpoint are unrelated." }, { "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Qwen3.5 chat streaming failure and an MPS histogram int-support error are different subsystems and symptoms." + "reason": "Qwen3.5 streaming inference failure and MPS histogram on Int in a MoE path are unrelated." }, { "left": "issue:43541", "right": "issue:45685", "accept": false, - "reason": "Mixtral grouped_mm tracing and MPS histogram int support are unrelated execution-path bugs." + "reason": "Grouped_mm tracing failure and MPS histogram error are different MoE-related symptoms, but not the same underlying bug." }, { "left": "issue:43541", "right": "issue:43828", "accept": false, - "reason": "Both touch MoE/runtime math, but grouped_mm tracing and autocast dtype mismatch are different concrete failures." + "reason": "Torch dynamo tracing grouped_mm failure and autocast dtype mismatch are distinct runtime issues." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Voxtral processor failure and 'Can't compile non template nodes' are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "ImportError for GenerationMixin and Gemma3 sliding_window_pattern are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are IndexErrors, but one is a mask/tensor shape mismatch and the other is a ZeRO-3 zero-sized-tensor load failure; not the same bug." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Torch 2.4.1/transformers 5.8 incompatibility and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "PaddleOCR-VL load failure and GGUF gpt-oss support are different problems." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig and the PretrainedConfig pydantic break are unrelated compatibility issues." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "Shape-mask IndexError and Qwen3.5 embedding unpacking with TRL SFT are similar-looking but distinct bugs." + "reason": "Both are loading/shape-related failures, but the mask/tensor-shape IndexError and Qwen3.5 embedding unpacking with TRL SFT trainer are different bugs." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Int4 OOM under v5 and tied-weights translation-model issues are different regressions." + "reason": "Int4 quantization OOM and tied-weights regression on translation models are unrelated." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Mixtral grouped_mm tracing and GPT-OSS MoE TP-only failure are both MoE-related but not the same code-path bug." + "reason": "Grouped_mm tracing failure and GPT-OSS MoE TP-only failure are different code-path problems." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE and Qwen3.5 streaming inference failure are distinct." + "reason": "Autocast dtype mismatch on Phi-tiny-MoE and Qwen3.5 streaming API failure are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, + "evaluator_result": { + "accept": false, + "feedback": "The summary overstates the evidence. The packet supports two accepted duplicate subgroups (ZeRO-3 loading IndexErrors and the v5 AutoTokenizer wrong-token-ID regression), but it does not support calling the mask/tensor shape IndexError cases 'duplicate-worthy' because the corresponding soft verdicts are explicitly rejected. That makes the cluster summary inconsistent with the soft-edge decisions. The other verdicts look conservative, but this overclaim is enough to reject." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -819,7 +864,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 3, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -890,14 +935,14 @@ "issue:39186|issue:44918", "issue:43716|issue:44918", "issue:40990|issue:44960", - "issue:45632|issue:45698", "issue:43645|issue:45399", + "issue:45632|issue:45698", "issue:43404|issue:45005", "issue:42915|issue:45399", "issue:44530|issue:45005", "issue:43873|issue:45005", - "issue:44360|issue:44512", "issue:44863|issue:45005", + "issue:44360|issue:44512", "issue:45183|issue:46463", "issue:46489|issue:46491", "issue:44918|issue:45507", @@ -909,229 +954,229 @@ "issue:43644|issue:43950", "issue:36331|issue:41762" ], - "prepared_review_unit_hash": "afe2e254ff63361c6e8ab854757252fcfc57d166bb02bd8e40f63580d0b3f0ec", + "prepared_review_unit_hash": "e12b4df5435879803efacc37c6a0cce9e1c76f7de7fb3bbe459a984e93512aa2", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36944, - "estimated_input_tokens": 9236, - "estimated_eval_tokens": 18728 + "serialized_chars": 36940, + "estimated_input_tokens": 9235, + "estimated_eval_tokens": 18726 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated Transformers regression reports. The only clear duplicate chain is the non-persistent-buffer corruption bug around 43644/43950/44534; the rest are mostly broad symptom or model-family similarities, not the same underlying defect.", - "confidence": 0.46, - "canonical_issue_reason": "43644 is the broadest and clearest statement of the non-persistent-buffer regression, and 43950/44534 read like direct rephrasings of that same bug.", + "summary": "This cluster is heterogeneous rather than one single duplicate group. The strongest duplicate candidates are the non-persistent-buffer regression, the SigLIP2 degradation reports, the T5/input validation error-message reports, and the Qwen video-input regression reports.", + "confidence": 0.72, + "canonical_issue_reason": "issue:43950 is the clearest, most actionable representative of the non-persistent-buffer regression, and it has direct duplicate support from nearby reports.", "canonical_pr_reason": null, - "best_issue_reason": "43644 is the best cluster representative because it captures the underlying regression in general terms and anchors the only strong duplicate chain in the set.", + "best_issue_reason": "issue:43950 is the best issue anchor in this set because it describes a concrete v5 regression with strong duplicate evidence and a clear fix surface.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "Different bugs: init_empty_weights/_is_hf_initialized argument handling vs a pydantic config regression." + "reason": "Different regressions: init_empty_weights argument handling vs pydantic PretrainedConfig field behavior." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "Unrelated features: gpt-oss GGUF support vs Qwen3.5 streaming chat/completions failure." + "reason": "Different areas: GGUF architecture support vs chat/completions streaming failure." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Different code paths: Qwen3-MoE sliding_window behavior vs pydantic model validation." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to the pydantic config regression." }, { "left": "issue:43716", "right": "issue:46294", "accept": false, - "reason": "Both are multimodal, but image-preprocessor dtype mismatch and video token/template mismatch are different bugs." + "reason": "Both are multimodal, but one is dtype mismatch and the other is a video token/feature count failure." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "Both mention compatibility, but remote_code breakage and torch/transformers version incompatibility are not the same defect." + "reason": "Remote_code breakage and torch/transformers version incompatibility are different bugs." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Different failures: torch dynamo grouped_mm tracing vs missing Float8 storage type." + "reason": "Different concrete failures: torch dynamo tracing grouped_mm vs missing Float8 storage type." }, { "left": "issue:44479", "right": "issue:46294", - "accept": false, - "reason": "Related Qwen3.5 video-input symptoms, but the evidence is too broad to call this the same concrete bug." + "accept": true, + "reason": "Both report the same Qwen video-input regression class, centered on video feature/token mismatch in Qwen3.5-family handling." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Both are shape/indexing errors, but the reported failure modes and loading contexts differ." + "reason": "Both are IndexErrors, but the masks-shape mismatch and zero3/BertModel loading failure are different code paths." }, { "left": "issue:43054", "right": "issue:43994", - "accept": false, - "reason": "Similar SigLIP2 degradation symptoms, but no clear evidence they are the same underlying issue." + "accept": true, + "reason": "Both are SigLIP2 reports describing bad or nonsensical outputs for the same model family, likely the same underlying model/processor issue." }, { "left": "issue:45759", "right": "issue:45902", "accept": false, - "reason": "Different defects: composite config unwrapping vs uninitialized rotary embedding memory after materialization." + "reason": "One is composite-config unwrapping, the other is uninitialized rotary embedding memory after materialization." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "Different model/runtime problems: FSDP weight-shape error vs TRL embedding unpacking failure." + "reason": "FSDP weight-shape failure and Qwen3.5/TRL embedding unpacking are unrelated." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Both involve model loading/training, but the failures are on different paths and unrelated." + "reason": "Both touch multimodal inputs, but the failures and model families are different." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "No concrete overlap beyond broad model-quality complaints." + "reason": "Perplexity issue on gpt-oss and GLM5 issue are unrelated." }, { - "left": "issue:45632", - "right": "issue:45698", + "left": "issue:43645", + "right": "issue:45399", "accept": false, - "reason": "Both involve custom-code loading, but the cache-path collision and wrong-module-after-save reports are not clearly the same bug." + "reason": "Custom model Jupyter initialization and flash-attn fallback gating are different bugs." }, { - "left": "issue:43645", - "right": "issue:45399", + "left": "issue:45632", + "right": "issue:45698", "accept": false, - "reason": "Jupyter custom-model initialization regression and flash-attn fallback gating are different issues." + "reason": "Both involve remote-code loading/caching, but the titles point to different failure modes and not the same concrete bug." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Both mention tied weights, but they affect different model families and likely different code paths." + "reason": "Both mention tied weights, but one is Mistral3 image-text generation and the other is translation-model v5 behavior; not enough to treat as the same bug." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "FP8 config failure vs flash-attn2 fallback checks are unrelated." + "reason": "Qwen3Moe FP8 failure and flash-attn fallback gating are unrelated." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttention/linear_attention crash is unrelated to tied-weight handling." + "reason": "PagedAttentionCache linear_attention crash is unrelated to tied-weights handling." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization offloading behavior is not the same as tied-weights regression." + "reason": "Quantization/offloading behavior is a different subsystem from tied-weight regressions." }, { - "left": "issue:44360", - "right": "issue:44512", + "left": "issue:44863", + "right": "issue:45005", "accept": false, - "reason": "A DSA indexer discussion and a docs typo are not duplicates." + "reason": "NemotronH checkpoint loading is unrelated to translation-model tied weights." }, { - "left": "issue:44863", - "right": "issue:45005", + "left": "issue:44360", + "right": "issue:44512", "accept": false, - "reason": "NemotronH checkpoint loading failure is unrelated to translation-model weight tying." + "reason": "DSA indexing discussion and docs cleanup are not duplicates." }, { "left": "issue:45183", "right": "issue:46463", - "accept": false, - "reason": "Both concern input validation messages, but the specific validators and model families differ too much to be a confident duplicate." + "accept": true, + "reason": "Both point at the same input validation/error-message path for decoder/input_ids vs inputs_embeds handling." }, { "left": "issue:46489", "right": "issue:46491", "accept": false, - "reason": "Different tokenizer regressions; only a broad round-trip symptom is shared." + "reason": "Both are tokenizer regressions, but they affect different models and different round-trip failures." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "Unpacking Qwen3.5 embeddings and GraniteMoEHybrid invalid-method calls are different bugs." + "reason": "TRL embedding unpacking and GraniteMoEHybrid invalid method calls are unrelated." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "Docs typos in different places are not the same issue." + "reason": "A typo in cache strategies and a removed CLI docs mention are unrelated." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Different tokenizer/runtime failures with no shared code path evident." + "reason": "DeepSeek tokenizer detokenization and Qwen-3-8B-NVFP4 runtime error are different issues." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Different model/runtime errors." + "reason": "Different model families and different failure modes." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Same non-persistent-buffer junk/corruption regression, just phrased differently." + "reason": "These are the same non-persistent-buffer corruption regression in v5, described from two angles." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "SigLIP2 docs example errors are different from the broader text-embedding quality complaint." + "reason": "Docs/example errors for SigLIP2 are not the same bug as degraded SigLIP2 text embeddings." }, { "left": "issue:43644", "right": "issue:43950", "accept": true, - "reason": "Same non-persistent-buffer corruption regression; 43644 is the broader report and 43950 is the same bug in alternate wording." + "reason": "Exact same v5 non-persistent-buffer corruption regression; one report is the general symptom, the other the specific from_pretrained corruption." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "CustomTrainer keyword-arg mismatch is unrelated to Gemma3 ZeRO-3 loading failures." + "reason": "Trainer loss-signature mismatch and Gemma3 ZeRO-3 loading failure are unrelated." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The cluster summary is mostly grounded, but the soft-edge handling is not conservative enough. In particular, several accepted pairs are justified mainly by broad symptom/model-family similarity rather than clear evidence of the same bug or a mergeable single fix (especially issue:43054\u2194issue:43994, and to a lesser extent issue:44479\u2194issue:46294, issue:45632\u2194issue:45698, and issue:45183\u2194issue:46463). The evidence in the packet supports relatedness, but not a confident duplicate verdict for all of those pairs." + "accept": true, + "feedback": "The summary is grounded in the packet: the cluster is clearly heterogeneous, and the highlighted duplicate groupings are supported by the issue titles. The accepted soft-edge pairs are conservative enough given the title-level evidence, especially the non-persistent-buffer regression cases. No obvious overstatement." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1141,7 +1186,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 4, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -1233,238 +1278,238 @@ "issue:45488|issue:46489", "issue:45020|issue:45356" ], - "prepared_review_unit_hash": "6a5b1d3ee8b3e9774710d673091255d9dcc005a6908b3284e561ac96ff80bf75", + "prepared_review_unit_hash": "944140d44ebe252a3a454a93305560b9107fde31233c3c16a8ca337c5fae33a1", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 38000, - "estimated_input_tokens": 9500, - "estimated_eval_tokens": 19256 + "serialized_chars": 37995, + "estimated_input_tokens": 9499, + "estimated_eval_tokens": 19254 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly heterogeneous; only the tokenizer v5 regression pair (45920/46489) looks like the same underlying bug. The rest are different model-loading, quantization, memory, or device-placement issues.", - "confidence": 0.71, - "canonical_issue_reason": "No single canonical issue cleanly represents the whole set because it spans many unrelated bugs; the closest umbrella is issue 45920, which covers the broad tokenizer-id regression family.", + "summary": "This cluster is mostly a mix of v5-era tokenizer/model-loading regressions plus a few unrelated buffer, quantization, and training issues. Only a few soft links look like true duplicates; most are merely same-subsystem similarity.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45920 is the best umbrella issue here: it describes a broad v5 tokenizer regression affecting multiple model families, which matches the densest shared theme in the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45920 is the best global representative because it is the broadest open report in the one clear duplicate family here, and it naturally subsumes the narrower DeepSeek-Coder manifestation in 46489.", + "best_issue_reason": "issue:45920 is the most representative and actionable issue in the set: it is concrete, cross-model, and clearly captures the recurring tokenizer-output regression pattern.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Different bugs: dtype mismatch under autocast vs Flash Attention/3D position_ids causing illegal access/NaNs." + "reason": "Different models and different failure modes: autocast dtype mismatch vs RLHF NaN/SDPA fallback." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Both are backend/quantization failures, but the concrete errors and code paths differ." + "reason": "Both are quantization-related TypeErrors, but they point to different incompatibilities (_is_hf_initialized vs Float8 storage lookup)." }, { "left": "issue:45920", "right": "issue:46489", "accept": true, - "reason": "Same tokenizer regression family in v5: 46489 is a narrower DeepSeek-Coder manifestation of the broader wrong-token-IDs bug in 45920." + "reason": "Both report a v5 tokenizer-output regression; 46489 looks like a model-specific gap in the same broader tokenization fix." }, { "left": "issue:43950", "right": "issue:45902", - "accept": false, - "reason": "Both involve loading/materialization, but one is generic non-persistent-buffer corruption and the other is Qwen3_5Moe meta->to_empty uninitialized memory." + "accept": true, + "reason": "Same underlying buffer/materialization bug: non-persistent buffer handling leading to corrupted or uninitialized buffer state." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Checkpoint corruption vs. noisy load messages; not the same bug." + "reason": "Unrelated: checkpoint corruption vs generic loading noise." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "Different problems: SigLIP2 embedding quality vs. tokenization changes across versions." + "reason": "SigLIP2 embedding quality regression vs a general tokenization-version issue; not the same bug." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Related model family, but zero probabilities and degraded embeddings are distinct symptoms with no clear shared code-path." + "reason": "Same model family, but the reported symptoms are too different to confidently treat as the same concrete bug." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad AutoTokenizer error reporting vs. missing all_tied_weights_keys on a different model class." + "reason": "Different subsystems: tokenizer error message vs missing tied-weights attribute on a model class." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Flash Attention crash on packed-sequence handling vs. CPU-offload incompatibility in rotary embedding." + "reason": "Different code paths: Flash Attention packed-sequence misuse vs CPU-offload incompatibility in rotary embedding." }, { "left": "issue:45663", "right": "issue:45902", "accept": false, - "reason": "Different forward-path bug: Gemma4 FSDP2 shared_kv_states KeyError vs. Qwen3_5Moe buffer materialization issue." + "reason": "Both involve loading/training internals, but one is an FSDP2 per-layer state bug and the other is buffer materialization corruption." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Gradient checkpointing/LoRA issue vs. int4 quantization memory regression; unrelated failure modes." + "reason": "Both are memory-related, but one is gradient checkpointing/LoRA behavior and the other is int4 quantization reservation growth." }, { "left": "issue:43054", "right": "issue:43493", - "accept": false, - "reason": "SigLIP2 text-embedding quality issue vs. HF/JAX implementation discrepancy; not the same bug." + "accept": true, + "reason": "Same SigLIP2 implementation discrepancy theme; 43493 reads like the root-cause report behind the bad embedding quality." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Tokenizer error-message issue vs. a generic glm-4v-9b loading failure." + "reason": "Different models and failures: bad tokenizer error message vs generic glm-4v-9b loading failure." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Loading noise/warnings vs. inability to load a specific model." + "reason": "Generic loading noise is not the same as a specific model load failure." }, { "left": "issue:24643", "right": "issue:28530", "accept": false, - "reason": "Training-with-DeepSpeed weight-shape error vs. early-stopping metric configuration warning." + "reason": "Completely different training problems: DeepSpeed weight shape error vs early-stopping metric configuration." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Tokenizer detokenization regression vs. flash_attn import failure on Python 3.13." + "reason": "Tokenizer detokenization bug is unrelated to the Python 3.13 flash_attn import issue." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Unnecessary load noise vs. model loading failure." + "reason": "One is about noisy loading logs; the other is a specific model-loading failure." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "LoRA gradient-checkpointing issue vs. Qwen3 MoE memory-usage regression." + "reason": "Both mention memory, but the code paths differ: LoRA checkpointing vs Qwen3 MoE training memory usage." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "Both mention remote_code, but one is missing fields in AutoConfig and the other is a broader compatibility regression; not clearly the same concrete bug." + "reason": "Remote-code broad regression vs missing fields in AutoConfig return object; too broad and not the same bug." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Different load-time failures on different model/config paths." + "reason": "Different failure classes and different model stacks: ZeRO-3 load crash vs Qwen3 MoE FP8 failure." }, { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "MPS histogram op limitation vs. Gemma4 device-map mismatch; unrelated." + "reason": "MPS histogram kernel limitation is unrelated to Gemma4 device placement mismatch." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Qwen3VL weight_only load error vs. a GLM5 issue with no demonstrated shared code path." + "reason": "Qwen3VL weight_only loading error is unrelated to GLM5." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "Generic loading failure vs. specific missing attribute in Molmo model internals." + "reason": "Different models and different errors: loading failure vs missing all_tied_weights_keys." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Missing model-class import vs. bitsandbytes Int8Params incompatibility." + "reason": "Import error for a missing model class is unrelated to the bitsandbytes storage incompatibility." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer loading regression vs. flash-attn fallback gating; different subsystems." + "reason": "Tokenizer load failure and flash-attn fallback gating are separate issues." }, { "left": "issue:45020", "right": "issue:45823", "accept": false, - "reason": "Remote_code compatibility breakage vs. Gemma4 device placement bug." + "reason": "Broad remote_code regressions and a Gemma4 device-map mismatch are different bugs." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES add-new-model-like failure vs. remote_code regression; not the same issue." + "reason": "A TOKENIZER_MAPPING_NAMES edge case is not the same as the broader remote_code regression." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "Both involve Qwen3 MoE/LoRA, but one is cross-version model usability and the other is a warning about tie_word_embeddings." + "reason": "Different LoRA/Qwen3-MoE compatibility complaints with no shared concrete failure path." }, { "left": "issue:43828", "right": "issue:45823", "accept": false, - "reason": "Dtype mismatch under autocast vs. device-map mismatch; different bugs." + "reason": "Autocast dtype mismatch and Gemma4 PLE device mismatch are unrelated." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Mistral regex patch crash vs. flash-attn2 fallback blocking; unrelated." + "reason": "Tokenizer regex crash and flash-attn fallback gating are separate loading paths." }, { "left": "issue:45715", "right": "issue:46315", "accept": false, - "reason": "Both are tokenizer performance regressions, but they affect opposite conversion directions and don't clearly share the same concrete defect." + "reason": "Both are tokenizer performance regressions, but they affect different conversion directions and different loops." }, { "left": "issue:45488", "right": "issue:46489", "accept": false, - "reason": "Both are tokenizer regressions in DeepSeek-related models, but the model families and code paths differ and the shared bug is not specific enough." + "reason": "Both are DeepSeek tokenizer regressions, but the concrete failures and implicated code paths are different enough to keep separate." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "Remote_code regression vs. Kimi-K2.5 tokenizer codec regression; different underlying failures." + "reason": "Kimi-K2.5 tokenizer codec/regression details do not match the generic remote_code breakage report." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative overall. The only soft accept, 45920/46489, is plausibly the same tokenizer regression family and is stated as a soft-edge match rather than a hard duplicate. The rest of the pairs are reasonably rejected as different failure modes." + "feedback": "The cluster summary is grounded in the packet: 45920 is indeed the broadest tokenizer-regression issue, and the accepted soft links stay within conservative same-bug/same-theme bounds. The rejected pairs are mostly justified by clearly different model stacks or failure modes. No obvious overstatement stands out." }, "evaluator_used": true, "retried": false, @@ -1477,7 +1522,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 5, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -1542,9 +1587,9 @@ "issue:42915|issue:43872", "issue:43792|issue:44220", "issue:42947|issue:45923", - "issue:43257|issue:45305", "issue:41553|issue:44991", "issue:36331|issue:41093", + "issue:43257|issue:45305", "issue:43940|issue:45313", "issue:43854|issue:44863", "issue:44928|issue:45588", @@ -1565,214 +1610,214 @@ "issue:43950|issue:45834", "issue:44560|issue:45685" ], - "prepared_review_unit_hash": "48f6ff24237dedd46430075020f511abe24f256226235c202d112bb530513605", + "prepared_review_unit_hash": "4f565990ec9c1c898172dc412bfc6cf6809d585291badaebdb39a31f22048722", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36674, - "estimated_input_tokens": 9169, - "estimated_eval_tokens": 18594 + "serialized_chars": 36679, + "estimated_input_tokens": 9170, + "estimated_eval_tokens": 18596 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous, but three duplicate subgroups stand out: the v5 docs/pipeline-removal reports, the `_is_hf_initialized` model-loading regression, and the generic mask-shape mismatch reports. The rest are too model- or subsystem-specific to merge.", - "confidence": 0.82, - "canonical_issue_reason": "No safe cluster-wide canonical: the items split into several unrelated subclusters (docs, tokenizer/loading, DeepSpeed, vision, audio, and model-specific bugs).", - "canonical_pr_reason": "No PRs in the cluster.", - "best_issue_reason": "If a single issue must anchor the closest duplicate subcluster, 44291 is the best choice: it states the `_is_hf_initialized` regression most clearly and directly overlaps with 43872.", - "best_pr_reason": "No PRs in the cluster.", + "summary": "Mostly heterogeneous issue set with a few true duplicate pairs; no single cluster-wide canonical issue/PR emerges.", + "confidence": 0.79, + "canonical_issue_reason": "No single canonical issue: the items split into several unrelated bug families, with only a few clear duplicate subgroups.", + "canonical_pr_reason": null, + "best_issue_reason": "No single best issue across the whole set; the strongest duplicate subgroups are separate and model-specific rather than one shared bug.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "Different problems: MobileLLM loading key error vs flash-attn fallback gating; same broad loading theme only." + "reason": "Different failures: tokenizer/model loading vs flash-attn fallback gating." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "Both involve loading/state handling, but one is a tied-weight bug in Mistral3 while the other is a non-persistent buffer regression; not the same code-path bug." + "reason": "Different bug classes: lm_head tying vs non-persistent buffer corruption." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Same docs regression: both say v5 docs still reference removed summarization/translation pipeline tasks." + "reason": "Same docs regression: both still reference removed v5 pipeline tasks." }, { "left": "issue:43872", "right": "issue:44291", "accept": true, - "reason": "Same underlying `_is_hf_initialized` compatibility bug during model initialization/loading; 44291 is the broader description." + "reason": "Same `_is_hf_initialized`/`Int8Params` incompatibility during empty-weight model loading." }, { "left": "issue:42915", "right": "issue:43872", "accept": false, - "reason": "Both are training/load failures, but FineGrainedFP8Config on Qwen3MoE is a different issue from the `_is_hf_initialized`/bitsandbytes regression." + "reason": "Qwen3-MoE FP8 failure is unrelated to the bitsandbytes init-arg incompatibility." }, { "left": "issue:43792", "right": "issue:44220", - "accept": false, - "reason": "Too little evidence that Whisper failing to run is the same root cause as the `_torch_extract_fbank_features()` issue; likely related but not clearly identical." + "accept": true, + "reason": "Both point to the Whisper/fbank feature extraction path failing." }, { "left": "issue:42947", "right": "issue:45923", "accept": false, - "reason": "Different gradient-checkpointing bugs: PEFT LoRA ineffectiveness vs missing supports_gradient_checkpointing on a trust_remote_code variant." - }, - { - "left": "issue:43257", - "right": "issue:45305", - "accept": false, - "reason": "Different DeepSpeed bugs: missing weight conversion for Qwen3 MoE vs gradient averaging with GAS and model_accepts_loss_kwargs." + "reason": "Gradient checkpointing behavior issue vs missing model flag; not the same concrete bug." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Different tokenizer-loading issues: Voxtral error messaging vs a concrete tokenizer load failure for EMBEDDIA/est-roberta." + "reason": "Different tokenizer-loading problems for different models." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Different failures: Trainer compute_loss signature mismatch vs tensor mask shape mismatch." + "reason": "Trainer kwarg mismatch is unrelated to the mask/tensor shape IndexError." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Both involve DeepSpeed but the failure modes are different: missing weights vs gradient averaging." }, { "left": "issue:43940", "right": "issue:45313", - "accept": false, - "reason": "Same general ZeRO-3 loading area, but different model variants and symptoms; not enough to say the same concrete bug." + "accept": true, + "reason": "Same ZeRO-3 weight-loading failure pattern in Qwen-family models with missing params." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "Different model-loading failures for different architectures; only the generic 'can't load' phrasing overlaps." + "reason": "Different model-loading failures for different architectures." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "Unrelated bugs: RLHF NaN explosion from position_ids/SDPA fallback vs flash_attention crash on s_aux=None." + "reason": "RLHF NaNs from position_id handling are unrelated to flash_attention.py crashing on s_aux=None." }, { "left": "issue:41093", "right": "issue:44805", "accept": true, - "reason": "Near-identical generic mask/index shape-mismatch reports; same underlying index/mask length bug class." + "reason": "Same mask-shape mismatch IndexError in tensor indexing, just with different reported sizes." }, { "left": "issue:44479", "right": "issue:44560", - "accept": false, - "reason": "Different video-input regressions affecting different Qwen variants and different failure modes." + "accept": true, + "reason": "Same v5.3.0 video-input regression family in Qwen VL models." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Different tokenizer-related problems; one is Voxtral error handling, the other is a model-specific load failure." + "reason": "A bad error message and a tokenizer load failure are not the same underlying bug." }, { "left": "issue:44863", "right": "issue:46402", "accept": false, - "reason": "Both are model-loading problems, but NemotronH checkpoint loading and Sam3 text-encoder loading are distinct issues." + "reason": "NemotronH checkpoint loading and Sam3Model text-encoder loading are unrelated." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "Completely different bugs: DeepSpeed training weight shape vs image processor void segmentation handling." + "reason": "DeepSpeed weight-dimension error is unrelated to segmentation-map preprocessing." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "Different subsystems and symptoms; paged attention cache / linear_attention is unrelated to NemotronH checkpoint loading." + "reason": "Qwen3.5 PagedAttention crash and NemotronH checkpoint loading are different problems." }, { "left": "issue:43596", "right": "issue:46515", "accept": false, - "reason": "Both mention loading/runtime failures, but ZeRO-3/BertModel index-out-of-bounds and torch 2.4.1 incompatibility are not the same bug." + "reason": "Specific ZeRO-3 BertModel index error is too different from a vague torch-version incompatibility report." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "Different domains: GGUF gpt-oss architecture support vs autocast dtype mismatch in Phi-tiny-MoE." + "reason": "GGUF architecture support and autocast dtype mismatch are unrelated." }, { "left": "issue:45325", "right": "issue:45902", "accept": false, - "reason": "Both concern rotary/position-related code, but the reported bugs are different models and different failure mechanisms." + "reason": "Different rotary-embedding issues with different root causes." }, { "left": "issue:43638", "right": "issue:46515", "accept": false, - "reason": "A specific ZeRO-3/BertModel index error is not the same as a broad torch-version incompatibility report." + "reason": "Non-pretrained Bert/ZeRO-3 index error is not the same as a generic torch incompatibility." }, { "left": "issue:45663", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma4 issues, but one is an FSDP2 shared_kv_states bug and the other is a device_map device mismatch; different code paths." + "reason": "FSDP2 shared_kv_states KeyError and PLE device mismatch are different Gemma4 bugs." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Different Qwen/GGUF issues: architecture support vs a sliding_window problem in Qwen3-MoE." + "reason": "GGUF support for gpt-oss is unrelated to Qwen3-MoE sliding_window behavior." }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "Both affect inference, but AudioFlamingo3 track leakage is unrelated to Whisper empty-transcription generation after align_special_tokens." + "reason": "Track-to-track embedding leak and empty-transcription generation failure are different audio bugs." }, { "left": "issue:43716", "right": "issue:44560", "accept": false, - "reason": "Different multimodal regressions: Mistral-3 image preprocessor dtype mismatch vs Qwen3-vl video StopIteration." + "reason": "Image-preprocessor dtype mismatch is unrelated to the Qwen3-VL video StopIteration regression." }, { "left": "issue:43950", "right": "issue:45834", "accept": false, - "reason": "Non-persistent buffer corruption during from_pretrained is unrelated to Kosmos2.5 long-ocr index errors." + "reason": "Non-persistent buffer corruption and Kosmos2.5 OCR index errors are unrelated." }, { "left": "issue:44560", "right": "issue:45685", "accept": false, - "reason": "Unrelated failures: Qwen3-vl video StopIteration vs MPS histogram_mps not implemented for Int." + "reason": "Video-processing regression and MPS histogram int support are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The three accepted duplicate pairs are well supported by the issue titles/bodies: docs v5 pipeline removals, the `_is_hf_initialized` loading regression, and the mask-shape mismatch reports. The rejected pairs are mostly appropriately cautious, and the cluster-wide non-canonical judgment fits the heterogeneous set of issues." + "feedback": "The summary is grounded in the packet and stays conservative overall. The rejected soft edges are mostly clearly distinct bug classes, and the accepted pairs are plausible duplicate families rather than overstatements. No single cluster-wide canonical issue emerges from these items." }, "evaluator_used": true, "retried": false, @@ -1785,7 +1830,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 6, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -1850,8 +1895,8 @@ "issue:42915|issue:43957", "issue:43550|issue:44910", "issue:44560|issue:44918", - "issue:45812|issue:46489", "issue:42617|issue:43366", + "issue:45812|issue:46489", "issue:45823|issue:46515", "issue:36010|issue:42915", "issue:44661|issue:45823", @@ -1874,227 +1919,251 @@ "issue:44960|issue:44977", "issue:42491|issue:45276", "issue:45507|issue:46468", - "issue:44488|issue:44991", "issue:43644|issue:43645", + "issue:44488|issue:44991", "issue:39692|issue:43994" ], - "prepared_review_unit_hash": "9b6c195d2fe7197efd014f2e9033e68be100bdf29b01b1ee656d9ec46dd89fb2", + "prepared_review_unit_hash": "66f93280ed0ba2437630870028f11e9c069ff3339f267a27e50997170f467a03", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37630, - "estimated_input_tokens": 9408, - "estimated_eval_tokens": 19072 + "serialized_chars": 37623, + "estimated_input_tokens": 9406, + "estimated_eval_tokens": 19068 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of Transformers v5-era regressions around model loading, initialization, tokenizers, and device/dtype handling. Only a couple of pairs look like true duplicates; most are related only at a broad subsystem level and should stay separate.", - "confidence": 0.61, - "canonical_issue_reason": "Issue 43646 is the best cluster anchor because it states the broadest and most central regression theme: Transformers 5.0.0 breaking custom model initialization. It covers the same family of initialization complaints more generally than the more specific notebook or flag-related reports.", + "summary": "The cluster is mostly heterogeneous. The strongest duplicate pair is 43644/44534 (near-exact same Transformers v5 buffer-corruption regression), and 43645/43646 also look like the same custom-model initialization bug. The SigLIP2 pair 38175/43994 is not conservatively duplicate from titles alone and should be rejected.", + "confidence": 0.88, + "canonical_issue_reason": "43644 is the clearest canonical issue: it states the broad Transformers v5 non-persistent-buffer regression and has a near-exact duplicate in 44534.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43646 is the strongest representative issue for this cluster: broad, clearly regression-focused, and likely to subsume nearby custom-model initialization reports without being too niche.", + "best_issue_reason": "43644 is the best representative because it captures the most reusable root-cause description and anchors the strongest duplicate pair in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Both are loading-related, but one is an import error for a specific Qwen2.5-VL class and the other is a meta-device loading failure across some models; different bugs." + "reason": "Different failures: a missing Qwen2.5-VL import vs meta-device loading breakage for other models." }, { "left": "issue:44683", "right": "issue:46515", "accept": false, - "reason": "Both mention torch/version compatibility, but one is a compiled flex_attention failure on torch>=2.9 and the other is a separate torch 2.4.1 incompatibility." + "reason": "Both involve torch, but the failure modes and affected code paths are unrelated." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "Unrelated problems: GGUF gpt-oss architecture support versus a Float8 storage lookup TypeError." + "reason": "GGUF/gpt-oss support and missing Float8 storage are different problems." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "These describe the same custom model initialization regression in Transformers 5.0.0; the notebook-specific report is a narrower manifestation of the broader issue." + "reason": "Both report the same custom-model initialization regression in Transformers 5.0.0; 43645 is just the notebook-specific wording of the same underlying bug." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "Different failures: Qwen3Moe FineGrainedFP8Config versus meta-device model loading." + "reason": "Qwen3Moe FP8 failures and meta-device loading failures are different underlying bugs." }, { "left": "issue:43550", "right": "issue:44910", "accept": false, - "reason": "Different model/code paths: Bamba torch.compile with SDPA versus Qwen3.5 flash-attention illegal-memory-access from packed-sequence misinterpretation." + "reason": "Different models and different attention/position-id failure modes." }, { "left": "issue:44560", "right": "issue:44918", "accept": false, - "reason": "Both involve Qwen models, but one is a video StopIteration error and the other is an input-embedding unpacking failure with TRL; not the same bug." + "reason": "Both involve Qwen3.x, but one is a video StopIteration error and the other is an embedding/unpacking trainer issue." }, { - "left": "issue:45812", - "right": "issue:46489", + "left": "issue:42617", + "right": "issue:43366", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and appear to be distinct code paths, not one shared bug." + "reason": "3d_parallel.py runtime failure is unrelated to GGUF gpt-oss support." }, { - "left": "issue:42617", - "right": "issue:43366", + "left": "issue:45812", + "right": "issue:46489", "accept": false, - "reason": "No overlap beyond general model/runtime failure; 3d_parallel.py execution is unrelated to GGUF gpt-oss support." + "reason": "Both are tokenizer regressions, but for different model families and likely different fixes." }, { "left": "issue:45823", "right": "issue:46515", "accept": false, - "reason": "A Gemma4 device-mismatch forward bug is unrelated to stale config dtype after from_config(dtype=...)." + "reason": "Gemma4 device mismatch and stale config dtype are not the same bug." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "Importing GenerationMixin and Qwen3Moe FP8 loading are unrelated issues." + "reason": "Importing GenerationMixin and Qwen3Moe FP8 loading are unrelated." }, { "left": "issue:44661", "right": "issue:45823", "accept": false, - "reason": "Tokenizer mapping validation and Gemma4 device placement are different subsystems and different failure modes." + "reason": "Tokenizer-mapping failure and Gemma4 PLE device mismatch are different issues." }, { "left": "issue:44560", "right": "issue:44805", "accept": false, - "reason": "Different errors: Qwen3-vl video StopIteration versus a generic mask/tensor shape mismatch." + "reason": "Video StopIteration and mask-shape IndexError are distinct failures." }, { "left": "issue:38175", "right": "issue:43994", - "accept": true, - "reason": "Same underlying SigLIP2 regression: both report broken outputs for google/siglip2-base-patch16-224, just with different symptom descriptions." + "accept": false, + "reason": "Both mention SigLIP2, but the evidence is too weak: one is about zero probabilities and the other about nonsensical outputs, which could be different symptoms or different root causes." }, { "left": "issue:43632", "right": "issue:45834", "accept": false, - "reason": "The _is_hf_initialized regression and Kosmos2.5 long-OCR index error are not the same code-path bug." + "reason": "_is_hf_initialized regression and Kosmos2.5 OCR index error are unrelated." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Both mention config validation, but the affected models and failure details differ; not a single duplicate bug." + "reason": "Both mention config validation, but they are different models and not clearly the same underlying bug." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Zero-3 Bert init failure and add-new-model-like tokenizer mapping failure are unrelated." + "reason": "Zero-3 BertModel indexing and add-new-model-like tokenizer mapping are unrelated." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "A DeepSpeed ZeRO-3 Bert index error is not the same as the tokenizer-mapping bug." + "reason": "Zero-3 BertModel indexing and tokenizer-mapping handling are different bugs." }, { "left": "issue:44661", "right": "issue:44805", "accept": false, - "reason": "Tokenizer-mapping validation and tensor mask shape mismatch are different issues." + "reason": "Tokenizer-mapping failure and mask-shape IndexError do not share a concrete code path." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "Device-map offload loading failure and Strix Halo mmap OOM are separate problems." + "reason": "gpu+cpu offload loading and Strix Halo mmap OOM are different memory/loading problems." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "These are the same non-persistent-buffer regression in Transformers v5, described with slightly different wording/version framing." + "reason": "Near-exact duplicate reports of the same Transformers v5 non-persistent-buffer junk-fill regression." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "Both concern SigLIP2, but one is an inference-output regression and the other is a broken documentation example with mismatched processor/model and quantization issues." + "reason": "SigLIP2 wrong outputs and a documentation example with model/processor mismatch are not the same report." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Qwen3_moe LoRA compatibility on hf5.x and BitNet packed-weight handling are different bugs." + "reason": "Qwen3 MoE LoRA incompatibility and BitNet packed-weight loading are unrelated." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Qwen3 auto device mapping on A800 and Qwen3Moe FP8 config failure are distinct regressions." + "reason": "Qwen3 auto-device-map CUDA assert and Qwen3Moe FP8 failure are different issues." }, { "left": "issue:46459", "right": "issue:46512", "accept": false, - "reason": "Both involve dtype handling, but one is composite-checkpoint loading ignoring dtype and the other is from_config leaving config.dtype stale." + "reason": "Composite-checkpoint dtype handling and stale config dtype after from_config are related in theme but not the same bug." }, { "left": "issue:42915", "right": "issue:44756", "accept": false, - "reason": "Qwen3Moe FP8 failure and mmap/OOM on Strix Halo are unrelated." + "reason": "FP8 config failure and mmap OOM are different code paths." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "ZeRO-3 loading for language_model and save_pretrained visual-encoder key regression are separate Qwen3.5 issues." + "reason": "One is ZeRO-3 weight loading for language_model, the other is a save_pretrained visual-encoder key regression." }, { "left": "issue:44451", "right": "issue:44488", "accept": false, - "reason": "Both are load failures, but they affect different Scandinavian BERT models and lack evidence of a single shared underlying bug." + "reason": "Both are model-loading failures, but for different models and without enough evidence of one root cause." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 and Qwen3.5 flash-attention generation failures are unrelated." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "LoRA training incompatibility and resize_token_embeddings behavior are different bugs." + }, + { + "left": "issue:45507", + "right": "issue:46468", + "accept": false, + "reason": "GraniteMoEHybrid invalid method calls and Nemotron example breakage are unrelated." }, { "left": "issue:43644", "right": "issue:43645", "accept": false, - "reason": "Non-persistent buffer corruption is unrelated to notebook custom model initialization." + "reason": "Related Transformers v5 regressions, but not the same concrete bug: buffer corruption vs custom model initialization." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Different model-loading regressions for different model families." }, { "left": "issue:39692", "right": "issue:43994", "accept": false, - "reason": "The SigLIP2 docs-example failure and the runtime nonsensical-output report are related only by model family, not the same bug." + "reason": "Documentation example errors are not the same as the SigLIP2 inference-output bug." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The cluster anchor choice is well grounded in the packet: issue 43646 is the broadest custom-model initialization regression, and the accepted soft edges are conservative duplicates with close title-level overlap (43645/43646, 43644/44534, 38175/43994). The rejected pairs are mostly clearly different bugs or at least not supported strongly enough to merge, so the overall summary stays precise and conservative." + "accept": false, + "feedback": "The summary is mostly grounded, but one soft-edge acceptance overreaches: issue:38175 vs issue:43994 are both about SigLIP2, yet the packet only gives titles and they describe different symptoms (zero probabilities vs nonsensical outputs). That is not enough evidence to conservatively call them a true duplicate/near-duplicate pair. The 43644/44534 pair is well supported, and 43645/43646 is plausible, but the SigLIP2 acceptance makes the overall verdict too strong." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -2104,7 +2173,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 7, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -2167,14 +2236,14 @@ "issue:43854|issue:43931", "issue:43257|issue:43854", "issue:43856|issue:44387", - "issue:45488|issue:45812", "issue:42491|issue:45292", "issue:44960|issue:46050", + "issue:45488|issue:45812", "issue:45362|issue:45406", "issue:42491|issue:43931", "issue:45276|issue:45335", - "issue:44910|issue:45588", "issue:44164|issue:44291", + "issue:44910|issue:45588", "issue:43404|issue:43454", "issue:41720|issue:44155", "issue:43425|issue:43957", @@ -2195,246 +2264,253 @@ "issue:45276|issue:45292", "issue:44863|issue:46468", "issue:45685|issue:46468", - "issue:43854|issue:45216" + "issue:43854|issue:45216", + "issue:44164|issue:44756" ], - "prepared_review_unit_hash": "d4085677a542a03b6c8e5e9f95f33c96275f72c14f8c6cf3a724313d5e4927b3", + "prepared_review_unit_hash": "7166774048116729263e3682822d60550b36b90ad49f56dc75f5963e9acd16d1", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 37963, - "estimated_input_tokens": 9491, - "estimated_eval_tokens": 19238 + "soft_pair_count": 35, + "serialized_chars": 38263, + "estimated_input_tokens": 9566, + "estimated_eval_tokens": 19388 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most candidate pairs are clearly different bugs or model-specific regressions and should not be merged. The only strong duplicate-looking pair is the resize_token_embeddings/output_embeddings regression between 45276 and 45292, with 45292 as the better canonical report.", - "confidence": 0.82, - "canonical_issue_reason": "Issue 45292 is the cleanest canonical issue: it states the underlying resize_token_embeddings/output_embeddings bug in a generic way, while 45276 looks like a Gemma4-specific manifestation of the same symptom.", - "canonical_pr_reason": null, - "best_issue_reason": "Issue 45292 is the best representative issue for this cluster because it is broad, concise, and directly describes the core regression without extra model-specific noise.", - "best_pr_reason": null, + "summary": "The set is mostly a collection of unrelated bug reports. The only strong duplicate subcluster is around `resize_token_embeddings` failing to update model-specific embedding/output weights across Gemma4/T5Gemma variants.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 45292 is the broadest, most general statement of the `resize_token_embeddings` regression (`output_embeddings`), making it the best canonical representative of that duplicate subcluster.", + "canonical_pr_reason": "No pull requests were provided in the input, so there is no canonical PR candidate.", + "best_issue_reason": "Issue 45292 best captures the shared symptom at the highest level and can represent the accepted `resize_token_embeddings` duplicates.", + "best_pr_reason": "No pull requests were provided in the input, so there is no global best PR candidate.", "soft_edge_verdicts": [ { "left": "issue:43708", "right": "issue:44075", "accept": false, - "reason": "Different subsystems: trainer checkpoint resumption/max_steps vs optimizer SGD argument handling." + "reason": "Different trainer resume-step calculation bug vs optimizer argument handling; unrelated code paths." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Both mention memory, but one is Qwen3 MoE training memory usage and the other is quantization/offloading behavior; not the same bug." + "reason": "One is Qwen3 MoE training memory use, the other is quantization/offloading behavior; not the same bug." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Different models and failures: GLM-4.7-Flash load test issue vs Qwen3-VL shape mismatch." + "reason": "Different model-loading failures for different models and symptoms; no shared concrete root cause visible from titles." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Qwen3 MoE accelerate/deepspeed weight conversion is unrelated to GLM-4.7-Flash loading in unit tests." + "reason": "Qwen3 MoE accelerate/deepspeed conversion issue vs GLM-4.7 loading problem; unrelated." }, { "left": "issue:43856", "right": "issue:44387", "accept": false, - "reason": "Both involve memory, but one is training memory inefficiency and the other is int4 quantization reserved-memory OOM; not the same concrete bug." - }, - { - "left": "issue:45488", - "right": "issue:45812", - "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and appear to have different root causes." + "reason": "Both mention memory, but one is Qwen3 MoE training usage and the other is int4 reserved-memory/OOM regression; not the same issue." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "LoRA qwen3_moe compatibility on hf5.x is unrelated to resize_token_embeddings not updating output embeddings." + "reason": "LoRA checkpoint compatibility on hf5.x vs `resize_token_embeddings` output embedding bug; different failures." }, { "left": "issue:44960", "right": "issue:46050", "accept": false, - "reason": "GLM5 model loading and sinq quantized save/load corruption are different problems." + "reason": "GLM5 loading/setup issue vs sinq quantized save/load regression; different code paths." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different model families and appear to have different root causes." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Qwen3.5 chat crash and Gemma4Processor missing _tokenizer are unrelated model/runtime failures." + "reason": "Qwen3.5 chat crash vs Gemma4Processor missing `_tokenizer`; unrelated." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "Different failure modes and models; one is MoE LoRA compatibility, the other is checkpoint shape mismatch on Qwen3-VL." + "reason": "Same broad area of model loading, but different models, symptoms, and likely causes." }, { "left": "issue:45276", "right": "issue:45335", - "accept": false, - "reason": "Similar resize_token_embeddings symptom, but different model families and likely separate implementations; not enough evidence of one underlying bug." + "accept": true, + "reason": "Both report `resize_token_embeddings` failing to propagate to model-specific embeddings/output layers; same regression family and plausibly one fix." }, { - "left": "issue:44910", - "right": "issue:45588", + "left": "issue:44164", + "right": "issue:44291", "accept": false, - "reason": "Different flash-attention failures with different triggers: packed-sequence misinterpretation vs s_aux=None AttributeError." + "reason": "Saving/from_pretrained extra_state handling vs `init_empty_weights` unexpected argument; different lifecycle bugs." }, { - "left": "issue:44164", - "right": "issue:44291", + "left": "issue:44910", + "right": "issue:45588", "accept": false, - "reason": "save/from_pretrained extra_state handling is unrelated to init_empty_weights/_is_hf_initialized constructor breakage." + "reason": "Flash-attention crash from 3D `position_ids` vs sink-less `s_aux=None` AttributeError; different failure modes." }, { "left": "issue:43404", "right": "issue:43454", "accept": false, - "reason": "Both concern lm_head tying, but they are different model classes and not clearly the same code-path bug." + "reason": "Both involve tying `lm_head`, but they are different model classes and not clearly the same underlying bug." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Qwen3 device-mapping cuda assert is unrelated to AudioFlamingo3 batched inference token leakage." + "reason": "Qwen3 device-map CUDA assert vs AudioFlamingo3 batched inference token leak; unrelated." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Torch version incompatibility is unrelated to meta-device model loading breakage." + "reason": "Torch version compatibility vs `torch.device('meta')` model-loading breakage; different problems." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Tokenizer vocab-size mismatch and a separate model-loading failure are not the same bug." + "reason": "Tokenizer vocab-size mismatch vs a specific model load failure; not enough evidence of one root cause." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "SigLIP2 doc-example errors are unrelated to GLM5 loading." + "reason": "Documentation example errors vs GLM5 model issue; unrelated." }, { "left": "issue:42947", "right": "issue:45704", "accept": false, - "reason": "Gradient checkpointing/LoRA behavior is unrelated to the apex FusedRMSNorm memory leak." + "reason": "Gradient checkpointing/PEFT behavior vs apex RMSNorm memory leak; different subsystems." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Both are memory-related, but one is offloading/quantization behavior and the other is int4 reserved-memory growth causing OOM." + "reason": "Quantization/offloading expectation mismatch vs CUDA reserved-memory OOM regression; related theme but not the same bug." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "ImportError for GenerationMixin and Whisper return_language pipeline regression are different issues." + "reason": "Import error for `GenerationMixin` vs Whisper pipeline `return_language`; different issues." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Gemma4 FSDP2 attention KeyError and Qwen multimodal config unwrapping are different model/config bugs." + "reason": "Gemma4 FSDP2 attention state bug vs Qwen multimodal `text_config` unwrapping bug; unrelated." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "init_empty_weights argument handling and disabling mmap on Strix Halo are unrelated." + "reason": "`init_empty_weights` argument handling vs mmap/OOM on Strix Halo; different code paths." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Gemma4 FSDP2 KeyError is unrelated to Qwen3.5 MoE CPU-offload incompatibility." + "reason": "Gemma4 FSDP2 state rebuild bug vs Qwen3_5Moe rotary embedding CPU-offload incompatibility; unrelated." }, { "left": "issue:44155", "right": "issue:45823", "accept": false, - "reason": "AudioFlamingo3 batched inference leakage is unrelated to Gemma4 device mismatch under device_map=auto." + "reason": "AudioFlamingo3 cross-track leakage vs Gemma4 PLE device mismatch; not the same bug." }, { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Qwen-3-8B-NVFP4 running error and git-oss-20b MXFP4 fallback are different quantization/model issues." + "reason": "Different quantized model-loading failures with different models and behavior." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Different OOM causes: int4 quantization reserved memory vs mmap on Strix Halo." + "reason": "Both can OOM, but one is an int4 memory regression and the other is mmap-related Strix Halo handling; distinct causes." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading regressions and Strix Halo mmap OOM are not the same bug." + "reason": "`meta` device loading regression vs mmap/OOM workaround; unrelated." }, { "left": "issue:43335", "right": "issue:45663", "accept": false, - "reason": "SwitchTransformers sparse-layer construction bug is unrelated to Gemma4 FSDP2 attention state handling." + "reason": "SwitchTransformers sparse-layer creation bug vs Gemma4 attention state bug; different models and failures." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Gradient checkpointing ineffective with PEFT LoRA is not the same as checkpointing being disallowed in eval mode." + "reason": "Both mention gradient checkpointing, but one is ineffectiveness with PEFT LoRA and the other is eval-mode restriction; not duplicates." }, { "left": "issue:45276", "right": "issue:45292", "accept": true, - "reason": "These describe the same resize_token_embeddings regression; 45276 is a Gemma4-specific report while 45292 states the general output_embeddings failure more cleanly." + "reason": "Same `resize_token_embeddings` regression: model-specific embeddings/output weights are not updated after resize." }, { "left": "issue:44863", "right": "issue:46468", "accept": false, - "reason": "Both mention Nemotron, but one is a checkpoint-loading implementation bug and the other is an official example failure; not enough evidence they are the same concrete defect." + "reason": "NemotronH checkpoint loading support vs a broken official example; too different." }, { "left": "issue:45685", "right": "issue:46468", "accept": false, - "reason": "MPS histogram_mps type error is unrelated to the Nemotron example issue." + "reason": "MPS MoE histogram dtype error vs Nemotron example breakage; unrelated." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "GLM-4.7-Flash loading in tests and Qwen3.5 save_pretrained regression are different bugs." + "reason": "Different Qwen/GLM load-vs-save regressions with no clear shared root cause." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "save/from_pretrained extra_state handling vs mmap/OOM workaround; different issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet, and the only accepted soft-edge pair (45276/45292) is supported by very strong title similarity while remaining reasonably conservative. The other soft-edge rejections are cautious and do not overstate equivalence. No mergeability issues apply here." + "feedback": "The summary is grounded in the packet, and the only accepted soft-edge pair is supported by the titles as the same `resize_token_embeddings` regression family. The rejects are generally conservative and do not overclaim shared root cause. No PRs were provided, so the PR fields are appropriately null." }, "evaluator_used": true, "retried": false, @@ -2447,7 +2523,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 8, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -2505,18 +2581,17 @@ ], "soft_pairs": [ "issue:42796|issue:46326", - "issue:44164|issue:44756", - "issue:43381|issue:44928", "issue:44479|issue:44877", + "issue:43381|issue:44928", "issue:45684|issue:45698", "issue:45588|issue:45663", "issue:43646|issue:43708", "issue:45292|issue:45335", "issue:43645|issue:44164", "issue:43828|issue:43957", - "issue:46315|issue:46491", "issue:44402|issue:44451", "issue:43425|issue:43828", + "issue:46315|issue:46491", "issue:45812|issue:46315", "issue:44589|issue:44756", "issue:44683|issue:45823", @@ -2526,10 +2601,10 @@ "issue:43425|issue:43716", "issue:44530|issue:45084", "issue:43906|issue:44530", - "issue:44265|issue:44898", "issue:44479|issue:44805", - "issue:39404|issue:44912", + "issue:44265|issue:44898", "issue:44589|issue:45084", + "issue:39404|issue:44912", "issue:43901|issue:44509", "issue:43425|issue:44918", "issue:43906|issue:46326", @@ -2537,235 +2612,229 @@ "issue:43994|issue:44079", "issue:43299|issue:43931" ], - "prepared_review_unit_hash": "60221d58b6508d8fa7aab7775a156fe354fe1dd81b73b91960702400c88f1a04", + "prepared_review_unit_hash": "e8e46bf3fc84a19ecce1eab7bde00e793d1f0a8183915214f622864b090fda4e", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36785, - "estimated_input_tokens": 9197, - "estimated_eval_tokens": 18650 + "soft_pair_count": 31, + "serialized_chars": 36482, + "estimated_input_tokens": 9121, + "estimated_eval_tokens": 18498 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of separate regressions, not one tight duplicate set. The strongest linkage is the `resize_token_embeddings` pair (`45292`/`45335`), which looks like the best-supported likely same-bug manifestation, but the packet does not explicitly prove duplication.", - "confidence": 0.69, - "canonical_issue_reason": "`issue:45292` is the best canonical issue because it states the generic `resize_token_embeddings`/output-embedding failure in the broadest form, while `issue:45335` reads as a model-specific report of the same behavior.", + "summary": "Most pairs are related only at a broad subsystem level (tokenizers, model loading, save/load, or training) and do not look like the same concrete bug. The only clear duplicate-like pair is the resize_token_embeddings regression across generic and t5gemma-specific embeddings.", + "confidence": 0.82, + "canonical_issue_reason": "issue:45292 is the broader representation of the resize_token_embeddings regression; issue:45335 looks like the t5gemma-specific manifestation of the same missing embedding-resize behavior.", "canonical_pr_reason": null, - "best_issue_reason": "`issue:45292` is the strongest anchor for the cluster: it captures the underlying `resize_token_embeddings` bug generically and is the most reusable representative for deduping.", + "best_issue_reason": "issue:45292 is the best issue representative here because it states the underlying API regression generically and can subsume the t5gemma variant.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42796", "right": "issue:46326", "accept": false, - "reason": "Different tokenizer bugs: BPE marker handling vs `continue_final_message` behavior." + "reason": "Both are tokenizer-related, but one is BPE marker handling and the other is continue_final_message behavior for mistral_common; different bugs." }, { - "left": "issue:44164", - "right": "issue:44756", + "left": "issue:44479", + "right": "issue:44877", "accept": false, - "reason": "Unrelated failures: save/load extra-state handling vs mmap/OOM on Strix Halo." + "reason": "Different model families and failure modes; video regression vs strict config loading." }, { "left": "issue:43381", "right": "issue:44928", "accept": false, - "reason": "Different code paths and symptoms: eval-mode gradient checkpointing vs RLHF NaN from 3D `position_ids`/SDPA fallback." - }, - { - "left": "issue:44479", - "right": "issue:44877", - "accept": false, - "reason": "Different subsystems and failures: Qwen VL video regression vs strict config loading for `granite_speech`." + "reason": "One is gradient checkpointing in eval mode, the other is Qwen3.5 RLHF NaN from 3D position_ids and SDPA fallback; unrelated." }, { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Both concern custom-model save/load, but the concrete bugs differ: permissions propagation vs loading the wrong custom module." + "reason": "Both touch save/load of custom models, but one is read-only permission propagation and the other is wrong module selection after save_pretrained; distinct bugs." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Different code paths: flash-attention sinkless-model crash vs Gemma4 FSDP2 shared-state rebuild issue." + "reason": "Different attention-path failures: flash_attention s_aux=None crash vs Gemma4 FSDP2 shared_kv_states KeyError." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model init in notebooks is unrelated to the trainer `resume_from_checkpoint` batch-size/max_steps bug." + "reason": "Custom model init in notebooks is unrelated to Trainer resume_from_checkpoint max_steps calculation." }, { "left": "issue:45292", "right": "issue:45335", "accept": true, - "reason": "Strong title and symptom match: both report `resize_token_embeddings` failing to update embeddings, with `45335` appearing to be a `t5gemma`-specific manifestation of the same bug." + "reason": "Same underlying regression: resize_token_embeddings fails to propagate to output embeddings/decoder.embed_tokens; the second is a model-specific instance." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Notebook-defined custom-model init and `extra_state` save/load handling are separate failures." + "reason": "Both are about custom-model workflows, but notebook initialization and extra_state save/from_pretrained handling are different issues." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "Both are dtype/device mismatch reports, but they affect different workflows and code paths." - }, - { - "left": "issue:46315", - "right": "issue:46491", - "accept": false, - "reason": "Both are tokenizer regressions, but one is a performance issue and the other is an encode/decode spacing bug." + "reason": "Both are dtype-mismatch reports, but one is autocast on Phi-tiny-MoE and the other is meta-device loading on different models." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Tokenizer vocab-size inconsistency and generic model-loading failure are not shown to be the same defect." + "reason": "Related to tokenizer/model loading, but pad/unk vocab-size mismatch is not the same as ScandiBERT load failure." }, { "left": "issue:43425", "right": "issue:43828", "accept": false, - "reason": "Torch version incompatibility is unrelated to a specific autocast dtype mismatch." + "reason": "Torch 2.10 compatibility is a version-level issue; Phi-tiny-MoE autocast dtype mismatch is a model-specific runtime bug." + }, + { + "left": "issue:46315", + "right": "issue:46491", + "accept": false, + "reason": "Both are tokenizer regressions, but one is performance of convert_tokens_to_ids and the other is a leading-space round-trip correctness bug." }, { "left": "issue:45812", "right": "issue:46315", "accept": false, - "reason": "Wrong token IDs for Granite models is unrelated to slow `convert_tokens_to_ids` behavior." + "reason": "Granite token ID corruption is a correctness regression; convert_tokens_to_ids slowdown is a performance regression." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Missing Float8 storage object and Strix Halo mmap/OOM are unrelated failures." + "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are unrelated." }, { "left": "issue:44683", "right": "issue:45823", "accept": false, - "reason": "Compiled `flex_attention` on torch>=2.9 and Gemma4 device mismatch are different bugs." + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device-map mismatch are different code paths and symptoms." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper pipeline `return_language` regression is unrelated to tokenization-version changes." + "reason": "Whisper pipeline return_language and a generic tokenization-version complaint do not describe the same bug." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "Both involve Transformers 5 custom-model loading, but they target different concrete failures." + "reason": "base_model_prefix loading failure and notebook custom-model initialization are separate custom-model regressions." }, { "left": "issue:43425", "right": "issue:45823", "accept": false, - "reason": "Torch compatibility is unrelated to a Gemma4 forward device mismatch." + "reason": "Torch version incompatibility is not the same as Gemma4 PLE device mismatch with device_map=auto." }, { "left": "issue:43425", "right": "issue:43716", "accept": false, - "reason": "Torch version incompatibility is unrelated to a Mistral-3 dtype mismatch." + "reason": "Torch 2.10 incompatibility and Mistral-3 dtype mismatch with the image preprocessor are unrelated." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "Different failures: Qwen3.5 PagedAttentionCache crash vs compile error on non-template nodes." + "reason": "PagedAttentionCache linear_attention crash is unrelated to the non-template-node compilation error." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "The repro-to-upstream link is too indirect; the titles point to different concrete failures." - }, - { - "left": "issue:44265", - "right": "issue:44898", - "accept": false, - "reason": "`torch.export.export`/`torch_compilable_check` failure is unrelated to the Perceiver interpolation failure." + "reason": "An isolated reproduction of another issue is not the same as the Qwen3.5 PagedAttentionCache crash." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "VL video-input regression is unrelated to a generic mask-shape `IndexError`." + "reason": "Video-input regression for Qwen VL models and a generic mask/tensor shape mismatch are different failures." }, { - "left": "issue:39404", - "right": "issue:44912", + "left": "issue:44265", + "right": "issue:44898", "accept": false, - "reason": "Whisper pipeline language-return regression and MXFP4 quantization fallback are unrelated." + "reason": "torch.export/torch_compilable_check failure and Perceiver interpolate_pos_encoding bug are distinct model/export problems." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Float8 storage lookup failure is unrelated to compile-time non-template-node errors." + "reason": "Float8 storage lookup and template-node compilation errors are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper return_language pipeline regression is unrelated to MXFP4 quantization fallback on git-oss-20b." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are docs mismatches, but they cover different APIs and removed tasks." + "reason": "Docs mentioning removed pipelines and return_all_scores behavior mismatch are related to docs/API drift but not the same concrete issue." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch version incompatibility is unrelated to Qwen3.5 SFT embedding unpacking." + "reason": "Torch 2.10 incompatibility is unrelated to Qwen3.5 embedding unpacking with TRL SFT trainer." }, { "left": "issue:43906", "right": "issue:46326", "accept": false, - "reason": "One is an isolated repro of a prior issue; the other is a tokenizer continuation bug." + "reason": "An isolated reproduction of another issue and a mistral_common continue_final_message tokenizer bug are not the same failure." }, { "left": "issue:46050", "right": "issue:46326", "accept": false, - "reason": "Quantized model save/load breakage is unrelated to tokenizer `continue_final_message` behavior." + "reason": "sinq save/load corruption and continue_final_message tokenization are unrelated bugs." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "Siglip2 pipeline outputs and `ModelOutput` key assignment are different bugs." + "reason": "Siglip2 pipeline nonsense outputs and ModelOutput key assignment are different problems." }, { "left": "issue:43299", "right": "issue:43931", "accept": false, - "reason": "Both involve Qwen3-VL loading, but the reports do not show the same concrete defect." + "reason": "Both are Qwen3-VL loading issues, but one targets MoE loading broadly and the other is a specific weight-shape mismatch; not clearly the same concrete bug." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The cluster summary is mostly grounded, but it overstates the evidence for the 45292/45335 pair by calling it a \u201cconfirmed duplicate\u201d / \u201csame underlying bug.\u201d The packet does not explicitly confirm duplication; it only shows strong title similarity and one high-jaccard soft pair. Please soften that language to something like \u201cbest-supported likely duplicate pair\u201d or \u201cmost probable same-bug manifestation.\u201d" + "accept": true, + "feedback": "The mapping is grounded in the packet, and the soft-edge decisions are conservative overall. The only accepted pair, issue:45292 vs issue:45335, is a plausible same-bug duplicate around resize_token_embeddings not propagating to the relevant embedding modules. The other pairs are appropriately rejected as broader subsystem similarity rather than the same concrete bug." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -2775,7 +2844,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 9, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -2805,9 +2874,8 @@ "issue:43646", "issue:43716", "issue:43856", + "issue:43883", "issue:44164", - "issue:44206", - "issue:44220", "issue:44451", "issue:44462", "issue:44479", @@ -2835,250 +2903,250 @@ "soft_pairs": [ "issue:36032|issue:43452", "issue:44671|issue:44991", - "issue:43596|issue:44683", "issue:44164|issue:45759", + "issue:43596|issue:44683", "issue:44991|issue:45081", "issue:43856|issue:44756", - "issue:44462|issue:46489", "issue:42796|issue:43404", "issue:44164|issue:44479", + "issue:44462|issue:46489", "issue:39404|issue:44671", "issue:44488|issue:45081", - "issue:44488|issue:44987", "issue:44661|issue:44683", + "issue:44488|issue:44987", "issue:44737|issue:45859", "issue:43611|issue:43644", "issue:32090|issue:39404", "issue:43582|issue:43632", "issue:46285|issue:46402", "issue:43716|issue:45084", - "issue:40990|issue:43421", "issue:36032|issue:42915", + "issue:40990|issue:43421", "issue:43643|issue:44530", "issue:34567|issue:40444", "issue:44451|issue:44987", "issue:42175|issue:46178", - "issue:44811|issue:45584", "issue:43541|issue:45084", + "issue:44811|issue:45584", "issue:42915|issue:44530", - "issue:40444|issue:46326", "issue:43644|issue:43646", - "issue:44206|issue:44220" + "issue:40444|issue:46326", + "issue:43883|issue:44488" ], - "prepared_review_unit_hash": "c7fe422564f1dbc44e4fba4b20c4ebbd4f6789498d5209a90dd8c7ec13927056", + "prepared_review_unit_hash": "37fab4528c5005152a0c48e6e95e2907038065d33a74d6f75375167ff48bf1bc", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 37016, - "estimated_input_tokens": 9254, - "estimated_eval_tokens": 18764 + "serialized_chars": 36474, + "estimated_input_tokens": 9119, + "estimated_eval_tokens": 18494 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No duplicate cluster here: the items are mostly unrelated issue reports across different regressions, models, and code paths. The soft edges are broad thematic similarities only, not the same underlying bug.", - "confidence": 0.94, - "canonical_issue_reason": "No single canonical issue is safe for this set; the issues span distinct bugs and subsystems, so picking one representative would be misleading.", + "summary": "This cluster is thematically related to Transformers v5 regressions, but the individual issues point to different concrete bugs across tokenizers, model loading, multimodal paths, and training/runtime edge cases. I did not find any soft pair that is clearly the same underlying bug.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43646 is the best representative issue because it describes a broad Transformers v5 initialization regression and sits closest to the cluster\u2019s recurring v5 breakage theme.", "canonical_pr_reason": null, - "best_issue_reason": "No best issue candidate stands out as a global representative because the cluster is heterogeneous rather than a true duplicate set.", + "best_issue_reason": "issue:43646 is the strongest canonical issue candidate: broad, version-specific, and more representative than the narrower tokenizer/model-specific reports in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "Both involve tokenizer/model loading, but one is a T5 tokenizer method conflict and the other is gguf loading behavior; different failure modes and fixes." + "reason": "Different failures: a T5 tokenizer method-name conflict vs gguf_file loading breakage for AutoTokenizer/AutoModelForCausalLM." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-related regressions, but one is incorrect masked-LM predictions in v5 and the other is loading a specific tokenizer; not the same bug." + "reason": "Both are tokenizer/model loading complaints, but one is CamemBERT MLM prediction correctness and the other is loading EMBEDDIA/est-roberta tokenizer." }, { - "left": "issue:43596", - "right": "issue:44683", + "left": "issue:44164", + "right": "issue:45759", "accept": false, - "reason": "Completely different code paths: DeepSpeed ZeRO-3/BertModel initialization vs compiled flex_attention on newer torch." + "reason": "Unrelated code paths: extra_state save/load handling vs AutoModelForCausalLM.from_config unwrapping composite Qwen text_config." }, { - "left": "issue:44164", - "right": "issue:45759", + "left": "issue:43596", + "right": "issue:44683", "accept": false, - "reason": "Save/from_pretrained extra_state handling and AutoModelForCausalLM.from_config text_config unwrapping are separate serialization/config issues." + "reason": "Different subsystems: DeepSpeed ZeRO-3/BertModel init error vs compiled flex_attention failure on torch >= 2.9." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Both affect tokenizer loading, but one is a generic tokenizer load regression and the other is a Mistral regex patch crash; different root causes." + "reason": "Different tokenizer bugs: loading a tokenizer for EMBEDDIA/est-roberta vs Mistral regex patch crashing on tokenizer backend access." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "Memory usage in Qwen3 MoE training is unrelated to mmap OOM on Strix Halo." - }, - { - "left": "issue:44462", - "right": "issue:46489", - "accept": false, - "reason": "Both mention tokenizer loading, but one ignores tokenizer.json while the other is a DeepSeek-Coder output regression tied to a previous fix gap." + "reason": "Different performance/memory issues in unrelated contexts: Qwen3 MoE training memory use vs disabling mmap on Strix Halo to avoid OOM." }, { "left": "issue:42796", "right": "issue:43404", "accept": false, - "reason": "Tokenizer BPE marker handling and lm_head weight tying in Mistral3 are distinct problems." + "reason": "One is a tokenizer BPE-marker handling bug; the other is a missing lm_head weight tie in Mistral3 multimodal generation." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "extra_state save/load issues are unrelated to the Qwen2.5/Qwen3.x video-input regression." + "reason": "save/from_pretrained extra_state handling is unrelated to the Qwen video-input regression." + }, + { + "left": "issue:44462", + "right": "issue:46489", + "accept": false, + "reason": "Both involve tokenizer loading/output, but one is ignoring tokenizer.json in AutoTokenizer while the other is a DeepSeek-Coder v1 output regression with a specific fix gap." }, { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper return_language/pipeline behavior is unrelated to CamemBERT masked-LM prediction drift." + "reason": "Whisper pipeline return_language regression is unrelated to CamemBERT masked LM prediction errors." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Both concern tokenizer loading, but one is a specific model failing to load and the other is a Mistral regex patch attribute error." + "reason": "Both are model-loading related, but one is a repository-loading failure for cjvt/sleng-bert and the other is a Mistral regex patch crash." }, { - "left": "issue:44488", - "right": "issue:44987", + "left": "issue:44661", + "right": "issue:44683", "accept": false, - "reason": "Different model-specific loading failures; no evidence they stem from the same underlying tokenizer bug." + "reason": "Different bugs: add-new-model-like failure in tokenizer mapping vs flex_attention compilation failure." }, { - "left": "issue:44661", - "right": "issue:44683", + "left": "issue:44488", + "right": "issue:44987", "accept": false, - "reason": "Add-new-model-like mapping logic is unrelated to compiled flex_attention failures on torch >= 2.9." + "reason": "Both are loading failures, but the affected models and failure modes are different; not the same underlying bug." }, { "left": "issue:44737", "right": "issue:45859", "accept": false, - "reason": "XLNet CPU placement bug and Qwen3_5Moe CPU offload incompatibility are separate device-handling issues in different models." + "reason": "XLNet CPU placement bug in relative positional encoding vs Qwen3_5 MoE rotary embedding incompatibility with CPU offload." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Both are v5 regressions, but base_model_prefix loading breaks and non-persistent buffer junk are different internal bugs." + "reason": "Both are Transformers v5 regressions, but one affects base_model_prefix loading and the other non-persistent buffer initialization; different concrete bugs." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer GPU broadcast NoneType error and Whisper return_language regression are unrelated subsystems." + "reason": "Trainer GPU broadcast NoneType error is unrelated to Whisper pipeline return_language behavior." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and _is_hf_initialized flag breakage do not share a concrete code path." + "reason": "Apple Silicon caching_allocator_warmup TypeError is unrelated to the _is_hf_initialized flag regression." }, { "left": "issue:46285", "right": "issue:46402", "accept": false, - "reason": "CLIP structure inconsistency and Sam3Model text encoder weight loading are separate model-architecture regressions." + "reason": "Different model families and issues: CLIP structure inconsistency vs Sam3Model text encoder weights not loading." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Image preprocessor/model dtype mismatch is unrelated to template compilation errors." + "reason": "Mistral-3 image preprocessor dtype mismatch is unrelated to the template-node compilation TypeError." }, { - "left": "issue:40990", - "right": "issue:43421", + "left": "issue:36032", + "right": "issue:42915", "accept": false, - "reason": "High perplexity on gpt-oss-20b is not the same bug as runtime special-token post-processor updates in TokenizersBackend." + "reason": "T5 tokenizer special-token conflict and Qwen3 MoE FineGrainedFP8Config failure are unrelated." }, { - "left": "issue:36032", - "right": "issue:42915", + "left": "issue:40990", + "right": "issue:43421", "accept": false, - "reason": "T5 tokenizer method conflict and Qwen3Moe FineGrainedFP8Config failure are different model/tokenizer issues." + "reason": "High perplexity on gpt-oss-20b is not the same bug as runtime post-processor updates when special tokens change." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "trust_remote_code AutoConfig field loss is unrelated to PagedAttentionCache's linear_attention group-type crash." + "reason": "trust_remote_code missing fields is unrelated to PagedAttentionCache crashing on invalid group type." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "TrainerState token counting and Qwen2.5-VL iterable multi-image finetuning failures are separate training concerns." + "reason": "TrainerState token counting bug and Qwen2.5-VL iterable dataset multi-image finetuning failure are different problems." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Loading a specific tokenizer repository and transformers>=5.1.0 loading a different model are not the same failure." + "reason": "Both are model-loading complaints, but they concern different models and likely different loader paths." }, { "left": "issue:42175", "right": "issue:46178", "accept": false, - "reason": "Both mention TensorFlow backend availability, but one is a packaging extra omission and the other is an invalid BACKENDS_MAPPING backend error." + "reason": "Packaging/backend installation issue is different from the BACKENDS_MAPPING error for tensorflow_text." }, { - "left": "issue:44811", - "right": "issue:45584", + "left": "issue:43541", + "right": "issue:45084", "accept": false, - "reason": "Whisper batch_decode skip_special_tokens handling and empty transcription after align_special_tokens are related only by model area, not the same bug." + "reason": "torch dynamo tracing failure in MixtralForCausalLM is unrelated to the non-template-node compilation error." }, { - "left": "issue:43541", - "right": "issue:45084", + "left": "issue:44811", + "right": "issue:45584", "accept": false, - "reason": "Mixtral grouped_mm tracing failure and template-node compilation error are unrelated runtime/compiler issues." + "reason": "Whisper batch_decode skip_special_tokens behavior is different from empty-transcription generation after align_special_tokens." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "Qwen3Moe FP8 config failure and Qwen3.5 PagedAttentionCache group-type crash are distinct model/runtime bugs." + "reason": "Qwen3 MoE training with FineGrainedFP8Config is unrelated to PagedAttentionCache group-type crashes in Qwen3.5 models." }, { - "left": "issue:40444", - "right": "issue:46326", + "left": "issue:43644", + "right": "issue:43646", "accept": false, - "reason": "Qwen2.5-VL multi-image finetuning and mistral_common continue_final_message behavior are different multimodal/tokenizer issues." + "reason": "Both are Transformers v5 regressions, but one is junk-filled non-persistent buffers and the other is custom model initialization; not the same concrete bug." }, { - "left": "issue:43644", - "right": "issue:43646", + "left": "issue:40444", + "right": "issue:46326", "accept": false, - "reason": "Non-persistent buffer junk on load and custom model initialization breakage are separate v5 regressions." + "reason": "Qwen2.5-VL iterable-dataset multi-image finetuning failure is unrelated to mistral_common continue_final_message behavior." }, { - "left": "issue:44206", - "right": "issue:44220", + "left": "issue:43883", + "right": "issue:44488", "accept": false, - "reason": "Unsupported center arg in LastFeatureExtractor and _torch_extract_fbank_features issues are related to audio preprocessing broadly, but the concrete bugs differ." + "reason": "Molmo all_tied_weights_keys attribute error is unrelated to loading cjvt/sleng-bert." } ] }, @@ -3094,7 +3162,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 10, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -3124,6 +3192,7 @@ "issue:43883", "issue:43994", "issue:44062", + "issue:44206", "issue:44220", "issue:44265", "issue:44291", @@ -3132,7 +3201,6 @@ "issue:44464", "issue:44466", "issue:44479", - "issue:44488", "issue:44534", "issue:44589", "issue:44610", @@ -3151,7 +3219,7 @@ "issue:46326" ], "soft_pairs": [ - "issue:43883|issue:44488", + "issue:44206|issue:44220", "issue:43452|issue:43482", "issue:36032|issue:43645", "issue:43317|issue:43482", @@ -3169,254 +3237,257 @@ "issue:42915|issue:45084", "issue:43994|issue:44610", "issue:43377|issue:43819", - "issue:43322|issue:43540", "issue:28530|issue:34567", + "issue:43322|issue:43540", "issue:43089|issue:45987", "issue:44451|issue:45081", "issue:44987|issue:45081", "issue:44265|issue:44610", "issue:42915|issue:44291", - "issue:45216|issue:45357", - "issue:44991|issue:46315", "issue:44220|issue:44683", + "issue:45216|issue:45357", "issue:36032|issue:43742", + "issue:44991|issue:46315", "issue:44062|issue:44987", "issue:42915|issue:44589", "issue:44479|issue:45325", "issue:36032|issue:41720" ], - "prepared_review_unit_hash": "4a84ad2ad2e36d81f7ed7fbfaa82370900699e3b7216c3db0ab012c7d11db102", + "prepared_review_unit_hash": "c5ca9f4e4080e9f5679f72c14b0ff9457fde5f9d5b88a4e567b76b560750358d", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 33, - "serialized_chars": 37162, - "estimated_input_tokens": 9291, - "estimated_eval_tokens": 18838 + "serialized_chars": 37193, + "estimated_input_tokens": 9299, + "estimated_eval_tokens": 18854 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly semantically similar bug reports about loading, generation, tokenizer, or multimodal regressions, but the pairwise evidence is too broad and model-specific to treat any soft edge as a true duplicate. I found no pair that clearly shares the same concrete underlying bug or fix path.", - "confidence": 0.61, - "canonical_issue_reason": "Issue 43452 is the most general API-level loading regression in the set (gguf_file/from_pretrained breakage) and is the closest thing to a representative anchor, though the cluster is too heterogeneous for a strong canonical duplicate.", + "summary": "This cluster is mostly a mix of unrelated Transformers v5 regression reports, with a common surface theme of loading/save_pretrained/tokenizer issues. Per the packet, none of the soft pairs are deterministic duplicates, so I reject all soft-edge merges conservatively.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 44991 is the best representative anchor for the cluster because it is a concrete, version-gated loading regression with a clear failure mode and broad enough scope to stand in for the loading-regression subcluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43452 is the best representative issue because it targets a broad, user-facing load-path regression rather than a single model-specific symptom.", + "best_issue_reason": "44991 has the clearest standalone regression boundary and the most directly triageable repro among the issue set, so it is the strongest issue to use as the cluster\u2019s representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43883", - "right": "issue:44488", + "left": "issue:44206", + "right": "issue:44220", "accept": false, - "reason": "Different failures: missing `all_tied_weights_keys` vs inability to load a specific model." + "reason": "Both are audio-related, but they describe different concrete failures and do not look like the same bug." }, { "left": "issue:43452", "right": "issue:43482", "accept": false, - "reason": "Both involve GGUF loading, but one is a generic `gguf_file` API break and the other is a model-specific Qwen2.5-GGUF failure." + "reason": "Both involve loading regressions, but one is a gguf_file/from_pretrained breakage and the other is a Qwen2.5-GGUF v5 failure; too weak to treat as duplicates." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "Tokenizer `add_special_tokens` conflict is unrelated to custom model initialization in notebooks." + "reason": "Tokenizer method-name conflict vs custom model initialization in Jupyter are different code paths and symptoms." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "Different load paths: dequantized model offload vs GGUF model loading." + "reason": "Device_map/offload dequantized-model loading and Qwen2.5-GGUF loading are distinct failures." }, { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Different problems and code paths; no shared concrete bug is evident." + "reason": "Different subsystems and failure modes; no evidence they share one underlying bug." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Qwen3.5 hidden-states behavior is unrelated to Qwen2.5-VL rope-index scaling." + "reason": "Hidden-states behavior and rope/position-id scaling are not the same concrete issue." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Both are loading regressions, but they affect different artifacts and likely different tokenizer/model code paths." + "reason": "Both are v5 loading regressions, but they affect different artifacts and failure mechanisms." }, { "left": "issue:44464", "right": "issue:45987", "accept": false, - "reason": "Compiled chunked generation inconsistency is not the same bug as `StaticCache.get_seq_length()` returning a tensor." + "reason": "Chunked generation inconsistency and StaticCache return-type issues are separate bugs." }, { "left": "issue:43746", "right": "issue:46326", "accept": false, - "reason": "Local checkpoint loading with PEFT adapters is unrelated to `continue_final_message` for mistral_common tokenizers." + "reason": "PEFT checkpoint loading and continue_final_message tokenizer handling are unrelated code paths." }, { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and MIMI padding-mask batching are different bugs." + "reason": "Audio chat-template sampling-rate defaults and MIMI batch padding-mask behavior are different issues." }, { "left": "issue:43645", "right": "issue:44534", "accept": false, - "reason": "Custom-model initialization regressions do not match non-persistent buffer corruption." + "reason": "Custom model init in notebooks and non-persistent buffer junk are distinct regressions." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Custom model init breakage is not the same issue as junk-filled buffers." + "reason": "Custom model initialization breakage and buffer initialization problems do not look like one bug." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "Different model families and failure modes; only the broad theme of loading regressions overlaps." + "reason": "Missing tied-weights attribute and tokenizer/model loading regression are different failure modes." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Both mention tied lm_head weights, but one is about missing tying and the other about serialization varying by device." + "reason": "Untied lm_head weights and serialization inconsistency are related in theme but not the same concrete defect." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "Loading noise/verbosity is unrelated to a tokenizer regex crash." + "reason": "Generic loading noise and a specific Mistral regex crash are not duplicates." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "CUDA assert on auto device mapping is unrelated to a template-node compilation TypeError." + "reason": "FP8 model loading failure and template compilation error are unrelated." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Different vision-model bugs: bad outputs from AutoModel/pipeline vs wrong processor image size." + "reason": "Wrong model outputs and wrong processor image size are separate bugs." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "MIMI batching/padding-mask issue is unrelated to DAC STE mismatch." + "reason": "Missing padding-mask support and missing STE in DAC are different algorithmic problems." }, { - "left": "issue:43322", - "right": "issue:43540", + "left": "issue:28530", + "right": "issue:34567", "accept": false, - "reason": "Both are multimodal, but a Llava Next segfault and Qwen3OmniMoe video ValueError are different bugs." + "reason": "Early-stopping metric selection and token-count tracking are unrelated trainer bugs." }, { - "left": "issue:28530", - "right": "issue:34567", + "left": "issue:43322", + "right": "issue:43540", "accept": false, - "reason": "Training early-stopping metric selection and `num_input_tokens_seen` updating are unrelated." + "reason": "Llava loading segfault and Qwen3OmniMoe video ValueError are different code paths." }, { "left": "issue:43089", "right": "issue:45987", "accept": false, - "reason": "Generation performance overhead is not the same as the cache sequence-length contract bug." + "reason": "Generation overhead/per-token syncs is not the same as the cache-length type contract issue." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "Model loading failure for ScandiBERT is unrelated to the Mistral regex patch crash." + "reason": "A model-specific load failure and a tokenizer regex crash are different bugs." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Different loading failures with different root causes and code paths." + "reason": "Tokenizer-loading regression and Mistral regex patch crash are unrelated." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export exportability and processor image-size mismatch are separate issues." + "reason": "torch.export failure and processor size mismatch are distinct issues." }, { "left": "issue:42915", "right": "issue:44291", "accept": false, - "reason": "CUDA device-mapping failure is unrelated to the unexpected `_is_hf_initialized` init_empty_weights argument." + "reason": "FP8 config loading and init_empty_weights argument mismatch are different failures." }, { - "left": "issue:45216", - "right": "issue:45357", + "left": "issue:44220", + "right": "issue:44683", "accept": false, - "reason": "Related Qwen3.5 save_pretrained regressions, but one is general checkpoint correctness and the other is specifically visual-encoder keys; not clearly the same bug." + "reason": "Audio fbank extraction and compiled flex_attention on torch>=2.9 are unrelated." }, { - "left": "issue:44991", - "right": "issue:46315", + "left": "issue:45216", + "right": "issue:45357", "accept": false, - "reason": "Tokenizer loading failure and convert_tokens_to_ids performance regression are different issues." + "reason": "Both concern Qwen3.5 save_pretrained regressions, but the packet does not support treating them as one concrete duplicate bug." }, { - "left": "issue:44220", - "right": "issue:44683", + "left": "issue:36032", + "right": "issue:43742", "accept": false, - "reason": "Fbank feature extraction and compiled flex_attention failures are unrelated." + "reason": "T5 tokenizer special-token conflict and MobileLLM key error have different root causes." }, { - "left": "issue:36032", - "right": "issue:43742", + "left": "issue:44991", + "right": "issue:46315", "accept": false, - "reason": "T5 tokenizer method conflict is unrelated to a KeyError loading MobileLLM-125M." + "reason": "Tokenizer loading regression and slow-tokenizer performance regression are separate problems." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "AddedToken constructor argument collision is not the same as a model loading failure." + "reason": "AddedToken special-argument duplication is a tokenizer-construction error, not the same as the model-loading regression." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "CUDA assert on A800 is unrelated to missing Float8 storage type resolution." + "reason": "FP8 config failure and missing Float8 storage are different failures." }, { "left": "issue:44479", "right": "issue:45325", "accept": false, - "reason": "A video-input regression and still-image temporal position-id scaling are different Qwen bugs." + "reason": "Video-input regression across Qwen variants and Qwen2.5-VL rope scaling bug are not the same underlying issue." }, { "left": "issue:36032", "right": "issue:41720", "accept": false, - "reason": "Tokenizer loading conflict and auto device-mapping CUDA assert do not share the same bug path." + "reason": "Tokenizer construction error and Qwen3 device-mapping cuda assert are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the analyst overstates evidence in the soft-edge verdicts. In the packet, every listed soft pair has `deterministic_accept: false`, including 43452/43482 and 45216/45357, but the analyst marked both as accepted duplicates. That is not conservative and is not grounded in the packet. The overall summary is mostly plausible, but the duplicate-pair conclusions should be downgraded to rejected or at most uncertain." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -3426,7 +3497,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 11, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -3484,19 +3555,19 @@ "issue:46491" ], "soft_pairs": [ - "issue:43632|issue:44756", "issue:43531|issue:44877", - "issue:45812|issue:46491", - "issue:42796|issue:43746", - "issue:40444|issue:45325", + "issue:43632|issue:44756", "issue:43611|issue:44534", + "issue:40444|issue:45325", + "issue:42796|issue:43746", "issue:43425|issue:44403", "issue:43296|issue:43531", + "issue:45812|issue:46491", "issue:43792|issue:44589", - "issue:40990|issue:44811", - "issue:33666|issue:40444", "issue:44403|issue:44991", + "issue:33666|issue:40444", "issue:43632|issue:44534", + "issue:40990|issue:44811", "issue:43381|issue:45923", "issue:44079|issue:45663", "issue:43421|issue:44568", @@ -3505,235 +3576,238 @@ "issue:35707|issue:40444", "issue:43482|issue:45538", "issue:43482|issue:43525", - "issue:44829|issue:45479", "issue:43317|issue:43856", + "issue:44829|issue:45479", "issue:42175|issue:45542", "issue:43818|issue:46402", - "issue:44462|issue:45812", "issue:43818|issue:45357", "issue:44079|issue:44534", + "issue:44462|issue:45812", "issue:43531|issue:44291", "issue:43377|issue:45491", "issue:41720|issue:45081" ], - "prepared_review_unit_hash": "dbfd382754950f904c101469b88d5bdbed2a9a84e59f78de494570a26c7279bb", + "prepared_review_unit_hash": "19ab70b8d7e5e2c3a4c0731c6cd787eb7e067845264f35d26524e7068641159b", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37140, - "estimated_input_tokens": 9285, - "estimated_eval_tokens": 18826 + "serialized_chars": 37141, + "estimated_input_tokens": 9286, + "estimated_eval_tokens": 18828 }, "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 and tokenizer/model-loading issues. I found one clear duplicate pair around the missing TensorFlow backend when installing `.[torch]`; the rest should stay separate.", - "confidence": 0.89, - "canonical_issue_reason": "Issue 45542 is the clearest representative of the duplicate TensorFlow-backend failure: it states the user-visible error directly, while 42175 is the same bug from the packaging/install angle.", + "summary": "This is a heterogeneous set of model- and framework-specific issues, not a true duplicate cluster. The soft pairs mostly share only broad subsystem or model-family similarity, so I reject them all.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45812 is the strongest anchor: it describes a broad, silent v4\u2192v5 tokenizer regression affecting all Granite models and has inbound references, making it the most representative issue in this set.", "canonical_pr_reason": null, - "best_issue_reason": "45542 is the most precise and actionable anchor for the shared TF-backend bug, with a clearer failure description than the sibling report.", + "best_issue_reason": "issue:45812 is the most actionable high-signal report in the group, with a clear regression description and broad impact across a model family.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43632", - "right": "issue:44756", - "accept": false, - "reason": "Different bugs: `_is_hf_initialized` loading regression vs mmap/OOM behavior on Strix Halo." - }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Different code paths: Qwen3-MoE sliding-window issue vs strict config loading for granite_speech." + "reason": "Different bugs: Qwen3-MoE sliding-window handling vs GraniteSpeech config strictness." }, { - "left": "issue:45812", - "right": "issue:46491", + "left": "issue:43632", + "right": "issue:44756", "accept": false, - "reason": "Both are tokenizer regressions, but for different models and different failure modes; not the same underlying bug." + "reason": "Unrelated failure modes: v5 init-flag regression vs Apple Silicon mmap/OOM behavior." }, { - "left": "issue:42796", - "right": "issue:43746", + "left": "issue:43611", + "right": "issue:44534", "accept": false, - "reason": "Tokenizer BPE-marker handling and PEFT local-checkpoint loading are unrelated failures." + "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer initialization; distinct code paths." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Both involve Qwen2.5-VL, but one is multi-image finetuning failure and the other is rope-index scaling; not the same bug." + "reason": "Both involve Qwen2.5-VL, but one is multi-image iterable training failure and the other is temporal position-id scaling in rope indexing." }, { - "left": "issue:43611", - "right": "issue:44534", + "left": "issue:42796", + "right": "issue:43746", "accept": false, - "reason": "Different regressions: `base_model_prefix` loading vs non-persistent buffer initialization." + "reason": "Tokenizer BPE marker handling vs GraniteSpeech PEFT checkpoint loading; not the same underlying bug." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch 2.10 compatibility and noisy model-loading output are unrelated." + "reason": "Torch 2.10 compatibility and load-time noise are different issues." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL load failure and Qwen3-MoE sliding-window behavior are unrelated." + "reason": "Different model families and symptoms: PaddleOCR-VL load failure vs Qwen3-MoE sliding_window bug." + }, + { + "left": "issue:45812", + "right": "issue:46491", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different behavior (Granite token IDs vs CodeLlama leading-space round-trip)." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "Whisper runtime failure and Float8 storage lookup error do not describe the same bug." + "reason": "Whisper runtime failure and Float8 storage lookup error are unrelated." }, { - "left": "issue:40990", - "right": "issue:44811", + "left": "issue:44403", + "right": "issue:44991", "accept": false, - "reason": "Perplexity regression and `batch_decode` special-token handling are unrelated." + "reason": "Unnecessary load noise is not the same as EMBEDDIA tokenizer loading failure." }, { "left": "issue:33666", "right": "issue:40444", "accept": false, - "reason": "Both mention Qwen vision models, but one is multi-GPU training and the other is iterable-dataset multi-image finetuning." + "reason": "Multi-GPU training support and iterable-dataset multi-image prompt failure are distinct Qwen2-VL training problems." }, { - "left": "issue:44403", - "right": "issue:44991", + "left": "issue:43632", + "right": "issue:44534", "accept": false, - "reason": "Unnecessary loading noise and tokenizer loading failure for EMBEDDIA/est-roberta are distinct issues." + "reason": "Different regressions: missing _is_hf_initialized handling vs junk-filled non-persistent buffers." }, { - "left": "issue:43632", - "right": "issue:44534", + "left": "issue:40990", + "right": "issue:44811", "accept": false, - "reason": "Same broad Transformers v5 area, but different concrete defects and code paths." + "reason": "High perplexity on a language-model benchmark is unrelated to Whisper skip_special_tokens behavior." }, { "left": "issue:43381", "right": "issue:45923", "accept": false, - "reason": "Eval-mode checkpointing guard and missing `supports_gradient_checkpointing` flag are related in theme only, not the same bug." + "reason": "Eval-mode checkpointing restriction and missing gradient-checkpointing flag are related broadly, but not the same bug." }, { "left": "issue:44079", "right": "issue:45663", "accept": false, - "reason": "ModelOutput key assignment and Gemma4 attention state rebuilding are unrelated." + "reason": "ModelOutput key assignment and Gemma-4 shared_kv_states KeyError are different issues." }, { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "Runtime special-token post-processor updates and mdeberta BOS/EOS insertion are different tokenizer bugs." + "reason": "Runtime post-processor updates and add_special_tokens BOS/EOS omission are separate tokenizer bugs." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "`_is_hf_initialized` init-empty-weights regression and Float8 storage lookup error are unrelated." + "reason": "Unexpected _is_hf_initialized argument and missing Float8 storage object are unrelated TypeErrors." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Bad tokenizer error messaging and AppleSilicon allocator TypeError are unrelated." + "reason": "Bad tokenizer error messaging is unrelated to the Apple Silicon warmup TypeError." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with `inputs_embeds`/`past_key_values` is unrelated to multi-image finetuning failures." + "reason": "Progressive generation with inputs_embeds/past_key_values is not the same as multi-image iterable dataset training failure." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "GGUF loading regression and CLIPTokenizer max-length handling are different bugs." + "reason": "GGUF loading failure and CLIPTokenizer model_max_length behavior are different bugs." }, { "left": "issue:43482", "right": "issue:43525", "accept": false, - "reason": "GGUF loading failure and missing `pad_token_id` on Llama4Config are unrelated." + "reason": "Qwen2.5-GGUF loading and missing pad_token_id on Llama4Config are unrelated issues." }, { - "left": "issue:44829", - "right": "issue:45479", + "left": "issue:43317", + "right": "issue:43856", "accept": false, - "reason": "Both affect training, but flash-attention-3 degeneration and single-label zero-loss behavior have different causes." + "reason": "Dequantized model offload loading failure and Qwen3 MoE training memory inefficiency are different code paths." }, { - "left": "issue:43317", - "right": "issue:43856", + "left": "issue:44829", + "right": "issue:45479", "accept": false, - "reason": "Device-map offload failure and MoE memory inefficiency are not the same issue." + "reason": "Similar training symptoms, but flash_attention_3 degeneration and num_labels=1 zero-loss are different root causes." }, { "left": "issue:42175", "right": "issue:45542", - "accept": true, - "reason": "Same missing-TensorFlow backend bug: both describe the undefined tf-backend failure after installing the torch-only extra." + "accept": false, + "reason": "Both touch TensorFlow availability, but one is missing TF from the torch extra and the other is a tensorboard-only backend error." }, { "left": "issue:43818", "right": "issue:46402", "accept": false, - "reason": "Video-LLaVA temporal-attention/weight-sharing defect and Sam3Model text-encoder loading are unrelated." + "reason": "Video-LLaVA architecture/weights issues are unrelated to Sam3Model text-encoder loading." }, { - "left": "issue:44462", - "right": "issue:45812", + "left": "issue:43818", + "right": "issue:45357", "accept": false, - "reason": "Ignoring `tokenizer.json` and wrong Granite token IDs are different tokenizer regressions." + "reason": "Different model families and different save/load problems." }, { - "left": "issue:43818", - "right": "issue:45357", + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer junk are unrelated regressions." + }, + { + "left": "issue:44462", + "right": "issue:45812", "accept": false, - "reason": "Different models and failures: Video-LLaVA vision tower vs Qwen3.5 visual-encoder key saving." + "reason": "Tokenizer.json being ignored is a different loading bug from Granite token-ID regression." }, { "left": "issue:43531", "right": "issue:44291", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and `_is_hf_initialized` TypeError are unrelated." + "reason": "Qwen3-MoE sliding_window and init_empty_weights _is_hf_initialized issues are unrelated." }, { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "MIMI padding-mask batching mismatch and Gemma3 mixed-length NaNs are different bugs." + "reason": "MIMI padding-mask mismatch and Gemma3 all-padding-window NaNs are both batching issues, but not the same bug." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "Qwen3 auto-device-map CUDA assert and Mistral regex patch AttributeError are unrelated." + "reason": "Qwen3 auto device mapping CUDA assert and Mistral regex patch crash are unrelated." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative overall. The one accepted pair, issue:42175 and issue:45542, is a plausible duplicate around the missing TensorFlow backend in the torch-only install path. The other soft-edge rejections are also sufficiently cautious given the issue titles available." - }, - "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 }, @@ -3742,7 +3816,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 12, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -3800,265 +3874,268 @@ "issue:46315" ], "soft_pairs": [ - "issue:40444|issue:45381", "issue:43644|issue:44079", + "issue:40444|issue:45381", "issue:44265|issue:45072", "issue:43611|issue:43646", "issue:43749|issue:44079", "issue:45325|issue:45381", "issue:42796|issue:45381", - "issue:43994|issue:44898", "issue:42175|issue:43296", + "issue:43994|issue:44898", "issue:43792|issue:43957", "issue:45325|issue:45834", "issue:43454|issue:44466", "issue:43632|issue:45072", "issue:42915|issue:43482", "issue:43701|issue:45325", - "issue:44479|issue:45823", "issue:44291|issue:44877", - "issue:43873|issue:45701", + "issue:44479|issue:45823", "issue:43296|issue:45800", + "issue:43873|issue:45701", "issue:43606|issue:44265", "issue:43262|issue:43746", "issue:40990|issue:45584", "issue:42796|issue:46209", - "issue:34634|issue:35707", "issue:43881|issue:44488", + "issue:34634|issue:35707", "issue:35707|issue:41720", - "issue:44186|issue:44265", "issue:43454|issue:43550", - "issue:43606|issue:43819", + "issue:44186|issue:44265", "issue:39692|issue:40444", + "issue:43606|issue:43819", "issue:41720|issue:45823", "issue:46305|issue:46315" ], - "prepared_review_unit_hash": "d51f462685710c5121c71f39eafb2019ba7c7a3aa76fb51fd6dc95be750759aa", + "prepared_review_unit_hash": "c23834375bea6f07da82e8ffd4d1b31a66f7727b9a5fe18e3f0f2b6a831b68e3", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37192, - "estimated_input_tokens": 9298, - "estimated_eval_tokens": 18852 + "serialized_chars": 37185, + "estimated_input_tokens": 9297, + "estimated_eval_tokens": 18850 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose similarity bucket, not a true duplicate set. Most pairs share only a broad subsystem or version-regression theme, but describe different concrete failures and code paths. If forced to pick a canonical issue, issue:43611 is the broadest v5-loading regression, though the overall cluster is heterogeneous.", - "confidence": 0.31, - "canonical_issue_reason": "issue:43611 is the broadest and most umbrella-like report in the set: it describes a general Transformers v5 loading regression around `base_model_prefix`, which is closer to several other v5-init/load breakages than the more model-specific reports.", + "summary": "Mostly a heterogeneous set of unrelated bugs. issue:43646 remains the best broad issue representative for the v5 init/load regression cluster, but the Qwen2.5-VL position-id reports are only related at a high level and should not be merged as the same bug.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43646 is the broadest representative of the repeated Transformers v5 initialization/loading regressions in this set, and it captures the shared failure mode better than the narrower model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43611 is the best representative of the cluster because it is the most general model-loading regression among the issues; the rest are narrower, model-specific, or unrelated symptoms.", + "best_issue_reason": "issue:43646 is the best overall issue candidate because it is the most central, general, and representative of the v5 init/load breakage cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:40444", - "right": "issue:45381", + "left": "issue:43644", + "right": "issue:44079", "accept": false, - "reason": "Both involve Qwen2.5-VL/video-image handling, but one is a finetuning + IterableDataset multi-image failure and the other is a video position-id regression; different failure modes." + "reason": "Different bugs: tied/non-persistent buffer serialization vs ModelOutput key assignment." }, { - "left": "issue:43644", - "right": "issue:44079", + "left": "issue:40444", + "right": "issue:45381", "accept": false, - "reason": "One is a v5 serialization/tied-weights regression, the other is ModelOutput key assignment when a field was previously None; not the same bug." + "reason": "Both are Qwen multimodal issues, but multi-image finetuning and video vision_position_ids are different code paths." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "Both touch export/compile-related behavior, but one is `torch.export.export` with `torch_compilable_check` and the other is bfloat16 dtype mismatches in specific models." + "reason": "torch.export with torch_compilable_check is unrelated to a bfloat16 dtype mismatch in inference." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both are v5 loading/init regressions, but `base_model_prefix` loading and custom model initialization are distinct code paths and symptoms." + "reason": "Related v5 regressions, but base_model_prefix loading and custom model initialization are distinct failures." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM efficient loading and ModelOutput key assignment are unrelated; only very broad Transformer internals overlap." + "reason": "FSDP CPU RAM efficient loading and ModelOutput key handling are unrelated." }, { "left": "issue:45325", "right": "issue:45381", "accept": false, - "reason": "Both concern Qwen2.5-VL RoPE/position IDs, but one is still-image temporal scaling and the other is video vision_position_ids; related area, different bug." + "reason": "Same Qwen2.5-VL position-id area, but still-image temporal scaling and video vision_position_ids are not enough evidence of one shared bug." }, { "left": "issue:42796", "right": "issue:45381", "accept": false, - "reason": "Tokenizer BPE-marker handling and Qwen2.5-VL video position IDs are unrelated issues." + "reason": "Tokenizer BPE marker handling is unrelated to Qwen2.5-VL video position IDs." }, { - "left": "issue:43994", - "right": "issue:44898", + "left": "issue:42175", + "right": "issue:43296", "accept": false, - "reason": "Both are Perceiver/SigLIP-ish vision reports, but one is nonsense outputs from model/pipeline mismatch and the other is non-default-resolution inference failure." + "reason": "Missing TensorFlow backend and PaddleOCR-VL loading in vLLM are different subsystems and failures." }, { - "left": "issue:42175", - "right": "issue:43296", + "left": "issue:43994", + "right": "issue:44898", "accept": false, - "reason": "TensorFlow backend packaging and PaddleOCR-VL loading in vLLM are unrelated; different packages and failure types." + "reason": "Both are vision-model regressions, but they affect different models and different failure mechanisms." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Whisper loading/runtime failure and meta-device loading regressions are not the same concrete bug." + "reason": "Whisper runtime failure and meta-device loading regression do not describe the same bug." }, { "left": "issue:45325", "right": "issue:45834", "accept": false, - "reason": "Both mention sequence/position issues, but Qwen2.5-VL RoPE scaling and Kosmos2.5 OCR index errors are different models and code paths." + "reason": "Qwen2.5-VL position IDs and Kosmos2.5 OCR long-input indexing are unrelated." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "AyaVision lm_head tying failure and inconsistent lm_head.weight serialization are related to tied weights, but one is a model config bug and the other is device-dependent serialization behavior." + "reason": "Both mention tied weights, but one is a tying bug and the other is serialization inconsistency." }, { "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "`_is_hf_initialized` flag breakage and export dtype mismatches are different regressions, even if both involve v5 compatibility." + "reason": "_is_hf_initialized breakage and bfloat16 dtype mismatch are separate problems." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Qwen3Moe FP8 config failure and Qwen2.5-GGUF loading failure are unrelated model/load paths." + "reason": "Different models and loading paths: Qwen3Moe FP8 vs Qwen2.5-GGUF v5 loading." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "Checkpoint key mismatch during resume and Qwen2.5-VL RoPE scaling are unrelated." - }, - { - "left": "issue:44479", - "right": "issue:45823", - "accept": false, - "reason": "Both are multimodal/device-regression reports, but one is a video-input regression for Qwen models and the other is a Gemma4 PLE device-map forward mismatch." + "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL position-id scaling." }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "Unexpected `_is_hf_initialized` argument in init-empty-weights loading and strict config blocking granite_speech loading are different loading failures." + "reason": "init_empty_weights argument handling and strict config loading are different issues." }, { - "left": "issue:43873", - "right": "issue:45701", + "left": "issue:44479", + "right": "issue:45823", "accept": false, - "reason": "Quantization/offloading behavior and tokenization-version changes are unrelated." + "reason": "Qwen video-position regression and Gemma4 device placement are not the same bug." }, { "left": "issue:43296", "right": "issue:45800", "accept": false, - "reason": "PaddleOCR-VL load failure in vLLM and torch/transformers incompatibility are different compatibility issues." + "reason": "PaddleOCR-VL load failure in vLLM is unrelated to torch/transformers version incompatibility." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Quantization offload behavior and tokenizer slowdown/version drift are unrelated." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "CPU offload device mismatch for bark-small and torch.export failures are different problems." + "reason": "CPU offload device mismatch and torch.export failure are different code paths." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio processor chat-template sampling-rate default and GraniteSpeech PEFT checkpoint loading are unrelated." + "reason": "Audio chat-template sampling rate default and GraniteSpeech PEFT checkpoint loading are unrelated." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "High perplexity on GPT-OSS and Whisper empty-transcription generation failure are unrelated model behaviors." + "reason": "High perplexity on one model and Whisper empty-transcription generation failure are unrelated." }, { "left": "issue:42796", "right": "issue:46209", "accept": false, - "reason": "Tokenizer BPE-marker handling and FP8 inference on B200 are unrelated." - }, - { - "left": "issue:34634", - "right": "issue:35707", - "accept": false, - "reason": "BarkProcessor voice_preset and inputs_embeds/past_key_values progressive generation are unrelated." + "reason": "Tokenizer marker handling and FP8 inference on B200 are unrelated." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "glm-4v-9b loading failure and cjvt/sleng-bert loading failure are distinct model-specific loading bugs." + "reason": "Two model-loading failures, but no evidence they share the same underlying bug." }, { - "left": "issue:35707", - "right": "issue:41720", + "left": "issue:34634", + "right": "issue:35707", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values and Qwen3 auto-device-map cuda assert are different failures." + "reason": "BarkProcessor voice preset and progressive generation with inputs_embeds/past_key_values are unrelated." }, { - "left": "issue:44186", - "right": "issue:44265", + "left": "issue:35707", + "right": "issue:41720", "accept": false, - "reason": "LayoutLMv2Tokenizer padding/truncation crash and torch.export compilation failure are unrelated." + "reason": "Different failures: generation cache handling vs Qwen3 device-map CUDA assert." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "AyaVision tied lm_head bug and Bamba torch.compile SDPA failure are different model/runtime issues." + "reason": "Tied-weight text generation bug and Bamba SDPA compile failure are unrelated." }, { - "left": "issue:43606", - "right": "issue:43819", + "left": "issue:44186", + "right": "issue:44265", "accept": false, - "reason": "Bark CPU-offload device mismatch and DAC.from_latents mismatch are unrelated audio-model bugs." + "reason": "Tokenizer padding/truncation crash and torch.export failure are different problems." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "SigLIP2 docs/example errors and Qwen2.5-VL multi-image finetuning failure are different problems." + "reason": "SigLIP2 doc example issues and Qwen2.5-VL multi-image finetuning failure are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "CPU offload mismatch and DAC from_latents/forward mismatch are unrelated." }, { "left": "issue:41720", "right": "issue:45823", "accept": false, - "reason": "Qwen3 cuda assert under auto device mapping and Gemma4 PLE device mismatch are unrelated model/device-map regressions." + "reason": "Both are device-related, but they affect different models and different failure modes." }, { "left": "issue:46305", "right": "issue:46315", "accept": false, - "reason": "MoonshineStreaming mask off-by-one and tokenizer slow-path performance regression are unrelated." + "reason": "Streaming mask off-by-one and tokenizer conversion slowdown are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "Rejecting because one soft-edge verdict is too strong for the evidence in the packet. `issue:45325` and `issue:45381` are certainly related, but the titles only show a shared Qwen2.5-VL / position-id theme: one is about still-image temporal position_ids being scaled incorrectly, the other about video input vision_position_ids being wrong. That is not enough on its own to conclude they are the same bug. The overall summary is mostly grounded, but this accepted duplicate-like pair is not conservative enough." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -4068,7 +4145,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 13, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -4094,6 +4171,7 @@ "issue:43450", "issue:43454", "issue:43526", + "issue:43550", "issue:43572", "issue:43577", "issue:43708", @@ -4102,7 +4180,6 @@ "issue:43873", "issue:43881", "issue:43883", - "issue:44008", "issue:44186", "issue:44387", "issue:44466", @@ -4119,19 +4196,18 @@ "issue:45701", "issue:45704", "issue:45759", - "issue:45823", "issue:45920", "issue:45923", "issue:46050", "issue:46402" ], "soft_pairs": [ - "issue:44568|issue:45920", "issue:42796|issue:44977", "issue:43377|issue:44610", + "issue:44568|issue:45920", "issue:43454|issue:45072", - "issue:43317|issue:43873", "issue:43257|issue:45923", + "issue:43317|issue:43873", "issue:43377|issue:44898", "issue:40990|issue:42947", "issue:43322|issue:46050", @@ -4142,236 +4218,236 @@ "issue:43526|issue:43883", "issue:42617|issue:42915", "issue:44466|issue:45357", - "issue:41720|issue:44977", "issue:43881|issue:44991", + "issue:41720|issue:44977", "issue:26413|issue:28530", "issue:44387|issue:45704", "issue:40990|issue:45701", "issue:36331|issue:42617", - "issue:44186|issue:45072", "issue:43450|issue:45701", "issue:36331|issue:42915", - "issue:43065|issue:46402", + "issue:44186|issue:45072", "issue:44877|issue:44991", + "issue:43065|issue:46402", "issue:43873|issue:44756", + "issue:43577|issue:43708", "issue:43404|issue:45357", - "issue:44008|issue:45823", - "issue:43577|issue:43708" + "issue:43550|issue:43720" ], - "prepared_review_unit_hash": "7d594088e5d5c9af11751fa1f5748009b5d0ee57df7d66f4f7daf59a2b51c832", + "prepared_review_unit_hash": "7866f5c7185d1ca17b5f68ed4b144495e5d3e5f0c74b94b0cbe10beea90aef80", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 37150, - "estimated_input_tokens": 9288, - "estimated_eval_tokens": 18832 + "serialized_chars": 36501, + "estimated_input_tokens": 9126, + "estimated_eval_tokens": 18508 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly heterogeneous: the soft pairs overlap on broad areas like tokenization, loading, quantization, or serialization, but they do not describe the same concrete bug/change. I would not merge any of the soft pairs.", - "confidence": 0.42, - "canonical_issue_reason": "issue:45701 is the broadest umbrella report in the tokenization-regression subgroup, but the cluster overall is too mixed for a strong canonical choice.", + "summary": "This cluster is a loose collection of mostly unrelated closed/open Transformers bugs, with only broad thematic overlap around tokenization, loading, offloading, and v5 regressions. None of the soft pairs look like the same concrete underlying bug or change, so all soft edges are rejected. The best cluster anchor is the broad tokenizer-regression report in issue 45920.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 45920 is the broadest, highest-signal regression in the set, covering wrong token IDs across multiple models and versions, so it is the best anchor for this loose cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 is the strongest concrete regression report among the tokenization-related items because it names multiple affected model families and has clearer reproduction detail.", + "best_issue_reason": "Issue 45920 best represents the cluster because it is a concrete, high-impact v5 tokenizer regression with multiple affected model families and the widest scope among the candidates.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44568", - "right": "issue:45920", - "accept": false, - "reason": "Related tokenization symptoms, but one is about add_special_tokens/BOS-EOS behavior and the other is a broader wrong-token-ids regression across models." - }, { "left": "issue:42796", "right": "issue:44977", "accept": false, - "reason": "Different bugs: BPE marker handling in a tokenizer versus flash-attention generation failure in Qwen3.5." + "reason": "Tokenizer/BPE-marker bug vs flash-attention generation bug; same general area but different failure modes and code paths." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "Different code paths and failures: MIMI padding-mask batching bug versus OmDet-Turbo processor image-size mismatch." + "reason": "MIMI padding-mask batching issue vs OmDet processor image-size mismatch; unrelated subsystems." }, { - "left": "issue:43454", - "right": "issue:45072", + "left": "issue:44568", + "right": "issue:45920", "accept": false, - "reason": "Different model issues: lm_head tying bug versus dtype mismatch during bfloat16 inference." + "reason": "One is special-token insertion for a specific tokenizer, the other is wrong token IDs across many models; not the same bug." }, { - "left": "issue:43317", - "right": "issue:43873", + "left": "issue:43454", + "right": "issue:45072", "accept": false, - "reason": "Both involve offload/quantization broadly, but the concrete failures differ: dequantized loading on cpu/gpu offload versus generic quantization offloading not working." + "reason": "Weight tying bug vs dtype mismatch in bf16 inference; different underlying problems." }, { "left": "issue:43257", "right": "issue:45923", "accept": false, - "reason": "Unrelated changes: Qwen3 MoE weight conversion under accelerate/deepspeed versus a missing gradient-checkpointing flag." + "reason": "MoE weight conversion/loading issue vs missing gradient-checkpointing flag on a trust_remote_code model." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both involve offloading/quantization, but one is a specific device_map load failure and the other is a broader offloading regression." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "Different components and symptoms: MIMI batching/padding mask bug versus Perceiver interpolation at non-default resolution." + "reason": "Batch padding-mask output mismatch vs Perceiver resolution/interpolation failure; unrelated model paths." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "No shared concrete bug: perplexity regression on a model/dataset versus gradient checkpointing being ineffective with PEFT LoRA." + "reason": "Perplexity regression vs gradient-checkpointing/LoRA behavior; no shared concrete bug." }, { "left": "issue:43322", "right": "issue:46050", "accept": false, - "reason": "Loading crash versus save/load quantization corruption; these are different failure modes and code paths." + "reason": "Llava Next load crash vs sinq save/load breakage; different model families and failure modes." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "Different underlying problems: tied lm_head weights versus packed BitNet weights not being unpacked." + "reason": "Weight tying problem vs packed-weight unpacking during accelerate loading; not the same issue." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "Different config/model-loading bugs: missing pad_token_idx versus not unwrapping text_config for composite multimodal configs." + "reason": "Missing config field vs composite multimodal text_config unwrapping; different config-loading bugs." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Different failures and models: Qwen3 auto device-map assert versus Llava Next load-time segmentation fault." + "reason": "CUDA assert with auto device mapping vs segmentation fault loading Llava Next; unrelated runtime failures." }, { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "Generic memory/loading regression versus non-persistent buffers being filled with junk; not the same bug." + "reason": "FSDP RAM-efficient loading breakage vs non-persistent buffers initialized with junk; both v5-related but distinct defects." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "Different subsystems entirely: image label reduction bug versus missing all_tied_weights_keys on a causal LM." + "reason": "BeitImageProcessor label reduction bug vs missing all_tied_weights_keys on Molmo; unrelated components." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "No concrete overlap beyond general failure to run: a 3d_parallel.py execution issue versus a Qwen3 MoE FP8 config failure." + "reason": "3d_parallel.py runtime issue vs Qwen3Moe FineGrainedFP8Config failure; unrelated." }, { "left": "issue:44466", "right": "issue:45357", "accept": false, - "reason": "Both mention saving/weights, but one is lm_head serialization tied to device while the other is incorrect visual-encoder keys in Qwen3.5 save_pretrained." + "reason": "lm_head serialization/tied-weights issue vs incorrect visual-encoder key saving; both serialization bugs but different code paths." }, { - "left": "issue:41720", - "right": "issue:44977", + "left": "issue:43881", + "right": "issue:44991", "accept": false, - "reason": "Different Qwen3 issues: device-mapping crash versus flash-attention generation correctness." + "reason": "Model loading failure vs tokenizer loading regression; different artifacts and failures." }, { - "left": "issue:43881", - "right": "issue:44991", + "left": "issue:41720", + "right": "issue:44977", "accept": false, - "reason": "Different artifacts and bugs: glm-4v model loading failure versus tokenizer loading regression for EMBEDDIA/est-roberta." + "reason": "Auto device-map CUDA assert vs flash-attention generation regression; same model family but different concrete bugs." }, { "left": "issue:26413", "right": "issue:28530", "accept": false, - "reason": "Checkpoint resume epoch_iterator failure is unrelated to early-stopping metric lookup/disable logic." + "reason": "Checkpoint resume iterator failure vs early-stopping metric lookup warning; unrelated trainer behaviors." }, { "left": "issue:44387", "right": "issue:45704", "accept": false, - "reason": "Both are memory-related, but one is int4 CUDA reserved-memory/OOM and the other is a T5 RMSNorm memory leak." + "reason": "Int4 quantization memory regression vs apex RMSNorm memory leak; not the same underlying cause." }, { "left": "issue:40990", "right": "issue:45701", "accept": false, - "reason": "One is a model perplexity issue and the other is a tokenizer-version regression; not the same bug." + "reason": "Perplexity on one model/dataset vs version-dependent tokenization changes; only vague tokenizer overlap." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "Compute_loss keyword-argument incompatibility is unrelated to the 3d_parallel.py run failure." - }, - { - "left": "issue:44186", - "right": "issue:45072", - "accept": false, - "reason": "Tokenizer padding/truncation crash versus dtype mismatch in inference; no shared code-path bug." + "reason": "CustomTrainer signature breakage vs 3d_parallel.py failure; unrelated." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Video processor shape bug is unrelated to tokenizer behavior changes." + "reason": "Batched video processor shape bug vs tokenization changes; different components." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Trainer API signature mismatch is unrelated to Qwen3 MoE FP8 loading failure." + "reason": "Trainer compute_loss kwarg regression vs Qwen3Moe FP8 failure; no shared bug." }, { - "left": "issue:43065", - "right": "issue:46402", + "left": "issue:44186", + "right": "issue:45072", "accept": false, - "reason": "Same model family but different components: dummy Conv2d in a pixel decoder versus missing text-encoder weight loading." + "reason": "LayoutLMv2 tokenizer crash vs bf16 dtype mismatch; unrelated issues." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config loading for granite_speech is unrelated to tokenizer loading for est-roberta." + "reason": "Strict config loading vs tokenizer loading regression; separate config/tokenizer problems." + }, + { + "left": "issue:43065", + "right": "issue:46402", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder vs missing Sam3 text-encoder weights; same model family but different code paths." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Both are memory/offload themed, but one concerns quantization offloading and the other mmap-induced OOM on Strix Halo." + "reason": "Quantization offloading regression vs mmap OOM on Strix Halo; different memory-management problems." }, { - "left": "issue:43404", - "right": "issue:45357", + "left": "issue:43577", + "right": "issue:43708", "accept": false, - "reason": "Different issues with weights: lm_head tying versus incorrect visual-encoder keys in save_pretrained." + "reason": "dtype propagation bug in Blip2 vs incorrect max_steps after resume; unrelated trainer/model-load bugs." }, { - "left": "issue:44008", - "right": "issue:45823", + "left": "issue:43404", + "right": "issue:45357", "accept": false, - "reason": "Different Gemma problems: variable-name collision in forward versus device mismatch during forward." + "reason": "lm_head weight tying bug vs incorrect visual encoder key saving; both multimodal model issues but not the same defect." }, { - "left": "issue:43577", - "right": "issue:43708", + "left": "issue:43550", + "right": "issue:43720", "accept": false, - "reason": "Different bugs: dtype propagation in Blip2 loading versus resume_from_checkpoint max_steps miscalculation." + "reason": "torch.compile/SDPA failure vs packed-weight loading bug; different runtime paths and fixes." } ] }, @@ -4387,7 +4463,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 14, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -4421,10 +4497,10 @@ "issue:43606", "issue:43653", "issue:43720", - "issue:43746", "issue:43782", "issue:43873", "issue:43881", + "issue:44008", "issue:44186", "issue:44361", "issue:44387", @@ -4437,15 +4513,15 @@ "issue:45059", "issue:45161", "issue:45406", - "issue:45464", "issue:45632", "issue:45684", "issue:45701", "issue:45704", + "issue:45823", "issue:46291" ], "soft_pairs": [ - "issue:43550|issue:43720", + "issue:44008|issue:45823", "issue:41720|issue:42617", "issue:44423|issue:45406", "issue:44610|issue:44898", @@ -4466,238 +4542,224 @@ "issue:42915|issue:43322", "issue:36331|issue:41720", "issue:43316|issue:44913", - "issue:29159|issue:36032", "issue:28530|issue:30064", + "issue:29159|issue:36032", "issue:33357|issue:35707", "issue:43388|issue:45059", - "issue:43550|issue:43653", "issue:30064|issue:34634", + "issue:43550|issue:43653", "issue:43606|issue:43720", "issue:43873|issue:45704", - "issue:44387|issue:45701", - "issue:41720|issue:43746", - "issue:45406|issue:45464" + "issue:44387|issue:45701" ], - "prepared_review_unit_hash": "20abdc73295695682e4cb0e5b8b28a42232a7f3198036cf43516540db99bc9f2", + "prepared_review_unit_hash": "7b0ddcf53707aae649f6972148537b0fb978e4a4a8e7510de0a0ea093f43f9f7", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 37234, - "estimated_input_tokens": 9309, - "estimated_eval_tokens": 18874 + "soft_pair_count": 30, + "serialized_chars": 36672, + "estimated_input_tokens": 9168, + "estimated_eval_tokens": 18592 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous and the soft-similarity pairs do not look like true duplicates. Most items are separate model-, tokenizer-, processor-, loading-, or serving-specific bugs across different code paths.", - "confidence": 0.83, - "canonical_issue_reason": "issue:43873 is the best umbrella candidate because it is broad, still open, and matches the recurring offload/quantization/device-mapping failure theme seen in several nearby reports.", + "summary": "This cluster is heterogeneous and does not contain any convincing duplicate pairs. The items span unrelated issues across tokenizers, multimodal processors, training, serving, quantization, and device/offload behavior.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "As a canonical issue, issue:43873 is the most useful consolidation point for related offloading/quantization regressions, even though the cluster overall is not a tight duplicate set.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43550", - "right": "issue:43720", + "left": "issue:44008", + "right": "issue:45823", "accept": false, - "reason": "Different concrete failures: torch.compile/SDPA on Bamba vs packed-weight unpacking during accelerate loading on BitNet." + "reason": "Both mention device mismatch, but they affect different Gemma variants and different code paths: a variable-name collision in forward vs auto device-map PLE forwarding." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Unrelated symptoms and code paths: Qwen3 device-map crash vs failure to run 3d_parallel.py." + "reason": "Different failures and contexts: Qwen3 auto device mapping cuda assert vs failure to run 3d_parallel.py." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Both are serving/multimodal crashes, but the models and root errors differ ('str'.to vs missing _tokenizer)." + "reason": "Both are serve crashes, but one is a multimodal batching str-to tensor issue and the other is Gemma4Processor missing _tokenizer." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Different model families and bugs: processor size mismatch vs interpolate_pos_encoding failure." + "reason": "Different models and bugs: OmDet-Turbo image size mismatch vs Perceiver interpolation failure." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "Tokenizer/NER padding crash is not the same bug as an image processor producing the wrong resolution." + "reason": "Tokenizer padding/NER crash vs OmDet processor input resolution mismatch; no shared underlying bug." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "Both mention version/regression effects, but one is processor/tokenizer API breakage and the other is tokenization behavior changes." + "reason": "A regression in processor/tokenizer API usage is not the same as a broad tokenization difference across versions." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Different failures: torch.compile with SDPA vs CPU offload device mismatch." + "reason": "Both involve runtime/device behavior, but one is torch.compile+SDPA and the other is CPU offload device mismatch for a different model." }, { "left": "issue:42915", "right": "issue:43317", "accept": false, - "reason": "Both are quantization/offload-adjacent, but one is FineGrainedFP8Config and the other is dequantized model loading with gpu+cpu offload." + "reason": "Different quantization/loading problems: FineGrainedFP8Config failure vs dequantized model loading with gpu+cpu offload." }, { "left": "issue:45632", "right": "issue:45684", "accept": false, - "reason": "Cache-path collision and read-only permission propagation are distinct save/load issues." + "reason": "Cache-path collision for trust_remote_code is unrelated to read-only permission propagation in save_pretrained." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to a script execution failure." + "reason": "ImportError for GenerationMixin is unrelated to 3d_parallel.py execution failure." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Different multi-GPU/device-parallel problems affecting different model setups." + "reason": "Both involve parallelism, but one is a cuda assert on Qwen3 auto device mapping and the other is a tensor-parallel-only failure on GPT-OSS MoE." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "Incorrect masked LM outputs and recurrent-state reset are separate model-behavior bugs." + "reason": "CamemBERT masked-LM prediction errors and Qwen3.5 recurrent-state resets are different model-specific bugs." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior is unrelated to StableLmConfig pad_token_idx regression." + "reason": "Qwen3-MoE sliding-window behavior and missing pad_token_idx in StableLmConfig are unrelated." }, { "left": "issue:42915", "right": "issue:46291", "accept": false, - "reason": "FP8 quantization failure is not the same as importing pipelines with a kernels version/revision requirement." + "reason": "Fine-grained FP8 model loading and transformers.pipelines import/version resolution are unrelated issues." }, { "left": "issue:28530", "right": "issue:33666", "accept": false, - "reason": "Early-stopping metric resolution is unrelated to Qwen2-VL multi-GPU training." + "reason": "Early stopping metric lookup and Qwen2-VL multi-GPU training are unrelated." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug and glm-4v loading failure are different code paths." + "reason": "BEiT label reduction bug and glm-4v-9b loading failure do not share a concrete code-path." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "Both are tokenizer-related, but the models and failure modes differ too much to be duplicates." + "reason": "Both are tokenizer failures, but they concern different tokenizers and different error mechanisms." }, { "left": "issue:43540", "right": "issue:43782", "accept": false, - "reason": "Video-input processing error and weight_only from_pretrained load error are not the same underlying bug." + "reason": "Different Qwen3 multimodal problems: video-processing ValueError vs Qwen3VL weight_only loading error." }, { "left": "issue:42915", "right": "issue:43322", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config and Llava Next segmentation fault are separate model-loading problems." + "reason": "Qwen3Moe FP8 config failure and Llava Next segmentation fault are distinct bugs." }, { "left": "issue:36331", "right": "issue:41720", "accept": false, - "reason": "Trainer API keyword mismatch is unrelated to a CUDA assert in auto device mapping." + "reason": "CustomTrainer compute_loss signature mismatch is unrelated to Qwen3 CUDA assertion on device mapping." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Gemma3TextConfig API discrepancy and GPTNeoX rotary_pct reload regression are different config issues." + "reason": "Gemma3TextConfig API discrepancy and GPTNeoX rotary_pct reload regression are different config bugs." }, { - "left": "issue:29159", - "right": "issue:36032", + "left": "issue:28530", + "right": "issue:30064", "accept": false, - "reason": "Slow/fast tokenizer inconsistency is not the same as the T5 add_special_tokens AttributeError." + "reason": "Early stopping metric selection and void segmentation-map processing are unrelated." }, { - "left": "issue:28530", - "right": "issue:30064", + "left": "issue:29159", + "right": "issue:36032", "accept": false, - "reason": "Early stopping/metric selection is unrelated to image processor failure on void segmentation maps." + "reason": "Tokenizer inconsistency and T5Tokenizer add_special_tokens conflict are both tokenizer-related but not the same bug." }, { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "MacOS bus error on a CLIP model and progressive generation with inputs_embeds/past_key_values are unrelated." + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are unrelated." }, { "left": "issue:43388", "right": "issue:45059", "accept": false, - "reason": "Gather-for-metrics label truncation and SAM3 PCS text/box behavior are different evaluation vs inference bugs." + "reason": "Metric gathering label truncation and SAM3 text/box behavior are unrelated." }, { - "left": "issue:43550", - "right": "issue:43653", + "left": "issue:30064", + "right": "issue:34634", "accept": false, - "reason": "Bamba torch.compile/SDPA failure and BigBirdTokenizer special-token registration bug are unrelated." + "reason": "Image processor void-segmentation failure and BarkProcessor voice_preset bug are unrelated." }, { - "left": "issue:30064", - "right": "issue:34634", + "left": "issue:43550", + "right": "issue:43653", "accept": false, - "reason": "Void-segmentation-map processing and BarkProcessor voice_preset are separate processor bugs." + "reason": "torch.compile SDPA failure and BigBirdTokenizer special-token registration are unrelated." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "Both involve offloading, but one is a device mismatch and the other is missing unpacking of packed weights." + "reason": "Both involve loading/offload, but Bark CPU-offload mismatch and BitNet packed-weight unpacking are different concrete code paths." }, { "left": "issue:43873", "right": "issue:45704", "accept": false, - "reason": "Quantization/offloading behavior and an apex FusedRMSNorm memory leak are different problems." + "reason": "Quantization/offloading behavior and T5 apex FusedRMSNorm memory leak are unrelated." }, { "left": "issue:44387", "right": "issue:45701", "accept": false, - "reason": "Int4 OOM from increased CUDA reserved memory is not the same as version-dependent tokenization changes." - }, - { - "left": "issue:41720", - "right": "issue:43746", - "accept": false, - "reason": "Auto device mapping CUDA assert and PEFT-adapter local checkpoint loading are unrelated." - }, - { - "left": "issue:45406", - "right": "issue:45464", - "accept": false, - "reason": "Gemma4Processor missing _tokenizer and Qwen3.5 streaming chat/completions failure are distinct serving issues." + "reason": "Int4 quantization OOM and tokenization changes across versions are unrelated." } ] }, @@ -4713,7 +4775,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 15, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -4759,18 +4821,20 @@ "issue:45072", "issue:45357", "issue:45381", + "issue:45406", + "issue:45464", "issue:45479", "issue:45588", "issue:45704", "issue:45759", "issue:45859", "issue:45923", - "issue:46032", "issue:46050", - "issue:46402", - "issue:46463" + "issue:46402" ], "soft_pairs": [ + "issue:41720|issue:43746", + "issue:45406|issue:45464", "issue:45357|issue:46402", "issue:44492|issue:44509", "issue:43531|issue:45759", @@ -4780,242 +4844,226 @@ "issue:34634|issue:41720", "issue:44451|issue:44991", "issue:43531|issue:44589", - "issue:44514|issue:44545", "issue:44625|issue:45479", - "issue:44898|issue:45072", "issue:44077|issue:44625", + "issue:44514|issue:44545", "issue:45759|issue:45859", + "issue:44898|issue:45072", "issue:44368|issue:45357", - "issue:44062|issue:44451", "issue:45588|issue:45759", + "issue:44062|issue:44451", "issue:44987|issue:45042", "issue:43421|issue:43927", "issue:39692|issue:44610", "issue:44918|issue:44936", "issue:43381|issue:43844", - "issue:43782|issue:45381", "issue:43746|issue:46050", + "issue:43782|issue:45381", "issue:42886|issue:44843", - "issue:45704|issue:45923", - "issue:46032|issue:46463", - "issue:43421|issue:43653", - "issue:45704|issue:45859", - "issue:41720|issue:45704" + "issue:45704|issue:45923" ], - "prepared_review_unit_hash": "ed92bb3c917d2600c10d6acd7563f2de77f0c977fee8cdd4a2ae5ca800b471f2", + "prepared_review_unit_hash": "a5fb58dd881f2b0f777fd17630fc62c05cd23eb1d21d4f10dff820e139826694", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 37021, - "estimated_input_tokens": 9256, - "estimated_eval_tokens": 18768 + "soft_pair_count": 28, + "serialized_chars": 36435, + "estimated_input_tokens": 9109, + "estimated_eval_tokens": 18474 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Only one soft pair looks like a true duplicate: the two Qwen2_5_VLProcessor.apply_chat_template crash reports. The rest are either same subsystem at a very broad level or clearly different bugs/features.", - "confidence": 0.93, - "canonical_issue_reason": "issue:44545 is the best issue candidate because it is the same crash report as 44514 (identical title and symptom: batched apply_chat_template with padding=False) and appears to be the more central duplicate target with more inbound references.", + "summary": "This cluster is mostly a set of unrelated issue-pair candidates; only the Qwen2_5_VLProcessor.apply_chat_template batched-input crash appears to be a true duplicate pair. No PRs are present in the input.", + "confidence": 0.97, + "canonical_issue_reason": "issue:44514 is the earlier and more complete report of the same Qwen2_5_VLProcessor.apply_chat_template crash on batched input with padding=False.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44545 is the strongest issue overall in this set because it most clearly represents a concrete, reproducible bug and is the only pair here that is effectively the same underlying failure as its counterpart.", + "best_issue_reason": "issue:44514 is the best representative issue because it cleanly describes the concrete failure mode and is duplicated by issue:44545.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Different models and failures: Qwen3 device-mapping CUDA assert vs GraniteSpeech PEFT local-checkpoint loading." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Different components and bugs: Gemma4Processor missing _tokenizer vs Qwen3.5 streaming chat/completions failure." + }, { "left": "issue:45357", "right": "issue:46402", "accept": false, - "reason": "Different models and different failure modes: Qwen3.5 visual encoder save_pretrained key regression vs Sam3Model text encoder weights not loading." + "reason": "Both concern multimodal weight loading/saving, but they target different models and different broken code paths." }, { "left": "issue:44492", "right": "issue:44509", "accept": false, - "reason": "Both are docs-related, but one is a typo and the other is outdated pipeline-task docs; not the same underlying issue." + "reason": "Unrelated documentation issues: a typo in cache strategies vs stale docs mentioning removed pipeline tasks." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Both mention Qwen, but one is a sliding_window bug in Qwen3-MoE and the other is from_config/text_config handling for composite multimodal configs." + "reason": "Different Qwen-related config problems: sliding-window behavior vs unwrapping text_config for multimodal from_config." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "Completely different bugs: Bark CPU-offload device mismatch vs BigBirdTokenizer special-token decode behavior." + "reason": "Completely different subsystems: Bark CPU offload device mismatch vs BigBirdTokenizer mask-token decode bug." }, { "left": "issue:42490", "right": "issue:43329", "accept": false, - "reason": "Unrelated tokenizer/path warning versus a multimodal token-counting video-branch NameError." + "reason": "Different bugs: import warning for trailing-slash paths vs undefined video-token helper names in multimodal token counting." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "TrainerState token accounting and progressive generation with inputs_embeds/past_key_values are different code paths." + "reason": "TrainerState token-count tracking and progressive generation with inputs_embeds/past_key_values are unrelated issues." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "BarkProcessor voice_preset and Qwen3 auto device map cuda assert are unrelated." + "reason": "BarkProcessor voice_preset handling is unrelated to Qwen3 auto device mapping CUDA assertions." }, { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "Both are model/tokenizer loading regressions, but they affect different models and likely different tokenizer paths." + "reason": "Both are tokenizer-loading regressions, but they affect different models and likely different failure paths." }, { "left": "issue:43531", "right": "issue:44589", "accept": false, - "reason": "Sliding-window Qwen3-MoE bug is unrelated to the Float8 storage TypeError." - }, - { - "left": "issue:44514", - "right": "issue:44545", - "accept": true, - "reason": "Same exact bug report: Qwen2_5_VLProcessor.apply_chat_template crashes on batched input when padding=False." + "reason": "Sliding-window Qwen3-MoE behavior is unrelated to the Float8 storage TypeError." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Qwen3.5 num_labels propagation and degenerate zero-loss classification behavior are distinct problems." - }, - { - "left": "issue:44898", - "right": "issue:45072", - "accept": false, - "reason": "Different subsystems and symptoms: Perceiver resize/interpolation vs dtype mismatch in inference." + "reason": "Different classification bugs: config propagation of num_labels vs degenerate zero-loss behavior with num_labels=1." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "Patchtsmixer post_init policy and Qwen3.5 num_labels propagation are unrelated." + "reason": "PatchTSMixer post_init policy is unrelated to Qwen3.5 num_labels propagation." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same title and same crash: Qwen2_5_VLProcessor.apply_chat_template fails on batched input when padding=False." }, { "left": "issue:45759", "right": "issue:45859", "accept": false, - "reason": "Both are Qwen3.5-related, but one is from_pretrained/config unwrapping and the other is CPU-offload rotary embedding incompatibility." + "reason": "Different Qwen3.5 multimodal issues: from_config text_config unwrapping vs CPU-offload incompatibility in rotary embedding." }, { - "left": "issue:44368", - "right": "issue:45357", + "left": "issue:44898", + "right": "issue:45072", "accept": false, - "reason": "One is a warning about tie_word_embeddings during LoRA fine-tuning; the other is a save_pretrained visual-key regression." + "reason": "Different model families and symptoms: Perceiver image resolution handling vs bfloat16 dtype mismatch in SwitchTransformers/TimmWrapperModel." }, { - "left": "issue:44062", - "right": "issue:44451", + "left": "issue:44368", + "right": "issue:45357", "accept": false, - "reason": "Tokenizer AddedToken keyword conflict is a different root cause than the ScandiBERT load failure." + "reason": "A training warning about tie_word_embeddings is unrelated to incorrect visual-encoder keys in save_pretrained." }, { "left": "issue:45588", "right": "issue:45759", "accept": false, - "reason": "Flash-attention s_aux handling and Qwen config unwrapping are unrelated." + "reason": "Flash-attention sink-less-model crash is unrelated to Qwen multimodal from_config text_config handling." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "AddedToken constructor TypeError is unrelated to the ScandiBERT loading failure." }, { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Different issues: loading a specific model vs PIL image processors incorrectly requiring torchvision." + "reason": "These are different regressions: loading a specific model vs PIL image processors incorrectly requiring torchvision." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Tokenizer post-processor refresh on special-token edits is unrelated to DiaConfig losing custom token IDs on save/load." + "reason": "Special-token post-processor updates and DiaConfig custom token ID persistence are related thematically but not the same bug." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "Both involve docs/model setup mismatches, but they concern different models and failures." + "reason": "Docs/example errors for SigLIP2 do not match the OmDet-Turbo processor/model resolution mismatch." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Different trainer failures: TRL SFT unpacking bug vs evaluate-after-train issue." + "reason": "Both mention trainers, but one is an SFT embedding-unpacking issue and the other is evaluate-after-train failure." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Gradient checkpointing eval-mode restriction and ZeRO-3 gradient growth are unrelated." + "reason": "Gradient checkpointing in eval mode and ZeRO-3 gradient growth are distinct training/runtime problems." }, { - "left": "issue:43782", - "right": "issue:45381", + "left": "issue:43746", + "right": "issue:46050", "accept": false, - "reason": "Different Qwen VL bugs: weight_only load failure vs wrong video vision_position_ids." + "reason": "Different checkpointing/loading mechanisms: PEFT local checkpoints vs sinq quantized save/load corruption." }, { - "left": "issue:43746", - "right": "issue:46050", + "left": "issue:43782", + "right": "issue:45381", "accept": false, - "reason": "Local PEFT checkpoint loading and sinq save/load quantization breakage are different serialization problems." + "reason": "Different Qwen VL bugs: weight_only loading error vs incorrect video vision_position_ids." }, { "left": "issue:42886", "right": "issue:44843", "accept": false, - "reason": "Both touch HF_HUB_OFFLINE behavior, but one is a general cache load failure and the other is a specific model_info call in _patch_mistral_regex." + "reason": "Both involve HF_HUB_OFFLINE, but one is cache loading and the other is an unconditional model_info call in tokenizer patching." }, { "left": "issue:45704", "right": "issue:45923", "accept": false, - "reason": "T5/apex RMSNorm memory leak and missing supports_gradient_checkpointing flag are unrelated." - }, - { - "left": "issue:46032", - "right": "issue:46463", - "accept": false, - "reason": "Mamba2 use_cache correctness bug and T5 decoder error-message bug are not the same issue." - }, - { - "left": "issue:43421", - "right": "issue:43653", - "accept": false, - "reason": "Tokenizer runtime special-token updates and BigBird mask-token decode output are different bugs." - }, - { - "left": "issue:45704", - "right": "issue:45859", - "accept": false, - "reason": "Different issues in different model paths: T5 RMSNorm leak vs Qwen3.5 MoE CPU-offload incompatibility." - }, - { - "left": "issue:41720", - "right": "issue:45704", - "accept": false, - "reason": "Qwen3 auto device mapping assert and T5 apex RMSNorm leak are unrelated." + "reason": "T5 RMSNorm memory-leak behavior is unrelated to a missing supports_gradient_checkpointing flag." } ] }, "evaluator_result": { "accept": true, - "feedback": "The evaluation is well grounded in the packet. The 44514/44545 pair is the only clearly exact duplicate by title and symptom, and choosing 44545 as the canonical issue is conservative given its higher inbound references. The remaining soft-edge rejections stay appropriately cautious and do not overstate identity across distinct models, symptoms, or code paths." + "feedback": "The summary is grounded in the packet: all listed items are issues, no PRs are present, and the only true duplicate pair shown is issue:44514 / issue:44545. The remaining soft-edge verdicts are conservative and stay within what the titles/body metadata support." }, "evaluator_used": true, "retried": false, @@ -5028,7 +5076,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 16, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -5038,12 +5086,11 @@ "nodes": [ "issue:30064", "issue:36010", - "issue:36331", "issue:39692", + "issue:41720", "issue:43295", "issue:43322", "issue:43335", - "issue:43388", "issue:43421", "issue:43531", "issue:43540", @@ -5054,7 +5101,6 @@ "issue:43645", "issue:43653", "issue:43697", - "issue:43716", "issue:43720", "issue:43746", "issue:43756", @@ -5078,20 +5124,27 @@ "issue:45479", "issue:45663", "issue:45685", + "issue:45704", "issue:45834", + "issue:45859", "issue:45907", "issue:45923", "issue:46032", - "issue:46294" + "issue:46294", + "issue:46463" ], "soft_pairs": [ + "issue:46032|issue:46463", + "issue:43421|issue:43653", + "issue:45704|issue:45859", + "issue:41720|issue:45704", "issue:43531|issue:43645", "issue:45663|issue:45923", "issue:43421|issue:44186", "issue:43994|issue:44671", "issue:43746|issue:43873", - "issue:45081|issue:45356", "issue:39692|issue:43873", + "issue:45081|issue:45356", "issue:43596|issue:44220", "issue:45356|issue:45907", "issue:43540|issue:46294", @@ -5107,230 +5160,221 @@ "issue:43540|issue:44008", "issue:44186|issue:44898", "issue:43653|issue:43819", - "issue:45183|issue:45479", "issue:43531|issue:43643", "issue:43322|issue:45685", + "issue:45183|issue:45479", "issue:43335|issue:43756", - "issue:44843|issue:45081", "issue:43643|issue:44877", - "issue:43716|issue:44008", - "issue:43388|issue:46032", - "issue:30064|issue:36331" + "issue:44843|issue:45081" ], - "prepared_review_unit_hash": "03c3148da6df7d0476bd635db8a55ee9d3c9d4a68df1ca8c467a407c6331236e", + "prepared_review_unit_hash": "b4f55f1d690b6f826e12640d3ec72b1787ac86a9668700e479ab98d46b6b664d", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 37021, - "estimated_input_tokens": 9256, - "estimated_eval_tokens": 18768 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37795, + "estimated_input_tokens": 9449, + "estimated_eval_tokens": 19154 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated issue reports spanning different models, tokenizer paths, training/runtime bugs, and doc regressions. None of the soft pairs look like the same underlying bug, though 44843 and 45081 are the closest because they touch the same mistral-regex codepath but describe different failures.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 44843 is the clearest representative: it has a specific, actionable regression in a named codepath (_patch_mistral_regex), a concrete offline-mode failure, and a well-scoped repro.", + "summary": "This is a heterogeneous set of transformer issues with only a loose tokenizer/Mistral-regex subtheme. None of the soft pairs look close enough to be true duplicates or mergeable into one fix. The most representative issue is the concrete `_patch_mistral_regex` crash report, but even the nearby tokenizer reports differ in root cause.", + "confidence": 0.83, + "canonical_issue_reason": "Issue 45081 is the clearest, most actionable report in the only obvious mini-cluster: it pinpoints `_patch_mistral_regex` with a concrete traceback and reproducer.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44843 is the best overall anchor because it is self-contained, clearly diagnosed, and easier to triage than the more model-specific or ambiguous reports.", + "best_issue_reason": "Issue 45081 is the best global representative because it is specific, reproducible, and closest to the other tokenizer-regex regressions in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:46032", + "right": "issue:46463", + "accept": false, + "reason": "Different subsystems and failure modes: Mamba2 cache correctness vs T5 decoder-input validation." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both involve tokenizers, but one is runtime special-token/post-processor sync and the other is BigBird mask-token registration; not the same bug." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "Unrelated code paths: T5 RMSNorm memory leak vs Qwen rotary embedding CPU-offload incompatibility." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "Qwen3 device-mapping crash and T5 RMSNorm leak are unrelated." + }, { "left": "issue:43531", "right": "issue:43645", "accept": false, - "reason": "Different bugs: a Qwen3-MoE sliding_window issue vs a Jupyter custom-model initialization regression." + "reason": "Different problems: Qwen3-MoE sliding-window behavior vs custom-model initialization in Jupyter." }, { "left": "issue:45663", "right": "issue:45923", "accept": false, - "reason": "Unrelated root causes: FSDP shared_kv_states rebuild bug vs a missing supports_gradient_checkpointing flag in a trust_remote_code model." + "reason": "Gemma4 FSDP2 shared-kv-state bug vs missing gradient-checkpointing flag; not the same defect." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Different tokenizer subsystems and failures: runtime post-processor updates vs LayoutLMv2 NER/padding crashes." + "reason": "Tokenizer special-token update issue vs LayoutLMv2 NER/padding crash; unrelated." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "Different model families and symptom classes: SigLIP2 AutoModel/pipeline behavior vs CamemBERT masked-LM predictions." + "reason": "Different model-specific correctness bugs: SigLIP2 pipeline outputs vs CamemBERT masked-LM predictions." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Different issues: local PEFT checkpoint loading for GraniteSpeech vs quantization/offloading behavior." + "reason": "PEFT/local-checkpoint loading bug vs quantization offloading issue; different fixes." }, { - "left": "issue:45081", - "right": "issue:45356", + "left": "issue:39692", + "right": "issue:43873", "accept": false, - "reason": "Both involve tokenizer regex handling, but the failures differ: backend_tokenizer AttributeError vs codec handling/warning regression for Kimi-K2.5." + "reason": "Documentation/example mismatch vs offloading-with-quantization runtime bug." }, { - "left": "issue:39692", - "right": "issue:43873", + "left": "issue:45081", + "right": "issue:45356", "accept": false, - "reason": "Doc/example errors for SigLIP2 are unrelated to quantization offloading behavior." + "reason": "Same helper area, but different concrete failures: backend_tokenizer AttributeError vs codec-handling/regression warning; too different to merge." }, { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "Different codepaths entirely: DeepSpeed ZeRO3/BertModel indexing vs audio fbank feature extraction." + "reason": "Deepspeed/BertModel initialization error and fbank feature extraction are unrelated." }, { "left": "issue:45356", "right": "issue:45907", "accept": false, - "reason": "Tokenizer regression vs Hub API forwarding bug; no shared underlying change." + "reason": "Tokenizer regression vs hub API revision forwarding; unrelated." }, { "left": "issue:43540", "right": "issue:46294", "accept": false, - "reason": "Both are video-related, but they affect different models and different stages of processing/chat templating." + "reason": "Both are video-related, but one is Qwen3OmniMoe input processing and the other is chat-template token/feature mismatch." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Image segmentation map processing vs an import failure for GenerationMixin are unrelated." + "reason": "Image segmentation-map processing vs a GenerationMixin import error; unrelated." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "Different model internals: packed-weight loading in BitNet vs DAC STE mismatch between from_latents and forward." + "reason": "BitNet accelerate-loading unpacking bug vs DAC latent/STE mismatch; different code paths." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "Tokenizer special-token registration bug vs BitNet accelerate-loading/unpacked-weights bug." + "reason": "BigBird special-token registration bug and BitNet weight-unpacking bug are unrelated." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Audio feature extraction bug vs add-new-model-like failing inside TOKENIZER_MAPPING_NAMES; different layers." + "reason": "Audio feature extraction bug vs add-new-model-like/tokenizer-mapping issue." }, { "left": "issue:43618", "right": "issue:43697", "accept": false, - "reason": "CLIPOutput attentions assignment regression is not the same as RTDetrV2 differing outputs in v5." + "reason": "CLIP attentions missing and RTDetrV2 output drift are separate regressions." }, { "left": "issue:43295", "right": "issue:45834", "accept": false, - "reason": "Processor/tokenizer regression vs long OCR input index error; different model paths and symptoms." + "reason": "Processor/tokenizer regression and Kosmos2.5 long-ocr index error are not the same bug." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "torch.compile+SDPA failure in Bamba-9B-v2 is unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "torch.compile SDPA crash vs bfloat16 dtype mismatch; different defects." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "AddedToken special-argument collision is unrelated to sleng-bert model loading failures." + "reason": "AddedToken keyword-argument error vs model-loading failure; unrelated." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Different failure modes and components: SDPA compile bug vs DAC latent reconstruction mismatch." - }, - { - "left": "issue:43540", - "right": "issue:44008", - "accept": false, - "reason": "Both are multimodal, but the root causes differ: Qwen3OmniMoe video-input validation vs Gemma3n variable-name collision in forward()." - }, - { - "left": "issue:44186", - "right": "issue:44898", - "accept": false, - "reason": "Tokenizer crash on NER/batching is unrelated to Perceiver image-classification interpolation failures." + "reason": "Bamba torch.compile/SDPA issue and DAC latent mismatch are unrelated." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "BigBirdTokenizer special-token registration and DAC STE behavior are unrelated bugs." - }, - { - "left": "issue:45183", - "right": "issue:45479", - "accept": false, - "reason": "Input validation error messaging vs a single-label classification zero-loss regression; different underlying logic." + "reason": "Tokenizer special-token bug vs DAC latent/forward mismatch; unrelated." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Qwen3-MoE sliding_window bug is unrelated to AutoConfig trust_remote_code missing fields." + "reason": "Qwen3-MoE sliding-window bug vs trust_remote_code AutoConfig missing-fields issue; unrelated." }, { "left": "issue:43322", "right": "issue:45685", "accept": false, - "reason": "Llava Next loading segfault and an MPS histogram op limitation are completely different issues." + "reason": "Llava Next loading segfault vs MPS histogram Int support error; unrelated." }, { - "left": "issue:43335", - "right": "issue:43756", + "left": "issue:45183", + "right": "issue:45479", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and Smollm3 RoPE-layer drop are distinct configuration/model-architecture bugs." + "reason": "Input-validation error messaging vs degenerate zero-loss classification behavior; different bugs." }, { - "left": "issue:44843", - "right": "issue:45081", + "left": "issue:43335", + "right": "issue:43756", "accept": false, - "reason": "Same mistral-regex area, but different bugs: offline model_info() call vs backend_tokenizer AttributeError with fix_mistral_regex=True." + "reason": "SwitchTransformers sparse-layer config bug vs Smollm3 RoPE-layer dropping issue." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "AutoConfig missing fields under trust_remote_code is not the same as strict config blocking granite_speech loading." - }, - { - "left": "issue:43716", - "right": "issue:44008", - "accept": false, - "reason": "Mistral-3 image dtype mismatch and Gemma3n audio_mel_mask AttributeError are unrelated." - }, - { - "left": "issue:43388", - "right": "issue:46032", - "accept": false, - "reason": "gather_for_metrics last-batch label truncation is unrelated to Mamba2Mixer use_cache correctness." + "reason": "Both touch config loading, but one is missing returned fields and the other is strict config rejection for granite_speech." }, { - "left": "issue:30064", - "right": "issue:36331", + "left": "issue:44843", + "right": "issue:45081", "accept": false, - "reason": "Void segmentation-map processing and CustomTrainer.compute_loss signature mismatch are unrelated failures." + "reason": "Both involve `_patch_mistral_regex`, but one is an unconditional hub call breaking offline mode and the other is a missing backend_tokenizer AttributeError; different failure points and fixes." } ] }, @@ -5346,7 +5390,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 17, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -5354,26 +5398,27 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:28218", - "issue:29159", + "issue:30064", "issue:34634", "issue:36010", + "issue:36331", "issue:39692", "issue:41720", + "issue:43388", "issue:43421", "issue:43540", "issue:43577", "issue:43582", "issue:43606", "issue:43644", - "issue:43720", + "issue:43716", "issue:43749", "issue:43782", "issue:43818", "issue:43825", - "issue:43827", "issue:43950", "issue:43994", + "issue:44008", "issue:44186", "issue:44442", "issue:44448", @@ -5381,7 +5426,6 @@ "issue:44485", "issue:44509", "issue:44561", - "issue:44610", "issue:44671", "issue:44805", "issue:44811", @@ -5394,9 +5438,9 @@ "issue:45084", "issue:45381", "issue:45584", - "issue:45701", "issue:45834", "issue:45987", + "issue:46032", "issue:46192", "issue:46291", "issue:46294", @@ -5404,245 +5448,228 @@ "issue:46512" ], "soft_pairs": [ + "issue:43716|issue:44008", + "issue:30064|issue:36331", + "issue:43388|issue:46032", "issue:41720|issue:44991", + "issue:44821|issue:45084", "issue:43950|issue:45987", "issue:44448|issue:44671", - "issue:44821|issue:45084", "issue:43582|issue:46512", "issue:39692|issue:45834", + "issue:45584|issue:45834", "issue:43421|issue:44811", + "issue:44186|issue:44442", "issue:43540|issue:45381", - "issue:45584|issue:45834", "issue:43644|issue:43749", - "issue:44186|issue:44442", "issue:43782|issue:44977", "issue:43577|issue:43818", "issue:43606|issue:45072", - "issue:46294|issue:46463", "issue:44479|issue:45834", + "issue:46294|issue:46463", "issue:43825|issue:43994", "issue:44485|issue:44509", - "issue:34634|issue:36010", "issue:44805|issue:44936", - "issue:45987|issue:46192", - "issue:44561|issue:45020", + "issue:34634|issue:36010", "issue:44991|issue:46291", - "issue:43577|issue:43582", - "issue:28218|issue:29159", - "issue:44610|issue:45072", - "issue:43720|issue:45072", - "issue:44671|issue:45701", - "issue:43825|issue:43827" + "issue:44561|issue:45020", + "issue:45987|issue:46192", + "issue:43577|issue:43582" ], - "prepared_review_unit_hash": "6ef9bfbb1b4031785acb17d985dbf40cba72110b2de8152dac4d031bfd496717", + "prepared_review_unit_hash": "e5570aae0337c6bddcc106db4a768802efbe7f07b44f53ae864d15c018a99cad", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36400, - "estimated_input_tokens": 9100, - "estimated_eval_tokens": 18456 + "soft_pair_count": 27, + "serialized_chars": 35971, + "estimated_input_tokens": 8993, + "estimated_eval_tokens": 18242 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Heterogeneous issue bucket with only loose semantic similarity; the soft pairs mostly span different models, subsystems, or root causes, so nearly all are false positives. No pull requests are present.", - "confidence": 0.43, - "canonical_issue_reason": "issue:45020 is the broadest umbrella in the set for recent v5/remote_code compatibility regressions, making it the best representative issue here.", + "summary": "Most pairs are clearly unrelated despite surface overlap; the only plausible duplicate is the v5 / `trust_remote_code` regression pair.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 45020 is the best umbrella issue in this set: it describes the broad remote-code breakage, and issue 44561 reads like a narrower concrete instance of the same regression.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the most reusable triage representative: broad enough to anchor the regression cluster, while the other items are narrower and more fragmented by model/subsystem.", + "best_issue_reason": "45020 is the strongest triage target because it is broader, more reusable as a canonical report, and most likely to subsume closely related remote-code failures.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Different models and different failure modes: image preprocessor dtype mismatch vs Gemma3n variable-name collision." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Unrelated areas: image processor void segmentation maps vs trainer `compute_loss` signature change." + }, + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Different code paths: metric gathering drops labels in the last batch vs Mamba2 `use_cache` producing wrong results." + }, { "left": "issue:41720", "right": "issue:44991", "accept": false, - "reason": "Different models and failure modes: CUDA assert during Qwen3 auto device mapping vs tokenizer load failure for EMBEDDIA/est-roberta." + "reason": "One is a CUDA assert in Qwen3 auto device mapping; the other is tokenizer loading failure for a specific model." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "Image processor URL loading and compiler template-node errors are unrelated." }, { "left": "issue:43950", "right": "issue:45987", "accept": false, - "reason": "Different bugs in different code paths: non-persistent buffer corruption vs StaticCache returning the wrong type." + "reason": "Both mention loading/cache behavior, but one is buffer corruption on `from_pretrained` while the other is a return-type contract bug." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Both are v4/v5 regression reports, but they target different model families and symptoms, not the same concrete bug." - }, - { - "left": "issue:44821", - "right": "issue:45084", - "accept": false, - "reason": "Unrelated subsystems: loading an AutoImageProcessor from URL vs a compile-time node/template error." + "reason": "Different models and symptoms: Pegasus output drift vs CamemBERT masked-LM prediction errors." }, { "left": "issue:43582", "right": "issue:46512", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError is unrelated to from_config leaving config.dtype stale." + "reason": "Apple Silicon warmup `TypeError` is unrelated to `from_config(dtype=...)` leaving `config.dtype` stale." }, { "left": "issue:39692", "right": "issue:45834", "accept": false, - "reason": "SigLIP2 documentation example errors and Kosmos2.5 OCR index error are different issues." + "reason": "Documentation/example issues for SigLIP2 vs an OCR index error in Kosmos2.5; not the same bug." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Whisper empty-transcription failure and Kosmos2.5 long-OCR index error are different concrete problems." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Runtime post-processor update bug vs Whisper batch_decode skip_special_tokens behavior; same broad tokenizer area but not the same bug." + "reason": "Tokenizer post-processor update behavior is distinct from Whisper `batch_decode` ignoring `skip_special_tokens`." }, { - "left": "issue:43540", - "right": "issue:45381", + "left": "issue:44186", + "right": "issue:44442", "accept": false, - "reason": "Different Qwen video regressions with different failure points and likely different fixes." + "reason": "Different tokenizer problems on different models: NER padding/truncation crash vs AutoTokenizer loading failure." }, { - "left": "issue:45584", - "right": "issue:45834", + "left": "issue:43540", + "right": "issue:45381", "accept": false, - "reason": "Whisper empty-transcription failure after align_special_tokens is unrelated to Kosmos2.5 long-input indexing." + "reason": "Both are Qwen video issues, but they involve different models and different broken paths." }, { "left": "issue:43644", "right": "issue:43749", "accept": false, - "reason": "Both concern loading, but one is non-persistent buffer junk and the other is FSDP CPU RAM efficient loading; not the same code-path." - }, - { - "left": "issue:44186", - "right": "issue:44442", - "accept": false, - "reason": "LayoutLMv2Tokenizer NER/padding crash and AutoTokenizer loading FastSpeech2ConformerTokenizer are distinct tokenizer failures." + "reason": "Non-persistent buffer junk on load is not the same as FSDP CPU RAM efficient loading being broken." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Different Qwen issues: weight_only loading error vs flash-attention generation failure." + "reason": "Load-time `weight_only=True` error and flash-attention generation failure are different failures." }, { "left": "issue:43577", "right": "issue:43818", "accept": false, - "reason": "BLIP2 dtype/state issue and Video-LLaVA temporal-attention/shared-weight issue are unrelated." + "reason": "BLIP2 dtype inconsistency is unrelated to Video-LLaVA missing temporal attention / weight-sharing." }, { "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "Both mention inference/loading problems, but bark CPU offload device mismatch is not the same as dtype mismatches in other models." + "reason": "CPU offload device mismatch and bfloat16 dtype mismatch are different regressions." }, { - "left": "issue:46294", - "right": "issue:46463", + "left": "issue:44479", + "right": "issue:45834", "accept": false, - "reason": "Video token/feature mismatch in Qwen3.5-9B is unrelated to T5 decoder-input validation messaging." + "reason": "Qwen video-input regression and Kosmos2.5 OCR index error are unrelated." }, { - "left": "issue:44479", - "right": "issue:45834", + "left": "issue:46294", + "right": "issue:46463", "accept": false, - "reason": "Qwen video regression and Kosmos2.5 OCR long-input index error are different failures." + "reason": "Qwen3.5 video chat-template mismatch is unrelated to T5 decoder-input validation messaging." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "One is an error-message regression around removed pipeline tasks; the other is a SigLIP2 AutoModel/pipeline quality issue." + "reason": "Pipeline task deprecation messaging does not match a SigLIP2 AutoModel/pipeline output bug." }, { "left": "issue:44485", "right": "issue:44509", "accept": false, - "reason": "GLM-5 RoPE implementation discussion and stale pipeline docs are not the same underlying bug." - }, - { - "left": "issue:34634", - "right": "issue:36010", - "accept": false, - "reason": "BarkProcessor voice_preset malfunction and GenerationMixin import failure are unrelated." + "reason": "GLM-5 RoPE implementation discussion is not the same as stale docs mentioning removed pipeline tasks." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "Mask/tensor shape IndexError and trainer.evaluate() after train are different issues." - }, - { - "left": "issue:45987", - "right": "issue:46192", - "accept": false, - "reason": "StaticCache seq-length contract bug is unrelated to return_tensors being ignored when text_kwargs is passed." + "reason": "Mask/index shape mismatch in model code is different from `trainer.evaluate()` failing after `train()`." }, { - "left": "issue:44561", - "right": "issue:45020", + "left": "issue:34634", + "right": "issue:36010", "accept": false, - "reason": "44561 pinpoints a specific removed helper; 45020 is a broad umbrella for remote_code breakage, so they are not clearly the same concrete bug." + "reason": "BarkProcessor `voice_preset` behavior and `GenerationMixin` import failure are unrelated." }, { "left": "issue:44991", "right": "issue:46291", "accept": false, - "reason": "Tokenizer load failure for a specific model and kernels v0.15 pipeline import failure are unrelated." - }, - { - "left": "issue:43577", - "right": "issue:43582", - "accept": false, - "reason": "Different failure modes: dtype handling in model loading vs a TypeError in caching_allocator_warmup." - }, - { - "left": "issue:28218", - "right": "issue:29159", - "accept": false, - "reason": "Added-token spacing bug is only loosely related to a general slow/fast tokenizer inconsistency report; not enough to prove the same root cause." - }, - { - "left": "issue:44610", - "right": "issue:45072", - "accept": false, - "reason": "Processor/model input-size mismatch is unrelated to bfloat16 dtype mismatches." - }, - { - "left": "issue:43720", - "right": "issue:45072", - "accept": false, - "reason": "Packed-weights unpacking during accelerate loading is a different problem from dtype mismatches." + "reason": "Tokenizer loading failure and `transformers.pipelines` import failure due to `kernels` versioning are separate issues." }, { - "left": "issue:44671", - "right": "issue:45701", - "accept": false, - "reason": "CamemBERT masked-LM regression and a generic tokenization-change report do not identify the same bug." + "left": "issue:44561", + "right": "issue:45020", + "accept": true, + "reason": "Both report the same v5/remote-code regression; 44561 is a specific instance of the broader breakage described in 45020." }, { - "left": "issue:43825", - "right": "issue:43827", + "left": "issue:45987", + "right": "issue:46192", "accept": false, - "reason": "Closely related topic area, but one is an error-message wording issue while the other is documentation stale after pipeline removals." + "reason": "`StaticCache.get_seq_length()` returning a tensor is unrelated to `return_tensors` being ignored when `text_kwargs` is passed." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The only accepted link is well-grounded in the packet: 45020 is the broad remote-code regression issue, and 44561 is a narrower v5 `trust_remote_code` failure that plausibly fits under it. The other soft-edge rejections stay conservative and are supported by the issue titles; I don\u2019t see overstatement." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -5651,7 +5678,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 18, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -5659,11 +5686,11 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:28218", + "issue:29159", "issue:34634", "issue:36331", - "issue:38175", "issue:42175", - "issue:43257", "issue:43262", "issue:43329", "issue:43452", @@ -5671,25 +5698,23 @@ "issue:43550", "issue:43582", "issue:43638", - "issue:43650", "issue:43653", "issue:43697", + "issue:43720", "issue:43742", - "issue:43746", "issue:43749", "issue:43782", "issue:43819", + "issue:43825", + "issue:43827", "issue:43906", - "issue:43927", "issue:44060", "issue:44291", "issue:44368", - "issue:44387", "issue:44554", "issue:44589", "issue:44610", - "issue:44617", - "issue:44811", + "issue:44671", "issue:44857", "issue:44912", "issue:44936", @@ -5697,6 +5722,7 @@ "issue:45003", "issue:45020", "issue:45042", + "issue:45072", "issue:45081", "issue:45084", "issue:45325", @@ -5704,11 +5730,17 @@ "issue:45406", "issue:45538", "issue:45600", + "issue:45701", "issue:46209", "issue:46285", "issue:46326" ], "soft_pairs": [ + "issue:44610|issue:45072", + "issue:28218|issue:29159", + "issue:43720|issue:45072", + "issue:44671|issue:45701", + "issue:43825|issue:43827", "issue:43782|issue:45325", "issue:43749|issue:44291", "issue:42175|issue:45081", @@ -5722,211 +5754,199 @@ "issue:45081|issue:45406", "issue:44060|issue:44368", "issue:45372|issue:46326", - "issue:45084|issue:45538", "issue:43742|issue:45020", + "issue:45084|issue:45538", "issue:43526|issue:44554", "issue:43329|issue:45600", "issue:44589|issue:45042", "issue:43452|issue:44991", "issue:43638|issue:44936", - "issue:43550|issue:44912", - "issue:43746|issue:44811", - "issue:43650|issue:43906", - "issue:43819|issue:43927", - "issue:43257|issue:43329", - "issue:44387|issue:44617", - "issue:38175|issue:44610" + "issue:43550|issue:44912" ], - "prepared_review_unit_hash": "1c628b17d266044fea95ae4b4f8da1d44fddb53fc2374e7924de27e09f311a17", + "prepared_review_unit_hash": "1543b4381e553766f85fd8b75d16799c3f8fcad4c80ea354423794bff590cd07", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 36059, - "estimated_input_tokens": 9015, - "estimated_eval_tokens": 18286 + "soft_pair_count": 26, + "serialized_chars": 35803, + "estimated_input_tokens": 8951, + "estimated_eval_tokens": 18158 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous and mostly not duplicate-safe: the soft-similarity pairs cover unrelated bugs across loading, tokenizers, processors, quantization, trainer behavior, and platform-specific runtime issues. No soft edge looks like the same concrete underlying bug/change.", - "confidence": 0.32, - "canonical_issue_reason": "issue:45020 is the broadest umbrella-style regression here, centered on recent model-loading / remote_code compatibility; it is the least bad representative of the noisy loading-related subset, though it is not a true duplicate of most other items.", + "summary": "This cluster is not a true duplicate set; it mixes many unrelated regression reports across tokenization, loading, processors, training, and backend-specific failures. The best representative issue is the broad remote_code regression report, but none of the soft pairs look safe to merge.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 45020 is the broadest, most representative regression report in the set and aligns with the general theme of recent-version breakages, though the cluster overall is heterogeneous.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the best cluster representative because it is broad, recent, and plausibly subsumes some of the loading-related noise better than the more narrow model-specific reports.", + "best_issue_reason": "Issue 45020 is the best single representative because it is high-level, clearly framed as a version regression, and can encompass several loading-related breakages better than the narrower model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Different models and different failures: OmDet-Turbo input-size mismatch vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": false, + "reason": "Both mention tokenizers, but one is a specific added-token spacing bug and the other is a broad slow/fast inconsistency report; not enough evidence they are the same defect." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Unrelated: BitNet packed-weight loading failure versus a bfloat16 dtype mismatch in other models." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "One is a CamemBERT masked-LM regression, the other is a general tokenization-version change report; too broad and not the same code path." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Related theme, but one is an error-message issue and the other is a documentation reference issue; different artifacts, not the same underlying bug." + }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "Different Qwen model families and different failures: weight_only loading error vs rope/position-id scaling bug." + "reason": "Different Qwen model bugs: a weight_only loading error versus a RoPE/position-id scaling bug." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "Both involve loading/init paths, but one is FSDP CPU RAM efficient loading and the other is an unexpected init_empty_weights argument error; not the same bug." + "reason": "Both are loading-related, but one is FSDP CPU RAM efficient loading and the other is an unexpected init_empty_weights argument; distinct failures." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "Unrelated subsystems: backend dependency packaging vs Mistral tokenizer regex patch crash." + "reason": "Completely different areas: backend installation metadata versus a Mistral regex patch crash." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "Different platforms and code paths: Apple Silicon allocator warmup TypeError vs CUDA AMP image-loss crash." + "reason": "Different runtime paths and symptoms: Apple Silicon allocator warmup TypeError versus CUDA AMP loss crash." }, { "left": "issue:43906", "right": "issue:46209", "accept": false, - "reason": "An isolated repro of an older issue is not the same as the FP8/B200 inference failure." + "reason": "No shared underlying bug: an isolated reproduction of an older issue versus an FP8 inference failure on B200." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sampling-rate default bug vs DAC latent reconstruction mismatch; different components." + "reason": "Audio chat-template sampling-rate default bug is unrelated to DAC latent reconstruction mismatch." }, { "left": "issue:45003", "right": "issue:46285", "accept": false, - "reason": "sys.modules access safety issue is unrelated to the CLIP text-model structure inconsistency." + "reason": "Different concerns: unsafe sys.modules access versus CLIP text model structure inconsistency." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "Different models and symptoms: Beit label reduction bug vs RTDetrV2 output divergence in v5." + "reason": "Different model components and symptoms: BeitImageProcessorFast label reduction versus RTDetrV2 output drift." }, { "left": "issue:43653", "right": "issue:44912", "accept": false, - "reason": "Tokenizer special-token registration bug vs MXFP4 quantization fallback; not the same underlying defect." + "reason": "Tokenizer special-token registration bug is unrelated to MXFP4 quantization fallback during model loading." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "Completely different areas: Bark voice preset handling vs trainer compute_loss signature mismatch." + "reason": "BarkProcessor voice_preset failure and CustomTrainer.compute_loss signature break are unrelated." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Both involve processor/tokenizer loading, but one is a Mistral regex patch crash and the other is Gemma4 serve missing _tokenizer." + "reason": "Both affect processor/tokenizer loading, but one is a Mistral regex patch crash and the other is a Gemma4Processor missing _tokenizer attribute; different causes." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "Similar warning theme, but different models and causes; not a shared code-path bug." + "reason": "Both mention tied-weights warnings, but one is an incorrect weight tying diagnosis and the other is a repeated warning during ms-swift fine-tuning; not clearly the same bug." }, { "left": "issue:45372", "right": "issue:46326", "accept": false, - "reason": "Different mistral_common-related failures: ReasoningEffort import error vs continue_final_message behavior bug." + "reason": "Different mistral_common-related failures: missing ReasoningEffort import versus continue_final_message behavior." }, { - "left": "issue:45084", - "right": "issue:45538", + "left": "issue:43742", + "right": "issue:45020", "accept": false, - "reason": "Template compilation TypeError and CLIPTokenizer max-length issue are unrelated." + "reason": "Specific MobileLLM load KeyError versus a broad remote_code regression; too different to be duplicates." }, { - "left": "issue:43742", - "right": "issue:45020", + "left": "issue:45084", + "right": "issue:45538", "accept": false, - "reason": "Both are loading failures, but one is a specific KeyError for MobileLLM and the other is a broad remote_code compatibility regression; not the same bug." + "reason": "Compile-time template-node error and CLIPTokenizer max-length behavior are unrelated." }, { "left": "issue:43526", "right": "issue:44554", "accept": false, - "reason": "Different model families and failure modes: Beit label processing vs MPS attention correctness." + "reason": "No overlap in code path: label reduction in BEiT versus MPS attention correctness with mismatched head dims." }, { "left": "issue:43329", "right": "issue:45600", "accept": false, - "reason": "Video-token counting undefined variables vs auto_mappings CI break from removed configs; distinct issues." + "reason": "Video multimodal token counting bug is unrelated to an auto_mappings CI break from removed configs." }, { "left": "issue:44589", "right": "issue:45042", "accept": false, - "reason": "Storage-type lookup failure on one hand, torchvision backend dependency requirement on the other; unrelated." + "reason": "Float8 storage lookup failure is unrelated to PIL backend processors incorrectly requiring torchvision." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "Different tokenizer/model-loading regressions: gguf_file handling vs EMBEDDIA tokenizer load failure." + "reason": "Both are tokenizer/model loading reports, but one is about gguf_file handling for from_pretrained and the other is a specific tokenizer load regression for est-roberta; not the same defect." }, { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "deepspeed zero3 empty-tensor IndexError vs trainer.evaluate() after train; different trainer bugs." + "reason": "Different training failures: DeepSpeed ZeRO-3 indexing error versus evaluate() failing after train()." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "torch.compile SDPA failure in Bamba is not the same as MXFP4 quantization falling back to bf16." - }, - { - "left": "issue:43746", - "right": "issue:44811", - "accept": false, - "reason": "PEFT local-checkpoint loading vs Whisper batch_decode skip_special_tokens handling; unrelated." - }, - { - "left": "issue:43650", - "right": "issue:43906", - "accept": false, - "reason": "A placeholder issue and an isolated repro are not duplicates." - }, - { - "left": "issue:43819", - "right": "issue:43927", - "accept": false, - "reason": "Both can affect generation, but one is DAC STE mismatch and the other is DiaConfig losing custom token IDs after save/load; different bugs." - }, - { - "left": "issue:43257", - "right": "issue:43329", - "accept": false, - "reason": "Qwen3 MoE weight conversion issue vs multimodal video token counting bug; not the same code path." - }, - { - "left": "issue:44387", - "right": "issue:44617", - "accept": false, - "reason": "Both mention OOM, but one is int4 quantization memory regression and the other is Sam3Video model OOM; too broad to merge." - }, - { - "left": "issue:38175", - "right": "issue:44610", - "accept": false, - "reason": "Siglip2 zero-probabilities issue and OmDet-Turbo processor image-size mismatch are different vision bugs." + "reason": "Bamba torch.compile SDPA failure and MXFP4 quantization fallback are unrelated model/runtime issues." } ] }, @@ -5942,7 +5962,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 19, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -5950,10 +5970,10 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:29159", "issue:34567", "issue:36010", "issue:36331", + "issue:38175", "issue:39692", "issue:40990", "issue:42175", @@ -5962,36 +5982,35 @@ "issue:43257", "issue:43262", "issue:43329", - "issue:43404", "issue:43450", "issue:43454", - "issue:43525", "issue:43531", "issue:43577", "issue:43606", "issue:43643", "issue:43645", + "issue:43650", "issue:43696", "issue:43720", "issue:43742", + "issue:43746", "issue:43749", + "issue:43819", "issue:43854", + "issue:43906", + "issue:43927", "issue:43931", "issue:44291", + "issue:44387", "issue:44479", - "issue:44568", "issue:44610", + "issue:44617", "issue:44756", "issue:44811", - "issue:44821", "issue:44912", "issue:44977", - "issue:44987", "issue:45042", - "issue:45127", - "issue:45356", "issue:45459", - "issue:45538", "issue:45759", "issue:45803", "issue:45859", @@ -6000,13 +6019,19 @@ "issue:46424" ], "soft_pairs": [ + "issue:43746|issue:44811", + "issue:43650|issue:43906", + "issue:43819|issue:43927", + "issue:43257|issue:43329", + "issue:44387|issue:44617", "issue:36010|issue:43645", "issue:42175|issue:46291", + "issue:38175|issue:44610", "issue:43720|issue:44610", - "issue:43854|issue:44977", "issue:36010|issue:36331", - "issue:39692|issue:44912", + "issue:43854|issue:44977", "issue:43531|issue:45042", + "issue:39692|issue:44912", "issue:46421|issue:46424", "issue:45459|issue:45803", "issue:42490|issue:43257", @@ -6020,211 +6045,205 @@ "issue:43257|issue:43931", "issue:34567|issue:40990", "issue:43606|issue:45859", - "issue:29159|issue:34567", - "issue:44568|issue:45356", "issue:42915|issue:43742", - "issue:43404|issue:45127", - "issue:44291|issue:45042", - "issue:43525|issue:46291", - "issue:42915|issue:44821", - "issue:44987|issue:45538" + "issue:44291|issue:45042" ], - "prepared_review_unit_hash": "f894d97fa2a3db53137c6e48bf6379b78a72d78d65c12a0c97fcd94b91b046b3", + "prepared_review_unit_hash": "c1f14f410a7d29ee8e10bc54d18e692318668ba9523ccd9c4f881eda8a14f6bc", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 36372, - "estimated_input_tokens": 9093, - "estimated_eval_tokens": 18442 + "serialized_chars": 35671, + "estimated_input_tokens": 8918, + "estimated_eval_tokens": 18092 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated one-off bugs across different subsystems; none of the soft-linked pairs look like the same underlying issue, and there are no PRs in the set.", - "confidence": 0.74, - "canonical_issue_reason": "issue:43257 is the closest hub among the provided issues: it is a concrete loading/dispatch bug and shows the strongest linkage to other candidates, even though the overall cluster is still very heterogeneous.", + "summary": "These are mostly unrelated standalone bug reports spanning different subsystems, models, and failure modes. None of the soft-paired items look like the same underlying bug/change, so I would not merge any pair as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43749 is the strongest standalone report: it has substantial discussion/inbound references and a clearly scoped broken loading mode, making it the best candidate to keep as a representative issue.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Different code paths and symptoms: GraniteSpeech PEFT/local checkpoint loading vs Whisper batch_decode skipping special tokens." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "One is a vague placeholder issue; the other is a reproduction of a specific upstream bug. No shared underlying defect." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "Different model bugs: DAC latents/STE mismatch vs DiaConfig losing custom token IDs on save/load." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MoE weight conversion/loading with accelerate+deepspeed is unrelated to multimodal token counting video-branch NameErrors." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but one is an int4 quantization memory regression and the other is a model-specific video OOM." + }, { "left": "issue:36010", "right": "issue:43645", "accept": false, - "reason": "Different problems: GenerationMixin import error vs custom model init in notebooks." + "reason": "Importing GenerationMixin and notebook custom-model initialization failures are different regression classes." }, { "left": "issue:42175", "right": "issue:46291", "accept": false, - "reason": "Unrelated dependency issues: backend extras packaging vs pipelines import failing on kernels version/revision requirements." + "reason": "Packaging/backend dependency omission is not the same as pipelines import failing on kernels v0.15 version requirements." }, { - "left": "issue:43720", + "left": "issue:38175", "right": "issue:44610", "accept": false, - "reason": "Different code paths: accelerate weight unpacking vs processor/model input-size mismatch." + "reason": "Different models and failures: zero probabilities in SigLIP2 vs image processor/model input-size mismatch in OmDet-Turbo." }, { - "left": "issue:43854", - "right": "issue:44977", + "left": "issue:43720", + "right": "issue:44610", "accept": false, - "reason": "Different models and failures: unit-test loading error for GLM vs flash-attention generation regression for Qwen3.5." + "reason": "BitNet packed-weight unpacking during accelerate loading is unrelated to OmDet-Turbo processor shape mismatch." }, { "left": "issue:36010", "right": "issue:36331", "accept": false, - "reason": "Both are trainer/import-related, but the concrete bugs differ: import path issue vs custom compute_loss signature breakage." + "reason": "GenerationMixin import error and compute_loss(num_items_in_batch) API break are separate trainer/import issues." }, { - "left": "issue:39692", - "right": "issue:44912", + "left": "issue:43854", + "right": "issue:44977", "accept": false, - "reason": "Documentation example errors vs quantization fallback bug; not the same underlying defect." + "reason": "Different model-specific failures; one is a GLM-4.7-Flash load issue, the other a Qwen3.5 flash-attention generation regression." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Different subsystems: Qwen3-MoE sliding-window behavior vs image processor torchvision dependency regression." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PIL backend processors requiring torchvision." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Both involve model/examples, but one is a SigLIP2 doc example with processor mismatch/quantization errors and the other is an MXFP4 loading regression for git-oss-20b." }, { "left": "issue:46421", "right": "issue:46424", "accept": false, - "reason": "Both are cache-related, but one is lazy-init recompilation and the other is max_cache_length semantics; not one bug." + "reason": "Both mention cache, but one is lazy initialization causing recompilation and the other is max_cache_length only applying to quantized cache." }, { "left": "issue:45459", "right": "issue:45803", "accept": false, - "reason": "Both involve exception handling, but they affect different functions and failure modes." + "reason": "Both involve exception handling, but they are different functions and bugs: tokenizer protobuf decode handling vs Fuyu tensor conversion bare except." }, { "left": "issue:42490", "right": "issue:43257", "accept": false, - "reason": "Local-path import warning vs Qwen3 MoE weight conversion during accelerate/deepspeed loading are unrelated." + "reason": "Local-path trailing-slash import warning is unrelated to Qwen3 MoE weight conversion under accelerate+deepspeed." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Audio processor default sampling-rate bug vs Whisper batch_decode special-token handling." + "reason": "Different audio processor bugs: apply_chat_template sample-rate default vs batch_decode skip_special_tokens handling." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Qwen3 MoE loading/weight conversion bug vs an OOM report on gpt-oss-20b." + "reason": "MoE weight conversion/loading problem is unrelated to GPT-oss-20b CUDA OOM." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "Both are config-loading issues, but one is missing fields with trust_remote_code and the other is composite-config unwrapping for multimodal models." + "reason": "Missing fields from trust_remote_code config objects is not the same as unwrapping text_config for composite multimodal configs." }, { "left": "issue:43454", "right": "issue:43577", "accept": false, - "reason": "Different model bugs: untied lm_head weights vs dtype remaining float32 on BLIP2 load." + "reason": "AyaVision weight tying/generation garbage and BLIP2 dtype propagation are distinct model issues." }, { "left": "issue:43329", "right": "issue:44479", "accept": false, - "reason": "Specific undefined video-token helper bug vs broader video-input regression; too different to be the same fix." + "reason": "Both touch video paths, but one is undefined helper names in token counting while the other is a broader v5.3.0 video input regression." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Undefined video token helper bug vs batched video processor shape bug." + "reason": "Undefined multimodal token-count helpers and batched video processor shape errors are different failure modes." }, { "left": "issue:43749", "right": "issue:44756", "accept": false, - "reason": "Both involve loading/memory, but one is FSDP RAM-efficient loading breakage and the other is mmap-induced OOM on Strix Halo." + "reason": "FSDP CPU RAM-efficient loading breakage is unrelated to disabling mmap on Strix Halo to avoid OOM." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Different Qwen3-related failures: accelerate/deepspeed weight conversion vs VL weight-shape mismatch." + "reason": "Qwen3 MoE accelerate/deepspeed weight conversion is unrelated to Qwen3-VL weight-shape mismatch." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "Trainer token-count tracking vs perplexity on gpt-oss-20b are unrelated." + "reason": "TrainerState token accounting and GPT-oss perplexity on WikiText are unrelated evaluation/training issues." }, { "left": "issue:43606", "right": "issue:45859", "accept": false, - "reason": "Both mention CPU offload, but they affect different models and code paths." - }, - { - "left": "issue:29159", - "right": "issue:34567", - "accept": false, - "reason": "Tokenizer fast/slow inconsistency vs TrainerState token counter update are unrelated." - }, - { - "left": "issue:44568", - "right": "issue:45356", - "accept": false, - "reason": "Both are tokenizer regressions, but for different models and different failure modes." + "reason": "Both involve CPU offload, but they affect different models and code paths; not the same underlying bug." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "FineGrainedFP8Config failure for Qwen3Moe vs key error loading MobileLLM-125M." - }, - { - "left": "issue:43404", - "right": "issue:45127", - "accept": false, - "reason": "Weight-tying bug in Mistral3 vs LoRA merge collapse with extended vocab are not the same concrete issue." + "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to a key error while loading MobileLLM-125M." }, { "left": "issue:44291", "right": "issue:45042", "accept": false, - "reason": "Init-empty-weights argument error vs torchvision dependency regression are unrelated." - }, - { - "left": "issue:43525", - "right": "issue:46291", - "accept": false, - "reason": "Missing pad_token_id on Llama4Config vs pipelines import failing on kernels version/revision requirements." - }, - { - "left": "issue:42915", - "right": "issue:44821", - "accept": false, - "reason": "Qwen3Moe FP8 loading bug vs AutoImageProcessor URL loading failure are different." - }, - { - "left": "issue:44987", - "right": "issue:45538", - "accept": false, - "reason": "Model loading regression for physical-intelligence/fast vs CLIPTokenizer max_length bug; no shared code path." + "reason": "init_empty_weights/_is_hf_initialized loading regression is unrelated to PIL backend torchvision dependency requirements." } ] }, @@ -6240,7 +6259,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 20, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -6260,9 +6279,10 @@ "issue:42175", "issue:42915", "issue:43317", + "issue:43404", "issue:43425", "issue:43452", - "issue:43493", + "issue:43525", "issue:43540", "issue:43577", "issue:43582", @@ -6271,21 +6291,20 @@ "issue:43749", "issue:43824", "issue:43927", - "issue:43994", "issue:44062", "issue:44291", "issue:44479", - "issue:44488", "issue:44560", "issue:44568", - "issue:44737", + "issue:44821", "issue:44849", "issue:44898", "issue:44912", "issue:44987", "issue:44991", - "issue:45072", + "issue:45127", "issue:45137", + "issue:45356", "issue:45357", "issue:45381", "issue:45405", @@ -6294,242 +6313,258 @@ "issue:45701", "issue:45907", "issue:46291", - "issue:46463", "issue:46512" ], "soft_pairs": [ + "issue:44568|issue:45356", + "issue:29159|issue:34567", + "issue:43525|issue:46291", + "issue:43404|issue:45127", + "issue:42915|issue:44821", "issue:43927|issue:44912", + "issue:44987|issue:45538", "issue:43697|issue:45357", "issue:44849|issue:45701", "issue:42915|issue:45538", "issue:43577|issue:46512", - "issue:43452|issue:43582", "issue:43425|issue:46291", + "issue:43452|issue:43582", "issue:42175|issue:43824", "issue:43317|issue:45538", "issue:33357|issue:34567", "issue:41720|issue:43540", - "issue:44062|issue:44991", - "issue:44479|issue:45381", "issue:28282|issue:34689", + "issue:44479|issue:45381", "issue:45137|issue:45405", + "issue:44062|issue:44991", "issue:43540|issue:44479", "issue:43540|issue:44560", - "issue:44568|issue:45488", "issue:43742|issue:44291", + "issue:44568|issue:45488", "issue:28282|issue:29127", "issue:44987|issue:45907", "issue:24643|issue:29159", "issue:42915|issue:43749", - "issue:44737|issue:46463", - "issue:38175|issue:44898", - "issue:42175|issue:44488", - "issue:43493|issue:43994", - "issue:43577|issue:45072" + "issue:38175|issue:44898" ], - "prepared_review_unit_hash": "2f33577c62a111c71c1c9582ea573bbbac198ede1ba016ff270178b6a18befd8", + "prepared_review_unit_hash": "dae6f411e0baaba8bdad1345aa4eb0841a5b8b4d4f9902ba8fc76a8ff0bd4f91", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 36174, - "estimated_input_tokens": 9044, - "estimated_eval_tokens": 18344 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36235, + "estimated_input_tokens": 9059, + "estimated_eval_tokens": 18374 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the suggested pairs are mostly broad topical similarities, not the same concrete bug or change. I would not merge any of the soft edges, and there is no clear single canonical issue for the whole set.", - "confidence": 0.91, - "canonical_issue_reason": null, + "summary": "The cluster is mostly a set of unrelated closed issues across many subsystems. The only likely duplicate is the Qwen video-input regression pair, where one report is a narrower instance of the broader v5.3.0 regression.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 44479 is the best canonical issue: it describes the broad v5.3.0 video-input regression across multiple Qwen model variants, and it naturally subsumes the narrower qwen2.5-vl-specific report.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "44479 is the strongest representative because it is broader, clearer about the regression scope, and more likely to attract all related discussion and fixes than the narrower follow-up report.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer-related, but they describe different regressions: special-token insertion vs codec/warning handling for a specific model family." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Inconsistent slow/fast tokenizer behavior is unrelated to TrainerState's token counter not updating." + }, + { + "left": "issue:43525", + "right": "issue:46291", + "accept": false, + "reason": "A missing pad_token_id on Llama4Config is unrelated to the kernels package now requiring version/revision." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both mention tied weights, but one is a model-class weight-tying bug and the other is a LoRA merge/extended-vocab collapse; not the same concrete bug." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "FineGrainedFP8Config loading failure is unrelated to AutoImageProcessor URL loading." + }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "Both involve model-loading/dtype behavior, but one is about DiaConfig token IDs causing generation errors and the other is about MXFP4 falling back to bf16; different bugs." + "reason": "Config save/load token-id loss causing IndexError is unrelated to MXFP4 quantization fallback behavior." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "A model-loading failure for physical-intelligence/fast is unrelated to CLIPTokenizer's max_length bug." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "RTDetrV2 output drift and Qwen3.5 save_pretrained key corruption are unrelated code paths and symptoms." + "reason": "RTDetrV2 output differences in v5 and Qwen3.5 save_pretrained key corruption are different regressions." }, { "left": "issue:44849", "right": "issue:45701", "accept": false, - "reason": "One is a Qwen3.5 hidden-states bug; the other is a broad tokenization-change report. Too generic to be the same issue." + "reason": "Qwen3.5 output_hidden_states behavior is not the same as the broad tokenization-change report." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Qwen3Moe FP8 loading and CLIPTokenizer max_length are unrelated." + "reason": "FineGrainedFP8Config loading and CLIPTokenizer max_length are unrelated." }, { "left": "issue:43577", "right": "issue:46512", "accept": false, - "reason": "Both mention dtype, but one is BLIP2 loading not casting correctly and the other is from_config leaving config.dtype stale; not the same failure." + "reason": "One is a model dtype propagation bug in Blip2 loading; the other is config.dtype staying stale after from_config. Similar theme, different bug." }, { - "left": "issue:43452", - "right": "issue:43582", + "left": "issue:43425", + "right": "issue:46291", "accept": false, - "reason": "gguf_file loading and Apple Silicon warmup TypeError are different subsystems and errors." + "reason": "Torch compatibility complaints are unrelated to kernels import requirements." }, { - "left": "issue:43425", - "right": "issue:46291", + "left": "issue:43452", + "right": "issue:43582", "accept": false, - "reason": "Torch compatibility and kernels revision/version requirements are different dependency issues." + "reason": "gguf_file loading errors and caching_allocator_warmup TypeError are unrelated." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Pip extra dependency packaging and missing Qwen2.5-VL import are unrelated." + "reason": "Missing TensorFlow backend in the torch extra is unrelated to the missing Qwen2_5_VLForConditionalGeneration import." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "GPU offload/loading failure is not the same as CLIPTokenizer max_length behavior." + "reason": "device_map/offload loading failure is unrelated to CLIPTokenizer's model_max_length bug." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "MacOS bus error in CLIP loading and TrainerState token counter not updating are unrelated." + "reason": "A MacOS bus error loading CLIP and a TrainerState counter bug are unrelated." }, { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Both are Qwen video-related, but one is an auto device-map cudaErrorAssert and the other is a video-processing ValueError; different concrete bugs." + "reason": "Both involve Qwen models, but one is an auto-device-map cuda assert and the other is a video-input ValueError; not the same code-path problem." }, { - "left": "issue:44062", - "right": "issue:44991", + "left": "issue:28282", + "right": "issue:34689", "accept": false, - "reason": "AddedToken special-argument duplication and tokenizer loading failure for EMBIDDIA/est-roberta are not the same issue." + "reason": "PyTorch missing import error is unrelated to Llama 3.2 Vision model-loading breakage." }, { "left": "issue:44479", "right": "issue:45381", - "accept": false, - "reason": "Both concern Qwen video input regressions, but one is a broad multi-model regression and the other specifically flags wrong vision_position_ids; not enough to treat as identical." + "accept": true, + "reason": "45381 is a narrower instance of the broader v5.3.0 Qwen video-input regression described in 44479, with the same underlying wrong-position-id behavior." }, { - "left": "issue:28282", - "right": "issue:34689", + "left": "issue:45137", + "right": "issue:45405", "accept": false, - "reason": "Missing PyTorch import and Llama 3.2 Vision loading breakage are unrelated." + "reason": "DeepSpeed ZeRO3 deque failure and an unreleased MIN_PEFT_VERSION bump are unrelated." }, { - "left": "issue:45137", - "right": "issue:45405", + "left": "issue:44062", + "right": "issue:44991", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error and MIN_PEFT_VERSION bump are unrelated." + "reason": "AddedToken constructor conflicts and tokenizer-loading failure for EMBEDDIA/est-roberta are different issues." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Different Qwen video failures with different symptoms and likely different fixes." + "reason": "Qwen3OmniMoe video processing ValueError is not the same bug as the broader Qwen video-input regression reported in 44479." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Video ValueError in Qwen3OmniMoe and StopIteration in Qwen3-vl-embedding are distinct bugs." + "reason": "These are separate video-input failures for different Qwen models and error modes." }, { - "left": "issue:44568", - "right": "issue:45488", + "left": "issue:43742", + "right": "issue:44291", "accept": false, - "reason": "mdeberta token addition bug and LlamaTokenizer pre-tokenizer override break DeepSeek are different tokenizer issues." + "reason": "A key error loading a specific model is unrelated to init_empty_weights rejecting an unexpected argument." }, { - "left": "issue:43742", - "right": "issue:44291", + "left": "issue:44568", + "right": "issue:45488", "accept": false, - "reason": "MobileLLM key error and init_empty_weights unexpected argument are separate loading problems." + "reason": "Both are tokenizer-related regressions, but one is special-token addition and the other is LlamaTokenizer pre-tokenizer override; different bugs." }, { "left": "issue:28282", "right": "issue:29127", "accept": false, - "reason": "Missing PyTorch import and LayoutLMv3 error-message clarity are unrelated." + "reason": "PyTorch missing import and LayoutLMv3 error-message clarity are unrelated." }, { "left": "issue:44987", "right": "issue:45907", "accept": false, - "reason": "Loading a specific model under newer transformers and list_repo_files revision forwarding are different API breakages." + "reason": "A model-loading failure is unrelated to list_repo_files needing to forward revision." }, { "left": "issue:24643", "right": "issue:29159", "accept": false, - "reason": "DeepSpeed weight shape RuntimeError and tokenizer slow/fast inconsistency are unrelated." + "reason": "DeepSpeed weight dimensionality error is unrelated to slow/fast tokenizer inconsistency." }, { "left": "issue:42915", "right": "issue:43749", "accept": false, - "reason": "FP8 config failure and FSDP CPU RAM efficient loading are different training/quantization issues." - }, - { - "left": "issue:44737", - "right": "issue:46463", - "accept": false, - "reason": "XLNet CPU arange bug and T5 decoder-input validation message are unrelated." + "reason": "FineGrainedFP8Config loading and FSDP CPU RAM efficient loading are unrelated training/loading bugs." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "SigLIP2 zero probabilities and Perceiver non-default-resolution failure are different vision-model bugs." - }, - { - "left": "issue:42175", - "right": "issue:44488", - "accept": false, - "reason": "TensorFlow backend packaging and a model-loading failure for cjvt/sleng-bert are unrelated." - }, - { - "left": "issue:43493", - "right": "issue:43994", - "accept": false, - "reason": "SigLIP2 implementation discrepancy and AutoModel/pipeline nonsense outputs are different symptom sets." - }, - { - "left": "issue:43577", - "right": "issue:45072", - "accept": false, - "reason": "BLIP2 dtype loading and SwitchTransformers/TimmWrapperModel bfloat16 mismatch are separate dtype bugs." + "reason": "Unexpected zero probabilities in SIGLIP2 and Perceiver non-default-resolution failure are different model-specific issues." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays grounded in the packet and is appropriately conservative. The only accepted soft-edge pair, issue:44479 with issue:45381, is reasonably supported by the titles as a broader/narrower Qwen video-input regression pair. The other rejections are phrased cautiously and do not overclaim duplicate evidence." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -6538,7 +6573,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 21, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -6550,34 +6585,34 @@ "issue:28282", "issue:34567", "issue:36331", + "issue:42175", "issue:42796", "issue:43257", "issue:43404", "issue:43452", + "issue:43493", "issue:43526", "issue:43531", + "issue:43577", "issue:43643", "issue:43697", - "issue:43720", "issue:43742", "issue:43756", "issue:43906", + "issue:43994", "issue:44062", "issue:44291", "issue:44488", "issue:44568", "issue:44589", - "issue:44625", "issue:44737", "issue:44792", "issue:44821", - "issue:44877", - "issue:44898", "issue:44913", "issue:44964", "issue:45042", + "issue:45072", "issue:45081", - "issue:45200", "issue:45216", "issue:45325", "issue:45357", @@ -6596,17 +6631,21 @@ "issue:46491" ], "soft_pairs": [ + "issue:44737|issue:46463", + "issue:42175|issue:44488", + "issue:43493|issue:43994", + "issue:43577|issue:45072", "issue:43257|issue:43756", "issue:44291|issue:44821", "issue:43906|issue:46468", "issue:34567|issue:36331", - "issue:43526|issue:45479", - "issue:44568|issue:46491", "issue:44062|issue:45081", "issue:27957|issue:28282", + "issue:44568|issue:46491", + "issue:43526|issue:45479", "issue:44737|issue:45588", - "issue:43404|issue:45987", "issue:44792|issue:44964", + "issue:43404|issue:45987", "issue:43756|issue:46399", "issue:42796|issue:43906", "issue:43756|issue:45325", @@ -6618,203 +6657,215 @@ "issue:45405|issue:45600", "issue:45357|issue:45920", "issue:44913|issue:46121", - "issue:43697|issue:45216", "issue:43643|issue:45375", - "issue:45200|issue:46463", - "issue:43720|issue:44898", - "issue:44625|issue:44877" + "issue:43697|issue:45216" ], - "prepared_review_unit_hash": "0bb5c9c8489baaf79d2fcbc70f324463e7fbc471c0f73d439a633d6cdff62fcf", + "prepared_review_unit_hash": "524aa80a6ec9962df3cb2407eec559dc999477402cb38cd0d12eebbed54e338e", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 36231, - "estimated_input_tokens": 9058, - "estimated_eval_tokens": 18372 + "soft_pair_count": 28, + "serialized_chars": 36562, + "estimated_input_tokens": 9141, + "estimated_eval_tokens": 18538 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The soft pairs are mostly lookalikes across the same broad subsystems (tokenizers, configs, multimodal loading, training) but they describe different concrete bugs or code paths, so none should be merged as duplicates. The strongest standalone issue in the set is the broad tokenizer regression affecting many models.", - "confidence": 0.84, - "canonical_issue_reason": "issue:45920 is the best representative issue: it describes a clear, broad regression in AutoTokenizer with multiple affected model families, strong user impact, and enough specificity to stand alone.", - "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 is the strongest overall issue candidate because it is broad, concrete, and regression-oriented, with multiple affected models and clear expected-vs-actual behavior.", + "summary": "The cluster is still mostly heterogeneous across unrelated tokenizer, trainer, config, model-loading, and backend issues. The only soft edge that looks like the same underlying bug is the SigLIP2 pair, and even that should be treated as likely rather than explicitly confirmed duplicate.", + "confidence": 0.82, + "canonical_issue_reason": "No single canonical issue fits the whole cluster because the items span many unrelated code paths and model families.", + "canonical_pr_reason": "No canonical PR: there are no PRs in this cluster.", + "best_issue_reason": "issue:43493 is the best representative for the likely SigLIP2 bug because it describes the implementation discrepancy more directly, while 43994 is mainly a downstream symptom report.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44737", + "right": "issue:46463", + "accept": false, + "reason": "Different subsystems and failures: XLNet CPU device placement vs T5 decoder input validation." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Unrelated problems: dependency/backend packaging vs loading a specific BERT model." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "Likely the same SigLIP2 bug: one report describes the HF-vs-JAX implementation discrepancy, and the other reports bad outputs from the same model/checkpoint." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Different dtype issues in different model families and code paths." + }, { "left": "issue:43257", "right": "issue:43756", "accept": false, - "reason": "Both are Qwen-related loading/config issues, but one is MOE weight conversion under accelerate+deepspeed and the other is Smollm3 RoPE layer dropping; different bugs." + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed is unrelated to Smollm3 RoPE-layer dropping." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "Unexpected `_is_hf_initialized` in `init_empty_weights` and `AutoImageProcessor` URL loading failure are different API paths and failure modes." + "reason": "Loading with init_empty_weights and loading an image processor from URL are different failures." }, { "left": "issue:43906", "right": "issue:46468", "accept": false, - "reason": "One is an isolated reproduction of issue 38071; the other is a broken official Nemotron example. Similar framing, not the same underlying bug." + "reason": "One is a reproduction of an old issue; the other is an official Nemotron example breakage with no clear same code path." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "`num_input_tokens_seen` not updating in TrainerState and `compute_loss()` getting an unexpected kwarg are distinct trainer bugs." + "reason": "Trainer state token counting and custom compute_loss signature compatibility are separate trainer regressions." }, { - "left": "issue:43526", - "right": "issue:45479", + "left": "issue:44062", + "right": "issue:45081", "accept": false, - "reason": "BeitImageProcessorFast label reduction and zero-loss classification with `num_labels=1` are unrelated behaviors." + "reason": "Both are tokenizer-related, but one is an AddedToken argument collision and the other is a Mistral regex patch crash." }, { - "left": "issue:44568", - "right": "issue:46491", + "left": "issue:27957", + "right": "issue:28282", "accept": false, - "reason": "Both are tokenizer regressions, but one is missing BOS/EOS insertion for mdeberta-v3 and the other is CodeLlama leading-space round-trip behavior." + "reason": "XLM-RoBERTa Flash Attention and missing PyTorch import are unrelated." }, { - "left": "issue:44062", - "right": "issue:45081", + "left": "issue:44568", + "right": "issue:46491", "accept": false, - "reason": "`AddedToken(special=...)` kwarg collision and `_patch_mistral_regex` crashing on missing `backend_tokenizer` are separate tokenizer internals bugs." + "reason": "Different tokenizer bugs with different models and behaviors." }, { - "left": "issue:27957", - "right": "issue:28282", + "left": "issue:43526", + "right": "issue:45479", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support and an `AutoModel` PyTorch import error are unrelated issues." + "reason": "Image processor label reduction and sequence-classification zero loss are unrelated." }, { "left": "issue:44737", "right": "issue:45588", "accept": false, - "reason": "XLNet CPU `arange` in rotary encoding and flash_attention `s_aux=None` crash are different code paths." + "reason": "XLNet CPU tensor creation and flash_attention sink handling are different bugs." }, { - "left": "issue:43404", - "right": "issue:45987", + "left": "issue:44792", + "right": "issue:44964", "accept": false, - "reason": "Tied `lm_head` weights in Mistral3 and `StaticCache.get_seq_length()` returning a tensor are unrelated bugs." + "reason": "Different multimodal model regressions with no shared concrete failure path." }, { - "left": "issue:44792", - "right": "issue:44964", + "left": "issue:43404", + "right": "issue:45987", "accept": false, - "reason": "Both involve multimodal models, but one is a janus test failure and the other is a Phi-4 multimodal load failure; not the same defect." + "reason": "Weight tying in Mistral3 and StaticCache sequence length return type are unrelated." }, { "left": "issue:43756", "right": "issue:46399", "accept": false, - "reason": "Smollm3 RoPE layer loss and Qwen3.5 kernel registration failure are different implementation issues." + "reason": "RoPE-layer dropping and missing apply_rotary_pos_emb registration are different defects." }, { "left": "issue:42796", "right": "issue:43906", "accept": false, - "reason": "Tokenizer BPE-marker handling and an isolated reproduction of issue 38071 do not match on root cause or target behavior." + "reason": "Ministral BPE-marker handling is not the same bug as the reproduced older issue." }, { "left": "issue:43756", "right": "issue:45325", "accept": false, - "reason": "Both mention RoPE, but one is missing layers in Smollm3 and the other is temporal position scaling in Qwen2.5-VL." + "reason": "Both involve RoPE/position handling, but the model, symptom, and code path differ." }, { "left": "issue:45823", "right": "issue:46463", "accept": false, - "reason": "Gemma4 device mismatch during forward and a T5 decoder-input validation error message are unrelated." + "reason": "Gemma4 device placement and T5 decoder-input error messaging are unrelated." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "PIL backend torchvision dependency and Mistral regex patch crash are different subsystems and failures." + "reason": "Image-processor torchvision dependency regression is unrelated to the Mistral tokenizer regex crash." }, { "left": "issue:43742", "right": "issue:44589", "accept": false, - "reason": "MobileLLM loading KeyError and missing Float8 storage are unrelated load-time errors." + "reason": "MobileLLM loading key error and Float8 storage lookup failure are different load-time issues." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "`gguf_file` breaking pretrained loads and inability to load `cjvt/sleng-bert` are different loading problems." + "reason": "gguf_file loading breakage and a specific model load failure do not look like the same underlying bug." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Qwen3-MoE `sliding_window` and `AutoImageProcessor` URL loading are unrelated." + "reason": "Qwen3-MoE sliding-window behavior and AutoImageProcessor URL loading are unrelated." }, { "left": "issue:45405", "right": "issue:45600", "accept": false, - "reason": "A PEFT version bump and stale `auto_mappings.py` references are separate CI/dependency issues." + "reason": "PEFT version bump and stale auto_mappings references are separate CI/dependency issues." }, { "left": "issue:45357", "right": "issue:45920", "accept": false, - "reason": "Qwen3.5 save_pretrained visual-encoder key regression and AutoTokenizer wrong token IDs are different bugs." + "reason": "Incorrect save_pretrained visual encoder keys is unrelated to tokenizer ID regressions." }, { "left": "issue:44913", "right": "issue:46121", "accept": false, - "reason": "GPTNeoXConfig `rotary_pct` reload behavior and `convert_rope_params_to_dict` TypeError are different rope-related failures." - }, - { - "left": "issue:43697", - "right": "issue:45216", - "accept": false, - "reason": "RTDetrV2 output differences and Qwen3.5 incorrect save_pretrained checkpoints are unrelated regressions." + "reason": "Config reload losing rotary_pct and JSON-list rope validation type errors are different config bugs." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Missing fields from `trust_remote_code=True` AutoConfig and a model-specific field dropped by `@strict` are similar in theme but not the same bug." - }, - { - "left": "issue:45200", - "right": "issue:46463", - "accept": false, - "reason": "Gemma 4 mm token type defaults and T5 decoder-input validation messaging are unrelated." + "reason": "Both mention missing config fields, but one is generic remote-code loading and the other is a specific strict-field drop; not clearly the same bug." }, { - "left": "issue:43720", - "right": "issue:44898", + "left": "issue:43697", + "right": "issue:45216", "accept": false, - "reason": "BitNet packed-weight unpacking during accelerate loading and Perceiver interpolation failure are different concrete defects." + "reason": "RTDetrV2 output drift and Qwen3.5 save_pretrained regression are different model-path regressions." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, + "evaluator_result": { + "accept": false, + "feedback": "Mostly grounded, but one claim is too strong for the packet: there is no explicit confirmation of a duplicate pair, so wording like \"confirmed duplicate pair\" overstates the evidence. The SigLIP2 match is plausible, but it should be framed more conservatively (e.g. likely same underlying bug / same area) rather than definitive." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -6823,7 +6874,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 22, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -6834,7 +6885,6 @@ "issue:24643", "issue:28282", "issue:29127", - "issue:29159", "issue:32090", "issue:33357", "issue:38175", @@ -6842,15 +6892,13 @@ "issue:42898", "issue:43295", "issue:43452", - "issue:43531", "issue:43577", "issue:43646", "issue:43653", "issue:43688", "issue:43697", - "issue:43742", + "issue:43720", "issue:43749", - "issue:43756", "issue:43819", "issue:44361", "issue:44442", @@ -6860,259 +6908,254 @@ "issue:44493", "issue:44514", "issue:44568", + "issue:44625", "issue:44704", "issue:44743", "issue:44811", "issue:44843", + "issue:44877", + "issue:44898", "issue:44936", "issue:44991", "issue:45005", + "issue:45200", "issue:45356", "issue:45372", "issue:45381", - "issue:45491", "issue:45542", "issue:45823", - "issue:46002", "issue:46050", "issue:46170", "issue:46178", - "issue:46402", + "issue:46463", "issue:46491" ], "soft_pairs": [ + "issue:43720|issue:44898", + "issue:44625|issue:44877", + "issue:45200|issue:46463", "issue:24643|issue:29127", "issue:41762|issue:44936", - "issue:42898|issue:46491", "issue:44493|issue:46050", "issue:44466|issue:45005", - "issue:44514|issue:45381", + "issue:42898|issue:46491", "issue:43452|issue:44843", "issue:43819|issue:44811", "issue:28282|issue:32090", - "issue:45372|issue:46170", + "issue:44514|issue:45381", "issue:28282|issue:33357", "issue:43688|issue:43749", - "issue:44484|issue:44704", + "issue:45372|issue:46170", "issue:44936|issue:45823", "issue:38175|issue:43295", "issue:44361|issue:44442", - "issue:43295|issue:45356", + "issue:44484|issue:44704", "issue:43653|issue:44361", "issue:45542|issue:46178", "issue:43452|issue:43577", + "issue:43295|issue:45356", "issue:43577|issue:43697", "issue:43653|issue:44568", "issue:44464|issue:44743", - "issue:44843|issue:44991", "issue:43646|issue:43749", - "issue:29159|issue:33357", - "issue:45491|issue:46002", - "issue:43756|issue:46402", - "issue:43531|issue:43742" + "issue:44843|issue:44991" ], - "prepared_review_unit_hash": "f8f06d4fbd4f9f7ec669b98b5816cc0d71c493a38c289d6b8381c945548f62c5", + "prepared_review_unit_hash": "73b23b5fc3c1c443f7ab99f40cc36556c80b8bb47a4f7cd4a74d153489eef278", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36668, - "estimated_input_tokens": 9167, - "estimated_eval_tokens": 18590 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35869, + "estimated_input_tokens": 8968, + "estimated_eval_tokens": 18192 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the paired items span unrelated tokenizer, loading, training, generation, and model-specific regressions, so there is no clear duplicate canonical issue or PR here. All soft edges should be rejected.", - "confidence": 0.94, - "canonical_issue_reason": null, + "summary": "This cluster is a set of mostly unrelated Hugging Face Transformers bugs/regressions. The titles span tokenizer loading, model init, processor behavior, backend/import issues, and several model-specific regressions, but none of the soft pairs look like the same concrete underlying bug.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43646 is the broadest and most representative umbrella bug in the set: a general Transformers v5 custom model initialization regression. It best matches the cluster\u2019s overall theme of v5 breakages without being tied to one niche model.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:43646 is the best single issue to anchor the cluster because it is general enough to cover the common regression pattern, while the others are narrower model/tokenizer/backend-specific cases.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "Different bugs: BitNet packed-weight loading vs Perceiver image-classification resizing/interpolation behavior." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Different models and failures: Qwen3.5 num_labels propagation vs Granite Speech strict-config loading." + }, + { + "left": "issue:45200", + "right": "issue:46463", + "accept": false, + "reason": "Different code paths: Gemma 4 mm_token_type_ids defaulting vs T5 decoder input validation messaging." + }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed weight-dimension runtime error vs. LayoutLMv3 box validation/error-message issue; different code paths and symptoms." + "reason": "Different problems: DeepSpeed training weight-shape RuntimeError vs LayoutLMv3 error-message clarity." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "Gemma3 ZeRO-3 load failure vs. trainer.evaluate() lifecycle bug; unrelated failures." - }, - { - "left": "issue:42898", - "right": "issue:46491", - "accept": false, - "reason": "Both are tokenizer regressions, but one is clean_up_tokenization_spaces behavior and the other is CodeLlama space handling on encode/decode; not the same bug." + "reason": "Both are training/loading failures, but they involve different models and failure modes: Gemma3 ZeRO-3 loading vs trainer.evaluate after train." }, { "left": "issue:44493", "right": "issue:46050", "accept": false, - "reason": "Position-id key warning vs. quantized model save/load corruption; different underlying issues." + "reason": "Different underlying issues: position-id unexpected keys vs quantized model save/load breakage after sinq." }, { "left": "issue:44466", "right": "issue:45005", "accept": false, - "reason": "Both mention tied weights, but one is device-dependent serialization and the other is translation-model tied-weights behavior; not clearly the same defect." + "reason": "Both concern tied weights, but the concrete behaviors differ: serialization inconsistency vs translation-model weight handling in v5." }, { - "left": "issue:44514", - "right": "issue:45381", + "left": "issue:42898", + "right": "issue:46491", "accept": false, - "reason": "Qwen2.5-VL chat-template batching crash vs. video vision_position_ids bug; different processor paths." + "reason": "Different tokenizer regressions: v5 clean_up_tokenization_spaces behavior vs CodeLlama leading-space round-trip loss." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Both affect loading/tokenizer paths, but one is gguf_file breakage and the other is an offline model_info call in _patch_mistral_regex; different root causes." + "reason": "Different failure modes: gguf_file breaking tokenizer/model loading vs model_info() being called offline in _patch_mistral_regex." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC latent/STE mismatch vs. Whisper batch_decode skip_special_tokens regression; unrelated functionality." + "reason": "Different components: DAC latent reconstruction mismatch vs Whisper processor batch_decode skip_special_tokens handling." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "Missing PyTorch import vs. Trainer GPU broadcast TypeError; unrelated." + "reason": "Different import/runtime paths: missing PyTorch for AutoModel vs Trainer passing NoneType to _gpu_broadcast_one." }, { - "left": "issue:45372", - "right": "issue:46170", + "left": "issue:44514", + "right": "issue:45381", "accept": false, - "reason": "Gemma 4 processor import error from mistral_common vs. Windows ROCm torch.distributed dependency issue; unrelated." + "reason": "Both are VL processor-related, but one is batched chat templating padding while the other is video vision_position_ids correctness." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "PyTorch not found ImportError vs. MacOS bus error on CLIP model; different failure modes." + "reason": "Different root causes: missing PyTorch import error vs MacOS bus error with a specific CLIP community model." }, { "left": "issue:43688", "right": "issue:43749", "accept": false, - "reason": "Aux-loss normalization bug in MoE models vs. FSDP CPU RAM efficient loading breakage; different problems." + "reason": "Related to loading/parallelism, but not the same bug: auxiliary-loss normalization vs FSDP CPU RAM efficient loading." }, { - "left": "issue:44484", - "right": "issue:44704", + "left": "issue:45372", + "right": "issue:46170", "accept": false, - "reason": "Save-pretrained shard-size question vs. AutoProcessor kwarg forwarding bug; not duplicates." + "reason": "Different import/dependency issues: Gemma 4 processor loading needs ReasoningEffort, while Windows ROCm issues stem from torch.distributed dependency handling." }, { "left": "issue:44936", "right": "issue:45823", "accept": false, - "reason": "Trainer evaluate/train lifecycle issue vs. Gemma4 device_map auto device mismatch; different code paths." + "reason": "Different failure modes: trainer lifecycle bug vs Gemma4 PLE device mismatch during forward with device_map=auto." }, { "left": "issue:38175", "right": "issue:43295", "accept": false, - "reason": "Siglip2 zero probabilities vs. v5 tokenizer cleanup behavior regression; unrelated." + "reason": "Different regressions: zero probabilities in SIGLIP2 vs processor.tokenizer access / image-passing regression in v4.57.5." }, { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer-loading failures for different models; same broad area but not the same bug." + "reason": "Both tokenizer-related, but one is MLukeTokenizer task AttributeError and the other is AutoTokenizer failing on FastSpeech2ConformerTokenizer." }, { - "left": "issue:43295", - "right": "issue:45356", + "left": "issue:44484", + "right": "issue:44704", "accept": false, - "reason": "Both are tokenizer regressions, but one concerns cleanup_spaces and the other codec handling plus a warning; separate issues." + "reason": "Different save/load API issues: max_shard_size default question vs AutoProcessor not forwarding kwargs to cached_file." }, { "left": "issue:43653", "right": "issue:44361", "accept": false, - "reason": "BigBird special-token registration bug vs. MLukeTokenizer AttributeError; unrelated tokenizers and symptoms." + "reason": "Different tokenizer bugs: BigBirdTokenizer special-token registration/empty decode vs MLukeTokenizer task AttributeError." }, { "left": "issue:45542", "right": "issue:46178", "accept": false, - "reason": "Backend detection/misconfiguration problems, but one is tensorboard without TensorFlow and the other is tensorflow_text backend mapping; different root causes." + "reason": "Different backend issues: TensorBoard without TensorFlow causing undefined tf backend vs unsupported tensorflow_text backend mapping." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "gguf_file loading breakage vs. BLIP2 dtype not propagating; unrelated." + "reason": "Different loading issues: gguf_file breaks tokenizer/model loading vs Blip2 dtype not propagating to model.qformer." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Different regressions: custom processor/tokenizer access broke in v4.57.5 vs Kimi-K2.5 tokenizer codec handling and warning regression." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "BLIP2 dtype propagation vs. RTDetrV2 output differences in v5; different model behaviors." + "reason": "Different model-specific issues: Blip2 dtype propagation vs RTDetrV2 producing different outputs in v5." }, { "left": "issue:43653", "right": "issue:44568", "accept": false, - "reason": "Both involve tokenizer special tokens, but BigBird mask-token registration and mdeberta BOS/EOS insertion are different bugs." + "reason": "Different tokenizer defects: BigBird mask token not registered vs mdeberta-v3-base not adding BOS/EOS with add_special_tokens." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Chunked generation compiled-forward inconsistency vs. Qwen3 recurrent state reset in cache; both generation-related but not the same concrete bug." - }, - { - "left": "issue:44843", - "right": "issue:44991", - "accept": false, - "reason": "Offline model_info patch issue vs. tokenizer loading failure for est-roberta; no clear shared code-path bug." + "reason": "Different generation/model-state bugs: chunked generation with compiled forward vs Qwen3.5 recurrent states resetting with cache." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Custom model initialization breakage in v5 vs. FSDP loading regression; unrelated." + "reason": "Custom model init regression is not the same bug as FSDP CPU RAM efficient loading." }, { - "left": "issue:29159", - "right": "issue:33357", - "accept": false, - "reason": "Slow/fast tokenizer inconsistency vs. MacOS bus error on CLIP model; unrelated." - }, - { - "left": "issue:45491", - "right": "issue:46002", - "accept": false, - "reason": "Both mention NaNs, but one is Gemma3 sliding-window attention and the other is classification-head init/checkout; different underlying failures." - }, - { - "left": "issue:43756", - "right": "issue:46402", - "accept": false, - "reason": "Smollm3 RoPE-layer omission vs. Sam3Model text-encoder weights not loaded; unrelated model-specific bugs." - }, - { - "left": "issue:43531", - "right": "issue:43742", + "left": "issue:44843", + "right": "issue:44991", "accept": false, - "reason": "Qwen3-MoE sliding_window issue vs. MobileLLM-125M loading KeyError; unrelated." + "reason": "Different tokenizer-loading regressions: offline mistral regex patch issue vs est-roberta tokenizer loading failure." } ] }, @@ -7128,7 +7171,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 23, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -7136,20 +7179,19 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:29159", "issue:32090", + "issue:33357", "issue:33666", "issue:36010", "issue:36331", "issue:41720", "issue:43262", - "issue:43278", "issue:43295", "issue:43452", - "issue:43493", "issue:43504", - "issue:43525", + "issue:43531", "issue:43550", - "issue:43572", "issue:43577", "issue:43606", "issue:43643", @@ -7157,7 +7199,7 @@ "issue:43688", "issue:43720", "issue:43742", - "issue:43825", + "issue:43756", "issue:43927", "issue:43994", "issue:44361", @@ -7172,24 +7214,29 @@ "issue:44625", "issue:44821", "issue:44849", - "issue:44964", "issue:44977", "issue:45290", - "issue:45405", "issue:45440", + "issue:45491", "issue:45701", "issue:45704", "issue:45823", + "issue:46002", "issue:46050", "issue:46192", "issue:46209", - "issue:46365" + "issue:46365", + "issue:46402" ], "soft_pairs": [ - "issue:43688|issue:45440", + "issue:43531|issue:43742", + "issue:29159|issue:33357", + "issue:45491|issue:46002", + "issue:43756|issue:46402", "issue:43688|issue:45704", - "issue:44514|issue:45290", + "issue:43688|issue:45440", "issue:43653|issue:43927", + "issue:44514|issue:45290", "issue:43550|issue:43927", "issue:43688|issue:45701", "issue:44479|issue:44625", @@ -7197,262 +7244,251 @@ "issue:44464|issue:44977", "issue:43643|issue:44561", "issue:41720|issue:44464", + "issue:33666|issue:36010", "issue:44466|issue:46050", + "issue:43742|issue:44821", + "issue:32090|issue:33666", "issue:43504|issue:43720", - "issue:33666|issue:36010", + "issue:43504|issue:43550", "issue:43577|issue:44554", - "issue:32090|issue:33666", "issue:44977|issue:46209", - "issue:43504|issue:43550", - "issue:43742|issue:44821", "issue:43504|issue:44610", "issue:44554|issue:46365", "issue:43262|issue:43295", "issue:43720|issue:43927", "issue:45823|issue:46050", - "issue:43452|issue:44451", "issue:43606|issue:43927", + "issue:43452|issue:44451", "issue:33666|issue:36331", "issue:43994|issue:46192", - "issue:44849|issue:46209", - "issue:43525|issue:43572", - "issue:43278|issue:43688", - "issue:43493|issue:43825", - "issue:43720|issue:44361", - "issue:44964|issue:45405" + "issue:44849|issue:46209" ], - "prepared_review_unit_hash": "d9fde29f80a2d4c4e910557ca97a8e644fadab9a58f11725467fd66132d0bca2", + "prepared_review_unit_hash": "e03b66501e0823942e4bbee2da68719d6e0e3ea4e85e9a1678344e4b731f5b7d", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 37936, - "estimated_input_tokens": 9484, - "estimated_eval_tokens": 19224 + "soft_pair_count": 33, + "serialized_chars": 37638, + "estimated_input_tokens": 9410, + "estimated_eval_tokens": 19076 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is highly heterogeneous; the soft pairs mostly connect superficially similar transformers/model-loading bugs rather than the same underlying issue. No soft edge looks safe to merge as a duplicate.", - "confidence": 0.9, - "canonical_issue_reason": "issue:43688 is the closest thing to a hub because it has multiple soft-similarity links and inbound references, but even it is not a true representative for the whole mixed cluster.", + "summary": "This is a heterogeneous bundle of unrelated issues; every soft pair looks like a topical similarity only, not the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: the cluster is heterogeneous and the soft links are mostly superficial title similarity rather than duplicates.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43688 is the strongest single issue candidate by connectivity, but the cluster is too diverse for it to serve as a meaningful global canonical beyond its own loss-normalization theme.", + "best_issue_reason": "No strong global best issue; none of the issues appears to represent the whole set.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43688", - "right": "issue:45440", + "left": "issue:43531", + "right": "issue:43742", "accept": false, - "reason": "Aux-loss normalization vs remote-implementation divergence; different model families and failure modes." + "reason": "Qwen3-MoE sliding-window behavior vs a MobileLLM loading key error; different bug classes." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency vs a MacOS bus error on CLIP; unrelated failures." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Both mention NaNs, but one is a sliding-window attention bug and the other is a classification-head init/checkpoint issue." + }, + { + "left": "issue:43756", + "right": "issue:46402", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping vs Sam3Model text-encoder weights not loading; different model families and fixes." }, { "left": "issue:43688", "right": "issue:45704", "accept": false, - "reason": "Both touch loss/normalization, but one is a model bug and the other is a T5 apex memory-leak issue." + "reason": "Aux-loss normalization bug vs an apex RMSNorm memory leak; unrelated code paths." }, { - "left": "issue:44514", - "right": "issue:45290", + "left": "issue:43688", + "right": "issue:45440", "accept": false, - "reason": "Both are chat/template-related, but one is a batched padding crash and the other is assistant tool-call handling." + "reason": "Different models and different semantics: loss normalization vs remote-implementation divergence." }, { "left": "issue:43653", "right": "issue:43927", "accept": false, - "reason": "Tokenizer special-token registration bug vs config custom-token persistence bug; not the same defect." + "reason": "BigBird special-token registration vs DiaConfig custom token IDs lost on save/load; only broadly token-related." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both crash in apply_chat_template, but on different triggers: batched padding=False vs assistant tool-call messages with no content." }, { "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "Unrelated: SDPA/torch.compile failure in Bamba vs DiaConfig token-ID persistence." + "reason": "Torch.compile SDPA failure vs config token-ID persistence; unrelated." }, { "left": "issue:43688", "right": "issue:45701", "accept": false, - "reason": "Aux-loss normalization is unrelated to version-dependent tokenization changes." + "reason": "Aux-loss normalization bug vs tokenization behavior changes across versions; not the same bug." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Video-input regression across Qwen variants vs num_labels propagation in Qwen3.5 config." + "reason": "Video-input regression in Qwen VL models vs num_labels propagation in Qwen3.5; different code paths." }, { "left": "issue:43550", "right": "issue:44361", "accept": false, - "reason": "Different subsystems: compiled attention in Bamba vs tokenizer AttributeError in MLuke." + "reason": "Bamba compile/SDPA failure vs MLukeTokenizer AttributeError; unrelated." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Chunked generation with compiled forward is not the same as flash-attention generation failure." + "reason": "Chunked generation with compiled forward vs flash-attention generation; both generation bugs, but not the same concrete failure." }, { "left": "issue:43643", "right": "issue:44561", "accept": false, - "reason": "Both affect trust_remote_code models, but one is missing config fields and the other is a removed helper function." + "reason": "Both involve trust_remote_code, but one is missing config fields and the other is removal of a helper API." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "Qwen3 device-map cuda assert vs chunked-generation output inconsistency; different code paths." + "reason": "Qwen3 device-map cudaErrorAssert vs compiled-forward chunked-generation inconsistency; different failures." }, { - "left": "issue:44466", - "right": "issue:46050", - "accept": false, - "reason": "v5 tied-weight serialization bug vs sinq quantization save/load breakage." - }, - { - "left": "issue:43504", - "right": "issue:43720", + "left": "issue:33666", + "right": "issue:36010", "accept": false, - "reason": "Legacy field in Beit config vs BitNet packed-weight unpacking during accelerate loading." + "reason": "Multi-GPU training issue vs GenerationMixin import error; unrelated." }, { - "left": "issue:33666", - "right": "issue:36010", + "left": "issue:44466", + "right": "issue:46050", "accept": false, - "reason": "Multi-GPU Qwen2-VL training vs GenerationMixin import error; not the same issue." + "reason": "lm_head serialization/tied-weights behavior vs quantized save/load breakage; both persistence-related but not the same bug." }, { - "left": "issue:43577", - "right": "issue:44554", + "left": "issue:43742", + "right": "issue:44821", "accept": false, - "reason": "Blip2 dtype mismatch on load vs MPS attention correctness bug with mismatched head dims." + "reason": "MobileLLM loading failure vs AutoImageProcessor URL loading; different APIs and symptoms." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Trainer GPU broadcast TypeError vs Qwen2-VL multi-GPU training request; different underlying failures." + "reason": "Trainer broadcast NoneType error vs Qwen2-VL multi-GPU training; same broad area only." }, { - "left": "issue:44977", - "right": "issue:46209", + "left": "issue:43504", + "right": "issue:43720", "accept": false, - "reason": "Flash-attention generation bug vs FP8 inference failure on B200 hardware." + "reason": "Legacy field loading in Beit vs BitNet packed-weight unpacking during accelerate loading; distinct model-loading bugs." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "Different models and failures: Beit pretrained-load legacy field vs Bamba torch.compile/SDPA." + "reason": "Beit pretrained-load legacy field issue vs Bamba compile/SDPA failure; unrelated." }, { - "left": "issue:43742", - "right": "issue:44821", + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Blip2 dtype propagation bug vs MPS attention correctness issue; unrelated." + }, + { + "left": "issue:44977", + "right": "issue:46209", "accept": false, - "reason": "MobileLLM config loading key error vs AutoImageProcessor URL loading issue." + "reason": "Qwen3.5 flash-attention generation bug vs FP8 inference failure on B200; different backend issues." }, { "left": "issue:43504", "right": "issue:44610", "accept": false, - "reason": "Beit legacy-field load bug vs OmDet processor/image-size mismatch." + "reason": "Beit loading legacy field vs OmDet processor/model size mismatch; unrelated." }, { "left": "issue:44554", "right": "issue:46365", "accept": false, - "reason": "Attention correctness on MPS is unrelated to the Trainer import conflict with datasets." + "reason": "MPS attention correctness issue vs inability to import Trainer with datasets installed; unrelated." }, { "left": "issue:43262", "right": "issue:43295", "accept": false, - "reason": "Audio sampling-rate default bug vs processor/tokenizer API regression for multimodal custom code." + "reason": "Audio sampling-rate default vs processor.tokenizer/images regression; both processor-related but different breakages." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "Packed-weight loading vs config custom-token persistence; different areas and fixes." + "reason": "BitNet accelerate loading vs DiaConfig token-ID persistence; unrelated." }, { "left": "issue:45823", "right": "issue:46050", "accept": false, - "reason": "Gemma4 device-map mismatch vs quantized save/load corruption; unrelated failures." + "reason": "Gemma4 device-map forward mismatch vs sinq quantized save/load breakage; different problems." }, { - "left": "issue:43452", - "right": "issue:44451", + "left": "issue:43606", + "right": "issue:43927", "accept": false, - "reason": "gguf_file loader regression vs inability to load a specific model; not the same code-path." + "reason": "CPU offload device mismatch in Bark vs DiaConfig token-ID loss; unrelated." }, { - "left": "issue:43606", - "right": "issue:43927", + "left": "issue:43452", + "right": "issue:44451", "accept": false, - "reason": "CPU offload device mismatch vs DiaConfig token-ID loss; different bug classes." + "reason": "GGUF-related tokenizer/model loading breakage vs ScandiBERT loading failure; different root causes." }, { "left": "issue:33666", "right": "issue:36331", "accept": false, - "reason": "Qwen2-VL multi-GPU training vs CustomTrainer.compute_loss signature break." + "reason": "Multi-GPU training issue vs custom Trainer compute_loss signature mismatch; unrelated." }, { "left": "issue:43994", "right": "issue:46192", "accept": false, - "reason": "SigLIP2 model/pipeline outputs vs return_tensors being ignored when text_kwargs is passed." + "reason": "Siglip2 bad outputs vs return_tensors being ignored with text_kwargs; unrelated." }, { "left": "issue:44849", "right": "issue:46209", "accept": false, - "reason": "Qwen3.5 hidden-states bug vs FP8 inference failure on B200." - }, - { - "left": "issue:43525", - "right": "issue:43572", - "accept": false, - "reason": "Similar missing-pad-field symptom, but different config classes, field names, and fixes." - }, - { - "left": "issue:43278", - "right": "issue:43688", - "accept": false, - "reason": "Embedding dtype drift in eval vs auxiliary-loss normalization; unrelated model behavior." - }, - { - "left": "issue:43493", - "right": "issue:43825", - "accept": false, - "reason": "SigLIP2 implementation discrepancy vs pipeline task-message regression; different components." - }, - { - "left": "issue:43720", - "right": "issue:44361", - "accept": false, - "reason": "BitNet packed-weight loading vs MLuke tokenizer task error; no shared underlying bug." - }, - { - "left": "issue:44964", - "right": "issue:45405", - "accept": false, - "reason": "Phi-4 multimodal load failure vs an outdated MIN_PEFT_VERSION pin; unrelated." + "reason": "Qwen3.5 hidden-states bug vs FP8 inference on B200; different features and failures." } ] }, @@ -7468,7 +7504,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 24, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -7487,63 +7523,68 @@ "issue:43334", "issue:43493", "issue:43504", + "issue:43525", "issue:43540", "issue:43572", "issue:43575", "issue:43606", "issue:43646", - "issue:43653", + "issue:43688", "issue:43701", + "issue:43720", "issue:43824", + "issue:43825", "issue:43844", "issue:43927", - "issue:44112", "issue:44291", "issue:44361", "issue:44464", "issue:44466", "issue:44488", - "issue:44493", "issue:44568", "issue:44589", "issue:44625", "issue:44792", "issue:44843", "issue:44849", + "issue:44964", "issue:44977", "issue:45200", "issue:45229", "issue:45357", "issue:45405", "issue:45520", - "issue:45584", "issue:45759", "issue:45812", "issue:45823", "issue:46002", - "issue:46285", "issue:46424", "issue:46489" ], "soft_pairs": [ + "issue:43525|issue:43572", + "issue:43278|issue:43688", + "issue:43493|issue:43825", + "issue:43720|issue:44361", + "issue:44964|issue:45405", "issue:39692|issue:43493", "issue:28218|issue:33357", "issue:44625|issue:46002", "issue:44849|issue:45357", + "issue:43493|issue:43844", "issue:44568|issue:46489", - "issue:43278|issue:43844", "issue:45200|issue:45229", + "issue:43278|issue:43844", "issue:39692|issue:43295", - "issue:43493|issue:43844", "issue:44488|issue:44843", "issue:43540|issue:44977", - "issue:44568|issue:45812", "issue:43646|issue:44466", + "issue:44568|issue:45812", "issue:44464|issue:46424", - "issue:33357|issue:33666", "issue:43824|issue:44589", - "issue:43257|issue:43299", + "issue:33357|issue:33666", "issue:44589|issue:45520", + "issue:43257|issue:43299", "issue:43334|issue:43572", "issue:43927|issue:44361", "issue:43504|issue:43606", @@ -7551,227 +7592,229 @@ "issue:45229|issue:45823", "issue:43278|issue:43701", "issue:44792|issue:45405", - "issue:44493|issue:46285", "issue:43575|issue:45405", - "issue:43572|issue:44291", - "issue:43606|issue:44112", - "issue:43504|issue:43653", - "issue:44464|issue:45584" + "issue:43572|issue:44291" ], - "prepared_review_unit_hash": "d1428ee9232e8cc89180a2cb1aa3946bab014ffda28cef9c7e2087abb2df45b9", + "prepared_review_unit_hash": "3840072b76f4cd9fd9347ab6d52f1b647c482128b9e95b07775e710f55ccfbbc", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 37082, - "estimated_input_tokens": 9271, - "estimated_eval_tokens": 18798 + "soft_pair_count": 32, + "serialized_chars": 37356, + "estimated_input_tokens": 9339, + "estimated_eval_tokens": 18934 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose collection of mostly unrelated Transformers regression reports. The issues share broad themes like tokenizers, model loading, and config compatibility, but the soft pairs do not look like the same underlying bug or change, so I would not merge any of them.", - "confidence": 0.34, - "canonical_issue_reason": "issue:45812 is the strongest cluster representative: it is a broad, clearly framed v4\u2192v5 tokenizer regression affecting all Granite models, with multiple inbound references and a concrete user-facing failure mode.", + "summary": "The items are mostly unrelated issues that only share broad transformer/model-regression vocabulary. None of the soft-edge pairs look like the same underlying bug or change, so all should be rejected.", + "confidence": 0.9, + "canonical_issue_reason": "No single true duplicate cluster emerges; issue 45812 is the closest representative because it is the broadest, highest-impact tokenizer regression with a clear v4\u2192v5 failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the best single issue to keep as the canonical anchor because it is the most general and well-scoped tokenizer regression in the set, with clearer impact and stronger discussion signal than the more niche reports.", + "best_issue_reason": "Issue 45812 is the best overall issue representative: concrete user impact, regression framing, and the broadest scope among the set.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both are config attribute regressions, but they affect different models and different missing fields (`pad_token_id` vs `pad_token_idx`) with no shared code path shown." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "One is an embedding dtype mismatch during evaluate; the other is auxiliary-loss normalization. Different symptoms and subsystems." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation mismatch and a pipeline translation error message are unrelated bugs." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "Accelerate loading of packed BitNet weights vs tokenizer task AttributeError are unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model-loading failure and a dependency-version bump are not the same bug." + }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "Both are SigLIP2-related, but one is a documentation/example failure and the other is an implementation discrepancy vs JAX; different bugs." + "reason": "Documentation example errors and SigLIP2 HF-vs-JAX discrepancy are different issues." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer spacing bug vs MacOS CLIP bus error; no shared code-path or failure mode." + "reason": "Tokenizer whitespace behavior and a MacOS bus error on CLIP are unrelated." }, { "left": "issue:44625", "right": "issue:46002", "accept": false, - "reason": "One is config propagation for num_labels, the other is NaNs in classification heads; unrelated symptoms and fixes." + "reason": "`num_labels` propagation in Qwen3.5 and NaNs in classification heads are different failure modes." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Both involve Qwen3.5, but one is output_hidden_states behavior and the other is save_pretrained key serialization; not the same bug." + "reason": "Output-hidden-states behavior and incorrect visual-encoder save keys are separate Qwen3.5 bugs." }, { - "left": "issue:44568", - "right": "issue:46489", + "left": "issue:43493", + "right": "issue:43844", "accept": false, - "reason": "Both are tokenizer regressions, but for different models and different failure points; too broad to treat as one bug." + "reason": "SigLIP2 implementation discrepancy and DeepSpeed gradient explosion are unrelated." }, { - "left": "issue:43278", - "right": "issue:43844", + "left": "issue:44568", + "right": "issue:46489", "accept": false, - "reason": "Embedding dtype drift in eval vs ZeRO-3 gradient growth are different training/runtime problems." + "reason": "Both are tokenizer regressions, but they concern different models and distinct tokenization failures." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Text-only mm_token_type_ids defaulting and multi-GPU OOM are unrelated issues." + "reason": "Text-only mm_token_type_ids defaults and multi-GPU CUDA OOM are unrelated." }, { - "left": "issue:39692", - "right": "issue:43295", + "left": "issue:43278", + "right": "issue:43844", "accept": false, - "reason": "SigLIP2 example errors and processor.tokenizer regression are distinct multimodal issues." + "reason": "Evaluate-time dtype drift and ZeRO-3 gradient growth are different bugs." }, { - "left": "issue:43493", - "right": "issue:43844", + "left": "issue:39692", + "right": "issue:43295", "accept": false, - "reason": "SigLIP2 implementation mismatch is unrelated to ZeRO-3 gradient instability." + "reason": "A broken doc example and a regression in processor/tokenizer behavior are not the same underlying issue." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "Loading a specific model vs offline-model_info regression in tokenizer patching; different root causes." + "reason": "Loading a specific model failing and an offline-mode tokenizer patch issue are unrelated." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Qwen3OmniMoe video processing and Qwen3.5 flash-attention generation are separate model-path bugs." - }, - { - "left": "issue:44568", - "right": "issue:45812", - "accept": false, - "reason": "Different tokenizer families and different regressions; one is missing BOS/EOS on add_special_tokens, the other is wrong Granite token IDs." + "reason": "Video-input processing in Qwen3OmniMoe and flash-attention generation in Qwen3.5 are different code paths." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Custom model initialization breakage and lm_head serialization inconsistency do not describe the same change." + "reason": "Custom model initialization breakage and model-device-dependent lm_head serialization are unrelated." }, { - "left": "issue:44464", - "right": "issue:46424", + "left": "issue:44568", + "right": "issue:45812", "accept": false, - "reason": "Chunked generation under compiled forward is a different code path than cache length handling for quantized cache." + "reason": "Both are tokenizer regressions, but they affect different model families and different concrete failures." }, { - "left": "issue:33357", - "right": "issue:33666", + "left": "issue:44464", + "right": "issue:46424", "accept": false, - "reason": "MacOS CLIP bus error and Qwen2-VL multi-GPU training are unrelated." + "reason": "Chunked generation with compiled forward and cache-length handling are separate concerns." }, { "left": "issue:43824", "right": "issue:44589", "accept": false, - "reason": "ImportError for a missing Qwen2_5_VL class vs Float8 storage lookup failure; different bugs." + "reason": "Missing Qwen2.5-VL import and Float8 storage lookup errors are unrelated." }, { - "left": "issue:43257", - "right": "issue:43299", + "left": "issue:33357", + "right": "issue:33666", "accept": false, - "reason": "Both mention Qwen3 MoE loading, but one is weight conversion under accelerate+deepspeed and the other is a Qwen3VL MoE loading breakage; not clearly the same concrete fix." + "reason": "MacOS bus error on a CLIP model and Qwen2-VL multi-GPU training are unrelated." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage error and flash_attn import KeyError are separate compatibility issues." + "reason": "Float8 storage lookup failure and Python 3.13 flash_attn import handling are different bugs." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed and Qwen3VL MoE loading are related only by model family, not the same failure." }, { "left": "issue:43334", "right": "issue:43572", "accept": false, - "reason": "Missing pad_token_id on Qwen3-VL text config vs missing pad_token_idx on StableLmConfig; different models and fields." + "reason": "Missing pad_token_id on Qwen3-VL and missing pad_token_idx on StableLmConfig are distinct model/config regressions." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "DiaConfig save/load token IDs and MLukeTokenizer AttributeError are unrelated tokenizer/config bugs." + "reason": "DiaConfig token-ID persistence and MLukeTokenizer task AttributeError are unrelated." }, { "left": "issue:43504", "right": "issue:43606", "accept": false, - "reason": "Legacy field load failure in Beit and CPU offload device mismatch in bark-small are unrelated." + "reason": "Beit semantic-segmentation legacy-field loading and Bark CPU-offload device mismatch are separate bugs." }, { "left": "issue:43334", "right": "issue:45759", "accept": false, - "reason": "Both touch Qwen composite configs, but one is a missing pad_token_id attribute and the other is from_config not unwrapping text_config; not enough to be the same bug." + "reason": "Both involve Qwen multimodal configs, but one is a missing config attribute and the other is from_config not unwrapping text_config." }, { "left": "issue:45229", "right": "issue:45823", "accept": false, - "reason": "Multi-GPU OOM and device mismatch with device_map='auto' are different failure modes." + "reason": "Multi-GPU OOM and PLE device mismatch during forward are different failure modes." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "Embedding dtype regression and resume_from_checkpoint key mismatch are unrelated." + "reason": "Embedding dtype drift and resume_from_checkpoint key mismatch are unrelated." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "Janus image-generation test failure and a peft version bump are unrelated." - }, - { - "left": "issue:44493", - "right": "issue:46285", - "accept": false, - "reason": "Unexpected position-id keys and CLIP text-model structure divergence are different API regressions." + "reason": "A failing janus image-generation test and a dependency version bump are not the same bug." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "Tensor-parallel OOM on a specific model and a dependency version bump are unrelated." + "reason": "TP-induced OOM on Qwen2-57B and a PEFT minimum-version bump are unrelated." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "Both are 5.0 compatibility problems, but one is a config attribute mismatch and the other is an init_empty_weights argument error; different code paths." - }, - { - "left": "issue:43606", - "right": "issue:44112", - "accept": false, - "reason": "CPU offload device mismatch vs stale GraniteSpeech CI override test; not the same underlying bug." - }, - { - "left": "issue:43504", - "right": "issue:43653", - "accept": false, - "reason": "Legacy preset loading and BigBirdTokenizer special-token registration are different issues." - }, - { - "left": "issue:44464", - "right": "issue:45584", - "accept": false, - "reason": "Compiled-forward chunked generation and Whisper empty-transcription behavior after token alignment are separate generation bugs." + "reason": "Missing pad_token_idx and init_empty_weights receiving an unexpected argument are different compatibility problems." } ] }, @@ -7787,7 +7830,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 25, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -7806,52 +7849,55 @@ "issue:38175", "issue:42490", "issue:43065", - "issue:43232", - "issue:43295", - "issue:43450", "issue:43493", + "issue:43504", "issue:43540", + "issue:43606", + "issue:43653", "issue:43696", "issue:43701", "issue:43844", "issue:43873", "issue:43957", "issue:43994", + "issue:44112", "issue:44186", "issue:44423", + "issue:44464", "issue:44493", "issue:44521", "issue:44554", - "issue:44610", "issue:44734", - "issue:44849", "issue:44964", "issue:45072", "issue:45084", "issue:45198", "issue:45200", "issue:45290", - "issue:45356", - "issue:45381", "issue:45406", "issue:45446", "issue:45520", + "issue:45584", "issue:45588", "issue:45823", "issue:46050", "issue:46121", "issue:46192", - "issue:46209", + "issue:46285", "issue:46369" ], "soft_pairs": [ + "issue:44493|issue:46285", + "issue:43606|issue:44112", + "issue:43504|issue:43653", + "issue:44464|issue:45584", "issue:43994|issue:44493", "issue:43873|issue:46050", "issue:45588|issue:46050", - "issue:44423|issue:44734", "issue:29942|issue:30333", - "issue:43065|issue:46369", + "issue:44423|issue:44734", "issue:43844|issue:44964", + "issue:43065|issue:46369", "issue:42490|issue:43696", "issue:45406|issue:46121", "issue:29159|issue:32090", @@ -7865,211 +7911,186 @@ "issue:43493|issue:43701", "issue:37428|issue:43957", "issue:44186|issue:46192", - "issue:44521|issue:45290", - "issue:43232|issue:44849", - "issue:45823|issue:46192", - "issue:43295|issue:43450", - "issue:43295|issue:44610", - "issue:45381|issue:46209", - "issue:43295|issue:43873", - "issue:44554|issue:45356" + "issue:44521|issue:45290" ], - "prepared_review_unit_hash": "5bf2ee48583e42a144df20adfd4f78974c4667295ba0ade30938fdcf51ef7ab6", + "prepared_review_unit_hash": "278f8e2b89352882816713f371019683b7177f89a5ecd484ae0d9e24416bccfd", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 36184, - "estimated_input_tokens": 9046, - "estimated_eval_tokens": 18348 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 25, + "serialized_chars": 34776, + "estimated_input_tokens": 8694, + "estimated_eval_tokens": 17644 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are mostly unrelated across tokenizer, serving, quantization, multimodal, and training subsystems; none of the proposed soft pairs looks like the same underlying bug or change.", - "confidence": 0.87, - "canonical_issue_reason": "No single canonical issue stands out: the set spans many distinct code paths and symptoms, so there is no shared root cause to anchor the cluster.", - "canonical_pr_reason": "No pull requests are present in this set.", - "best_issue_reason": "No clear best issue for deduplication: while some reports are high-signal, none subsumes the others or represents the same bug family.", - "best_pr_reason": "No pull requests are present in this set.", + "summary": "This cluster is mostly a mix of unrelated Transformers bugs. None of the soft pairs look like the same underlying defect/change, so I would not merge any of them as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43957 is the best anchor: it describes a broad, concrete model-loading regression with `torch.device(\"meta\")` and clear examples, making it the most representative issue in this mixed set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43957 has the strongest global triage value because it is reproducible, well-scoped, and impacts multiple models through one loading path.", + "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44493", + "right": "issue:46285", + "accept": false, + "reason": "Different bugs: one is a positional-id/key mismatch across many models, the other is a CLIP text-model structure inconsistency." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to a stale GraniteSpeech CI test." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Different components and failure modes: BEiT legacy-field loading vs BigBirdTokenizer special-token decoding." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Compiled-forward chunked generation and Whisper empty-transcription handling are separate code paths." + }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "Different symptoms and code paths: SigLIP2 nonsensical outputs vs a position-id key warning/regression." + "reason": "SigLIP2 bad outputs are not the same issue as the generic unexpected key/position-id problem." }, { "left": "issue:43873", "right": "issue:46050", "accept": false, - "reason": "Both touch quantization, but one is about offloading behavior and the other about save/load breakage after sinq quantization." + "reason": "Both touch quantization/offload, but one is an offloading behavior bug and the other is a save/load regression for sinq quantization." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "Flash-attention sink-less crash and quantization serialization failure are unrelated bugs." - }, - { - "left": "issue:44423", - "right": "issue:44734", - "accept": false, - "reason": "Both are serve crashes, but they affect different endpoints and different failure mechanisms." + "reason": "Flash-attention sinkless-model crash is unrelated to sinq quantized model persistence." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "Flash Attention 2 test failures and MLflow reporting on failed training are unrelated." + "reason": "Flash Attention 2 test failures and MLflow job-reporting are unrelated subsystems." }, { - "left": "issue:43065", - "right": "issue:46369", + "left": "issue:44423", + "right": "issue:44734", "accept": false, - "reason": "SAM3 pixel-decoder architecture stub vs memory overflow in a SAM3 litetext model; not the same bug." + "reason": "Both are serve-related, but one is multimodal input handling and the other is KV-cache continuation indexing." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "DeepSpeed/ZeRO gradient anomaly and a multimodal model loading failure do not share a concrete root cause." + "reason": "DeepSpeed gradient growth and Phi-4 multimodal loading are different failure classes." + }, + { + "left": "issue:43065", + "right": "issue:46369", + "accept": false, + "reason": "A dummy Conv2d implementation is not the same as a memory-overflow report for the same model family." }, { "left": "issue:42490", "right": "issue:43696", "accept": false, - "reason": "Import warning on trailing-slash paths and CUDA OOM on GPT-oss-20b are unrelated." + "reason": "Local-path import warnings are unrelated to a GPT-oss-20b OOM." }, { "left": "issue:45406", "right": "issue:46121", "accept": false, - "reason": "Gemma4Processor missing _tokenizer and a rope-validation TypeError are different code paths." + "reason": "Gemma4 serve tokenizer state and rope-parameter JSON type handling are separate bugs." }, { "left": "issue:29159", "right": "issue:32090", "accept": false, - "reason": "Tokenizer slow/fast inconsistency is unrelated to the trainer _gpu_broadcast_one NoneType error." + "reason": "Tokenizer slow/fast inconsistency is unrelated to Trainer GPU broadcast NoneType errors." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Extra space after added token and num_input_tokens_seen not updating are different tokenizer vs trainer issues." + "reason": "Added-token spacing and TrainerState token-count tracking are different defects." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Compile error for non-template nodes is unrelated to the flash_attn import KeyError on Python 3.13." + "reason": "Torch compile template-node errors are unrelated to flash_attn import-key errors." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "MPS attention correctness bug and a PyTorch version-check import bug are not the same problem." + "reason": "An MPS attention correctness issue is not the same as a flex_attention import-version check bug." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "TorchScript generator-expression export failure and SigLIP2 zero-probability outputs are unrelated." + "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero probabilities are unrelated." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to Gemma4 mm_token_type_ids defaults." + "reason": "Video-input processing in Qwen3OmniMoe is unrelated to Gemma4 text-only mm_token_type_ids defaults." }, { "left": "issue:45200", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma4, but one is missing default multimodal token ids and the other is a device-map forward mismatch." + "reason": "Both involve Gemma4, but one is token-type defaults and the other is a forward device-mismatch bug." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "bfloat16 dtype mismatch in inference and Wav2Vec2 save/tokenization failure are different bugs." + "reason": "bfloat16 dtype mismatches and Wav2Vec2 save/tokenization failures are different code paths." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 implementation discrepancy and resume-from-checkpoint key mismatch do not share the same root cause." + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Missing flash-attention import symbol and meta-device model-loading failures are separate issues." + "reason": "Flash-support import errors are separate from meta-device model-loading regressions." }, { "left": "issue:44186", "right": "issue:46192", "accept": false, - "reason": "LayoutLMv2 NER padding/truncation crashes and return_tensors being ignored are different tokenizer behaviors." + "reason": "LayoutLMv2 tokenizer crashes on padding/truncation are not the same as return_tensors being ignored with text_kwargs." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "All-zero assistant masks for multimodal inputs and a crash on tool-call assistant messages are distinct apply_chat_template bugs." - }, - { - "left": "issue:43232", - "right": "issue:44849", - "accept": false, - "reason": "Generation kwargs update after sync_gpus and output_hidden_states behavior in Qwen3.5 are unrelated." - }, - { - "left": "issue:45823", - "right": "issue:46192", - "accept": false, - "reason": "Gemma4 device_map auto mismatch and return_tensors being ignored are unrelated API issues." - }, - { - "left": "issue:43295", - "right": "issue:43450", - "accept": false, - "reason": "Processor.tokenizer regression and batched video shape errors are different multimodal processing problems." - }, - { - "left": "issue:43295", - "right": "issue:44610", - "accept": false, - "reason": "Processor/tokenizer access regression does not match the OmDet-Turbo input-size mismatch." - }, - { - "left": "issue:45381", - "right": "issue:46209", - "accept": false, - "reason": "Qwen2.5-VL video position-id bug and FP8 inference failure on B200 are unrelated." - }, - { - "left": "issue:43295", - "right": "issue:43873", - "accept": false, - "reason": "Processor regression and quantization offloading problems are unrelated." - }, - { - "left": "issue:44554", - "right": "issue:45356", - "accept": false, - "reason": "MPS attention correctness and Kimi-K2.5 tokenizer codec regression are different code paths." + "reason": "Both involve apply_chat_template, but one is all-zero multimodal assistant masks and the other is a crash on tool-call messages with no content; too different to merge." } ] }, @@ -8085,7 +8106,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 26, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -8098,11 +8119,10 @@ "issue:37428", "issue:38175", "issue:39692", - "issue:41720", "issue:42490", "issue:42915", - "issue:43262", - "issue:43278", + "issue:43232", + "issue:43295", "issue:43334", "issue:43388", "issue:43450", @@ -8116,275 +8136,266 @@ "issue:43653", "issue:43824", "issue:43844", - "issue:43856", + "issue:43873", "issue:43950", "issue:43957", "issue:43975", "issue:44112", - "issue:44442", "issue:44466", - "issue:44561", + "issue:44554", + "issue:44610", "issue:44792", + "issue:44849", "issue:44898", "issue:44987", "issue:45127", - "issue:45198", "issue:45200", "issue:45325", + "issue:45356", + "issue:45381", "issue:45412", "issue:45520", "issue:45563", - "issue:46170", + "issue:45823", "issue:46178", "issue:46192", "issue:46209", - "issue:46234", "issue:46285", "issue:46512" ], "soft_pairs": [ + "issue:43232|issue:44849", + "issue:45823|issue:46192", + "issue:43295|issue:43450", + "issue:43295|issue:44610", + "issue:45381|issue:46209", + "issue:43295|issue:43873", + "issue:44554|issue:45356", "issue:39692|issue:46192", - "issue:45520|issue:46178", "issue:43572|issue:44987", - "issue:44466|issue:45127", + "issue:45520|issue:46178", "issue:43550|issue:43844", + "issue:44466|issue:45127", "issue:43572|issue:43957", - "issue:44898|issue:45412", "issue:37428|issue:43824", + "issue:44898|issue:45412", "issue:43950|issue:46285", - "issue:45200|issue:45325", "issue:43450|issue:43575", + "issue:45200|issue:45325", "issue:42490|issue:43493", "issue:44792|issue:45563", - "issue:43334|issue:43525", "issue:43452|issue:46512", + "issue:43334|issue:43525", "issue:42915|issue:43824", - "issue:43388|issue:43526", "issue:38175|issue:43493", "issue:43824|issue:43975", "issue:28218|issue:35141", - "issue:43653|issue:44112", - "issue:43526|issue:46192", - "issue:43278|issue:43856", - "issue:41720|issue:46209", - "issue:44442|issue:45198", - "issue:43262|issue:43450", - "issue:45520|issue:46170", - "issue:38175|issue:45412", - "issue:44561|issue:46234" + "issue:43388|issue:43526", + "issue:43653|issue:44112" ], - "prepared_review_unit_hash": "64934d0b8ef4adc5d36092d6f2252efbb7c13ef51ea8dda7b2636a19edeeefd7", + "prepared_review_unit_hash": "13454338c9c4aa0e9eb45f3c6b382e29efc0972eddfab5bc2ed0542084b67f45", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36809, - "estimated_input_tokens": 9203, - "estimated_eval_tokens": 18662 + "soft_pair_count": 28, + "serialized_chars": 36535, + "estimated_input_tokens": 9134, + "estimated_eval_tokens": 18524 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated transformers regressions loosely sharing v5-era API/loading themes. Only two soft pairs look like plausible duplicates: the SigLIP2 implementation discrepancy pair and the pad_token_id-missing config pair. No PRs are present.", - "confidence": 0.46, - "canonical_issue_reason": "issue:43950 is the cleanest representative: it describes a concrete, broadly applicable v5 loading regression with a clear failure mode and reproducible impact.", + "summary": "This cluster is mostly a set of unrelated Transformers bug reports with a few superficial theme overlaps (tokenizers, model loading, multimodal processors). I don\u2019t see any strong duplicate group; the closest items still describe different code paths and model families.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 43295 is the most central candidate by soft-link connectivity and is broad enough to serve as the cluster\u2019s representative, even though it is not a true duplicate of the others.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the best single issue to anchor the cluster because it is specific, high-signal, and more generally representative of the v5 regression/load-path problems than the model-specific reports.", + "best_issue_reason": "Issue 43295 is the best single representative: it has the most adjacency in the soft graph and describes a user-facing regression rather than a narrowly model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:39692", - "right": "issue:46192", + "left": "issue:43232", + "right": "issue:44849", "accept": false, - "reason": "Different bugs: SigLIP2 docs/example errors vs return_tensors being ignored in text_kwargs handling." + "reason": "Different bugs: generation kwarg handling after sync_gpus vs output_hidden_states behavior in Qwen3.5." }, { - "left": "issue:45520", - "right": "issue:46178", + "left": "issue:45823", + "right": "issue:46192", "accept": false, - "reason": "Both are import/backend errors, but they involve different missing backends and different code paths." + "reason": "Device-map forward mismatch in Gemma4 PLE is unrelated to return_tensors being ignored when text_kwargs is passed." }, { - "left": "issue:43572", - "right": "issue:44987", + "left": "issue:43295", + "right": "issue:43450", "accept": false, - "reason": "StableLm pad_token_id regression and a separate model-loading failure for physical-intelligence/fast are not the same bug." + "reason": "Processor/tokenizer regression and batched video processor shape bug affect different code paths and symptoms." }, { - "left": "issue:44466", - "right": "issue:45127", + "left": "issue:43295", + "right": "issue:44610", "accept": false, - "reason": "One is tied-weight serialization inconsistency; the other is LoRA merge collapse with extended vocab. Related area, but not the same failure." + "reason": "Custom processor.tokenizer regression is not the same as OmDet-Turbo\u2019s 640px vs 224px preprocessing mismatch." }, { - "left": "issue:43550", - "right": "issue:43844", + "left": "issue:45381", + "right": "issue:46209", "accept": false, - "reason": "torch.compile+SDPA failure in Bamba is unrelated to the ZeRO-3 gradient growth report." + "reason": "Qwen2.5-VL vision_position_ids bug is unrelated to FP8 inference failure on B200." }, { - "left": "issue:43572", - "right": "issue:43957", + "left": "issue:43295", + "right": "issue:43873", "accept": false, - "reason": "Different loading regressions: missing pad_token_idx vs meta-device loading breakage." + "reason": "Processor/tokenizer access regression and quantization offloading behavior are different problems." }, { - "left": "issue:44898", - "right": "issue:45412", + "left": "issue:44554", + "right": "issue:45356", "accept": false, - "reason": "Perceiver interpolation failure and RT-DETR memory leak are distinct issues." + "reason": "MPS attention correctness issue is unrelated to Kimi-K2.5 tokenizer codec handling." }, { - "left": "issue:37428", - "right": "issue:43824", + "left": "issue:39692", + "right": "issue:46192", "accept": false, - "reason": "Both are ImportErrors, but for different symbols and different modules; not the same underlying breakage." + "reason": "Doc example/model mismatch and return_tensors being ignored are different bugs." }, { - "left": "issue:43950", - "right": "issue:46285", + "left": "issue:43572", + "right": "issue:44987", "accept": false, - "reason": "Non-persistent buffer corruption on from_pretrained is unrelated to CLIP text-model structure changes." + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to loading physical-intelligence/fast on 5.1.0." }, { - "left": "issue:45200", - "right": "issue:45325", + "left": "issue:45520", + "right": "issue:46178", "accept": false, - "reason": "Both touch multimodal processing, but one is mm_token_type_ids defaults and the other is Qwen2.5-VL rope indexing." + "reason": "flash_attn import lookup on Python 3.13 is unrelated to tensorflow_text backend mapping." }, { - "left": "issue:43450", - "right": "issue:43575", + "left": "issue:43550", + "right": "issue:43844", "accept": false, - "reason": "Video processor shape bug and Qwen2 OOM under tensor parallelism are unrelated." + "reason": "torch.compile+SDPA failure and DeepSpeed ZeRO-3 gradient growth are distinct training issues." }, { - "left": "issue:42490", - "right": "issue:43493", + "left": "issue:44466", + "right": "issue:45127", "accept": false, - "reason": "Trailing-slash path import warning is unrelated to the SigLIP2 implementation discrepancy." + "reason": "Both mention tied weights, but one is serialization/device-dependent and the other is LoRA merge with extended vocab." }, { - "left": "issue:44792", - "right": "issue:45563", + "left": "issue:43572", + "right": "issue:43957", "accept": false, - "reason": "Janus image-generation test failure and paged generate warning are different generation issues." - }, - { - "left": "issue:43334", - "right": "issue:43525", - "accept": true, - "reason": "Both report the same class of regression: model/config objects missing pad_token_id and failing during load." + "reason": "Config missing pad_token_idx and meta-device loading breakage are different regressions." }, { - "left": "issue:43452", - "right": "issue:46512", + "left": "issue:37428", + "right": "issue:43824", "accept": false, - "reason": "gguf_file loading failure and stale config.dtype after from_config affect different APIs and code paths." + "reason": "These are separate import errors in different modules and for different symbols/models." }, { - "left": "issue:42915", - "right": "issue:43824", + "left": "issue:44898", + "right": "issue:45412", "accept": false, - "reason": "Qwen3Moe FP8 failure and Qwen2.5-VL import error are unrelated." + "reason": "Perceiver interpolation failure and RT-DETR memory release are unrelated." }, { - "left": "issue:43388", - "right": "issue:43526", + "left": "issue:43950", + "right": "issue:46285", "accept": false, - "reason": "Both are label-handling bugs, but they affect different pipelines and different transformations of labels." + "reason": "Non-persistent buffer corruption on load and CLIP text-model structure inconsistency are different loading/serialization bugs." }, { - "left": "issue:38175", - "right": "issue:43493", - "accept": true, - "reason": "Same model family and likely same underlying SigLIP2 implementation mismatch; one symptom is zero probabilities, the other is a broader HF-vs-JAX discrepancy." + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Batched video processor shape bug and tensor-parallel OOM are not the same underlying issue." }, { - "left": "issue:43824", - "right": "issue:43975", + "left": "issue:45200", + "right": "issue:45325", "accept": false, - "reason": "Importing a missing class and incorrect detokenization are unrelated." + "reason": "Gemma4 mm_token_type_ids defaulting and Qwen2.5-VL temporal scaling are distinct multimodal bugs." }, { - "left": "issue:28218", - "right": "issue:35141", + "left": "issue:42490", + "right": "issue:43493", "accept": false, - "reason": "Tokenizer extra-space behavior and embedding reinitialization are distinct regressions." + "reason": "Trailing-slash import warning is unrelated to the SigLIP2 implementation discrepancy." }, { - "left": "issue:43653", - "right": "issue:44112", + "left": "issue:44792", + "right": "issue:45563", "accept": false, - "reason": "BigBirdTokenizer special-token registration and GraniteSpeech CI device-override failure are unrelated." + "reason": "Janus generate-image test failure and stale paged-generate warning are different generate-path issues." }, { - "left": "issue:43526", - "right": "issue:46192", + "left": "issue:43452", + "right": "issue:46512", "accept": false, - "reason": "BeitImageProcessorFast label reduction and return_tensors handling are different preprocessing bugs." + "reason": "gguf_file breaking from_pretrained and dtype metadata going stale in from_config are separate loading behaviors." }, { - "left": "issue:43278", - "right": "issue:43856", + "left": "issue:43334", + "right": "issue:43525", "accept": false, - "reason": "Embedding dtype drift in eval and Qwen3 MoE memory usage are not the same issue." + "reason": "Both mention missing pad_token_id, but they affect different model families and there is no evidence of one shared duplicate bug." }, { - "left": "issue:41720", - "right": "issue:46209", + "left": "issue:42915", + "right": "issue:43824", "accept": false, - "reason": "Different model/hardware failures: Qwen3 device-mapping cuda assert vs FP8 inference on B200." + "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to the Qwen2_5_VL import error." }, { - "left": "issue:44442", - "right": "issue:45198", + "left": "issue:38175", + "right": "issue:43493", "accept": false, - "reason": "Two separate tokenizer/save-pretrained failures on different models; not one bug." + "reason": "Both involve SigLIP2, but one reports zero probabilities while the other is a broader HF-vs-JAX discrepancy; not enough to confirm the same bug." }, { - "left": "issue:43262", - "right": "issue:43450", + "left": "issue:43824", + "right": "issue:43975", "accept": false, - "reason": "Audio chat-template sampling-rate default and video batched-shape handling are unrelated processor bugs." + "reason": "Different model families and different failure modes: import error vs incorrect detokenization." }, { - "left": "issue:45520", - "right": "issue:46170", + "left": "issue:28218", + "right": "issue:35141", "accept": false, - "reason": "Both are environment/import issues, but they stem from different dependency problems and platforms." + "reason": "Tokenizer spacing after added token and embedding reinitialization after resize are unrelated." }, { - "left": "issue:38175", - "right": "issue:45412", + "left": "issue:43388", + "right": "issue:43526", "accept": false, - "reason": "SigLIP2 probability anomaly and RT-DETR memory leak are unrelated." + "reason": "Label dropping in gather_for_metrics and BeitImageProcessorFast reduce_labels are different label-handling bugs." }, { - "left": "issue:44561", - "right": "issue:46234", + "left": "issue:43653", + "right": "issue:44112", "accept": false, - "reason": "Removal of is_torch_fx_available and missing is_falcon_mamba_ssm_available are different import-utils regressions." + "reason": "BigBirdTokenizer special-token registration and a stale CI device-override test are unrelated." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative. The canonical issue choice is defensible, and the accepted soft-edge pairs are both supported by the issue titles/bodies without overstating certainty. I did not see any unsupported merge claims or clear evidence inflation." - }, - "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 }, @@ -8393,7 +8404,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 27, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -8401,57 +8412,65 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:26413", - "issue:29127", "issue:29942", - "issue:31501", "issue:32090", "issue:37428", - "issue:39692", + "issue:38175", + "issue:41720", + "issue:43262", "issue:43278", "issue:43335", "issue:43425", - "issue:43504", + "issue:43450", "issue:43526", "issue:43550", "issue:43606", "issue:43611", "issue:43653", - "issue:43720", "issue:43819", - "issue:43825", "issue:43827", + "issue:43856", "issue:43901", "issue:44112", "issue:44206", - "issue:44351", + "issue:44442", "issue:44448", - "issue:44493", "issue:44514", "issue:44521", + "issue:44561", "issue:44610", "issue:44743", - "issue:44855", "issue:44857", "issue:45083", - "issue:45127", + "issue:45198", "issue:45200", - "issue:45245", "issue:45290", + "issue:45412", + "issue:45520", "issue:45584", "issue:45588", "issue:45593", "issue:45759", - "issue:45823", - "issue:45859", + "issue:46170", + "issue:46192", + "issue:46209", + "issue:46234", "issue:46285", "issue:46326", "issue:46402", "issue:46512" ], "soft_pairs": [ - "issue:44206|issue:45584", + "issue:43278|issue:43856", + "issue:43526|issue:46192", + "issue:41720|issue:46209", + "issue:44442|issue:45198", + "issue:43262|issue:43450", + "issue:45520|issue:46170", + "issue:38175|issue:45412", "issue:43819|issue:44112", + "issue:44561|issue:46234", + "issue:44206|issue:45584", "issue:45759|issue:46285", "issue:45290|issue:46326", "issue:43827|issue:43901", @@ -8463,246 +8482,196 @@ "issue:43611|issue:46402", "issue:43606|issue:45083", "issue:37428|issue:43425", - "issue:29942|issue:32090", "issue:45759|issue:46512", + "issue:29942|issue:32090", "issue:44448|issue:44743", "issue:45200|issue:45588", - "issue:43335|issue:43653", - "issue:45127|issue:45245", - "issue:43550|issue:44112", - "issue:44855|issue:45083", - "issue:43335|issue:43504", - "issue:37428|issue:44351", - "issue:43819|issue:44493", - "issue:43278|issue:43825", - "issue:44206|issue:45823", - "issue:39692|issue:43526", - "issue:26413|issue:31501", - "issue:29127|issue:32090", - "issue:45588|issue:45859", - "issue:43335|issue:45083", - "issue:43720|issue:44112" + "issue:43335|issue:43653" ], - "prepared_review_unit_hash": "91381e24ffc99789ac1eb94f872deca13ff4fdcf1da3eff3fa44ffa07e79fe21", + "prepared_review_unit_hash": "0cc1120fa0d1aa4ff343abf005aba2279411baba8ded7b8e844be9892fc46496", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 37032, - "estimated_input_tokens": 9258, - "estimated_eval_tokens": 18772 + "soft_pair_count": 26, + "serialized_chars": 35069, + "estimated_input_tokens": 8768, + "estimated_eval_tokens": 17792 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is broadly heterogeneous: mostly unrelated Transformers bug reports spanning model loading, chat templates, processors, CI regressions, and config edge cases. The soft edges are largely semantic false positives; I would not merge any of the suggested pairs as duplicates.", - "confidence": 0.86, - "canonical_issue_reason": "issue:44521 is the strongest representative of the cluster\u2019s recurring chat-template/multimodal theme: it is concrete, reproducible, high-activity, and sits in the most broadly relevant code path among the nearby tokenizer/template reports.", + "summary": "This cluster is mostly a grab-bag of unrelated issue reports. I did not find any true duplicate pair; the closest match is the Qwen2.5-VL apply_chat_template pair, but the failure modes are still distinct enough to keep separate.", + "confidence": 0.31, + "canonical_issue_reason": "issue:44521 is the broadest, highest-signal representative of the multimodal apply_chat_template/assistant_mask thread and has the most follow-up discussion among the related subset.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the best single issue to anchor this cluster because it has the most discussion/inbound interest, a clear failure mode, and broad user impact compared with the more niche config/import/CI reports.", + "best_issue_reason": "issue:44521 is the best single issue anchor because it is open, active, and spans the widest version of the chat-template masking problem, even though the overall cluster is heterogeneous.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44206", - "right": "issue:45584", + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Different subsystems and symptoms: dtype regression in evaluation vs MoE training memory usage." + }, + { + "left": "issue:43526", + "right": "issue:46192", + "accept": false, + "reason": "Unrelated bugs: BeitImageProcessor label reduction vs return_tensors being ignored in explicit text_kwargs." + }, + { + "left": "issue:41720", + "right": "issue:46209", + "accept": false, + "reason": "Different model/runtime failures: Qwen3 auto device mapping cuda assert vs FP8 inference on B200." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Both are tokenizer/loading problems, but they affect different models and appear to have different root causes." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate bug vs batched video shape bug; no shared underlying code path is evident." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Unrelated import/environment failures: flash_attn lookup on Python 3.13 vs Windows ROCm distributed dependency issue." + }, + { + "left": "issue:38175", + "right": "issue:45412", "accept": false, - "reason": "Different bugs: one is a feature-extractor API regression, the other is Whisper generation failing after special-token alignment." + "reason": "Different models and behaviors: zero probabilities in SigLIP2 vs RT-DETR memory not being released." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "Unrelated code paths: DAC latent conversion mismatch vs a GraniteSpeech CI device-override test failure." + "reason": "DAC latent/STE mismatch is unrelated to GraniteSpeech's stale device override test failure." + }, + { + "left": "issue:44561", + "right": "issue:46234", + "accept": false, + "reason": "Both are missing/removed utility symbols, but they affect different functions and code paths." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Different audio bugs: unsupported center arg in feature extraction vs empty-transcription failure after align_special_tokens." }, { "left": "issue:45759", "right": "issue:46285", "accept": false, - "reason": "Both involve model structure, but the concrete problems differ: composite Qwen config unwrapping vs CLIP text-model attribute inconsistency." + "reason": "Different model-construction regressions: Qwen multimodal config unwrapping vs CLIP text-model structural inconsistency." }, { "left": "issue:45290", "right": "issue:46326", "accept": false, - "reason": "Both touch chat templates, but they fail in different scenarios and code paths: tool-call tokenization crash vs mistral_common continue_final_message handling." + "reason": "Both involve apply_chat_template, but one is a tool-call/content crash and the other is mistral_common continue_final_message behavior." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are docs-related, but they describe different API mismatches and would not be fixed by one PR." + "reason": "Documentation regressions in the same area, but they refer to different deprecated fields and are not the same issue." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "Different vision bugs: OmDet-Turbo input-size mismatch vs LwDetrImageLoss AMP/CUDA crash." + "reason": "Unrelated vision bugs: processor input size mismatch vs float16 AMP loss crash." }, { "left": "issue:44514", "right": "issue:44521", "accept": false, - "reason": "Same general area, but not the same underlying bug: one crashes on batched padding=False input, the other returns incorrect all-zero assistant masks." + "reason": "Same general API, but one is a batched padding=False crash and the other is all-zero assistant masks; not clearly the same bug." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "Unrelated failures: DAC forward/latents mismatch vs qwen3_omni_moe feature-length helper behavior." + "reason": "Different audio/model concerns: DAC latent decoding mismatch vs helper length computation in qwen3_omni_moe." }, { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "Different training issues: AMP crash in LwDetrImageLoss vs missing auxiliary losses when denoising is off." + "reason": "Different detection training issues: AMP crash vs denoising-off auxiliary-loss logic." }, { "left": "issue:43335", "right": "issue:43550", "accept": false, - "reason": "Distinct model bugs in different subsystems: SwitchTransformers config construction vs Bamba torch.compile/SDPA failure." + "reason": "Different models and failure modes: SwitchTransformers config bug vs Bamba torch.compile SDPA failure." }, { "left": "issue:43611", "right": "issue:46402", "accept": false, - "reason": "Both concern model loading, but one is a Transformers 5 base_model_prefix regression and the other is Sam3 text-encoder weights not being loaded." + "reason": "Both concern loading regressions, but the specific breakages and affected models are different." }, { "left": "issue:43606", "right": "issue:45083", "accept": false, - "reason": "Different areas: CPU offload device mismatch in bark-small vs qwen3_omni_moe feature-length computation." + "reason": "Device-mismatch offload bug is unrelated to qwen3_omni_moe feature-length computation." }, { "left": "issue:37428", "right": "issue:43425", "accept": false, - "reason": "Import error vs version compatibility issue; no shared concrete code-path." + "reason": "Flash-attention import breakage and Torch 2.10 incompatibility are different environment/version issues." }, { - "left": "issue:29942", - "right": "issue:32090", + "left": "issue:45759", + "right": "issue:46512", "accept": false, - "reason": "CI Flash Attention test failure and a Trainer broadcast TypeError are unrelated." + "reason": "Both touch from_config/config state, but one is multimodal text_config unwrapping and the other is dtype staleness." }, { - "left": "issue:45759", - "right": "issue:46512", + "left": "issue:29942", + "right": "issue:32090", "accept": false, - "reason": "Both involve from_config behavior, but one is about composite config unwrapping while the other is about stale dtype metadata." + "reason": "Flash Attention 2 test failures and Trainer broadcast NoneType errors are different bugs." }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Different model behaviors: Pegasus output drift across versions vs recurrent state reset with cache." + "reason": "Pegasus output drift between v4/v5 is unrelated to Qwen3.5 recurrent-state reset behavior." }, { "left": "issue:45200", "right": "issue:45588", "accept": false, - "reason": "Gemma4 multimodal token-type defaults and flash_attention sink-less-model handling are separate bugs." + "reason": "Gemma mm_token_type_ids defaulting and flash_attention sink-less-model crash are unrelated." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug and BigBirdTokenizer special-token registration are unrelated." - }, - { - "left": "issue:45127", - "right": "issue:45245", - "accept": false, - "reason": "LoRA/weight-tying collapse and category-cardinality limits are different failures." - }, - { - "left": "issue:43550", - "right": "issue:44112", - "accept": false, - "reason": "Bamba compile/SDPA failure is unrelated to GraniteSpeech stale device override CI behavior." - }, - { - "left": "issue:44855", - "right": "issue:45083", - "accept": false, - "reason": "Python 3.13 import/torch.jit parsing issue and qwen3_omni_moe feature-length helper behavior are unrelated." - }, - { - "left": "issue:43335", - "right": "issue:43504", - "accept": false, - "reason": "Config-generated sparse layers and Beit preset loading with a legacy field are different problems." - }, - { - "left": "issue:37428", - "right": "issue:44351", - "accept": false, - "reason": "Both are import errors, but they affect different symbols and likely different fixes." - }, - { - "left": "issue:43819", - "right": "issue:44493", - "accept": false, - "reason": "DAC latent mismatch and unexpected position-id keys are not the same bug." - }, - { - "left": "issue:43278", - "right": "issue:43825", - "accept": false, - "reason": "Embedding dtype drift and pipeline error-message wording are unrelated." - }, - { - "left": "issue:44206", - "right": "issue:45823", - "accept": false, - "reason": "Different model/processor bugs: unsupported center arg vs device mismatch with Gemma4 PLE." - }, - { - "left": "issue:39692", - "right": "issue:43526", - "accept": false, - "reason": "Docs example errors for SigLIP2 and BeitImageProcessorFast label reduction are unrelated." - }, - { - "left": "issue:26413", - "right": "issue:31501", - "accept": false, - "reason": "Checkpoint resume iterator failure and FineWeb training startup failure are not the same underlying issue." - }, - { - "left": "issue:29127", - "right": "issue:32090", - "accept": false, - "reason": "LayoutLMv3 error-message clarity and Trainer NoneType broadcast failure are distinct." - }, - { - "left": "issue:45588", - "right": "issue:45859", - "accept": false, - "reason": "Flash-attention s_aux handling and Qwen3_5Moe CPU-offload compatibility are separate code paths." - }, - { - "left": "issue:43335", - "right": "issue:45083", - "accept": false, - "reason": "SwitchTransformers sparse-layer config and qwen3_omni_moe feature-length helper behavior are unrelated." - }, - { - "left": "issue:43720", - "right": "issue:44112", - "accept": false, - "reason": "BitNet packed-weight loading and a GraniteSpeech CI test issue are not the same bug." + "reason": "Different components and symptoms: SwitchTransformers sparse-layer creation vs BigBirdTokenizer special-token registration." } ] }, @@ -8718,7 +8687,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 28, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -8726,13 +8695,11 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:10000", - "issue:15354", - "issue:27957", + "issue:26413", "issue:28218", - "issue:29942", + "issue:29127", + "issue:31501", "issue:32090", - "issue:33357", "issue:34689", "issue:35141", "issue:37428", @@ -8740,16 +8707,19 @@ "issue:39692", "issue:40444", "issue:42490", + "issue:43278", "issue:43335", + "issue:43504", "issue:43526", "issue:43550", - "issue:43575", "issue:43606", - "issue:43701", - "issue:43746", + "issue:43720", + "issue:43819", + "issue:43825", "issue:43854", - "issue:43927", "issue:44112", + "issue:44206", + "issue:44351", "issue:44360", "issue:44448", "issue:44493", @@ -8757,273 +8727,272 @@ "issue:44779", "issue:44855", "issue:44857", - "issue:44977", - "issue:44987", "issue:45072", "issue:45083", - "issue:45229", + "issue:45127", + "issue:45245", "issue:45290", "issue:45310", "issue:45381", "issue:45459", "issue:45478", + "issue:45588", "issue:45593", - "issue:45698", + "issue:45823", + "issue:45859", "issue:45920", "issue:46192", - "issue:46209", - "issue:46234", - "issue:46399" + "issue:46234" ], "soft_pairs": [ + "issue:45127|issue:45245", + "issue:43550|issue:44112", + "issue:44855|issue:45083", + "issue:37428|issue:44351", + "issue:43335|issue:43504", + "issue:43819|issue:44493", + "issue:43278|issue:43825", + "issue:39692|issue:43526", + "issue:44206|issue:45823", + "issue:29127|issue:32090", + "issue:26413|issue:31501", + "issue:45588|issue:45859", + "issue:43335|issue:45083", + "issue:43720|issue:44112", "issue:45459|issue:46234", "issue:45310|issue:45478", "issue:43335|issue:44112", - "issue:45290|issue:45381", "issue:43335|issue:45593", + "issue:45290|issue:45381", "issue:39692|issue:44493", "issue:44448|issue:44779", - "issue:34689|issue:35141", "issue:43606|issue:44857", + "issue:34689|issue:35141", "issue:42490|issue:43854", "issue:44360|issue:44561", "issue:44857|issue:45072", "issue:38175|issue:44493", - "issue:45920|issue:46192", "issue:28218|issue:32090", - "issue:40444|issue:43526", - "issue:44448|issue:44977", - "issue:43701|issue:45083", - "issue:44977|issue:45229", - "issue:29942|issue:35141", - "issue:43746|issue:46209", - "issue:32090|issue:33357", - "issue:44855|issue:46399", - "issue:43550|issue:45083", - "issue:42490|issue:43575", - "issue:43927|issue:45698", - "issue:10000|issue:15354", - "issue:44561|issue:44779", - "issue:27957|issue:34689", - "issue:37428|issue:44987" + "issue:45920|issue:46192", + "issue:40444|issue:43526" ], - "prepared_review_unit_hash": "e447abb019466aaf61ef256dffaf27d7e2d8c0e44c2c3dbff5024d0307e34c46", + "prepared_review_unit_hash": "75758e6ae6071f16f485c0ad99a716406a6e3b84595508fcc8b6501e19d99541", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36725, - "estimated_input_tokens": 9182, - "estimated_eval_tokens": 18620 + "serialized_chars": 36962, + "estimated_input_tokens": 9241, + "estimated_eval_tokens": 18738 }, "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 issues. The strongest relationship is 45310/45478, which look like a likely duplicate or near-duplicate pair for the same Qwen3.5 MoE `from_pretrained` regression, but the packet alone does not prove they are an exact duplicate with high certainty.", - "confidence": 0.86, - "canonical_issue_reason": "Issue 45310 is the best canonical issue because it is the earlier report and the titles match exactly with 45478, making it the strongest likely duplicate/near-duplicate for the Qwen3.5 MoE `from_pretrained` regression. I would not claim exact duplication from the packet alone.", + "summary": "The cluster is mostly a set of unrelated issue reports across different bugs and subsystems. Only the two identical Qwen3.5 MoE from_pretrained reports look like true duplicates; the rest are too semantically distant to merge.", + "confidence": 0.95, + "canonical_issue_reason": "Pick issue:45310 as the canonical issue: it is the earlier report of the Qwen3.5 MoE from_pretrained regression and has essentially the same title/root problem as 45478, making it the best anchor for the duplicate pair.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45310 is the strongest global issue representative: it is specific, earlier, and the closest match to 45478. The evidence supports a likely duplicate/near-duplicate relationship, not a guaranteed exact same failure.", + "best_issue_reason": "Issue:45310 is the best representative because it states the concrete regression clearly and is the original of the only true duplicate pair in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45459", - "right": "issue:46234", + "left": "issue:45127", + "right": "issue:45245", "accept": false, - "reason": "Both are import/utility-related, but they describe different bugs in different code paths: tokenizer error masking vs missing availability helper." - }, - { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "Very strong similarity: identical title, same Qwen3.5 MoE `from_pretrained` regression, and same version context. This is best treated as a likely duplicate/near-duplicate pair, though the packet alone does not prove exact identity." + "reason": "Different failure modes: LoRA/vocabulary merge collapse vs category-count limit in labeling." }, { - "left": "issue:43335", + "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug vs GraniteSpeech CI stale-device test failure; unrelated failures." + "reason": "One is an SDPA/torch.compile model bug; the other is a CI stale device-override test failure." }, { - "left": "issue:45290", - "right": "issue:45381", + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Unrelated models and code paths: DebertaV2 import parsing vs qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:37428", + "right": "issue:44351", "accept": false, - "reason": "Chat template tool-call crash vs Qwen2.5-VL vision position-id issue; different functionality and failure mode." + "reason": "Both are import errors, but for different missing symbols and different modules; not the same underlying bug." }, { "left": "issue:43335", - "right": "issue:45593", + "right": "issue:43504", "accept": false, - "reason": "Sparse-layer configuration bug vs D-FINE auxiliary-loss behavior; no shared underlying bug." + "reason": "SwitchTransformers sparse-layer config bug vs BEiT legacy-field loading issue." }, { - "left": "issue:39692", + "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 doc example errors vs position-id key warnings; distinct issues." + "reason": "DAC latent decoding mismatch is unrelated to position-id key warnings." }, { - "left": "issue:44448", - "right": "issue:44779", + "left": "issue:43278", + "right": "issue:43825", "accept": false, - "reason": "Pegasus version-output mismatch vs DeepSeek tokenizer regression; different models and code paths." + "reason": "Embedding dtype drift during eval vs a pipeline error-message regression." }, { - "left": "issue:34689", - "right": "issue:35141", + "left": "issue:39692", + "right": "issue:43526", "accept": false, - "reason": "Llama 3.2 vision loading problem vs token-embedding reinitialization after resize; unrelated." + "reason": "SigLIP2 docs/quantization issues are unrelated to BEiT fast image processor label handling." }, { - "left": "issue:43606", - "right": "issue:44857", + "left": "issue:44206", + "right": "issue:45823", "accept": false, - "reason": "CPU offload device mismatch vs AMP/CUDA loss crash; different subsystems." + "reason": "Different models and bugs: unsupported feature extractor arg vs device-mismatch forward pass." }, { - "left": "issue:42490", - "right": "issue:43854", + "left": "issue:29127", + "right": "issue:32090", "accept": false, - "reason": "Local-path import warning vs GLM-4.7-Flash load failure; not the same bug." + "reason": "LayoutLMV3 error-message clarity vs Trainer NoneType broadcast failure." }, { - "left": "issue:44360", - "right": "issue:44561", + "left": "issue:26413", + "right": "issue:31501", "accept": false, - "reason": "DSA indexer architecture comment vs removal of `is_torch_fx_available`; unrelated." + "reason": "Checkpoint resume iterator exhaustion is not the same as FineWeb training not starting." }, { - "left": "issue:44857", - "right": "issue:45072", + "left": "issue:45588", + "right": "issue:45859", "accept": false, - "reason": "Different numerical/mixed-precision failures in different models; not mergeable as one fix." + "reason": "Flash-attention sinkless-model crash vs Qwen3_5Moe rotary embedding CPU-offload incompatibility." }, { - "left": "issue:38175", - "right": "issue:44493", + "left": "issue:43335", + "right": "issue:45083", "accept": false, - "reason": "SigLIP2 zero-probabilities issue vs position-id key warnings; different symptoms and likely different causes." + "reason": "Different bugs in different subsystems; config creation vs helper length computation." }, { - "left": "issue:45920", - "right": "issue:46192", + "left": "issue:43720", + "right": "issue:44112", "accept": false, - "reason": "Tokenizer regression affecting many models vs `return_tensors` handling when `text_kwargs` is passed; separate API bugs." + "reason": "BitNet packed-weight loading bug vs GraniteSpeech stale CI override test." }, { - "left": "issue:28218", - "right": "issue:32090", + "left": "issue:45459", + "right": "issue:46234", "accept": false, - "reason": "Tokenizer extra-space behavior vs Trainer `_gpu_broadcast_one` NoneType error; no common code path." + "reason": "Tokenizer protobuf error masking vs missing import_utils symbol; different fixes." }, { - "left": "issue:40444", - "right": "issue:43526", - "accept": false, - "reason": "Qwen2.5-VL iterable dataset/multi-image bug vs BEiT fast processor label reduction bug; unrelated model components." + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Exact same Qwen3.5 MoE from_pretrained regression title and underlying failure." }, { - "left": "issue:44448", - "right": "issue:44977", + "left": "issue:43335", + "right": "issue:44112", "accept": false, - "reason": "Pegasus output differences vs Qwen3.5 flash-attention generation failure; different models and mechanisms." + "reason": "SwitchTransformers sparse-layer config bug and GraniteSpeech CI failure are unrelated." }, { - "left": "issue:43701", - "right": "issue:45083", + "left": "issue:43335", + "right": "issue:45593", "accept": false, - "reason": "Checkpoint-key mismatch vs qwen3_omni_moe feature-length helper behavior; unrelated." + "reason": "Config bug vs missing auxiliary losses in D-FINE; different code paths." }, { - "left": "issue:44977", - "right": "issue:45229", + "left": "issue:45290", + "right": "issue:45381", "accept": false, - "reason": "Flash-attention generation bug vs multi-GPU OOM; one is correctness, the other capacity." + "reason": "Chat-template tool-call crash vs Qwen2.5-VL video position-id issue." }, { - "left": "issue:29942", - "right": "issue:35141", + "left": "issue:39692", + "right": "issue:44493", "accept": false, - "reason": "Flash-attention test failures vs embedding reinit on resize; no shared underlying bug." + "reason": "Documentation/model-processor mismatch is not the same as unexpected position-id keys." }, { - "left": "issue:43746", - "right": "issue:46209", + "left": "issue:44448", + "right": "issue:44779", "accept": false, - "reason": "PEFT local checkpoint loading vs FP8 inference on B200; different failure modes." + "reason": "Pegasus v4/v5 output divergence vs DeepSeek tokenizer regression; distinct tokenizer/model behavior." }, { - "left": "issue:32090", - "right": "issue:33357", + "left": "issue:43606", + "right": "issue:44857", "accept": false, - "reason": "Trainer broadcast NoneType error vs MacOS bus error for CLIP; unrelated." + "reason": "CPU-offload device mismatch vs AMP float16 loss crash; different runtime paths." }, { - "left": "issue:44855", - "right": "issue:46399", + "left": "issue:34689", + "right": "issue:35141", "accept": false, - "reason": "Python 3.13 `torch.jit.script` parsing/IndentationError vs missing decorator registration in Qwen3.5; different code paths." + "reason": "Llama 3.2 Vision load regression vs post_init reinitializing output embeddings." }, { - "left": "issue:43550", - "right": "issue:45083", + "left": "issue:42490", + "right": "issue:43854", "accept": false, - "reason": "torch.compile/SDPA issue vs feature-length helper bug; unrelated." + "reason": "Local-path import warning vs GLM-4.7-Flash unit-test load failure." }, { - "left": "issue:42490", - "right": "issue:43575", + "left": "issue:44360", + "right": "issue:44561", "accept": false, - "reason": "Import warning on slash-suffixed local path vs tensor-parallel OOM loading Qwen2-57B; different problems." + "reason": "DSA indexer activation issue is unrelated to removal of is_torch_fx_available." }, { - "left": "issue:43927", - "right": "issue:45698", + "left": "issue:44857", + "right": "issue:45072", "accept": false, - "reason": "Custom token IDs lost on save/load vs wrong custom module selected after save_pretrained; both involve loading, but not the same concrete bug." + "reason": "Different dtype/device mismatches affecting different models and execution modes." }, { - "left": "issue:10000", - "right": "issue:15354", + "left": "issue:38175", + "right": "issue:44493", "accept": false, - "reason": "German DistilBert issue vs generator expression unsupported by torch.jit.script; unrelated." + "reason": "Zero-probability SigLIP2 issue is not the same as unexpected position-id keys." }, { - "left": "issue:44561", - "right": "issue:44779", + "left": "issue:28218", + "right": "issue:32090", "accept": false, - "reason": "Missing compatibility helper breaks trust_remote_code vs DeepSeek tokenizer regression; different breakages." + "reason": "Tokenizer spacing bug is unrelated to Trainer broadcast NoneType error." }, { - "left": "issue:27957", - "right": "issue:34689", + "left": "issue:45920", + "right": "issue:46192", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support vs Llama 3.2 vision model loading; different models and fixes." + "reason": "Wrong AutoTokenizer IDs across models vs return_tensors being ignored when text_kwargs is passed." }, { - "left": "issue:37428", - "right": "issue:44987", + "left": "issue:40444", + "right": "issue:43526", "accept": false, - "reason": "Missing flash-attention helper symbol vs generic load failure for a specific model package; too little evidence of a shared underlying bug." + "reason": "IterableDataset multi-image fine-tuning failure vs BEiT Fast reduce_labels bug." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The overall summary is reasonable, but it overstates the 45310/45478 relationship. The packet only gives a soft-similarity pair with `deterministic_accept: false`; identical titles make them look closely related, but not enough to call them an exact duplicate with high certainty. I\u2019d soften the language to \u2018likely duplicate/near-duplicate\u2019 and avoid asserting an exact same failure from the packet alone." + "accept": true, + "feedback": "The analysis is grounded in the packet and stays conservative on the soft-edge calls. The only clearly duplicate pair here is issue:45310 and issue:45478, which have identical titles and a strong similarity score, so choosing 45310 as canonical is reasonable. The remaining pair verdicts are appropriately cautious given the titles/body summaries available." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -9033,7 +9002,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 29, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -9041,302 +9010,302 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:10000", + "issue:15354", "issue:25251", - "issue:29127", + "issue:27957", + "issue:29942", "issue:32090", "issue:33290", "issue:33357", - "issue:34567", - "issue:39692", + "issue:34689", + "issue:35141", + "issue:37428", "issue:41720", "issue:42490", - "issue:42915", - "issue:42994", - "issue:43065", "issue:43278", "issue:43329", "issue:43381", "issue:43412", + "issue:43550", "issue:43575", - "issue:43611", - "issue:43653", + "issue:43701", + "issue:43746", "issue:43873", + "issue:43927", "issue:43976", - "issue:43994", - "issue:44186", - "issue:44206", - "issue:44361", "issue:44448", "issue:44466", - "issue:44479", + "issue:44561", "issue:44568", - "issue:44743", + "issue:44779", + "issue:44855", "issue:44857", - "issue:44871", "issue:44977", + "issue:44987", + "issue:45083", "issue:45198", "issue:45229", "issue:45245", "issue:45372", "issue:45406", - "issue:45412", "issue:45440", - "issue:45459", "issue:45491", "issue:45563", "issue:45588", - "issue:45712", + "issue:45698", "issue:45803", "issue:45823", - "issue:46121" + "issue:46209", + "issue:46399" ], "soft_pairs": [ - "issue:43278|issue:43329", + "issue:43701|issue:45083", + "issue:44977|issue:45229", + "issue:44448|issue:44977", + "issue:32090|issue:33357", + "issue:29942|issue:35141", + "issue:43746|issue:46209", + "issue:44855|issue:46399", + "issue:43550|issue:45083", + "issue:42490|issue:43575", + "issue:43927|issue:45698", + "issue:10000|issue:15354", + "issue:27957|issue:34689", + "issue:44561|issue:44779", + "issue:37428|issue:44987", "issue:43873|issue:45229", + "issue:43278|issue:43329", "issue:44857|issue:45588", "issue:41720|issue:45229", - "issue:44568|issue:45491", "issue:44977|issue:45563", + "issue:44568|issue:45491", "issue:43278|issue:43381", + "issue:45588|issue:45823", "issue:43976|issue:45440", "issue:45372|issue:45588", "issue:43412|issue:45588", - "issue:45588|issue:45823", "issue:45372|issue:45406", "issue:44448|issue:44466", "issue:45198|issue:45803", "issue:25251|issue:32090", "issue:45245|issue:45563", - "issue:43653|issue:44743", - "issue:32090|issue:33290", - "issue:42490|issue:43873", - "issue:43065|issue:45412", - "issue:44206|issue:44479", - "issue:43065|issue:43994", - "issue:44361|issue:45198", - "issue:44186|issue:45198", - "issue:33357|issue:39692", - "issue:43611|issue:46121", - "issue:29127|issue:34567", - "issue:44871|issue:44977", - "issue:42915|issue:42994", - "issue:45459|issue:45712", - "issue:43278|issue:43575" + "issue:32090|issue:33290" ], - "prepared_review_unit_hash": "afa0777d3446ed206721f8bde2e81c5d8751fa58b3ae22d175ec0c6722a59569", + "prepared_review_unit_hash": "a0eb75a583b492787f9b8ed3f56d76175f1baeaeebdf15e0a5c9be010ba0f35b", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37223, - "estimated_input_tokens": 9306, - "estimated_eval_tokens": 18868 + "serialized_chars": 37166, + "estimated_input_tokens": 9292, + "estimated_eval_tokens": 18840 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the soft pairs mostly share only broad vocabulary (trainer, quantization, tokenizer, Gemma/Qwen, etc.) and do not look like the same underlying bug or change. I would not merge any of the proposed soft edges.", - "confidence": 0.97, - "canonical_issue_reason": "issue:43873 is the best anchor for the set because it is open, discussion-heavy, and has inbound references, giving it the strongest triage signal; however, it is still a specific quantization/offloading bug, not a duplicate of the other reports.", - "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the strongest representative issue: it is open, well-discussed, and has the clearest cross-link signal among the candidates. It is still only a representative, not a true duplicate target for the rest.", - "best_pr_reason": null, + "summary": "This cluster is not a duplicate set: the items span unrelated bugs in tokenization, flash-attention, model loading, quantization, export, and training/runtime behavior. None of the soft pairs look like the same underlying bug or change, so no mergeable duplicate canon emerges.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue fits the cluster: the soft-linked items describe different code paths and failure modes, so choosing one would incorrectly collapse unrelated reports.", + "canonical_pr_reason": "No pull requests are present in the cluster.", + "best_issue_reason": "No issue is a strong global canonical candidate because there is no coherent duplicate group; the pairwise similarities are title-level only and the underlying problems differ.", + "best_pr_reason": "No pull requests are present in the cluster.", "soft_edge_verdicts": [ { - "left": "issue:43278", - "right": "issue:43329", + "left": "issue:43701", + "right": "issue:45083", "accept": false, - "reason": "Different bug paths: BF16/evaluate dtype drift vs undefined video-token helper variables." + "reason": "Checkpoint key mismatch on resume_from_checkpoint is unrelated to qwen3_omni_moe feature-extraction length behavior." }, { - "left": "issue:43873", + "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Both are memory/quantization themed, but one is offloading failure and the other is Gemma4 multi-GPU OOM." + "reason": "Flash-attention generation issues for Qwen3.5 are not the same bug as Gemma4 multi-GPU CUDA OOM." }, { - "left": "issue:44857", - "right": "issue:45588", + "left": "issue:44448", + "right": "issue:44977", "accept": false, - "reason": "AMP crash in image loss code vs flash-attention AttributeError; unrelated components." + "reason": "Pegasus v4/v5 output differences are a different problem from Qwen3.5 flash-attention generation failure." }, { - "left": "issue:41720", - "right": "issue:45229", + "left": "issue:32090", + "right": "issue:33357", "accept": false, - "reason": "Qwen3 device-map assert vs Gemma4 OOM; different models and failure modes." + "reason": "Trainer _gpu_broadcast_one NoneType error is unrelated to the MacOS bus error with a community CLIP model." }, { - "left": "issue:44568", - "right": "issue:45491", + "left": "issue:29942", + "right": "issue:35141", "accept": false, - "reason": "Tokenizer special-token handling vs NaN embeddings in attention; different root causes." + "reason": "Flash-Attention 2 test failures are unrelated to token embedding reinitialization after resizing." }, { - "left": "issue:44977", - "right": "issue:45563", + "left": "issue:43746", + "right": "issue:46209", "accept": false, - "reason": "Generation regression vs stale warning; not the same bug." + "reason": "PEFT local checkpoint loading is not the same as FP8 inference failure on B200/SM100." }, { - "left": "issue:43278", - "right": "issue:43381", + "left": "issue:44855", + "right": "issue:46399", "accept": false, - "reason": "Training/eval checkpointing guard vs embedding dtype drift; no overlap." + "reason": "Python 3.13 IndentationError in a JIT-scripted model is unrelated to a rotary-pos-emb registration decorator bug in Qwen3.5." }, { - "left": "issue:43976", - "right": "issue:45440", + "left": "issue:43550", + "right": "issue:45083", "accept": false, - "reason": "Python version compatibility vs DeepSeekMoE model divergence; unrelated." + "reason": "Bamba torch.compile/SDPA failure is not the same as qwen3_omni_moe feature-length calculation behavior." }, { - "left": "issue:45372", - "right": "issue:45588", + "left": "issue:42490", + "right": "issue:43575", "accept": false, - "reason": "Gemma4 processor import issue vs flash-attention crash; different code paths." + "reason": "Trailing-slash local path import warning is unrelated to tensor-parallel loading OOM for Qwen2-57B-A14B." }, { - "left": "issue:43412", - "right": "issue:45588", + "left": "issue:43927", + "right": "issue:45698", "accept": false, - "reason": "Executorch export guard/erfinv issue vs flash-attention None handling; unrelated." + "reason": "DiaConfig custom token ID loss after save/load is a different bug from loading the wrong custom module after save_pretrained." }, { - "left": "issue:45588", - "right": "issue:45823", + "left": "issue:10000", + "right": "issue:15354", "accept": false, - "reason": "Same Gemma4 family, but missing _tokenizer in serve is a different concrete failure than s_aux=None in flash attention." + "reason": "German DistilBertModel error and GeneratorExp/JIT export error are different model/runtime failures." }, { - "left": "issue:45372", - "right": "issue:45406", + "left": "issue:27957", + "right": "issue:34689", "accept": false, - "reason": "Gemma4 processor import failure vs serve-time missing _tokenizer; related model family but distinct bugs." + "reason": "XLMRoberta Flash Attention 2 support and Llama 3.2 90B Vision model-loading regression are unrelated." }, { - "left": "issue:44448", - "right": "issue:44466", + "left": "issue:44561", + "right": "issue:44779", "accept": false, - "reason": "Pegasus v4/v5 output difference vs lm_head.weight serialization; different regression surfaces." + "reason": "Removal of is_torch_fx_available affecting trust_remote_code is unrelated to DeepSeek tokenizer output changes." }, { - "left": "issue:45198", - "right": "issue:45803", + "left": "issue:37428", + "right": "issue:44987", "accept": false, - "reason": "Wav2Vec2 save/tokenization failure vs Fuyu exception handling in tensor conversion; unrelated." + "reason": "Missing _flash_supports_window_size import and transformers>=5.1.0 load failure for physical-intelligence/fast are different import/load issues." }, { - "left": "issue:25251", - "right": "issue:32090", + "left": "issue:43873", + "right": "issue:45229", "accept": false, - "reason": "Pipeline top_k nesting bug vs Trainer broadcast NoneType error; different subsystems." + "reason": "Quantization offloading behavior and Gemma4 multi-GPU CUDA OOM are not the same failure mode." }, { - "left": "issue:45245", - "right": "issue:45563", + "left": "issue:43278", + "right": "issue:43329", "accept": false, - "reason": "Category-count runtime error vs stale generate warning; unrelated." + "reason": "Embedding dtype drift between train/eval is unrelated to undefined variables in the video branch of _get_num_multimodal_tokens." }, { - "left": "issue:43653", - "right": "issue:44743", + "left": "issue:44857", + "right": "issue:45588", "accept": false, - "reason": "BigBirdTokenizer special-token registration vs modular Qwen3_5 recurrent-state reset; different components." + "reason": "LwDetrImageLoss AMP/CUDA crash is not the same as the flash_attention.py s_aux=None AttributeError." }, { - "left": "issue:32090", - "right": "issue:33290", + "left": "issue:41720", + "right": "issue:45229", "accept": false, - "reason": "Trainer NoneType broadcast error vs Adafactor+DeepSpeed OOM; both training-related but not the same failure." + "reason": "Qwen3 device-mapping cudaErrorAssert and Gemma4 multi-GPU CUDA OOM are different GPU placement/runtime problems." }, { - "left": "issue:42490", - "right": "issue:43873", + "left": "issue:44977", + "right": "issue:45563", "accept": false, - "reason": "Local-path trailing-slash warning vs quantization offloading behavior; unrelated." + "reason": "Qwen3.5 flash-attention generation failure is unrelated to the stale num_return_sequences warning in paged generate()." }, { - "left": "issue:43065", - "right": "issue:45412", + "left": "issue:44568", + "right": "issue:45491", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d vs RT-DETR memory leak; different problems." + "reason": "mdeberta-v3-base special-token handling is unrelated to Gemma3 sliding-window NaN embeddings." }, { - "left": "issue:44206", - "right": "issue:44479", + "left": "issue:43278", + "right": "issue:43381", "accept": false, - "reason": "Feature extractor center-arg crash vs Qwen VL video regression; unrelated." + "reason": "Training-to-eval embedding dtype change is not the same as gradient checkpointing being disallowed in eval mode." }, { - "left": "issue:43065", - "right": "issue:43994", + "left": "issue:45588", + "right": "issue:45823", "accept": false, - "reason": "Dummy Conv2d cleanup vs SigLIP2 model/pipeline mismatch; different issues." + "reason": "A flash_attention.py crash on sink-less models is unrelated to Gemma4 PLE device mismatch under device_map=auto." }, { - "left": "issue:44361", - "right": "issue:45198", + "left": "issue:43976", + "right": "issue:45440", "accept": false, - "reason": "MLUke tokenizer AttributeError vs Wav2Vec2 save/tokenization failure; different tokenizers and causes." + "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to DeepseekV3 MoE divergence from the remote implementation." }, { - "left": "issue:44186", - "right": "issue:45198", + "left": "issue:45372", + "right": "issue:45588", "accept": false, - "reason": "LayoutLMv2 tokenizer NER/padding crash vs Wav2Vec2 save/tokenization failure; unrelated." + "reason": "Gemma4 processor loading import failure is unrelated to the flash_attention.py s_aux=None crash." }, { - "left": "issue:33357", - "right": "issue:39692", + "left": "issue:43412", + "right": "issue:45588", "accept": false, - "reason": "MacOS bus error in CLIP vs SigLIP2 docs example errors; different platforms and symptoms." + "reason": "Gemma3n executorch export failures are unrelated to the flash_attention.py AttributeError." }, { - "left": "issue:43611", - "right": "issue:46121", + "left": "issue:45372", + "right": "issue:45406", "accept": false, - "reason": "Base_model_prefix loading regression vs rope-params JSON list TypeError; unrelated." + "reason": "Gemma4 processor import failure is not the same as transformers serve crashing because Gemma4Processor lacks _tokenizer." }, { - "left": "issue:29127", - "right": "issue:34567", + "left": "issue:44448", + "right": "issue:44466", "accept": false, - "reason": "LayoutLMv3 error messaging vs TrainerState token counter not updating; different features." + "reason": "Pegasus output changes and lm_head.weight serialization inconsistency are different v5 regressions." }, { - "left": "issue:44871", - "right": "issue:44977", + "left": "issue:45198", + "right": "issue:45803", "accept": false, - "reason": "Gemma-3 eos_token_id mismatch vs Qwen3.5 flash-attention generation; unrelated." + "reason": "Wav2Vec2 save/tokenization failure is unrelated to FuyuBatchFeature bare-except error handling." }, { - "left": "issue:42915", - "right": "issue:42994", + "left": "issue:25251", + "right": "issue:32090", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure vs quantized model saving failure; both quantization-related, but different stages and code paths." + "reason": "Pipeline top_k nesting behavior is unrelated to the Trainer NoneType broadcast error." }, { - "left": "issue:45459", - "right": "issue:45712", + "left": "issue:45245", + "right": "issue:45563", "accept": false, - "reason": "Tokenizer protobuf import error handling vs leftover dummy classes; unrelated." + "reason": "Category-count RuntimeError is unrelated to a stale warning in paged generate()." }, { - "left": "issue:43278", - "right": "issue:43575", + "left": "issue:32090", + "right": "issue:33290", "accept": false, - "reason": "BF16/evaluate dtype issue vs Qwen2-57B tensor-parallel OOM; different models and failure modes." + "reason": "Trainer broadcast NoneType error is unrelated to Adafactor OOM under deepspeed." } ] }, @@ -9352,7 +9321,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 30, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -9365,289 +9334,303 @@ "issue:29942", "issue:30064", "issue:33357", + "issue:34567", "issue:35141", + "issue:39692", "issue:42490", "issue:42915", + "issue:42994", + "issue:43065", "issue:43066", - "issue:43334", + "issue:43278", "issue:43381", "issue:43421", "issue:43526", - "issue:43531", "issue:43575", + "issue:43611", "issue:43653", "issue:43701", "issue:43756", - "issue:43761", "issue:43824", - "issue:43931", + "issue:43873", "issue:43994", "issue:44186", + "issue:44206", "issue:44265", "issue:44279", - "issue:44387", + "issue:44361", "issue:44442", "issue:44462", - "issue:44464", - "issue:44625", - "issue:44857", + "issue:44479", + "issue:44743", + "issue:44871", "issue:44977", - "issue:45072", "issue:45198", - "issue:45245", "issue:45372", "issue:45412", "issue:45440", "issue:45459", "issue:45491", - "issue:45584", "issue:45593", + "issue:45712", "issue:45834", - "issue:45859", "issue:46121", - "issue:46369", - "issue:46399" + "issue:46369" ], "soft_pairs": [ + "issue:43653|issue:44743", + "issue:42490|issue:43873", + "issue:43065|issue:45412", + "issue:44206|issue:44479", + "issue:33357|issue:39692", + "issue:43065|issue:43994", + "issue:44361|issue:45198", + "issue:29127|issue:34567", + "issue:44186|issue:45198", + "issue:43611|issue:46121", + "issue:42915|issue:42994", + "issue:44871|issue:44977", + "issue:45459|issue:45712", + "issue:43278|issue:43575", "issue:45412|issue:46369", - "issue:43066|issue:44462", - "issue:45372|issue:45459", "issue:29942|issue:30064", + "issue:45372|issue:45459", "issue:43994|issue:45412", "issue:43381|issue:43701", - "issue:44265|issue:45198", + "issue:43066|issue:44462", "issue:43756|issue:45440", + "issue:44265|issue:45198", "issue:43653|issue:44442", "issue:43421|issue:45491", "issue:27957|issue:29942", - "issue:45440|issue:45593", "issue:29127|issue:33357", + "issue:45440|issue:45593", "issue:43994|issue:46369", "issue:43526|issue:45834", "issue:43526|issue:44186", "issue:29127|issue:35141", - "issue:44625|issue:45584", - "issue:43824|issue:44279", - "issue:43575|issue:43701", - "issue:44442|issue:45072", - "issue:42915|issue:43931", - "issue:42490|issue:43381", - "issue:43575|issue:44387", - "issue:43761|issue:45859", - "issue:46121|issue:46399", - "issue:43526|issue:44857", - "issue:43334|issue:43531", - "issue:44977|issue:45245", - "issue:44464|issue:45491" + "issue:43824|issue:44279" ], - "prepared_review_unit_hash": "4390c578f2e4b089f50054e5e15f3ac720afb702086fe44f575499e20efe0ef5", + "prepared_review_unit_hash": "fcb533f13885723d28301aad01678cf1ed00841d23027992e7bb8dbfd357644a", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 30, - "serialized_chars": 36050, - "estimated_input_tokens": 9013, - "estimated_eval_tokens": 18282 + "soft_pair_count": 32, + "serialized_chars": 36636, + "estimated_input_tokens": 9159, + "estimated_eval_tokens": 18574 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated Transformers issues with a few superficial topic overlaps (tokenizers, Qwen, memory, Flash Attention). None of the soft pairs look like the same underlying bug/change closely enough to merge as duplicates.", - "confidence": 0.93, - "canonical_issue_reason": null, + "summary": "The set is highly heterogeneous, and none of the soft pairs look like the same underlying bug or change with enough specificity to merge them. The strongest anchor issue is the broad model-loading regression in transformers 5.0.0.", + "confidence": 0.78, + "canonical_issue_reason": "issue:43611 is the broadest and most central bug report in the set: a concrete Transformers 5.0.0 model-loading regression affecting the common `base_model_prefix` path, making it a reasonable dedupe anchor.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44462 is the clearest self-contained bug report: a concrete AutoTokenizer regression with a specific failure mode and likely broader impact.", + "best_issue_reason": "issue:43611 is the most generally useful issue to keep as the representative bug: it is specific, user-facing, and plausibly impacts many models through a shared loading code path.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45412", - "right": "issue:46369", + "left": "issue:43653", + "right": "issue:44743", "accept": false, - "reason": "Both mention memory problems, but one is RT-DETR GC/memory retention and the other is SAM3-LiteText overflow; different models and failure modes." + "reason": "Tokenizer special-token decode bug vs Qwen recurrent-state reset bug; unrelated code paths and symptoms." }, { - "left": "issue:43066", - "right": "issue:44462", + "left": "issue:42490", + "right": "issue:43873", "accept": false, - "reason": "Both concern tokenizers, but one is a v5 decoder-type mismatch and the other is AutoTokenizer ignoring tokenizer.json; distinct bugs." + "reason": "Path-with-slash import warning is unrelated to quantization/offloading behavior." }, { - "left": "issue:45372", - "right": "issue:45459", + "left": "issue:43065", + "right": "issue:45412", "accept": false, - "reason": "Both are loading/import errors, but one is a mistral_common import breakage for Gemma 4 and the other is protobuf exception handling hiding tokenizer errors." + "reason": "Dummy Conv2d placeholder in Sam3PixelDecoder is not the same as RT-DETR memory not being released." }, { - "left": "issue:29942", - "right": "issue:30064", + "left": "issue:44206", + "right": "issue:44479", "accept": false, - "reason": "Flash Attention 2 tests failing is unrelated to void segmentation map processing in an image processor." + "reason": "Different regressions in different components/models; not the same concrete failure." }, { - "left": "issue:43994", - "right": "issue:45412", + "left": "issue:33357", + "right": "issue:39692", "accept": false, - "reason": "Siglip2 producing wrong outputs is a model-inference bug, not the same as RT-DETR memory not being released." + "reason": "MacOS bus error with a community CLIP model is distinct from SigLIP2 documentation example errors." }, { - "left": "issue:43381", - "right": "issue:43701", + "left": "issue:43065", + "right": "issue:43994", "accept": false, - "reason": "Gradient checkpointing in eval mode and resume_from_checkpoint key mismatch are different training/control-flow issues." + "reason": "Placeholder layer implementation is unrelated to SigLIP2 producing nonsensical outputs." }, { - "left": "issue:44265", + "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "torch.export with torch_compilable_check and Wav2Vec2 save_pretrained/tokenization failures do not share the same code path." + "reason": "MLukeTokenizer AttributeError and Wav2Vec2 save/tokenization failure are different tokenizer/model bugs." }, { - "left": "issue:43756", - "right": "issue:45440", + "left": "issue:29127", + "right": "issue:34567", "accept": false, - "reason": "Smollm3 RoPE-layer loss and DeepseekV3MoE divergence are different architecture/implementation mismatches." + "reason": "Clarity of a LayoutLMv3 error message is not the same bug as TrainerState token counter not updating." }, { - "left": "issue:43653", - "right": "issue:44442", + "left": "issue:44186", + "right": "issue:45198", "accept": false, - "reason": "BigBirdTokenizer special-token registration and FastSpeech2ConformerTokenizer loading are both tokenizer issues, but not the same defect." + "reason": "LayoutLMv2 NER/padding crash is unrelated to Wav2Vec2 save_pretrained/tokenization failure." }, { - "left": "issue:43421", - "right": "issue:45491", + "left": "issue:43611", + "right": "issue:46121", "accept": false, - "reason": "Runtime special-token post-processor updates and Whisper empty-transcription generation after align_special_tokens are related only loosely." + "reason": "Base-model-prefix loading regression is a different problem from rope-parameter JSON list handling." }, { - "left": "issue:27957", - "right": "issue:29942", + "left": "issue:42915", + "right": "issue:42994", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support and Flash Attention 2 test failures are not the same concrete bug." + "reason": "Qwen3Moe FP8 failure and quantized model saving failure are different quantization issues." }, { - "left": "issue:45440", - "right": "issue:45593", + "left": "issue:44871", + "right": "issue:44977", "accept": false, - "reason": "DeepseekV3MoE implementation divergence and D-FINE auxiliary-loss behavior are unrelated model bugs." + "reason": "EOS token config mismatch is not the same as flash-attention generation regression." }, { - "left": "issue:29127", - "right": "issue:33357", + "left": "issue:45459", + "right": "issue:45712", "accept": false, - "reason": "LayoutLMv3 box error messaging and MacOS CLIP bus error are entirely different problems." + "reason": "protobuf-hidden tokenizer errors and leftover dummy classes are separate maintenance/runtime issues." }, { - "left": "issue:43994", + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Embedding dtype drift in eval is unrelated to tensor-parallel OOM on Qwen2-57B." + }, + { + "left": "issue:45412", "right": "issue:46369", "accept": false, - "reason": "Wrong/nonsensical Siglip2 outputs and SAM3-LiteText memory overflow are unrelated failures." + "reason": "Memory leak on deletion is not the same as model memory overflow during SAM3-liteText inference." }, { - "left": "issue:43526", - "right": "issue:45834", + "left": "issue:29942", + "right": "issue:30064", "accept": false, - "reason": "BeitImageProcessorFast label reduction and Kosmos2.5 long OCR input index errors do not match." + "reason": "Flash Attention 2 test failures and void segmentation-map processing are different feature areas." }, { - "left": "issue:43526", - "right": "issue:44186", + "left": "issue:45372", + "right": "issue:45459", "accept": false, - "reason": "BeitImageProcessorFast reduce_labels bug and LayoutLMv2Tokenizer NER padding/truncation crash are different components." + "reason": "ReasoningEffort import failure affects Gemma 4 processor loading; protobuf exception masking is a separate tokenizer error-handling bug." }, { - "left": "issue:29127", - "right": "issue:35141", + "left": "issue:43994", + "right": "issue:45412", "accept": false, - "reason": "LayoutLMv3 box validation messaging and embedding reinitialization after resize are unrelated." + "reason": "SigLIP2 inference quality bug is unrelated to RT-DETR object lifetime/memory release." }, { - "left": "issue:44625", - "right": "issue:45584", + "left": "issue:43381", + "right": "issue:43701", "accept": false, - "reason": "Qwen3.5 num_labels propagation and Whisper empty transcription after token alignment are different bugs." + "reason": "Eval-mode gradient checkpointing restriction is not the same as checkpoint key mismatch on resume." }, { - "left": "issue:43824", - "right": "issue:44279", + "left": "issue:43066", + "right": "issue:44462", "accept": false, - "reason": "Missing Qwen2.5-VL import and a generic dependency issue are not the same underlying problem." + "reason": "Wrong tokenizer decoder type in v5 and ignoring repository tokenizer.json are distinct tokenizer-loading issues." }, { - "left": "issue:43575", - "right": "issue:43701", + "left": "issue:43756", + "right": "issue:45440", "accept": false, - "reason": "Qwen2-57B-A14B-Instruct TP OOM and checkpoint key mismatch are unrelated." + "reason": "Smollm3 RoPE-layer count mismatch is unrelated to DeepSeekV3MoE divergence from remote implementation." }, { - "left": "issue:44442", - "right": "issue:45072", + "left": "issue:44265", + "right": "issue:45198", "accept": false, - "reason": "FastSpeech2ConformerTokenizer loading and bfloat16 dtype mismatches affect different subsystems and code paths." + "reason": "torch.export failure on torch_compilable_check is unrelated to Wav2Vec2 save/tokenization problems." }, { - "left": "issue:42915", - "right": "issue:43931", + "left": "issue:43653", + "right": "issue:44442", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure and Qwen3-VL weight-shape mismatch are separate model/config issues." + "reason": "BigBird mask-token decode issue and AutoTokenizer loading FastSpeech2ConformerTokenizer are different failures." }, { - "left": "issue:42490", - "right": "issue:43381", + "left": "issue:43421", + "right": "issue:45491", "accept": false, - "reason": "Import warning for local paths with trailing slash and gradient checkpointing eval-mode restrictions are not duplicates." + "reason": "Runtime special-token post-processor update bug is unrelated to Gemma3 NaN embeddings from sliding-window attention." }, { - "left": "issue:43575", - "right": "issue:44387", + "left": "issue:27957", + "right": "issue:29942", "accept": false, - "reason": "Both mention memory/OOM, but one is TP load OOM for a specific model and the other is int4 CUDA reserved-memory regression." + "reason": "XLMRoberta Flash Attention 2 support issue is not the same as generic Flash Attention 2 test failures." }, { - "left": "issue:43761", - "right": "issue:45859", + "left": "issue:29127", + "right": "issue:33357", "accept": false, - "reason": "CLIPVision hidden_states regression and Qwen3.5 CPU-offload rotary embedding incompatibility are different regressions." + "reason": "LayoutLMv3 error-message clarity and MacOS CLIP bus error are unrelated." }, { - "left": "issue:46121", - "right": "issue:46399", + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "Remote-vs-native DeepSeekV3MoE divergence is unrelated to D-FINE auxiliary-loss behavior." + }, + { + "left": "issue:43994", + "right": "issue:46369", "accept": false, - "reason": "JSON-loaded list handling in rope validation and missing apply_rotary_pos_emb decorator are unrelated rope-related bugs." + "reason": "SigLIP2 incorrect outputs and SAM3-liteText memory overflow are distinct model bugs." }, { "left": "issue:43526", - "right": "issue:44857", + "right": "issue:45834", "accept": false, - "reason": "BeitImageProcessorFast label reduction and LwDetrImageLoss AMP crash are different image-processing/training failures." + "reason": "BeitImageProcessorFast reduce_labels bug and Kosmos2.5 OCR index error are not the same underlying issue." }, { - "left": "issue:43334", - "right": "issue:43531", + "left": "issue:43526", + "right": "issue:44186", "accept": false, - "reason": "Qwen3-VL pad_token_id loading failure and Qwen3-MoE sliding_window issue are separate configuration/runtime problems." + "reason": "reduce_labels returning one label and LayoutLMv2Tokenizer crashes are different processor/tokenizer defects." }, { - "left": "issue:44977", - "right": "issue:45245", + "left": "issue:29127", + "right": "issue:35141", "accept": false, - "reason": "Qwen3.5 flash-attention generation bug and category-count overflow are unrelated." + "reason": "LayoutLMv3 error clarity and output embedding reinitialization after resize are unrelated bugs." }, { - "left": "issue:44464", - "right": "issue:45491", + "left": "issue:43824", + "right": "issue:44279", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency and Gemma3 NaN embeddings with sliding-window attention are different failure modes." + "reason": "Specific Qwen2.5-VL import error and a generic dependency issue are not the same concrete failure." } ] }, @@ -9663,7 +9646,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 31, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -9677,282 +9660,274 @@ "issue:34567", "issue:34689", "issue:35141", - "issue:42175", "issue:42490", - "issue:42796", - "issue:42907", - "issue:43232", + "issue:42915", "issue:43262", "issue:43316", + "issue:43334", "issue:43381", - "issue:43388", - "issue:43493", "issue:43525", - "issue:43611", + "issue:43526", + "issue:43531", + "issue:43575", "issue:43646", - "issue:43720", + "issue:43701", "issue:43723", "issue:43761", - "issue:43827", "issue:43873", "issue:43931", "issue:44062", "issue:44077", - "issue:44079", - "issue:44188", "issue:44206", + "issue:44387", "issue:44442", - "issue:44479", + "issue:44464", + "issue:44625", "issue:44743", "issue:44811", + "issue:44857", "issue:44898", - "issue:44987", + "issue:44977", "issue:45003", - "issue:45216", - "issue:45290", + "issue:45072", + "issue:45245", "issue:45440", "issue:45479", + "issue:45491", + "issue:45584", "issue:45588", "issue:45600", - "issue:45698", - "issue:45803", + "issue:45859", "issue:45907", - "issue:46178", - "issue:46234" + "issue:46121", + "issue:46399" ], "soft_pairs": [ + "issue:44625|issue:45584", + "issue:43575|issue:43701", + "issue:44442|issue:45072", + "issue:42915|issue:43931", + "issue:42490|issue:43381", + "issue:43575|issue:44387", + "issue:43761|issue:45859", + "issue:46121|issue:46399", + "issue:43526|issue:44857", + "issue:43334|issue:43531", + "issue:44977|issue:45245", "issue:43262|issue:44206", + "issue:44464|issue:45491", "issue:27957|issue:29127", "issue:27957|issue:33290", - "issue:44743|issue:44811", "issue:45588|issue:45600", + "issue:44743|issue:44811", "issue:43646|issue:44077", "issue:33290|issue:34689", "issue:33290|issue:35141", "issue:33290|issue:34567", - "issue:44062|issue:45907", "issue:43381|issue:43873", "issue:44442|issue:44898", + "issue:44062|issue:45907", "issue:45003|issue:45440", "issue:43723|issue:45479", - "issue:43761|issue:44079", - "issue:42175|issue:46234", "issue:42490|issue:44743", - "issue:43316|issue:43525", - "issue:34689|issue:43611", - "issue:43720|issue:44442", - "issue:43232|issue:43388", - "issue:43931|issue:45216", - "issue:43761|issue:44188", - "issue:44077|issue:44479", - "issue:43493|issue:43827", - "issue:42907|issue:45698", - "issue:44442|issue:45803", - "issue:42796|issue:45290", - "issue:44987|issue:46178" + "issue:43316|issue:43525" ], - "prepared_review_unit_hash": "f72f3311db0082962a8e7c74cfa28deacfa61a42a3b1f60ea8a857ce4a307b3e", + "prepared_review_unit_hash": "490b759b4e8b230283067065914aed512df836ddcc070ce079e70feef869a7f8", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36533, - "estimated_input_tokens": 9134, - "estimated_eval_tokens": 18524 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35465, + "estimated_input_tokens": 8867, + "estimated_eval_tokens": 17990 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are mostly unrelated single-bug reports across different models and subsystems, so none of the soft candidate pairs looks like a true duplicate. The broadest representative issue here is the save/load regression in issue:45698.", - "confidence": 0.79, - "canonical_issue_reason": "issue:45698 is the broadest, most reusable representative in this set: a high-discussion, cross-cutting from_pretrained/save_pretrained bug with inbound references.", + "summary": "This set does not look like a single duplicate cluster; it is a grab bag of unrelated issues across models, tokenizer/config loading, training behavior, audio, and quantization. I would not merge any of the suggested soft pairs.", + "confidence": 0.21, + "canonical_issue_reason": "issue:43873 is the most discussion-heavy and externally referenced item in the set, so it is the best available anchor if one must pick a representative issue. That said, the overall group is too heterogeneous to have a strong canonical duplicate.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45698 best fits as the cluster representative because it is broad, actionable, and already shows strong user impact/engagement.", + "best_issue_reason": "issue:43873 is the strongest single issue candidate by activity and inbound references, but it is not truly representative of the whole set because the set is not a coherent duplicate cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43262", - "right": "issue:44206", - "accept": false, - "reason": "Different audio-related failures: chat-template sampling-rate defaulting vs an unsupported `center` argument in `LasrFeatureExtractor`." - }, - { - "left": "issue:27957", - "right": "issue:29127", + "left": "issue:44625", + "right": "issue:45584", "accept": false, - "reason": "Unrelated models and bugs: Flash Attention 2 support for XLM-Roberta vs a LayoutLMv3 error-message clarity issue." + "reason": "Different models and failures: Qwen3.5 config propagation vs Whisper empty-transcription generation." }, { - "left": "issue:27957", - "right": "issue:33290", + "left": "issue:43575", + "right": "issue:43701", "accept": false, - "reason": "Different code paths: Flash Attention support vs DeepSpeed/Adafactor OOM." + "reason": "OOM on tensor-parallel model load is unrelated to resume_from_checkpoint key mismatch." }, { - "left": "issue:44743", - "right": "issue:44811", + "left": "issue:44442", + "right": "issue:45072", "accept": false, - "reason": "Different generation/audio issues: Qwen recurrent-state reset vs Whisper `batch_decode()` skipping special tokens." + "reason": "Tokenizer loading bug vs dtype mismatch in bf16 inference; different code paths." }, { - "left": "issue:45588", - "right": "issue:45600", + "left": "issue:42915", + "right": "issue:43931", "accept": false, - "reason": "One is a flash-attention crash on `s_aux=None`; the other is an auto-mapping reference to removed configs." + "reason": "Qwen3Moe FP8 config failure and Qwen3-VL weight-shape mismatch are different model/config issues." }, { - "left": "issue:43646", - "right": "issue:44077", + "left": "issue:42490", + "right": "issue:43381", "accept": false, - "reason": "Both involve model initialization, but the concrete failures differ: custom model init regression vs `patchtsmixer` allowing optional `post_init`." + "reason": "Local-path import warning is unrelated to gradient checkpointing in eval mode." }, { - "left": "issue:33290", - "right": "issue:34689", + "left": "issue:43575", + "right": "issue:44387", "accept": false, - "reason": "OOM in optimizer/deepspeed is unrelated to a model-loading regression for Llama 3.2 Vision." + "reason": "Both mention memory, but one is TP load OOM and the other is int4 reserved-memory regression; not the same bug." }, { - "left": "issue:33290", - "right": "issue:35141", + "left": "issue:43761", + "right": "issue:45859", "accept": false, - "reason": "Different problems: optimizer memory blowup vs output embedding reinitialization after resizing token embeddings." + "reason": "CLIPVisionModel hidden-states regression and Qwen3.5Moe CPU-offload incompatibility are unrelated." }, { - "left": "issue:33290", - "right": "issue:34567", + "left": "issue:46121", + "right": "issue:46399", "accept": false, - "reason": "Different subsystems: DeepSpeed/Adafactor memory issue vs TrainerState token counting not updating." + "reason": "Rope-validation TypeError and missing decorator for apply_rotary_pos_emb are different implementation problems." }, { - "left": "issue:44062", - "right": "issue:45907", + "left": "issue:43526", + "right": "issue:44857", "accept": false, - "reason": "Tokenizer `AddedToken(..., special=...)` argument duplication is unrelated to `list_repo_files` not forwarding `revision`." + "reason": "BeitImageProcessorFast label reduction bug is unrelated to LwDetrImageLoss AMP/CUDA crash." }, { - "left": "issue:43381", - "right": "issue:43873", + "left": "issue:43334", + "right": "issue:43531", "accept": false, - "reason": "Gradient checkpointing in eval mode and quantization offloading are separate behaviors with different code paths." + "reason": "Missing pad_token_id on Qwen3-VL text config is different from sliding-window behavior in Qwen3-MoE." }, { - "left": "issue:44442", - "right": "issue:44898", + "left": "issue:44977", + "right": "issue:45245", "accept": false, - "reason": "AutoTokenizer loading a speech tokenizer is unrelated to Perceiver image-classification resizing failures." + "reason": "Flash-attention generation regression and category-cardinality RuntimeError are unrelated." }, { - "left": "issue:45003", - "right": "issue:45440", + "left": "issue:43262", + "right": "issue:44206", "accept": false, - "reason": "Unsafe `sys.modules` access in modeling utils is unrelated to DeepSeekV3MoE implementation divergence." + "reason": "Audio chat-template sampling-rate default and unsupported center argument in feature extraction are different bugs." }, { - "left": "issue:43723", - "right": "issue:45479", + "left": "issue:44464", + "right": "issue:45491", "accept": false, - "reason": "Tokenizer loading in v5 is unrelated to the zero-loss classification bug when `num_labels=1`." + "reason": "Compiled-forward chunked generation inconsistency is not the same as Gemma3 NaN embeddings from sliding-window padding." }, { - "left": "issue:43761", - "right": "issue:44079", + "left": "issue:27957", + "right": "issue:29127", "accept": false, - "reason": "CLIP hidden-states regression and `ModelOutput` key assignment are different bugs." + "reason": "XLMRoberta flash-attention support and LayoutLMv3 box-validation messaging are unrelated." }, { - "left": "issue:42175", - "right": "issue:46234", + "left": "issue:27957", + "right": "issue:33290", "accept": false, - "reason": "Backend packaging for TensorFlow in extras is unrelated to a missing `is_falcon_mamba_ssm_available` import utility." + "reason": "Flash Attention 2 support and deepspeed Adafactor OOM are different problems." }, { - "left": "issue:42490", - "right": "issue:44743", + "left": "issue:45588", + "right": "issue:45600", "accept": false, - "reason": "Local-path import warning and Qwen recurrent-state reset are unrelated." + "reason": "Flash-attention sinkless-model crash and removed-config CI breakage do not share the same code path." }, { - "left": "issue:43316", - "right": "issue:43525", + "left": "issue:44743", + "right": "issue:44811", "accept": false, - "reason": "Both are config/API mismatches, but they affect different models and different missing attributes, so not the same bug." + "reason": "Qwen3.5 recurrent-state reset and Whisper batch_decode skip_special_tokens are unrelated." }, { - "left": "issue:34689", - "right": "issue:43611", + "left": "issue:43646", + "right": "issue:44077", "accept": false, - "reason": "Both are loading regressions, but they stem from different compatibility breaks and different model classes." + "reason": "General Transformers 5.0 custom initialization breakage is broader than the PatchTSMixer post_init-specific change." }, { - "left": "issue:43720", - "right": "issue:44442", + "left": "issue:33290", + "right": "issue:34689", "accept": false, - "reason": "BitNet packed-weight loading is unrelated to a tokenizer loading failure." + "reason": "Adafactor/deepspeed OOM and Llama 3.2 Vision model-loading regression are unrelated." }, { - "left": "issue:43232", - "right": "issue:43388", + "left": "issue:33290", + "right": "issue:35141", "accept": false, - "reason": "Generation kwargs syncing and metrics-label gathering are different distributed-training issues." + "reason": "OOM during optimizer use is unrelated to output embedding reinitialization after resize." }, { - "left": "issue:43931", - "right": "issue:45216", + "left": "issue:33290", + "right": "issue:34567", "accept": false, - "reason": "Weight-shape mismatch on loading a model is unrelated to an incorrect checkpoint saved by `save_pretrained`." + "reason": "Memory exhaustion and TrainerState token-count tracking are different issues." }, { - "left": "issue:43761", - "right": "issue:44188", + "left": "issue:43381", + "right": "issue:43873", "accept": false, - "reason": "A CLIP forward-output regression is unrelated to attention-kernel divergence under `torch.compile`." + "reason": "Eval-mode gradient checkpointing restriction is unrelated to quantization offloading behavior." }, { - "left": "issue:44077", - "right": "issue:44479", + "left": "issue:44442", + "right": "issue:44898", "accept": false, - "reason": "`post_init` policy in one model is unrelated to the video-input regression in Qwen variants." + "reason": "Tokenizer loading failure and Perceiver image-resolution failure are different subsystems." }, { - "left": "issue:43493", - "right": "issue:43827", + "left": "issue:44062", + "right": "issue:45907", "accept": false, - "reason": "SigLIP2 implementation drift is unrelated to stale docs referencing removed pipelines." + "reason": "AddedToken special-argument duplication is unrelated to list_repo_files revision forwarding." }, { - "left": "issue:42907", - "right": "issue:45698", + "left": "issue:45003", + "right": "issue:45440", "accept": false, - "reason": "Both touch save/load behavior, but one is dequantized model saving and the other is wrong module resolution on load; different underlying bugs." + "reason": "Unsafe sys.modules access and DeepseekV3MoE divergence are unrelated." }, { - "left": "issue:44442", - "right": "issue:45803", + "left": "issue:43723", + "right": "issue:45479", "accept": false, - "reason": "Tokenizer loading and a bare-except swallowing errors in Fuyu tensor conversion are unrelated." + "reason": "Tokenizer loading in v5 and zero-loss single-label classification are different bugs." }, { - "left": "issue:42796", - "right": "issue:45290", + "left": "issue:42490", + "right": "issue:44743", "accept": false, - "reason": "Ministral BPE-marker handling and chat-template tool-call crashes are different tokenizer/chat-template bugs." + "reason": "Import-path trailing-slash warning is unrelated to Qwen3.5 recurrent-state reset." }, { - "left": "issue:44987", - "right": "issue:46178", + "left": "issue:43316", + "right": "issue:43525", "accept": false, - "reason": "A v5 model-loading regression is unrelated to the missing `tensorflow_text` backend mapping error." + "reason": "Gemma3TextConfig API discrepancy and Llama4Config missing pad_token_id are similar in theme but not the same underlying bug." } ] }, @@ -9968,7 +9943,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 32, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -9976,56 +9951,69 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:16998", - "issue:22355", "issue:25251", - "issue:27957", "issue:28282", - "issue:32090", + "issue:34689", "issue:42175", "issue:42617", + "issue:42796", + "issue:42907", + "issue:43232", "issue:43295", "issue:43316", - "issue:43540", + "issue:43388", + "issue:43493", "issue:43550", "issue:43575", "issue:43606", + "issue:43611", "issue:43643", - "issue:43650", "issue:43701", - "issue:43716", - "issue:43742", + "issue:43720", "issue:43749", + "issue:43761", "issue:43825", + "issue:43827", "issue:43873", "issue:43931", "issue:44077", - "issue:44162", + "issue:44079", + "issue:44188", "issue:44190", + "issue:44442", "issue:44464", + "issue:44479", "issue:44496", - "issue:44514", - "issue:44534", - "issue:44610", - "issue:44792", "issue:44855", "issue:44871", "issue:44913", + "issue:44987", "issue:45137", - "issue:45584", - "issue:45632", + "issue:45216", + "issue:45290", + "issue:45698", "issue:45759", "issue:45803", - "issue:46202", + "issue:46178", "issue:46234", "issue:46291", - "issue:46365", - "issue:46369", "issue:46421", - "issue:46424", "issue:46506" ], "soft_pairs": [ + "issue:42175|issue:46234", + "issue:43761|issue:44079", + "issue:34689|issue:43611", + "issue:43720|issue:44442", + "issue:43232|issue:43388", + "issue:43931|issue:45216", + "issue:43761|issue:44188", + "issue:44077|issue:44479", + "issue:43493|issue:43827", + "issue:42907|issue:45698", + "issue:44442|issue:45803", + "issue:44987|issue:46178", + "issue:42796|issue:45290", "issue:43873|issue:44871", "issue:43749|issue:43931", "issue:44464|issue:46421", @@ -10039,218 +10027,198 @@ "issue:25251|issue:28282", "issue:43550|issue:43701", "issue:42617|issue:43575", - "issue:43606|issue:43701", - "issue:43575|issue:43742", - "issue:45632|issue:46424", - "issue:16998|issue:22355", - "issue:27957|issue:32090", - "issue:44610|issue:46369", - "issue:43316|issue:43716", - "issue:43650|issue:46202", - "issue:43295|issue:44871", - "issue:42175|issue:43316", - "issue:45803|issue:46365", - "issue:43540|issue:44514", - "issue:44162|issue:44792", - "issue:44871|issue:45584", - "issue:44077|issue:44534" + "issue:43606|issue:43701" ], - "prepared_review_unit_hash": "5d1d827b1a8b499660065f703959ebaffd806d19b9ca57c012e4369086fe39a0", + "prepared_review_unit_hash": "ca94a51e7dd40ee98b32550a795c6220dafe650bc65974709b8473b1087df836", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 35795, - "estimated_input_tokens": 8949, - "estimated_eval_tokens": 18154 + "soft_pair_count": 27, + "serialized_chars": 35961, + "estimated_input_tokens": 8991, + "estimated_eval_tokens": 18238 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are mostly unrelated Transformers bugs spanning configs, loading, generation, distributed training, and multimodal processors. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates. The best representative issue is the detailed, open quantization/offloading report.", - "confidence": 0.93, - "canonical_issue_reason": "Issue #43873 is the most broadly relevant and actionable report in this set: it is open, detailed, and centered on a concrete offloading/quantization failure with clear repro context.", + "summary": "This cluster is heterogeneous: the issues span unrelated loading, generation, config, tokenizer, backend, and distributed-training bugs. None of the soft pairs look like the same underlying defect, so I would not merge any of them as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "#43873 is the strongest standalone issue to represent the cluster because it has substantial discussion, clear impact, and a specific reproducible code-path problem.", + "best_issue_reason": "No true canonical issue for the whole cluster; if forced, issue:43749 is the strongest standalone report because it is concrete, high-signal, and has the most discussion around a specific regression ('FSDP_CPU_RAM_EFFICIENT_LOADING broken').", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43873", - "right": "issue:44871", + "left": "issue:42175", + "right": "issue:46234", "accept": false, - "reason": "Different problems: quantized offloading behavior vs Gemma eos_token_id config inconsistency." + "reason": "Different dependency/back-end problems: Torch backend packaging vs a missing import utility helper." }, { - "left": "issue:43749", - "right": "issue:43931", + "left": "issue:43761", + "right": "issue:44079", "accept": false, - "reason": "Both are loading-related, but one is FSDP CPU RAM efficient loading and the other is a Qwen3-VL weight-shape mismatch." + "reason": "One is a hidden_states regression in forward; the other is ModelOutput key assignment behavior." }, { - "left": "issue:44464", - "right": "issue:46421", + "left": "issue:34689", + "right": "issue:43611", "accept": false, - "reason": "Chunked generation output inconsistency and chunked-prefill recompilation are related settings but not the same bug path." + "reason": "Both concern loading regressions, but they affect different models and different failure modes." }, { - "left": "issue:43316", - "right": "issue:45759", - "accept": false, - "reason": "Both concern config API behavior, but one is a Gemma3TextConfig discrepancy and the other is unwrapping text_config for composite Qwen configs." - }, - { - "left": "issue:43701", - "right": "issue:46506", + "left": "issue:43720", + "right": "issue:44442", "accept": false, - "reason": "Checkpoint resume key mismatch is unrelated to CPU/CUDA result divergence in PPDocLayoutV3." + "reason": "Packed-weight loading under accelerate is unrelated to AutoTokenizer failing for a specific tokenizer class." }, { - "left": "issue:44190", - "right": "issue:44496", + "left": "issue:43232", + "right": "issue:43388", "accept": false, - "reason": "Local dataset loading failure in a script is unrelated to an unrecognized model/config.json error." + "reason": "Generation kwargs syncing and last-batch metric gathering are different code paths and bugs." }, { - "left": "issue:43643", - "right": "issue:44913", + "left": "issue:43931", + "right": "issue:45216", "accept": false, - "reason": "Both involve config persistence, but trust_remote_code field loss is a different issue from GPTNeoXConfig rotary_pct reloading correctly." + "reason": "Shape-mismatch loading and save_pretrained checkpoint corruption are opposite directions and distinct bugs." }, { - "left": "issue:44855", - "right": "issue:45137", + "left": "issue:43761", + "right": "issue:44188", "accept": false, - "reason": "Python 3.13 torch.jit parsing error and DeepSpeed ZeRO3 deque underflow are separate failures." + "reason": "Hidden_states output regression is unrelated to attention-kernel divergence under torch.compile." }, { - "left": "issue:46234", - "right": "issue:46291", + "left": "issue:44077", + "right": "issue:44479", "accept": false, - "reason": "Missing import_utils symbol is unrelated to kernels v0.15 breaking transformers.pipelines import." + "reason": "A post_init API constraint is unrelated to a video-input regression in Qwen VL models." }, { - "left": "issue:43295", - "right": "issue:43825", + "left": "issue:43493", + "right": "issue:43827", "accept": false, - "reason": "Processor/tokenizer regression in v4.57.5 is not the same as a pipeline v5 translation-task error message issue." + "reason": "Model implementation discrepancy vs documentation still mentioning pipeline(); not the same bug." }, { - "left": "issue:25251", - "right": "issue:28282", + "left": "issue:42907", + "right": "issue:45698", "accept": false, - "reason": "top_k output nesting in pipelines is unrelated to an ImportError about missing PyTorch." + "reason": "Saving dequantized models and loading the wrong custom module after save_pretrained are different save/load failures." }, { - "left": "issue:43550", - "right": "issue:43701", + "left": "issue:44442", + "right": "issue:45803", "accept": false, - "reason": "torch.compile/SDPA failure for Bamba is unrelated to checkpoint resume key mismatch." + "reason": "Tokenizer class loading and tensor conversion exception handling are unrelated." }, { - "left": "issue:42617", - "right": "issue:43575", + "left": "issue:44987", + "right": "issue:46178", "accept": false, - "reason": "3d_parallel.py startup failure and Qwen2 tensor-parallel OOM are different parallelism problems." + "reason": "Generic model-loading failure is not the same as a backend-mapping error for tensorflow_text." }, { - "left": "issue:43606", - "right": "issue:43701", + "left": "issue:42796", + "right": "issue:45290", "accept": false, - "reason": "CPU offload device mismatch in bark-small is unrelated to resume_from_checkpoint key handling." + "reason": "Tokenizer BPE marker handling is unrelated to chat-template crashes with tool calls." }, { - "left": "issue:43575", - "right": "issue:43742", + "left": "issue:43873", + "right": "issue:44871", "accept": false, - "reason": "Tensor-parallel OOM and MobileLLM key error are distinct model-loading failures." + "reason": "Quantization/offloading behavior and eos_token_id config inconsistency are different bugs." }, { - "left": "issue:45632", - "right": "issue:46424", + "left": "issue:43749", + "right": "issue:43931", "accept": false, - "reason": "trust_remote_code cache path collisions are unrelated to cache max length being honored only for quantized cache." + "reason": "FSDP RAM-efficient loading and Qwen3-VL shape mismatches are unrelated loading problems." }, { - "left": "issue:16998", - "right": "issue:22355", + "left": "issue:44464", + "right": "issue:46421", "accept": false, - "reason": "Model max length behavior for DeBERTa-V3 has nothing to do with missing transformers.onnx." + "reason": "Both mention chunked/compiled generation, but one is inconsistent outputs and the other is recompilation from cache init; not the same concrete defect." }, { - "left": "issue:27957", - "right": "issue:32090", + "left": "issue:43316", + "right": "issue:45759", "accept": false, - "reason": "Flash Attention 2 support for XLMRoberta is unrelated to a Trainer NoneType broadcast error." + "reason": "Both are config/API issues, but they concern different model families and different failures." }, { - "left": "issue:44610", - "right": "issue:46369", + "left": "issue:43701", + "right": "issue:46506", "accept": false, - "reason": "Processor image-size mismatch and SAM memory overflow are different model/runtime issues." + "reason": "Checkpoint resume key mismatch is unrelated to CPU/CUDA output divergence." }, { - "left": "issue:43316", - "right": "issue:43716", + "left": "issue:44190", + "right": "issue:44496", "accept": false, - "reason": "Composite config API discrepancy and image-preprocessor dtype mismatch are not the same bug." + "reason": "Local dataset loading in an example script is unrelated to an unrecognized model/config error." }, { - "left": "issue:43650", - "right": "issue:46202", + "left": "issue:43643", + "right": "issue:44913", "accept": false, - "reason": "Placeholder issue title and Gemma pipeline roundtrip failure are unrelated." + "reason": "Both involve config persistence, but one loses fields with trust_remote_code while the other resets rotary_pct on reload." }, { - "left": "issue:43295", - "right": "issue:44871", + "left": "issue:44855", + "right": "issue:45137", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to Gemma eos_token_id list-vs-scalar inconsistency." + "reason": "Python 3.13 import parsing error and DeepSpeed ZeRO3 deque underflow are unrelated." }, { - "left": "issue:42175", - "right": "issue:43316", + "left": "issue:46234", + "right": "issue:46291", "accept": false, - "reason": "Packaging backend selection is unrelated to a config API discrepancy." + "reason": "Missing import utility function and kernels package version/revision requirement are different import failures." }, { - "left": "issue:45803", - "right": "issue:46365", + "left": "issue:43295", + "right": "issue:43825", "accept": false, - "reason": "Bare except swallowing errors in Fuyu processing is unrelated to importing Trainer failing because of datasets." + "reason": "Processor/tokenizer regression and a v5 translation-task error message are not the same bug." }, { - "left": "issue:43540", - "right": "issue:44514", + "left": "issue:25251", + "right": "issue:28282", "accept": false, - "reason": "Video-processing ValueError in Qwen3OmniMoe is unrelated to batched apply_chat_template crashing with padding=False." + "reason": "Pipeline top_k nesting behavior is unrelated to an ImportError for missing PyTorch." }, { - "left": "issue:44162", - "right": "issue:44792", + "left": "issue:43550", + "right": "issue:43701", "accept": false, - "reason": "ESM2 being broken and a janus generate-images test failure are different model/test issues." + "reason": "torch.compile SDPA failure and resume_from_checkpoint key mismatch are different runtime paths." }, { - "left": "issue:44871", - "right": "issue:45584", + "left": "issue:42617", + "right": "issue:43575", "accept": false, - "reason": "Gemma eos_token_id inconsistency is unrelated to Whisper empty transcription generation after align_special_tokens." + "reason": "A 3d_parallel script failure and a tensor-parallel OOM are both distributed issues but not the same concrete bug." }, { - "left": "issue:44077", - "right": "issue:44534", + "left": "issue:43606", + "right": "issue:43701", "accept": false, - "reason": "Optional post_init policy for patchtsmixer is unrelated to non-persistent buffers being filled with junk in v5." + "reason": "CPU offload device mismatch is unrelated to checkpoint key mismatch during resume." } ] }, @@ -10266,7 +10234,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 33, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -10274,281 +10242,267 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:35141", + "issue:16998", + "issue:22355", + "issue:27957", + "issue:32090", "issue:41950", - "issue:42490", + "issue:42175", "issue:43065", + "issue:43295", + "issue:43316", "issue:43388", "issue:43404", "issue:43441", "issue:43452", "issue:43475", - "issue:43504", - "issue:43525", - "issue:43531", + "issue:43540", "issue:43572", + "issue:43575", "issue:43646", + "issue:43650", "issue:43673", "issue:43688", + "issue:43716", + "issue:43742", "issue:43761", - "issue:43931", - "issue:44060", - "issue:44190", - "issue:44242", + "issue:44077", + "issue:44162", "issue:44479", "issue:44496", + "issue:44514", "issue:44534", + "issue:44610", "issue:44617", - "issue:44734", "issue:44743", "issue:44792", "issue:44871", "issue:44913", - "issue:44964", - "issue:44977", "issue:45127", "issue:45200", - "issue:45381", - "issue:45406", - "issue:45542", - "issue:45563", "issue:45584", - "issue:45600", - "issue:45698", + "issue:45632", "issue:45803", "issue:46050", "issue:46170", "issue:46178", - "issue:46234", + "issue:46202", + "issue:46365", "issue:46369", - "issue:46421" + "issue:46424" ], "soft_pairs": [ + "issue:43575|issue:43742", + "issue:27957|issue:32090", + "issue:16998|issue:22355", + "issue:43316|issue:43716", + "issue:45632|issue:46424", + "issue:44610|issue:46369", + "issue:43295|issue:44871", + "issue:42175|issue:43316", + "issue:43650|issue:46202", + "issue:45803|issue:46365", + "issue:44871|issue:45584", + "issue:44077|issue:44534", "issue:43065|issue:44617", + "issue:43540|issue:44514", + "issue:44162|issue:44792", "issue:45803|issue:46170", "issue:44871|issue:46050", "issue:41950|issue:43475", - "issue:43404|issue:43441", "issue:43646|issue:44913", + "issue:43404|issue:43441", "issue:43452|issue:44496", "issue:43761|issue:44479", "issue:45127|issue:45200", "issue:43388|issue:43688", "issue:43673|issue:44743", - "issue:43572|issue:46178", - "issue:43504|issue:45698", - "issue:45542|issue:46234", - "issue:44734|issue:45406", - "issue:44617|issue:45584", - "issue:43673|issue:46421", - "issue:43525|issue:43531", - "issue:44792|issue:44977", - "issue:45381|issue:46369", - "issue:43931|issue:44060", - "issue:43761|issue:44242", - "issue:44792|issue:45600", - "issue:45563|issue:45584", - "issue:44190|issue:44964", - "issue:44534|issue:44913", - "issue:35141|issue:42490", - "issue:44479|issue:44734" + "issue:43572|issue:46178" ], - "prepared_review_unit_hash": "f1886b81d3bd8ecdd77ce9d42309f2a1ab9e448508b5aec66b6845118bd4b840", + "prepared_review_unit_hash": "60bb6fa6239a3b6ba62891230466edcaa3f8307351f31176672644ceb90b7c8c", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 36288, - "estimated_input_tokens": 9072, - "estimated_eval_tokens": 18400 + "soft_pair_count": 26, + "serialized_chars": 35326, + "estimated_input_tokens": 8832, + "estimated_eval_tokens": 17920 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a duplicate set; it contains many unrelated Transformers issues across model loading, configs, generation, vision/video, SAM3, and backend/import paths. None of the soft pairs look like the same underlying bug or change.", + "summary": "This cluster is not a duplicate set: it mixes unrelated bugs across model configs, processors, training, packaging/imports, and backend-specific regressions. None of the soft pairs look like the same underlying defect or change.", "confidence": 0.96, - "canonical_issue_reason": null, + "canonical_issue_reason": "No single issue is a true canonical for this cluster because the items are heterogeneous. If a placeholder is needed, issue 45584 is the most active/open entry with the most inbound references, but it is not a thematic match for the rest.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a good canonical representative because the cluster is heterogeneous and the soft pairs are mostly false positives across different subsystems and models.", + "best_issue_reason": "Issue 45584 is the best placeholder only by engagement and recency; it is not a strong representative of the overall set because the cluster lacks a shared bug theme.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43065", - "right": "issue:44617", + "left": "issue:43575", + "right": "issue:43742", "accept": false, - "reason": "Both are SAM3-related, but one is about a dummy Conv2d in Sam3PixelDecoder while the other is a CUDA OOM in Sam3Video; different failure modes and code paths." + "reason": "Different failures: tensor-parallel OOM on Qwen2-57B vs a KeyError while loading MobileLLM." }, { - "left": "issue:45803", - "right": "issue:46170", + "left": "issue:27957", + "right": "issue:32090", "accept": false, - "reason": "Unrelated bugs: one is a broad bare-except in FuyuBatchFeature tensor conversion, the other is a Windows ROCm import/dependency issue involving torch.distributed." + "reason": "Unrelated problems: FlashAttention support for XLM-Roberta vs Trainer distributed broadcast NoneType error." }, { - "left": "issue:44871", - "right": "issue:46050", + "left": "issue:16998", + "right": "issue:22355", "accept": false, - "reason": "Different problems: Gemma eos_token_id config inconsistency versus quantized model save/load breakage after sinq." + "reason": "A model max-length question is unrelated to a missing transformers.onnx import." }, { - "left": "issue:41950", - "right": "issue:43475", + "left": "issue:43316", + "right": "issue:43716", "accept": false, - "reason": "Video-classification processor lookup and Sam3VisionEncoderOutput missing attribute are separate bugs in different pipelines/components." + "reason": "Different models and bugs: Gemma3 config API mismatch vs Mistral-3 dtype mismatch." }, { - "left": "issue:43404", - "right": "issue:43441", + "left": "issue:45632", + "right": "issue:46424", "accept": false, - "reason": "Mistral3 tied lm_head weights and Ministral-3 FlashAttention failure are distinct model-specific issues with different root causes." + "reason": "Cache-path collision for trust_remote_code is unrelated to max_cache_length behavior in quantized cache." }, { - "left": "issue:43646", - "right": "issue:44913", + "left": "issue:44610", + "right": "issue:46369", "accept": false, - "reason": "General custom model initialization regression is not the same as GPTNeoX rotary_pct resetting on reload; one is broad, the other is a specific config field persistence bug." + "reason": "Processor image-size bug vs Sam3 LiteText memory overflow; different code paths and symptoms." }, { - "left": "issue:43452", - "right": "issue:44496", + "left": "issue:43295", + "right": "issue:44871", "accept": false, - "reason": "GGUF loading failures for tokenizer/model APIs are different from an unrecognized-model config missing model_type." + "reason": "Processor/tokenizer regression is not the same as Gemma eos_token_id inconsistency." }, { - "left": "issue:43761", - "right": "issue:44479", + "left": "issue:42175", + "right": "issue:43316", "accept": false, - "reason": "CLIPVisionModel hidden_states regression is unrelated to the Qwen2.5/3/3.5 video input regression." + "reason": "Packaging/backend dependency issue vs model config API discrepancy; not the same bug." }, { - "left": "issue:45127", - "right": "issue:45200", + "left": "issue:43650", + "right": "issue:46202", "accept": false, - "reason": "LoRA merge with extended vocabulary on tied embeddings is a different change path from Gemma 4 mm_token_type_ids defaulting behavior." + "reason": "Placeholder issue title is unrelated to Gemma pipeline roundtrip tokenization failure." }, { - "left": "issue:43388", - "right": "issue:43688", + "left": "issue:45803", + "right": "issue:46365", "accept": false, - "reason": "One is batch gathering dropping tuple labels; the other is auxiliary loss normalization in MoE models. Different training-time bugs." + "reason": "Bare except swallowing errors is unrelated to Trainer import failing due to datasets." }, { - "left": "issue:43673", - "right": "issue:44743", + "left": "issue:44871", + "right": "issue:45584", "accept": false, - "reason": "Missing GenerationMixin cache in v5.0.0 during chunked prefill is not the same as recurrent states resetting when seq_len>1." + "reason": "Gemma eos-token config mismatch and Whisper empty-transcription failure are different bugs." }, { - "left": "issue:43572", - "right": "issue:46178", + "left": "issue:44077", + "right": "issue:44534", "accept": false, - "reason": "StableLmConfig missing pad_token_idx and a backend registry error for tensorflow_text are unrelated configuration/import issues." + "reason": "Optional post_init restriction is unrelated to non-persistent buffer initialization junk." }, { - "left": "issue:43504", - "right": "issue:45698", + "left": "issue:43065", + "right": "issue:44617", "accept": false, - "reason": "Legacy field handling for Beit preset loading is separate from from_pretrained selecting the wrong custom module after save_pretrained." + "reason": "Dummy Conv2d in Sam3PixelDecoder is a model-architecture issue, not the Sam3Video OOM problem." }, { - "left": "issue:45542", - "right": "issue:46234", + "left": "issue:43540", + "right": "issue:44514", "accept": false, - "reason": "TensorBoard/TensorFlow backend detection and missing is_falcon_mamba_ssm_available are different import-utils/backend availability bugs." + "reason": "Video-input processing error in Qwen3OmniMoe is different from batched chat-template padding crash." }, { - "left": "issue:44734", - "right": "issue:45406", + "left": "issue:44162", + "right": "issue:44792", "accept": false, - "reason": "Serving KV-cache continuation tensor indexing and Gemma4Processor missing _tokenizer are separate serve failures with different causes." + "reason": "ESM2 model breakage is unrelated to a janus image-generation test failure." }, { - "left": "issue:44617", - "right": "issue:45584", + "left": "issue:45803", + "right": "issue:46170", "accept": false, - "reason": "SAM3 video OOM is unrelated to Whisper empty-transcription failure after align_special_tokens." + "reason": "Exception handling bug is unrelated to Windows ROCm import/type dependency issues." }, { - "left": "issue:43673", - "right": "issue:46421", + "left": "issue:44871", + "right": "issue:46050", "accept": false, - "reason": "Both mention cache/chunked prefill, but one is missing GenerationMixin cache and the other is lazy cache initialization causing recompilation; different bugs and fixes." + "reason": "EOS-token config mismatch is not the same as sinq quantized save/load corruption." }, { - "left": "issue:43525", - "right": "issue:43531", + "left": "issue:41950", + "right": "issue:43475", "accept": false, - "reason": "Llama4Config missing pad_token_id is not the same as Qwen3-MoE sliding_window behavior." + "reason": "Video-classification processor lookup bug is unrelated to Sam3Video missing output attribute." }, { - "left": "issue:44792", - "right": "issue:44977", + "left": "issue:43646", + "right": "issue:44913", "accept": false, - "reason": "A failed janus image-generation test is not the same as Qwen3.5 flash-attention generation failure." + "reason": "Broad v5 initialization regression is not the same as GPTNeoX rotary_pct not persisting." }, { - "left": "issue:45381", - "right": "issue:46369", + "left": "issue:43404", + "right": "issue:43441", "accept": false, - "reason": "Qwen2.5-VL video position ids regression and SAM3-LiteText memory overflow are unrelated model-specific issues." + "reason": "Weight tying bug in Mistral3ForConditionalGeneration is different from FlashAttention failure in Ministral-3." }, { - "left": "issue:43931", - "right": "issue:44060", + "left": "issue:43452", + "right": "issue:44496", "accept": false, - "reason": "Weight shape mismatch in Qwen3-VL-30B-A3B is a load-time incompatibility, while the Qwen3-Next issue is an incorrect tied-weights warning tying the wrong parameters." + "reason": "Tokenizer/gguf loading breakage is unrelated to an unrecognized model_type config error." }, { "left": "issue:43761", - "right": "issue:44242", - "accept": false, - "reason": "CLIPVisionModel hidden_states regression and missing load-balancing loss when output_router_logits=False are unrelated code paths." - }, - { - "left": "issue:44792", - "right": "issue:45600", - "accept": false, - "reason": "A janus generate-images test failure and stale auto_mappings references to removed Sam3LiteText configs are not the same issue." - }, - { - "left": "issue:45563", - "right": "issue:45584", + "right": "issue:44479", "accept": false, - "reason": "Paged generate stale warning about num_return_sequences is unrelated to Whisper empty transcription failure." + "reason": "CLIPVision hidden_states regression is unrelated to Qwen video-input regression." }, { - "left": "issue:44190", - "right": "issue:44964", + "left": "issue:45127", + "right": "issue:45200", "accept": false, - "reason": "Local dataset loading in a training script and loading a multimodal model preset are different loading paths with different causes." + "reason": "LoRA extended-vocab collapse and Gemma 4 mm_token_type_ids are distinct model behaviors." }, { - "left": "issue:44534", - "right": "issue:44913", + "left": "issue:43388", + "right": "issue:43688", "accept": false, - "reason": "Non-persistent buffers filled with junk and GPTNeoX rotary_pct reloading are different persistence bugs." + "reason": "Metric gathering last-batch label truncation is unrelated to auxiliary-loss normalization." }, { - "left": "issue:35141", - "right": "issue:42490", + "left": "issue:43673", + "right": "issue:44743", "accept": false, - "reason": "Token embedding resizing reinitialization and import warning for local model paths ending with a slash are unrelated." + "reason": "Missing GenerationMixin cache in chunked_prefill is not the same as recurrent-state reset in Qwen3.5." }, { - "left": "issue:44479", - "right": "issue:44734", + "left": "issue:43572", + "right": "issue:46178", "accept": false, - "reason": "Qwen video input regression and serve KV-cache continuation indexing crash are separate regressions in different subsystems." + "reason": "StableLm pad_token_idx config regression is unrelated to tensorflow_text backend registration." } ] }, @@ -10564,7 +10518,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 34, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -10572,276 +10526,300 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:25251", - "issue:27957", + "issue:35141", "issue:41720", - "issue:42796", + "issue:42490", "issue:43065", - "issue:43116", - "issue:43295", - "issue:43335", + "issue:43504", + "issue:43525", + "issue:43531", "issue:43540", "issue:43550", "issue:43644", + "issue:43673", + "issue:43761", "issue:43827", + "issue:43931", "issue:43976", "issue:43994", + "issue:44060", "issue:44077", - "issue:44112", "issue:44190", - "issue:44462", + "issue:44242", "issue:44464", + "issue:44479", "issue:44521", + "issue:44534", "issue:44610", "issue:44617", "issue:44734", - "issue:44737", - "issue:44779", + "issue:44792", "issue:44855", "issue:44877", "issue:44913", "issue:44933", - "issue:44938", - "issue:45127", + "issue:44964", + "issue:44977", "issue:45245", - "issue:45276", - "issue:45357", "issue:45362", "issue:45381", + "issue:45406", "issue:45542", "issue:45563", "issue:45584", "issue:45588", "issue:45600", - "issue:45632", - "issue:45803", - "issue:45907", + "issue:45698", "issue:46234", "issue:46369", - "issue:46489" + "issue:46421" ], "soft_pairs": [ - "issue:43827|issue:43994", + "issue:43504|issue:45698", + "issue:45542|issue:46234", + "issue:44734|issue:45406", + "issue:44617|issue:45584", + "issue:43673|issue:46421", + "issue:43525|issue:43531", + "issue:44792|issue:44977", + "issue:45381|issue:46369", + "issue:43931|issue:44060", + "issue:44792|issue:45600", + "issue:44190|issue:44964", + "issue:44534|issue:44913", + "issue:43761|issue:44242", + "issue:45563|issue:45584", + "issue:35141|issue:42490", + "issue:44479|issue:44734", "issue:44855|issue:45588", + "issue:43827|issue:43994", "issue:43065|issue:44610", "issue:41720|issue:45245", "issue:44933|issue:45362", - "issue:44464|issue:45563", "issue:44617|issue:46369", - "issue:44521|issue:45381", + "issue:44464|issue:45563", "issue:43644|issue:44913", - "issue:43976|issue:44855", + "issue:44521|issue:45381", "issue:43550|issue:44190", + "issue:43976|issue:44855", "issue:44077|issue:44877", - "issue:45127|issue:45357", - "issue:43540|issue:44734", - "issue:25251|issue:27957", - "issue:44779|issue:46489", - "issue:43335|issue:44190", - "issue:44462|issue:45907", - "issue:42796|issue:44521", - "issue:45632|issue:46234", - "issue:44112|issue:45600", - "issue:44737|issue:45803", - "issue:43116|issue:44190", - "issue:44938|issue:45542", - "issue:43295|issue:45276", - "issue:45276|issue:45584" + "issue:43540|issue:44734" ], - "prepared_review_unit_hash": "55e79cef42957a06384917f4a831f6f56f29515874943817fadbca7c47ab182f", + "prepared_review_unit_hash": "57372d897f6cc4008a2150a0dad8897beb2cc137dad5e779e317dd7661730f6a", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 26, - "serialized_chars": 34849, - "estimated_input_tokens": 8713, - "estimated_eval_tokens": 17682 + "soft_pair_count": 29, + "serialized_chars": 35636, + "estimated_input_tokens": 8909, + "estimated_eval_tokens": 18074 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issue set; only the two DeepSeek tokenizer regressions look like the same underlying v5+ tokenizer bug. No PRs are present in the cluster.", + "summary": "The cluster is mostly heterogeneous, but there is one clear duplicate issue pair about non-persistent buffers being filled with junk in Transformers v5. The rest of the soft pairs describe different model-, feature-, or subsystem-specific failures and should not be merged.", "confidence": 0.91, - "canonical_issue_reason": "issue:44779 is the broader DeepSeek tokenizer regression report, while issue:46489 is a narrower follow-up that explicitly identifies a gap in PR #44801's fix; together they describe the same tokenizer breakage on v5+.", + "canonical_issue_reason": "issue:43644 is the better canonical issue: it is earlier, states the underlying bug clearly, and issue:44534 reads like the same regression restated with a slightly different version label.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44779 is the best anchor for the duplicate set because it states the general regression clearly and can subsume the more specific DeepSeek-Coder v1 follow-up.", + "best_issue_reason": "issue:43644 is the best representative issue in this set because it cleanly captures the shared non-persistent-buffer regression and has the strongest duplicate evidence with issue:44534.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43827", - "right": "issue:43994", + "left": "issue:43504", + "right": "issue:45698", "accept": false, - "reason": "Docs/pipeline-removal reference issue vs a model/pipeline output-quality bug; different underlying problems." + "reason": "Different problems: legacy field preset loading vs. wrong custom module resolution after save/load." }, { - "left": "issue:44855", - "right": "issue:45588", + "left": "issue:45542", + "right": "issue:46234", "accept": false, - "reason": "Python 3.13 import/IndentationError in DebertaV2 vs flash-attention crash on sink-less models; unrelated code paths." + "reason": "Unrelated failures: TensorBoard/TF backend installation issue vs. missing import_utils helper symbol." }, { - "left": "issue:43065", - "right": "issue:44610", + "left": "issue:44734", + "right": "issue:45406", "accept": false, - "reason": "Sam3PixelDecoder dummy conv vs OmDet-Turbo processor/model image-size mismatch; not the same bug." + "reason": "Different serve crashes with different root causes: KV-cache tensor indexing vs. missing _tokenizer on Gemma4Processor." }, { - "left": "issue:41720", - "right": "issue:45245", + "left": "issue:44617", + "right": "issue:45584", "accept": false, - "reason": "CUDA assert during auto device mapping vs category-count limit error; different failure modes and causes." + "reason": "Different models and symptoms: Sam3Video OOM vs. Whisper empty-transcription generation failure." }, { - "left": "issue:44933", - "right": "issue:45362", + "left": "issue:43673", + "right": "issue:46421", "accept": false, - "reason": "Missing/nonexistent import from image_utils vs Qwen3.5 chat crash; unrelated." + "reason": "Both touch chunked prefill/cache, but one is a missing cache in v5.0.0 and the other is spurious recompilation from lazy init; not clearly the same bug." }, { - "left": "issue:44464", - "right": "issue:45563", + "left": "issue:43525", + "right": "issue:43531", "accept": false, - "reason": "Compiled-forward chunked generation mismatch vs stale warning for num_return_sequences; not the same concrete bug." + "reason": "Llama4Config missing pad_token_id and Qwen3-MoE sliding_window behavior are distinct configuration issues." }, { - "left": "issue:44617", + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Unrelated: janus image-generation test failure vs. Qwen3.5 flash-attention generation issue." + }, + { + "left": "issue:45381", "right": "issue:46369", "accept": false, - "reason": "OOM in Sam3Video vs memory overflow in sam3-litetext; same family name but different model and scenario." + "reason": "Different multimodal regressions: Qwen2.5-VL video position ids vs. Sam3-LiteText memory overflow." }, { - "left": "issue:44521", - "right": "issue:45381", + "left": "issue:43931", + "right": "issue:44060", "accept": false, - "reason": "apply_chat_template assistant mask bug vs Qwen2.5-VL video vision_position_ids bug; different multimodal logic." + "reason": "Different model-loading/tied-weight problems with different error surfaces and code paths." }, { - "left": "issue:43644", - "right": "issue:44913", + "left": "issue:44792", + "right": "issue:45600", "accept": false, - "reason": "Non-persistent buffer junk initialization vs GPTNeoXConfig reload resetting rotary_pct; unrelated." + "reason": "Janus test failure is unrelated to stale auto_mappings references to removed Sam3LiteText configs." }, { - "left": "issue:43976", - "right": "issue:44855", + "left": "issue:44190", + "right": "issue:44964", "accept": false, - "reason": "Python version support/load failure vs Python 3.13 parsing/import error; different issues." + "reason": "Dataset loading failure in a training example is unrelated to a multimodal model loading error." }, { - "left": "issue:43550", - "right": "issue:44190", + "left": "issue:44534", + "right": "issue:44913", "accept": false, - "reason": "Bamba torch.compile/SDPA failure vs example script dataset loading problem; unrelated." + "reason": "Non-persistent buffer initialization bug is unrelated to GPTNeoXConfig rotary_pct not persisting on reload." }, { - "left": "issue:44077", - "right": "issue:44877", + "left": "issue:43761", + "right": "issue:44242", "accept": false, - "reason": "Optional post_init restriction vs strict config loading for granite_speech; both config-related but not the same bug." + "reason": "Different regressions: CLIPVisionModel hidden_states output vs. MoE load-balancing loss computation." }, { - "left": "issue:45127", - "right": "issue:45357", + "left": "issue:45563", + "right": "issue:45584", "accept": false, - "reason": "LoRA merge with tied embeddings collapse vs incorrect visual encoder keys on save_pretrained; different code paths." + "reason": "Paged-generate warning behavior is unrelated to Whisper failing on empty transcription after token alignment." }, { - "left": "issue:43540", + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Different issues: embedding reinitialization after resize vs. path-with-trailing-slash import warning." + }, + { + "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Qwen3OmniMoe video input ValueError vs serve KV-cache continuation tensor indexing crash; unrelated." + "reason": "Both involve multimodal/serve code, but the Qwen video-input regression and KV-cache continuation tensor-indexing crash are different bugs." }, { - "left": "issue:25251", - "right": "issue:27957", + "left": "issue:44855", + "right": "issue:45588", "accept": false, - "reason": "Pipeline top_k nesting bug vs XLMRoberta Flash Attention 2; no shared underlying change." + "reason": "Different failures: TorchScript parsing/IndentationError on Python 3.13 vs. flash-attention crash on sink-less models." }, { - "left": "issue:44779", - "right": "issue:46489", - "accept": true, - "reason": "Both report the same DeepSeek tokenizer regression on transformers v5+; 46489 is a narrower follow-up to PR #44801's incomplete fix." + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() is not the same as SigLIP2 producing nonsensical outputs with AutoModel/pipeline." }, { - "left": "issue:43335", - "right": "issue:44190", + "left": "issue:43065", + "right": "issue:44610", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug vs image-classification dataset loading failure; unrelated." + "reason": "Dummy Conv2d in Sam3PixelDecoder is not the same as processor/model input-size mismatch in OmDet-Turbo." }, { - "left": "issue:44462", - "right": "issue:45907", + "left": "issue:41720", + "right": "issue:45245", "accept": false, - "reason": "AutoTokenizer ignoring tokenizer.json vs list_repo_files revision forwarding; different APIs and bugs." + "reason": "Completely different failures: Qwen3 auto device mapping cudaErrorAssert vs. category-count limit error." }, { - "left": "issue:42796", - "right": "issue:44521", + "left": "issue:44933", + "right": "issue:45362", "accept": false, - "reason": "Ministral tokenizer BPE-marker handling vs multimodal assistant-mask generation; unrelated." + "reason": "Nonexistent import from image_utils is unrelated to Qwen3.5 chat crashing." }, { - "left": "issue:45632", - "right": "issue:46234", + "left": "issue:44617", + "right": "issue:46369", "accept": false, - "reason": "Local-model cache-path collision vs missing import_utils helper; unrelated." + "reason": "Related family name only; Sam3Video OOM and Sam3-LiteText memory overflow are different model-specific resource issues." }, { - "left": "issue:44112", - "right": "issue:45600", + "left": "issue:44464", + "right": "issue:45563", "accept": false, - "reason": "GraniteSpeech stale CI test vs Sam3LiteText auto_mappings CI break; both touch CI, but not the same underlying issue." + "reason": "Chunked generation inconsistency under compiled forward is not the same as a stale num_return_sequences warning." }, { - "left": "issue:44737", - "right": "issue:45803", + "left": "issue:43644", + "right": "issue:44913", "accept": false, - "reason": "XLNet device-placement bug vs bare-except error swallowing in FuyuBatchFeature; unrelated." + "reason": "Non-persistent buffer initialization regression is unrelated to GPTNeoXConfig rotary_pct persistence." }, { - "left": "issue:43116", + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Assistant-mask generation for multimodal chat templates is different from Qwen2.5-VL video vision_position_ids." + }, + { + "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "Multi-label classification example returning empty results vs no-trainer image-classification dataset loading; different scripts and failures." + "reason": "Bamba SDPA torch.compile failure is unrelated to local dataset loading in a training script." }, { - "left": "issue:44938", - "right": "issue:45542", + "left": "issue:43976", + "right": "issue:44855", "accept": false, - "reason": "Python 3.14 import/load failure vs TensorFlow backend detection error when only tensorboard is installed; unrelated." + "reason": "Python-version support issue is unrelated to the DebertaV2 TorchScript parse error." }, { - "left": "issue:43295", - "right": "issue:45276", + "left": "issue:44077", + "right": "issue:44877", "accept": false, - "reason": "processor.tokenizer regression vs gemma4 embedding resize propagation bug; different model code paths." + "reason": "Optional post_init allowance for patchtsmixer and strict config rejection for granite_speech are different config-loading issues." }, { - "left": "issue:45276", - "right": "issue:45584", + "left": "issue:43540", + "right": "issue:44734", "accept": false, - "reason": "Gemma4 resize_token_embeddings propagation bug vs Whisper empty-transcription generation failure; unrelated." + "reason": "Qwen3OmniMoe video processing ValueError is unrelated to serve KV-cache indexing crashes." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Same underlying regression: both report Transformers v5 filling non-persistent buffers with junk, with only minor wording/version differences." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: the only clearly related pair is issue:44779 and issue:46489, which both describe the DeepSeek tokenizer regression on transformers v5+ and the follow-up gap in PR #44801. The note that no PRs are present is also correct. The other soft-edge rejections are conservative and do not overclaim identity from weak title similarity." - }, - "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 }, @@ -10850,7 +10828,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 35, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -10858,281 +10836,266 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:25251", "issue:27957", - "issue:28282", - "issue:32090", - "issue:35141", "issue:36246", "issue:38617", "issue:41628", "issue:41950", "issue:42175", + "issue:42796", + "issue:43116", "issue:43295", + "issue:43335", "issue:43404", - "issue:43479", - "issue:43526", - "issue:43531", "issue:43643", - "issue:43650", - "issue:43673", - "issue:43854", - "issue:43901", - "issue:43976", "issue:44016", "issue:44112", "issue:44186", - "issue:44220", + "issue:44190", "issue:44246", "issue:44279", "issue:44295", + "issue:44462", "issue:44496", + "issue:44521", "issue:44683", "issue:44734", - "issue:44964", - "issue:44987", + "issue:44737", + "issue:44779", + "issue:44938", "issue:45003", + "issue:45127", "issue:45216", "issue:45245", - "issue:45335", - "issue:45375", + "issue:45276", + "issue:45357", "issue:45446", + "issue:45542", "issue:45584", "issue:45600", "issue:45632", + "issue:45803", + "issue:45907", "issue:46082", "issue:46121", - "issue:46123", - "issue:46369", + "issue:46234", "issue:46421", - "issue:46424", + "issue:46489", "issue:46506" ], "soft_pairs": [ + "issue:45127|issue:45357", + "issue:25251|issue:27957", + "issue:44779|issue:46489", + "issue:43335|issue:44190", + "issue:44462|issue:45907", + "issue:42796|issue:44521", + "issue:44112|issue:45600", + "issue:44737|issue:45803", + "issue:43116|issue:44190", + "issue:45632|issue:46234", + "issue:44938|issue:45542", + "issue:43295|issue:45276", + "issue:45276|issue:45584", "issue:43404|issue:45216", - "issue:45245|issue:45584", "issue:36246|issue:41628", + "issue:45245|issue:45584", "issue:46082|issue:46121", - "issue:45632|issue:46421", "issue:36246|issue:38617", - "issue:44186|issue:44734", - "issue:44683|issue:45446", + "issue:45632|issue:46421", "issue:42175|issue:44279", + "issue:44683|issue:45446", "issue:44246|issue:45003", - "issue:44295|issue:46506", + "issue:44186|issue:44734", "issue:41950|issue:44016", - "issue:44964|issue:45600", - "issue:43673|issue:46424", - "issue:43643|issue:44496", - "issue:43854|issue:44964", - "issue:43650|issue:44220", - "issue:44112|issue:44964", - "issue:44016|issue:44220", - "issue:27957|issue:35141", - "issue:43479|issue:45375", - "issue:43531|issue:45600", - "issue:43295|issue:45335", - "issue:43976|issue:44987", - "issue:46123|issue:46369", - "issue:43526|issue:43901", - "issue:28282|issue:35141", - "issue:32090|issue:35141" + "issue:44295|issue:46506", + "issue:43643|issue:44496" ], - "prepared_review_unit_hash": "e85ad0791153ed1fe4ea431e68df76ce7a32c02b797e5d39177d8b22e0d9ffdc", + "prepared_review_unit_hash": "d6b450ce632b4354cc747bcb6c0449d6dd58084945c219feaa0125bdd083180b", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 36124, - "estimated_input_tokens": 9031, - "estimated_eval_tokens": 18318 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35135, + "estimated_input_tokens": 8784, + "estimated_eval_tokens": 17824 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated Transformers bug reports, not true duplicates. All soft pairs should be rejected; they only share broad themes like imports, cache behavior, or multimodal models. If one issue has to stand in as the best representative, issue 46424 is the clearest concrete runtime bug, but it is not a duplicate target for the rest.", - "confidence": 0.96, - "canonical_issue_reason": "No tight duplicate cluster emerges from these items. If forced to pick one representative issue, 46424 is the most concrete and actionable cache-path bug, with a clear behavior regression and a focused code path.", + "summary": "This cluster is heterogeneous: the titles overlap on broad themes like tokenizers, imports, pipelines, and multimodal models, but none of the pairs point to the same concrete bug or fix.", + "confidence": 0.94, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 46424 is the strongest standalone report here: it describes a specific, reproducible cache configuration bug rather than a vague import/model-loading failure or a test/doc issue.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43404", - "right": "issue:45216", - "accept": false, - "reason": "Both mention model weight tying/saving, but they are different models and different failure modes, so not the same underlying bug." - }, - { - "left": "issue:45245", - "right": "issue:45584", - "accept": false, - "reason": "Completely different problems: category cardinality error versus Whisper empty-transcription handling." - }, - { - "left": "issue:36246", - "right": "issue:41628", + "left": "issue:45127", + "right": "issue:45357", "accept": false, - "reason": "Both are import errors, but for different missing symbols and different modules; not one shared defect." + "reason": "Both involve Qwen-related saving/weights, but one is LoRA merge collapse with tied embeddings while the other is incorrect visual encoder keys in save_pretrained; different failure modes and code paths." }, { - "left": "issue:46082", - "right": "issue:46121", + "left": "issue:25251", + "right": "issue:27957", "accept": false, - "reason": "One is LlamaConfig head_dim validation, the other is rope-parameter JSON handling; unrelated code paths." + "reason": "Pipeline top_k output nesting and XLMRoberta Flash Attention 2 are unrelated; one is an output-shape bug, the other a model/attention compatibility issue." }, { - "left": "issue:45632", - "right": "issue:46421", + "left": "issue:44779", + "right": "issue:46489", "accept": false, - "reason": "Both involve cache, but one is a trust_remote_code path collision and the other is chunked-prefill recompilation behavior." + "reason": "Both mention DeepSeek tokenizer regressions, but they target different model families and reports; the titles do not identify the same tokenizer bug or fix." }, { - "left": "issue:36246", - "right": "issue:38617", + "left": "issue:43335", + "right": "issue:44190", "accept": false, - "reason": "Different import breakages in different modules; no shared concrete bug." + "reason": "SwitchTransformers sparse-layer configuration and loading a local dataset in an image-classification script are unrelated issues." }, { - "left": "issue:44186", - "right": "issue:44734", + "left": "issue:44462", + "right": "issue:45907", "accept": false, - "reason": "Tokenizer padding/truncation crash is unrelated to server KV-cache tensor indexing." + "reason": "AutoTokenizer ignoring tokenizer.json and list_repo_files not forwarding revision are separate repository-loading bugs with different APIs." }, { - "left": "issue:44683", - "right": "issue:45446", + "left": "issue:42796", + "right": "issue:44521", "accept": false, - "reason": "Same flex_attention area, but one is a torch>=2.9 failure and the other is a version-check/import guard bug." + "reason": "Tokenizer BPE-marker handling and multimodal assistant_masks generation are different tokenizer behaviors and do not share a concrete underlying bug." }, { - "left": "issue:42175", - "right": "issue:44279", + "left": "issue:44112", + "right": "issue:45600", "accept": false, - "reason": "One is a missing TensorFlow backend in extras install, the other is a vague dependency issue; too broad to be the same bug." + "reason": "One is a stale GraniteSpeech device-override test, the other is auto_mappings referencing removed configs; both affect CI but not the same defect." }, { - "left": "issue:44246", - "right": "issue:45003", + "left": "issue:44737", + "right": "issue:45803", "accept": false, - "reason": "Import performance regression and unsafe sys.modules access are distinct issues." + "reason": "XLNet device placement in relative positional encoding and Fuyu's bare except swallowing errors are unrelated code-path problems." }, { - "left": "issue:44295", - "right": "issue:46506", + "left": "issue:43116", + "right": "issue:44190", "accept": false, - "reason": "Buffer registration/read error and CPU-vs-CUDA output divergence are different defects." + "reason": "A multi-label classification example returning empty results is unrelated to a script failing to load a local dataset." }, { - "left": "issue:41950", - "right": "issue:44016", + "left": "issue:45632", + "right": "issue:46234", "accept": false, - "reason": "Pipeline processor lookup bug versus a notebook syntax error; no duplication." + "reason": "A trust_remote_code cache-path collision and a missing import_utils availability flag are different import/cache bugs." }, { - "left": "issue:44964", - "right": "issue:45600", + "left": "issue:44938", + "right": "issue:45542", "accept": false, - "reason": "Model loading failure and stale auto_mappings references are different problems." + "reason": "Python 3.14 import failure and missing TensorFlow backend handling are separate environment/dependency issues." }, { - "left": "issue:43673", - "right": "issue:46424", + "left": "issue:43295", + "right": "issue:45276", "accept": false, - "reason": "Both mention cache, but one is missing cache in chunked prefill and the other is max_cache_length being ignored except for quantized cache." + "reason": "The regression around processor.tokenizer/images passed to tokenizer is different from gemma4 resize_token_embeddings not updating output/embed layers." }, { - "left": "issue:43643", - "right": "issue:44496", + "left": "issue:45276", + "right": "issue:45584", "accept": false, - "reason": "Missing fields from trust_remote_code config loading is unrelated to unrecognized model_type handling." + "reason": "Gemma4 embedding resize propagation and Whisper empty-transcription behavior after align_special_tokens are not the same bug." }, { - "left": "issue:43854", - "right": "issue:44964", + "left": "issue:43404", + "right": "issue:45216", "accept": false, - "reason": "Different multimodal model-load failures with different repros and likely different fixes." + "reason": "Mistral3 lm_head tying/save behavior and Qwen3.5 save_pretrained checkpoint correctness are model-specific serialization/weight issues, not one defect." }, { - "left": "issue:43650", - "right": "issue:44220", + "left": "issue:36246", + "right": "issue:41628", "accept": false, - "reason": "No meaningful overlap; one is a placeholder title and the other is an audio feature extraction bug." + "reason": "Both are import errors, but one is a missing Qwen2_5_VLImageProcessor export and the other is a missing top-level AutoImageProcessor export; different symbols and fixes." }, { - "left": "issue:44112", - "right": "issue:44964", + "left": "issue:45245", + "right": "issue:45584", "accept": false, - "reason": "CI test flake in GraniteSpeech is not the same as a Phi-4 multimodal loading bug." + "reason": "The 2^24 categories limit error and Whisper empty transcription after token alignment are unrelated." }, { - "left": "issue:44016", - "right": "issue:44220", + "left": "issue:46082", + "right": "issue:46121", "accept": false, - "reason": "Notebook syntax error is unrelated to _torch_extract_fbank_features." + "reason": "LlamaConfig head_dim validation and convert_rope_params_to_dict rejecting a JSON-loaded list are separate config-validation bugs." }, { - "left": "issue:27957", - "right": "issue:35141", + "left": "issue:36246", + "right": "issue:38617", "accept": false, - "reason": "Flash Attention support for XLM-Roberta and post_init reinitialization after resize_token_embeddings are different code paths." + "reason": "Both are import failures, but they concern different missing names in different modules and do not indicate the same underlying change." }, { - "left": "issue:43479", - "right": "issue:45375", + "left": "issue:45632", + "right": "issue:46421", "accept": false, - "reason": "Different config bugs: default sub-config initialization versus a missing field dropped by strict validation." + "reason": "A cache collision for trust_remote_code and spurious recompilation from lazy cache initialization are unrelated caching problems." }, { - "left": "issue:43531", - "right": "issue:45600", + "left": "issue:42175", + "right": "issue:44279", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE and removed auto_mappings references are unrelated." + "reason": "TensorFlow missing from a torch install and a vague dependency issue with transformers are not the same specific defect." }, { - "left": "issue:43295", - "right": "issue:45335", + "left": "issue:44683", + "right": "issue:45446", "accept": false, - "reason": "Processor/tokenizer regression and t5gemma embedding resize mismatch are different bugs." + "reason": "Compiled flex_attention failing on torch >= 2.9 and an incorrect version check for AuxRequest import are related to flex_attention but are distinct concrete bugs." }, { - "left": "issue:43976", - "right": "issue:44987", + "left": "issue:44246", + "right": "issue:45003", "accept": false, - "reason": "Python version compatibility failure is unrelated to a specific model-loading regression." + "reason": "Slow import time and unsafe sys.modules access are different problems in different parts of the import path." }, { - "left": "issue:46123", - "right": "issue:46369", + "left": "issue:44186", + "right": "issue:44734", "accept": false, - "reason": "Final partial-batch dropping results and sam3-litetext memory overflow are different symptoms and likely different fixes." + "reason": "LayoutLMv2Tokenizer NER/padding crashes and transformers serve KV-cache continuation indexing are unrelated failures." }, { - "left": "issue:43526", - "right": "issue:43901", + "left": "issue:41950", + "right": "issue:44016", "accept": false, - "reason": "Processor label reduction bug and stale docs are not the same underlying issue." + "reason": "A video-classification pipeline looking for image processors and a syntax error in a notebook section are wholly different issues." }, { - "left": "issue:28282", - "right": "issue:35141", + "left": "issue:44295", + "right": "issue:46506", "accept": false, - "reason": "ImportError due to missing PyTorch is unrelated to embedding reinitialization after resizing." + "reason": "Reading position_ids after buffer registration and CPU/CUDA output divergence in PPDocLayoutV3 are separate runtime bugs." }, { - "left": "issue:32090", - "right": "issue:35141", + "left": "issue:43643", + "right": "issue:44496", "accept": false, - "reason": "Trainer gpu_broadcast_one NoneType error is a different failure mode from token embedding resizing." + "reason": "Missing fields when trust_remote_code=True in AutoConfig and an unrecognized model lacking model_type are different config-loading failures." } ] }, @@ -11148,7 +11111,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 36, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -11156,281 +11119,316 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:38175", - "issue:41950", + "issue:27957", + "issue:28282", + "issue:32090", + "issue:35141", "issue:42796", "issue:42907", + "issue:43295", "issue:43299", "issue:43381", "issue:43441", + "issue:43479", + "issue:43526", + "issue:43531", "issue:43577", "issue:43650", "issue:43673", "issue:43824", + "issue:43854", "issue:43901", "issue:43976", "issue:44016", "issue:44062", "issue:44075", - "issue:44242", + "issue:44112", + "issue:44220", "issue:44279", "issue:44336", - "issue:44351", "issue:44368", "issue:44496", - "issue:44521", - "issue:44617", "issue:44792", "issue:44933", - "issue:45020", + "issue:44964", + "issue:44987", "issue:45042", "issue:45083", - "issue:45103", - "issue:45200", "issue:45216", - "issue:45278", - "issue:45325", "issue:45335", - "issue:45357", - "issue:45372", - "issue:45405", + "issue:45375", "issue:45584", "issue:45588", "issue:45593", + "issue:45600", "issue:45759", - "issue:46082", - "issue:46209", - "issue:46399", + "issue:46123", + "issue:46369", "issue:46424", "issue:46506", "issue:46515" ], "soft_pairs": [ + "issue:43854|issue:44964", + "issue:44964|issue:45600", + "issue:43673|issue:46424", + "issue:44112|issue:44964", + "issue:43650|issue:44220", + "issue:44016|issue:44220", + "issue:27957|issue:35141", + "issue:43479|issue:45375", + "issue:43531|issue:45600", + "issue:43295|issue:45335", + "issue:43976|issue:44987", + "issue:46123|issue:46369", + "issue:43526|issue:43901", + "issue:32090|issue:35141", + "issue:28282|issue:35141", "issue:42907|issue:43441", "issue:43441|issue:45588", - "issue:44016|issue:44062", "issue:44496|issue:45042", - "issue:45083|issue:46506", + "issue:44016|issue:44062", "issue:43299|issue:45216", - "issue:43673|issue:44336", + "issue:45083|issue:46506", "issue:44792|issue:46424", + "issue:43673|issue:44336", "issue:43824|issue:44933", - "issue:43976|issue:44279", "issue:43381|issue:45593", + "issue:43976|issue:44279", "issue:43901|issue:44496", "issue:42796|issue:43650", "issue:45335|issue:45584", "issue:44496|issue:45759", "issue:44279|issue:46515", "issue:44075|issue:44368", - "issue:43577|issue:45216", - "issue:44242|issue:45200", - "issue:45103|issue:45372", - "issue:46082|issue:46399", - "issue:43299|issue:45357", - "issue:44279|issue:45405", - "issue:43650|issue:46209", - "issue:38175|issue:44617", - "issue:44351|issue:45278", - "issue:41950|issue:45020", - "issue:44521|issue:45325" + "issue:43577|issue:45216" ], - "prepared_review_unit_hash": "6a4b4fe4847bf9e1d8338873bab2c637654698847042864d04063f7c0ed9e39f", + "prepared_review_unit_hash": "4845aba75cd122bda1703a57dd2a5c4da00b1731669ed30965cf9b4ca3820c25", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 35991, - "estimated_input_tokens": 8998, - "estimated_eval_tokens": 18252 + "soft_pair_count": 33, + "serialized_chars": 37525, + "estimated_input_tokens": 9382, + "estimated_eval_tokens": 19020 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated issue reports with only superficial lexical overlap (many model- or subsystem-specific regressions, import errors, and docs bugs). No soft pair looks like the same underlying bug/change, so there is no strong duplicate cluster here.", - "confidence": 0.3, - "canonical_issue_reason": "No true canonical duplicate emerges from this heterogeneous set; if one must be chosen, issue 44521 is the broadest and most active representative (open, higher inbound references/discussion), but it is not a genuine duplicate hub.", - "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 is the best representative by activity and cross-references, but the set is too diverse to treat it as a real canonical bug report.", + "summary": "This is a loose similarity bucket, not a true duplicate cluster. The issues cover many unrelated model-loading, config, tokenizer, cache, and test failures across different models and code paths, so none of the soft links should be merged.", + "confidence": 0.91, + "canonical_issue_reason": "No single canonical issue: the items are too heterogeneous and do not appear to describe the same underlying bug.", + "canonical_pr_reason": "No PRs are present.", + "best_issue_reason": "issue:44964 is the clearest high-level regression report in the set, with concrete reproduction and broad impact on loading a specific multimodal model in recent Transformers versions.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:42907", - "right": "issue:43441", + "left": "issue:43854", + "right": "issue:44964", "accept": false, - "reason": "Same model family mention, but one is about saving dequantized Ministral models and the other is a FlashAttention load/runtime failure; different code paths." + "reason": "Both are model-loading failures, but for different models and different failure modes (missing import vs bad load behavior)." }, { - "left": "issue:43441", - "right": "issue:45588", + "left": "issue:44964", + "right": "issue:45600", "accept": false, - "reason": "Both mention flash attention, but one is Ministral-specific and the other is a sink-less-models AttributeError in `integrations/flash_attention.py`; not the same bug." + "reason": "Different causes: one is a model load regression, the other is a broken auto-mapping reference in CI." }, { - "left": "issue:44016", - "right": "issue:44062", + "left": "issue:43673", + "right": "issue:46424", "accept": false, - "reason": "Notebook syntax error vs tokenizer `AddedToken` argument collision; unrelated failures." + "reason": "Both mention cache, but one is a generation cache regression and the other is cache-length behavior for quantized cache only." }, { - "left": "issue:44496", - "right": "issue:45042", - "accept": false, - "reason": "Unrecognized model config vs image processor torchvision dependency issue; different underlying problems." + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "Unrelated: GraniteSpeech CI test failure vs a multimodal model loading regression." }, { - "left": "issue:45083", - "right": "issue:46506", + "left": "issue:43650", + "right": "issue:44220", "accept": false, - "reason": "Different models and different failures: a helper-length calculation bug vs CPU/CUDA result divergence." + "reason": "No shared bug: the first is a placeholder-like issue, the second is an audio feature extraction error." }, { - "left": "issue:43299", - "right": "issue:45216", + "left": "issue:44016", + "right": "issue:44220", "accept": false, - "reason": "Both are Qwen3.5-related, but one is model loading and the other is incorrect `save_pretrained` checkpoint output; not the same concrete bug." + "reason": "Notebook syntax error and feature-extraction TypeError are unrelated." }, { - "left": "issue:43673", - "right": "issue:44336", + "left": "issue:27957", + "right": "issue:35141", "accept": false, - "reason": "Cache/missing GenerationMixin issue vs ANSI escape output in loading_report; unrelated." + "reason": "Different subsystems: FlashAttention support for XLM-Roberta vs embedding reinitialization after resize." }, { - "left": "issue:44792", - "right": "issue:46424", + "left": "issue:43479", + "right": "issue:45375", "accept": false, - "reason": "Janus test failure vs cache-length semantics; no shared code-path or bug." + "reason": "Both are config-related, but they concern different composite-model fields and different initialization bugs." }, { - "left": "issue:43824", - "right": "issue:44933", + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Different problems: Qwen3-MoE sliding-window behavior vs stale auto-mapping references." + }, + { + "left": "issue:43295", + "right": "issue:45335", "accept": false, - "reason": "Both are import errors, but for different symbols/modules and different root causes." + "reason": "Both are regression-style model issues, but one is processor/tokenizer API breakage and the other is embedding resize behavior." }, { "left": "issue:43976", - "right": "issue:44279", + "right": "issue:44987", "accept": false, - "reason": "Python version compatibility regression vs vague dependency issue; too broad and not the same bug." + "reason": "These are separate version-compatibility/load failures affecting different environments and models." }, { - "left": "issue:43381", - "right": "issue:45593", + "left": "issue:46123", + "right": "issue:46369", "accept": false, - "reason": "Gradient checkpointing eval-mode guard vs D-FINE auxiliary-loss behavior; unrelated." + "reason": "Different code paths: mask generation batching vs SAM3-lite memory overflow." }, { - "left": "issue:43901", - "right": "issue:44496", + "left": "issue:43526", + "right": "issue:43901", "accept": false, - "reason": "Docs wording mismatch vs model-loading config error; no overlap in underlying defect." + "reason": "Beit processor label handling is unrelated to pipeline documentation wording." }, { - "left": "issue:42796", - "right": "issue:43650", + "left": "issue:32090", + "right": "issue:35141", "accept": false, - "reason": "Tokenizer BPE-marker handling vs nonspecific low-signal issue title; no duplicate evidence." + "reason": "Trainer broadcast error and token-embedding resize regression are unrelated." }, { - "left": "issue:45335", - "right": "issue:45584", + "left": "issue:28282", + "right": "issue:35141", "accept": false, - "reason": "Token embedding resize bug vs Whisper empty-transcription generation failure; different code paths." + "reason": "ImportError due to missing PyTorch is not the same as the embedding reinit bug." }, { - "left": "issue:44496", - "right": "issue:45759", + "left": "issue:42907", + "right": "issue:43441", "accept": false, - "reason": "Both concern model config/loading, but one is an unrecognized model type and the other is `from_config` not unwrapping `text_config`; distinct bugs." + "reason": "Same model family, but different failures: saving dequantized Ministrals vs FlashAttention load/runtime failure." }, { - "left": "issue:44279", - "right": "issue:46515", + "left": "issue:43441", + "right": "issue:45588", "accept": false, - "reason": "Generic dependency complaint vs torch 2.4.1 incompatibility; not the same concrete issue." + "reason": "Both touch FlashAttention, but they describe different concrete crashes and code paths." }, { - "left": "issue:44075", - "right": "issue:44368", + "left": "issue:44496", + "right": "issue:45042", "accept": false, - "reason": "Optimizer argument handling vs a model config warning about `tie_word_embeddings`; unrelated." + "reason": "Different failures: unrecognized model config vs PIL image processors wrongly requiring torchvision." }, { - "left": "issue:43577", + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error and AddedToken keyword conflict are unrelated." + }, + { + "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "Both are Qwen/BLIP-style save/load issues, but one is dtype propagation in BLIP2 loading and the other is Qwen3.5 checkpoint serialization; not mergeable as one fix." + "reason": "Different Qwen regressions: model loading failure vs incorrect checkpoint saved by save_pretrained." }, { - "left": "issue:44242", - "right": "issue:45200", + "left": "issue:45083", + "right": "issue:46506", "accept": false, - "reason": "Moe load-balancing loss gating vs Gemma 4 multimodal token-type defaults; different model behaviors." + "reason": "Different model families and different bug types: helper length computation vs CPU/CUDA divergence." }, { - "left": "issue:45103", - "right": "issue:45372", + "left": "issue:44792", + "right": "issue:46424", "accept": false, - "reason": "Auto-docstring annotation parsing crash vs missing `ReasoningEffort` import breaking Gemma 4 processor loading; unrelated." + "reason": "Janus image-generation test failure is unrelated to cache-length configuration behavior." }, { - "left": "issue:46082", - "right": "issue:46399", + "left": "issue:43673", + "right": "issue:44336", "accept": false, - "reason": "LlamaConfig validation vs Qwen3.5 rotary-kernel registration error; different components and failure modes." + "reason": "Generation cache regression and ANSI color output bug are unrelated." }, { - "left": "issue:43299", - "right": "issue:45357", + "left": "issue:43824", + "right": "issue:44933", "accept": false, - "reason": "Same family of Qwen models, but loading failure vs incorrect visual-encoder keys during save; separate regressions." + "reason": "Both are import errors, but they refer to different modules and missing symbols." }, { - "left": "issue:44279", - "right": "issue:45405", + "left": "issue:43381", + "right": "issue:45593", "accept": false, - "reason": "Generic dependency issue vs a specific minimum PEFT version bump; not the same bug." + "reason": "Gradient checkpointing eval-mode limitation is unrelated to D-FINE auxiliary-loss handling." }, { - "left": "issue:43650", - "right": "issue:46209", + "left": "issue:43976", + "right": "issue:44279", "accept": false, - "reason": "Low-information issue title vs FP8 inference on B200; no evidence of sameness." + "reason": "Python version incompatibility and generic dependency issue are not the same underlying bug." }, { - "left": "issue:38175", - "right": "issue:44617", + "left": "issue:43901", + "right": "issue:44496", "accept": false, - "reason": "Siglip2 zero-probabilities bug vs Sam3Video OOM; completely different issues." + "reason": "Pipeline docs mismatch is unrelated to model config import failure." }, { - "left": "issue:44351", - "right": "issue:45278", + "left": "issue:42796", + "right": "issue:43650", "accept": false, - "reason": "Single missing import vs broad import errors after upgrade; one does not imply the other." + "reason": "Tokenizer BPE-marker handling and a placeholder issue are unrelated." }, { - "left": "issue:41950", - "right": "issue:45020", + "left": "issue:45335", + "right": "issue:45584", "accept": false, - "reason": "Video pipeline processor lookup bug vs remote_code breakage; different subsystems and failure causes." + "reason": "Different behaviors: decoder embedding resize vs Whisper empty-transcription generation failure." }, { - "left": "issue:44521", - "right": "issue:45325", + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Both involve model/config loading, but one is an unknown-model error and the other is missing unwrapping of composite text_config." + }, + { + "left": "issue:44279", + "right": "issue:46515", + "accept": false, + "reason": "Generic dependency issue is not the same as a torch 2.4.1 incompatibility." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Optimizer argument handling and a tie_word_embeddings warning are different issues." + }, + { + "left": "issue:43577", + "right": "issue:45216", "accept": false, - "reason": "Both touch multimodal helpers, but one is assistant mask generation and the other is rope-index scaling for Qwen2.5-VL; different concrete bugs." + "reason": "Blip2 dtype propagation and Qwen checkpoint saving are unrelated bugs." } ] }, @@ -11446,7 +11444,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 37, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -11454,10 +11452,9 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:33290", + "issue:38175", "issue:39692", "issue:41950", - "issue:42175", "issue:42994", "issue:43295", "issue:43299", @@ -11465,276 +11462,256 @@ "issue:43441", "issue:43479", "issue:43525", + "issue:43650", "issue:43653", - "issue:43723", "issue:43854", "issue:43901", "issue:43976", - "issue:43994", "issue:44016", - "issue:44162", - "issue:44261", - "issue:44263", + "issue:44242", + "issue:44279", "issue:44291", "issue:44297", - "issue:44361", + "issue:44351", "issue:44462", "issue:44466", - "issue:44623", - "issue:44671", + "issue:44521", + "issue:44617", "issue:44683", - "issue:44749", "issue:44779", + "issue:45020", "issue:45081", + "issue:45103", + "issue:45200", "issue:45276", - "issue:45397", + "issue:45278", + "issue:45325", + "issue:45357", + "issue:45372", "issue:45405", "issue:45542", "issue:45593", - "issue:45600", - "issue:45656", - "issue:45712", - "issue:45800", "issue:45812", "issue:45823", "issue:45834", "issue:45907", "issue:46082", + "issue:46209", + "issue:46399", "issue:46512" ], "soft_pairs": [ - "issue:43295|issue:44779", + "issue:46082|issue:46399", + "issue:45103|issue:45372", + "issue:44279|issue:45405", + "issue:44242|issue:45200", + "issue:43299|issue:45357", + "issue:41950|issue:45020", + "issue:43650|issue:46209", + "issue:38175|issue:44617", + "issue:44351|issue:45278", + "issue:44521|issue:45325", "issue:39692|issue:43901", + "issue:43295|issue:44779", "issue:41950|issue:45823", "issue:44683|issue:45542", - "issue:44779|issue:45812", "issue:43976|issue:45405", + "issue:44779|issue:45812", "issue:46082|issue:46512", "issue:43299|issue:43854", "issue:43479|issue:43525", "issue:44297|issue:44462", "issue:44466|issue:45276", "issue:44016|issue:44291", - "issue:45081|issue:45907", "issue:43441|issue:45834", + "issue:45081|issue:45907", "issue:43653|issue:45593", - "issue:45600|issue:45712", - "issue:42994|issue:43317", - "issue:33290|issue:45656", - "issue:44623|issue:44749", - "issue:44623|issue:45397", - "issue:44261|issue:44263", - "issue:44361|issue:45593", - "issue:45542|issue:45800", - "issue:44016|issue:45823", - "issue:42175|issue:43976", - "issue:43441|issue:44683", - "issue:44671|issue:44779", - "issue:43723|issue:44162", - "issue:43901|issue:43994" + "issue:42994|issue:43317" ], - "prepared_review_unit_hash": "2af24d4cb4274dbb2b2f448365314dae0b39f7a96f7d507d9218979d8773295f", + "prepared_review_unit_hash": "2f83f538b94bee428d1089c32f71f335d51f87964db0414217302fd33a7b2ae0", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 35898, - "estimated_input_tokens": 8975, - "estimated_eval_tokens": 18206 + "soft_pair_count": 26, + "serialized_chars": 35026, + "estimated_input_tokens": 8757, + "estimated_eval_tokens": 17770 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous; the soft pairs mostly share superficial keywords (tokenizer, loading, v5, deepspeed, docs) but describe different bugs, regressions, or cleanup items. I would not merge any of the proposed pairs.", - "confidence": 0.93, - "canonical_issue_reason": null, + "summary": "This cluster is mostly a grab bag of unrelated Transformers bugs and regressions across tokenizers, multimodal processors, quantization, config handling, and docs. None of the soft pairs look like the same concrete bug or change, so all are rejected.", + "confidence": 0.58, + "canonical_issue_reason": "No true duplicate hub stands out; 45278 is only the broadest umbrella-like regression report (many import errors after upgrade), but it still does not subsume the rest of the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44779 is the most concrete user-facing regression report among the set, but it is still not a true duplicate of the other tokenizer/model issues, so the cluster has no strong canonical issue.", + "best_issue_reason": "45278 is the best triage anchor among these issues because it is the most general, user-facing regression report and closest to the small import-error subtheme, though it is still not a duplicate of the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43295", - "right": "issue:44779", + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "Different failures: LlamaConfig head_dim validation vs a Qwen3.5 kernel registration crash." + }, + { + "left": "issue:45103", + "right": "issue:45372", "accept": false, - "reason": "Both are tokenizer-related regressions, but one is a processor/tokenizer API regression and the other is DeepSeek token ID corruption; different failure modes and likely different fixes." + "reason": "Unrelated code paths: auto_docstring annotation handling vs mistral_common import break in Gemma processor loading." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Generic dependency issue vs one specific unreleased PEFT version bump; not the same bug." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "MoE load-balancing loss gating and Gemma4 mm_token_type_ids defaults are different model behaviors." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Loading Qwen3VL MoE models and saving Qwen3.5 visual encoder keys are separate code paths." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video pipeline processor lookup bug is not the same as remote_code compatibility regressions." + }, + { + "left": "issue:43650", + "right": "issue:46209", + "accept": false, + "reason": "Placeholder/noisy issue title vs FP8 inference on B200; no shared bug." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "SigLIP2 zero probabilities and Sam3Video OOM are unrelated model/runtime problems." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "A missing HybridCache import is not the same as a broad set of import errors after upgrade." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "All-zero assistant masks in chat templating and Qwen2.5-VL rope indexing are different multimodal bugs." }, { "left": "issue:39692", "right": "issue:43901", "accept": false, - "reason": "Both mention docs/examples, but one is a SigLIP2 example with model/processor mismatch and quantization failure while the other is a pipeline docs wording mismatch." + "reason": "Both are docs issues, but they concern different examples/pages and different failures." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Processor.tokenizer regression and Deepseek tokenizer output regression are distinct tokenizer problems." }, { "left": "issue:41950", "right": "issue:45823", "accept": false, - "reason": "Unrelated problems: video-classification processor lookup vs Gemma4 device_map forward mismatch." + "reason": "Video-classification processor mismatch and Gemma4 PLE device mismatch are different bugs." }, { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "Different subsystems and bugs: compiled flex_attention on torch>=2.9 vs TensorBoard/TensorFlow backend installation issue." + "reason": "Torch flex_attention compilation failure and tensorboard-only backend install issue are unrelated." }, { - "left": "issue:44779", - "right": "issue:45812", + "left": "issue:43976", + "right": "issue:45405", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different code paths (DeepSeek tokenizer vs Granite AutoTokenizer IDs)." + "reason": "Python version support break is not the same as the unreleased PEFT dependency bump." }, { - "left": "issue:43976", - "right": "issue:45405", + "left": "issue:44779", + "right": "issue:45812", "accept": false, - "reason": "Python version compatibility issue vs a PEFT version pin bump; no common bug." + "reason": "Different model families and different tokenizer regressions; not mergeable as one fix." }, { "left": "issue:46082", "right": "issue:46512", "accept": false, - "reason": "LlamaConfig head_dim validation and dtype staleness in from_config are separate config bugs." + "reason": "Config validation for head_dim and stale config.dtype after from_config are separate initialization bugs." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Both are model-loading failures, but for different models and different underlying loader paths." + "reason": "Qwen3.5 loading regression and GLM-4.7-Flash unit-test loading failure are different concrete problems." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Distinct config initialization bugs in different model families; not the same underlying defect." + "reason": "Default multimodal config initialization and missing pad_token_id on Llama4Config are not the same issue." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Both concern tokenizer persistence/loading, but one is save_pretrained writing the wrong tokenizer_class while the other ignores tokenizer.json during AutoTokenizer loading." + "reason": "Tokenizer class serialization mismatch and AutoTokenizer ignoring tokenizer.json affect different load/save paths." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "Different weight-serialization and resize_embedding behaviors; not the same code-path problem." + "reason": "Inconsistent lm_head serialization by device and resize_token_embeddings not propagating are different weight-management bugs." }, { "left": "issue:44016", "right": "issue:44291", "accept": false, - "reason": "Notebook syntax error is unrelated to init_empty_weights/_is_hf_initialized model-loading TypeError." + "reason": "Notebook syntax error vs init_empty_weights unexpected-arg TypeError; unrelated." }, { - "left": "issue:45081", - "right": "issue:45907", + "left": "issue:43441", + "right": "issue:45834", "accept": false, - "reason": "Tokenizer regex patch crash vs Hub list_repo_files revision forwarding; completely different APIs." + "reason": "FlashAttention failure for Ministral-3 and Kosmos2.5 OCR index error are separate model-specific issues." }, { - "left": "issue:43441", - "right": "issue:45834", + "left": "issue:45081", + "right": "issue:45907", "accept": false, - "reason": "FlashAttention failure for Ministral vs OCR index error in Kosmos2.5; unrelated bugs." + "reason": "Mistral tokenizer patch crash and list_repo_files revision forwarding are unrelated repository/client bugs." }, { "left": "issue:43653", "right": "issue:45593", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug vs D-FINE auxiliary loss logic; different components." - }, - { - "left": "issue:45600", - "right": "issue:45712", - "accept": false, - "reason": "Both touch repo maintenance/CI, but one is stale auto_mappings references and the other is leftover dummy classes; not the same change." + "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss omission are different subsystems." }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Both involve quantization/offload, but one is saving a quantized model and the other is loading a dequantized model with device_map=auto; different failure points." - }, - { - "left": "issue:33290", - "right": "issue:45656", - "accept": false, - "reason": "Both mention deepspeed/optimizer behavior, but one is OOM with Adafactor and the other is double optimizer.step calls; not the same bug." - }, - { - "left": "issue:44623", - "right": "issue:44749", - "accept": false, - "reason": "Processor save_pretrained missing files is unrelated to a Chinese-language data-filtering performance regression." - }, - { - "left": "issue:44623", - "right": "issue:45397", - "accept": false, - "reason": "processor.save_pretrained file emission bug vs gemma-4 zero3 from_pretrained failure; different code paths." - }, - { - "left": "issue:44261", - "right": "issue:44263", - "accept": false, - "reason": "MLA rms_norm_eps precision mismatch and GlmMoeDsaIndexer torch.split return handling are separate issues." - }, - { - "left": "issue:44361", - "right": "issue:45593", - "accept": false, - "reason": "MLukeTokenizer AttributeError on tasks vs D-FINE auxiliary loss behavior; unrelated." - }, - { - "left": "issue:45542", - "right": "issue:45800", - "accept": false, - "reason": "TensorBoard/TensorFlow dependency issue vs torch/transformers compatibility problem; different root causes." - }, - { - "left": "issue:44016", - "right": "issue:45823", - "accept": false, - "reason": "Notebook syntax error and Gemma4 PLE device mismatch are unrelated." - }, - { - "left": "issue:42175", - "right": "issue:43976", - "accept": false, - "reason": "TensorFlow extra/backend packaging issue vs Python version support issue; not the same bug." - }, - { - "left": "issue:43441", - "right": "issue:44683", - "accept": false, - "reason": "Different backend failures: Ministral FlashAttention vs compiled flex_attention on newer torch." - }, - { - "left": "issue:44671", - "right": "issue:44779", - "accept": false, - "reason": "Both are v5 regressions affecting outputs, but one is CamemBERT masked-LM predictions and the other is DeepSeek tokenizer encoding; different models and fixes." - }, - { - "left": "issue:43723", - "right": "issue:44162", - "accept": false, - "reason": "AutoTokenizer loading issue vs ESM2 model breakage; too broad and not the same underlying defect." - }, - { - "left": "issue:43901", - "right": "issue:43994", - "accept": false, - "reason": "Docs text about return_all_scores does not duplicate a SigLIP2 model/pipeline correctness regression." + "reason": "Quantized model save failure and dequantized model offload/load failure are related only broadly, not the same concrete bug." } ] }, @@ -11750,7 +11727,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 38, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -11758,41 +11735,40 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:36246", + "issue:33290", "issue:41628", - "issue:41950", - "issue:42491", - "issue:42757", - "issue:42886", - "issue:43299", + "issue:42175", + "issue:43441", "issue:43525", - "issue:43526", - "issue:43531", - "issue:43550", "issue:43643", "issue:43673", + "issue:43723", "issue:43761", "issue:43784", "issue:43824", "issue:43881", "issue:43901", "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", "issue:44188", - "issue:44220", "issue:44242", "issue:44246", + "issue:44261", + "issue:44263", "issue:44295", "issue:44297", "issue:44336", - "issue:44351", + "issue:44361", "issue:44373", - "issue:44485", "issue:44496", "issue:44589", - "issue:44734", - "issue:44857", - "issue:44871", - "issue:45003", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", "issue:45103", "issue:45362", "issue:45397", @@ -11800,259 +11776,254 @@ "issue:45561", "issue:45563", "issue:45593", - "issue:46132", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45800", + "issue:45823", "issue:46291", - "issue:46399", - "issue:46424", - "issue:46506" + "issue:46399" ], "soft_pairs": [ - "issue:43673|issue:43761", + "issue:45600|issue:45712", + "issue:33290|issue:45656", + "issue:44623|issue:44749", + "issue:44261|issue:44263", + "issue:44623|issue:45397", + "issue:44361|issue:45593", + "issue:45542|issue:45800", + "issue:44016|issue:45823", + "issue:42175|issue:43976", + "issue:43441|issue:44683", + "issue:44671|issue:44779", + "issue:43723|issue:44162", + "issue:43901|issue:43994", "issue:46291|issue:46399", - "issue:43824|issue:43976", + "issue:43673|issue:43761", "issue:43525|issue:43881", - "issue:44297|issue:45397", + "issue:43824|issue:43976", "issue:43901|issue:44246", "issue:43525|issue:43643", + "issue:44297|issue:45397", "issue:44589|issue:45362", - "issue:41628|issue:43976", "issue:43784|issue:45542", + "issue:41628|issue:43976", "issue:44188|issue:44336", "issue:44336|issue:44373", - "issue:44242|issue:45593", "issue:45103|issue:46399", "issue:45561|issue:45563", + "issue:44242|issue:45593", "issue:44246|issue:44496", - "issue:44295|issue:45397", - "issue:42491|issue:43299", - "issue:43531|issue:45362", - "issue:43550|issue:46506", - "issue:43526|issue:45003", - "issue:41950|issue:44220", - "issue:44246|issue:44871", - "issue:42886|issue:46132", - "issue:44336|issue:44485", - "issue:36246|issue:42757", - "issue:36246|issue:44351", - "issue:44734|issue:46424", - "issue:44336|issue:46424", - "issue:45003|issue:45561", - "issue:44242|issue:44857" + "issue:44295|issue:45397" ], - "prepared_review_unit_hash": "41e21d4b11a76370d121b66f946c47b040829bfaf968c179bd72f86ee73dee40", + "prepared_review_unit_hash": "c13163f29bd505d34e2d14c0bc77708f4ed4cef380d4b85782ca133b9d180779", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 36288, - "estimated_input_tokens": 9072, - "estimated_eval_tokens": 18400 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36531, + "estimated_input_tokens": 9133, + "estimated_eval_tokens": 18522 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly topical noise: the soft pairs share broad themes like import errors, model-loading regressions, or cache behavior, but they do not look like the same concrete bug or change. I would not merge any of the proposed soft edges.", + "summary": "No soft edge looks like a true duplicate; the linked issues are mostly superficial similarities across different bugs, regressions, packaging problems, and docs/test fallout.", "confidence": 0.95, - "canonical_issue_reason": "No single issue is a clear canonical duplicate here; the items span unrelated regressions across imports, model config, kernels, caching, docs, and testing.", + "canonical_issue_reason": "No single canonical issue stands out: the soft-linked items are heterogeneous and do not describe the same underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "No issue stands out as a global representative for the whole set because the cluster is heterogeneous rather than one underlying bug family.", + "best_issue_reason": "No strong best issue for deduplication; the set is too mixed to use one issue as a representative canonical.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43673", - "right": "issue:43761", + "left": "issue:45600", + "right": "issue:45712", "accept": false, - "reason": "Both are v5 regressions, but one is cache/chunked_prefill and the other is CLIP hidden_states output; different code paths." + "reason": "Different subsystems: config mapping/CI break vs leftover dummy classes in testing stubs." }, { - "left": "issue:46291", - "right": "issue:46399", + "left": "issue:33290", + "right": "issue:45656", "accept": false, - "reason": "One is a pipelines import failure from kernels metadata, the other is a Qwen3.5 kernel registration crash; not the same bug." + "reason": "Both mention DeepSpeed/optimizer, but one is Adafactor OOM while the other is double optimizer.step()." }, { - "left": "issue:43824", - "right": "issue:43976", + "left": "issue:44623", + "right": "issue:44749", "accept": false, - "reason": "Qwen2.5-VL import error vs Python version compatibility regression; unrelated failures." + "reason": "Unrelated: missing processor save files vs tokenizer slowdown on Chinese text." }, { - "left": "issue:43525", - "right": "issue:43881", + "left": "issue:44261", + "right": "issue:44263", "accept": false, - "reason": "Different models and symptoms: missing pad_token_id vs a glm-4v-9b load failure." + "reason": "Different code paths: MLA rms_norm_eps precision issue vs torch.split return values in a MoE indexer." }, { - "left": "issue:44297", + "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "Tokenizer save metadata mismatch and gemma-4 zero3 from_pretrained are separate issues." + "reason": "processor.save_pretrained file output bug vs gemma-4 ZeRO-3 from_pretrained failure are different problems." }, { - "left": "issue:43901", - "right": "issue:44246", + "left": "issue:44361", + "right": "issue:45593", "accept": false, - "reason": "Docs inconsistency and slow import performance are unrelated." + "reason": "Tokenizer AttributeError on tasks vs D-FINE auxiliary-loss behavior are unrelated." }, { - "left": "issue:43525", - "right": "issue:43643", + "left": "issue:45542", + "right": "issue:45800", "accept": false, - "reason": "Llama4Config pad_token_id and AutoConfig trust_remote_code field loss are different config bugs." + "reason": "Backend packaging issue vs torch/transformers version incompatibility." }, { - "left": "issue:44589", - "right": "issue:45362", + "left": "issue:44016", + "right": "issue:45823", "accept": false, - "reason": "Float8 storage lookup failure and Qwen3.5 chat crash do not share the same underlying code path." + "reason": "Notebook syntax error vs Gemma4 device-mapping mismatch." }, { - "left": "issue:41628", + "left": "issue:42175", "right": "issue:43976", "accept": false, - "reason": "ImportError for AutoImageProcessor is unrelated to Python 3.9/3.10 compatibility." + "reason": "Install/backend packaging issue vs Python version support regression." }, { - "left": "issue:43784", - "right": "issue:45542", + "left": "issue:43441", + "right": "issue:44683", "accept": false, - "reason": "Sentence-transformers import NameError and tensorboard-without-TF backend error have different causes." + "reason": "Model FlashAttention failure vs compiled flex_attention on newer torch; different code paths." }, { - "left": "issue:44188", - "right": "issue:44336", + "left": "issue:44671", + "right": "issue:44779", "accept": false, - "reason": "Attention kernel divergence and ANSI escape leakage are unrelated." + "reason": "CamemBERT MLM outputs vs DeepSeek tokenizer correctness are distinct bugs." }, { - "left": "issue:44336", - "right": "issue:44373", + "left": "issue:43723", + "right": "issue:44162", "accept": false, - "reason": "Loading report terminal formatting and a position_ids docstring bug are not the same issue." + "reason": "AutoTokenizer loading failure vs ESM2 model breakage." }, { - "left": "issue:44242", - "right": "issue:45593", + "left": "issue:43901", + "right": "issue:43994", "accept": false, - "reason": "Both mention missing losses, but they affect different models and gating conditions." + "reason": "Docs mismatch for return_all_scores vs SigLIP2 model/pipeline wrong outputs." }, { - "left": "issue:45103", + "left": "issue:46291", "right": "issue:46399", "accept": false, - "reason": "Auto-docstring future-annotations crash and Qwen3.5 kernel registration are unrelated." + "reason": "Pipeline import dependency validation vs Qwen3.5 rotary-pos-embedding registration error." }, { - "left": "issue:45561", - "right": "issue:45563", + "left": "issue:43673", + "right": "issue:43761", "accept": false, - "reason": "xdist race in testing utils vs stale warning in paged generate; different subsystems." + "reason": "Generation cache/chunked_prefill vs CLIPVisionModel hidden_states regression." }, { - "left": "issue:44246", - "right": "issue:44871", + "left": "issue:43525", + "right": "issue:43881", "accept": false, - "reason": "Slow import performance and Gemma-3 EOS config mismatch are unrelated." + "reason": "Missing pad_token_id in config vs glm-4v model load failure." }, { - "left": "issue:44295", - "right": "issue:45397", + "left": "issue:43824", + "right": "issue:43976", "accept": false, - "reason": "position_ids buffer access and gemma-4 zero3 loading are different bugs." + "reason": "Import error for a model class vs Python version compatibility; not the same issue." }, { - "left": "issue:42491", - "right": "issue:43299", + "left": "issue:43901", + "right": "issue:44246", "accept": false, - "reason": "Both involve Qwen3 MoE, but one is a LoRA/version compatibility problem and the other is a model-loading regression." + "reason": "Pipeline doc mismatch vs import slowness complaint." }, { - "left": "issue:43531", - "right": "issue:45362", + "left": "issue:43525", + "right": "issue:43643", "accept": false, - "reason": "Sliding_window behavior in Qwen3-MoE and Qwen3.5 chat failure are distinct failures." + "reason": "Missing config attribute on Llama4Config vs trust_remote_code returning incomplete fields." }, { - "left": "issue:43550", - "right": "issue:46506", + "left": "issue:44297", + "right": "issue:45397", "accept": false, - "reason": "torch.compile/SDPA failure in Bamba is unrelated to CPU/CUDA output divergence in PPDocLayoutV3." + "reason": "Tokenizer save_pretrained metadata bug vs gemma-4 Zero3 from_pretrained issue." }, { - "left": "issue:43526", - "right": "issue:45003", + "left": "issue:44589", + "right": "issue:45362", "accept": false, - "reason": "BeitImageProcessorFast label reduction and sys.modules access in modeling_utils are unrelated." + "reason": "Float8 storage deserialization vs Qwen3.5 chat crash." }, { - "left": "issue:41950", - "right": "issue:44220", + "left": "issue:43784", + "right": "issue:45542", "accept": false, - "reason": "Video-classification pipeline misrouting to image processors and fbank feature extraction are different code paths." + "reason": "Import-time NameError in sentence-transformers vs missing tf backend package." }, { - "left": "issue:44246", - "right": "issue:44871", + "left": "issue:41628", + "right": "issue:43976", "accept": false, - "reason": "Import-time slowdown and EOS-token mismatch do not indicate the same bug." + "reason": "AutoImageProcessor import failure vs Python 3.9/3.10 support issue." }, { - "left": "issue:42886", - "right": "issue:46132", + "left": "issue:44188", + "right": "issue:44336", "accept": false, - "reason": "Offline cache loading and AttentionInterface.register behavior changes are unrelated." + "reason": "torch.compile attention-kernel divergence vs ANSI codes in loading_report." }, { "left": "issue:44336", - "right": "issue:44485", + "right": "issue:44373", "accept": false, - "reason": "ANSI code leakage and GLM-5 RoPE implementation are not the same issue." + "reason": "ANSI codes in loading_report vs wrong docstring for position_ids." }, { - "left": "issue:36246", - "right": "issue:42757", + "left": "issue:45103", + "right": "issue:46399", "accept": false, - "reason": "Different missing symbols from different packages: qwen2_5_vl processor export vs huggingface_hub offline helper." + "reason": "auto_docstring crash under future annotations vs Qwen3.5 kernels registration error." }, { - "left": "issue:36246", - "right": "issue:44351", + "left": "issue:45561", + "right": "issue:45563", "accept": false, - "reason": "Both are import errors, but they expose different APIs and failure modes." + "reason": "xdist race on captured_info.txt vs stale generation warning." }, { - "left": "issue:44734", - "right": "issue:46424", + "left": "issue:44242", + "right": "issue:45593", "accept": false, - "reason": "Server KV-cache continuation crash and cache max length config semantics are different cache bugs." + "reason": "MoE load-balancing loss omission vs D-FINE auxiliary-loss behavior." }, { - "left": "issue:44336", - "right": "issue:46424", + "left": "issue:44246", + "right": "issue:44496", "accept": false, - "reason": "Terminal ANSI formatting and cache configuration are unrelated." + "reason": "Import slowness vs unrecognized model/config error." }, { - "left": "issue:45003", - "right": "issue:45561", + "left": "issue:44295", + "right": "issue:45397", "accept": false, - "reason": "Unsafe sys.modules access and a pytest-xdist file race are unrelated." - }, - { - "left": "issue:44242", - "right": "issue:44857", - "accept": false, - "reason": "MoE load-balancing loss gating and LwDetrImageLoss AMP/CUDA crash are different issues." + "reason": "position_ids buffer access error vs Gemma4 ZeRO-3 loading failure." } ] }, @@ -12068,7 +12039,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 39, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -12077,287 +12048,287 @@ "cluster_id": "cluster-10000-463", "nodes": [ "issue:36246", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:42886", "issue:43257", "issue:43262", + "issue:43299", "issue:43352", "issue:43475", - "issue:43540", + "issue:43526", + "issue:43531", + "issue:43550", "issue:43596", - "issue:43653", "issue:43673", "issue:43901", - "issue:44038", - "issue:44060", - "issue:44077", "issue:44162", + "issue:44220", + "issue:44242", "issue:44246", "issue:44295", - "issue:44466", - "issue:44479", + "issue:44336", + "issue:44351", + "issue:44485", "issue:44496", "issue:44561", + "issue:44734", "issue:44749", "issue:44779", - "issue:44877", - "issue:44987", - "issue:44995", + "issue:44857", + "issue:44871", "issue:45003", "issue:45137", "issue:45200", "issue:45250", "issue:45292", - "issue:45335", + "issue:45362", "issue:45375", - "issue:45397", - "issue:45412", - "issue:45459", + "issue:45561", "issue:45584", - "issue:45600", - "issue:45656", - "issue:45685", "issue:45701", "issue:45759", "issue:45859", - "issue:46123", - "issue:46170", + "issue:46132", "issue:46369", "issue:46399", - "issue:46431", + "issue:46424", "issue:46506" ], "soft_pairs": [ - "issue:43475|issue:46369", + "issue:42491|issue:43299", + "issue:43531|issue:45362", + "issue:43550|issue:46506", + "issue:43526|issue:45003", + "issue:41950|issue:44220", + "issue:42886|issue:46132", + "issue:44246|issue:44871", + "issue:44336|issue:44485", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:44734|issue:46424", + "issue:44336|issue:46424", + "issue:45003|issue:45561", + "issue:44242|issue:44857", "issue:44779|issue:45292", - "issue:44295|issue:44749", + "issue:43475|issue:46369", "issue:45859|issue:46399", "issue:36246|issue:43596", + "issue:44295|issue:44749", "issue:43673|issue:44561", + "issue:43901|issue:45003", "issue:44162|issue:45200", "issue:43352|issue:45250", - "issue:43901|issue:45003", "issue:43257|issue:45137", "issue:44496|issue:45003", "issue:45292|issue:45584", "issue:45292|issue:45701", - "issue:44038|issue:46506", "issue:43262|issue:44246", - "issue:45412|issue:46123", - "issue:45375|issue:45759", - "issue:44877|issue:45375", - "issue:44561|issue:46399", - "issue:45685|issue:46431", - "issue:44987|issue:45459", - "issue:44466|issue:45335", - "issue:44995|issue:45200", - "issue:45137|issue:45600", - "issue:43475|issue:43540", - "issue:44060|issue:44077", - "issue:45397|issue:45656", - "issue:43653|issue:44162", - "issue:44479|issue:46170" + "issue:45375|issue:45759" ], - "prepared_review_unit_hash": "5e62a91e005f46f96859bf7517b921f564f2d8dc96881e96f8859024db9f605a", + "prepared_review_unit_hash": "1df48a692fa140e6641d9afd3a8b59211f20c3b74f65e8cb1aab52957a2567a7", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36460, - "estimated_input_tokens": 9115, - "estimated_eval_tokens": 18486 + "serialized_chars": 36318, + "estimated_input_tokens": 9080, + "estimated_eval_tokens": 18416 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct regressions or feature gaps across Qwen, SAM, DeepSpeed, tokenizers, docs, and config handling. None of the soft pairs look like the same underlying bug/change, so all should stay separate.", - "confidence": 0.96, - "canonical_issue_reason": "No single issue is a reliable canonical duplicate target; the set is too mixed, with only loose theme-level similarity between a few embedding/tokenization reports.", + "summary": "This cluster is highly heterogeneous: most items are superficially similar only at the level of model family, version bumps, caches, or generic import/runtime failures. I would reject every soft pair as a duplicate/merge candidate.", + "confidence": 0.83, + "canonical_issue_reason": "No single canonical issue stands out; the items span unrelated bugs across model loading, tokenization, cache handling, audio/image pipelines, and test utilities.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45292 is the closest thing to a reusable representative because it sits in the token-embedding-resize subtheme, but it still does not cover the broader cluster.", + "best_issue_reason": "If forced to pick one representative issue, 44561 is the broadest cross-model compatibility regression and the most generally useful triage anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43475", - "right": "issue:46369", + "left": "issue:42491", + "right": "issue:43299", "accept": false, - "reason": "Both involve SAM 3, but one is a missing output attribute in the vision encoder and the other is a memory overflow in sam3-litetext; different failures and code paths." + "reason": "Both are Qwen3-MoE version regressions, but one is LoRA/accelerate loading and the other is multimodal model loading; not the same bug." }, { - "left": "issue:44779", - "right": "issue:45292", + "left": "issue:43531", + "right": "issue:45362", "accept": false, - "reason": "DeepSeek tokenizer correctness vs resize_token_embeddings not updating output embeddings; unrelated subsystems and symptoms." + "reason": "Same model family, different failures: sliding_window behavior vs a chat crash. Different code paths." }, { - "left": "issue:44295", - "right": "issue:44749", + "left": "issue:43550", + "right": "issue:46506", "accept": false, - "reason": "Position_ids buffer/read error is a model-state bug; the Chinese report is about a v4.57.3-to-v5.3.0 slowdown while filtering data. Not the same issue." + "reason": "Completely different models and symptoms: torch.compile+SDPA failure vs CPU/CUDA output mismatch." }, { - "left": "issue:45859", - "right": "issue:46399", + "left": "issue:43526", + "right": "issue:45003", "accept": false, - "reason": "Both are Qwen3.5-related, but one is CPU-offload incompatibility in rotary embedding and the other is a kernel registration/decorator crash. Different concrete bugs." + "reason": "Unrelated bugs: BeitImageProcessorFast label reduction vs unsafe sys.modules access in modeling_utils." }, { - "left": "issue:36246", - "right": "issue:43596", + "left": "issue:41950", + "right": "issue:44220", "accept": false, - "reason": "Qwen2.5-VL import error and a DeepSpeed ZeRO3 BertModel IndexError are unrelated." + "reason": "Different subsystems and failures; video-classification processor lookup is not the same as fbank feature extraction." }, { - "left": "issue:43673", - "right": "issue:44561", + "left": "issue:42886", + "right": "issue:46132", "accept": false, - "reason": "Chunked-prefill cache bug vs removal of is_torch_fx_available breaking trust_remote_code; different code paths and causes." + "reason": "Offline cache loading and AttentionInterface.register behavior are unrelated." }, { - "left": "issue:44162", - "right": "issue:45200", + "left": "issue:44246", + "right": "issue:44871", "accept": false, - "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaulting are different model-family bugs." + "reason": "Import latency and Gemma eos_token_id config mismatch are unrelated issues." }, { - "left": "issue:43352", - "right": "issue:45250", + "left": "issue:44336", + "right": "issue:44485", "accept": false, - "reason": "A specific Nemotron FA2 unsupported error is not the same as a generic Flash Attention 2.0 topic." + "reason": "ANSI log formatting and GLM-5 RoPE implementation are unrelated." }, { - "left": "issue:43901", - "right": "issue:45003", + "left": "issue:36246", + "right": "issue:42757", "accept": false, - "reason": "Docs mismatch for return_all_scores vs unsafe sys.modules access in modeling_utils; unrelated issues." + "reason": "Both are import errors, but for different missing symbols in different modules." }, { - "left": "issue:43257", - "right": "issue:45137", + "left": "issue:36246", + "right": "issue:44351", "accept": false, - "reason": "Both mention DeepSpeed/ZeRO3, but one is Qwen3 MOE weight conversion and the other is an empty-deque IndexError. Not the same bug." + "reason": "Different missing exports from different packages; no shared underlying defect." }, { - "left": "issue:44496", - "right": "issue:45003", + "left": "issue:44734", + "right": "issue:46424", "accept": false, - "reason": "Unrecognized model/config loading failure is unrelated to sys.modules access in modeling_utils." + "reason": "Both mention cache, but one is KV-cache continuation indexing and the other is max_cache_length semantics." }, { - "left": "issue:45292", - "right": "issue:45584", + "left": "issue:44336", + "right": "issue:46424", "accept": false, - "reason": "resize_token_embeddings/output_embeddings behavior and Whisper empty-transcription after align_special_tokens are different model-specific problems." + "reason": "No real overlap beyond being general bugs; unrelated features and failures." }, { - "left": "issue:45292", - "right": "issue:45701", + "left": "issue:45003", + "right": "issue:45561", "accept": false, - "reason": "A token-embedding resize bug is not the same as a broad version-dependent tokenization change." + "reason": "Modeling_utils sys.modules access is unrelated to the xdist captured_info.txt race." }, { - "left": "issue:44038", - "right": "issue:46506", + "left": "issue:44242", + "right": "issue:44857", "accept": false, - "reason": "Qwen3-VL-Moe video-input regression and PPDocLayoutV3 CPU/CUDA differences are separate multimodal issues." + "reason": "Different model/training bugs: router loss omission vs AMP/CUDA crash in a loss class." }, { - "left": "issue:43262", - "right": "issue:44246", + "left": "issue:44779", + "right": "issue:45292", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and intermittent slow import are unrelated." + "reason": "Tokenizer output regression and resize_token_embeddings not updating output embeddings are different bugs." }, { - "left": "issue:45412", - "right": "issue:46123", + "left": "issue:43475", + "right": "issue:46369", "accept": false, - "reason": "RT-DETR memory leak on deletion and mask generation partial-batch dropping are unrelated." + "reason": "Both are SAM 3-related, but one is a missing attribute in video vision output and the other is memory overflow in litetext." }, { - "left": "issue:45375", - "right": "issue:45759", + "left": "issue:45859", + "right": "issue:46399", "accept": false, - "reason": "Missing strict field propagation in Qwen3.5MoeVisionConfig and from_config not unwrapping text_config are both config issues but different concrete failures." + "reason": "Different Qwen3.5 MoE failures: CPU offload incompatibility vs missing decorator during kernel registration." }, { - "left": "issue:44877", - "right": "issue:45375", + "left": "issue:36246", + "right": "issue:43596", "accept": false, - "reason": "Strict config rejection for granite_speech and a missing Qwen3.5MoeVisionConfig field are not the same bug." + "reason": "Different model families and failure modes: missing Qwen processor export vs DeepSpeed/BertModel zero3 IndexError." }, { - "left": "issue:44561", - "right": "issue:46399", + "left": "issue:44295", + "right": "issue:44749", "accept": false, - "reason": "A v5 removal breaking trust_remote_code and a Qwen3.5 rotary-registration crash are different regressions." + "reason": "One is a position_ids buffer access error; the other is a performance regression after upgrade." }, { - "left": "issue:45685", - "right": "issue:46431", + "left": "issue:43673", + "right": "issue:44561", "accept": false, - "reason": "MPS histogram failure in MOE and missing chat template in a processor are unrelated." + "reason": "Both are v5 regressions, but cache-missing in chunked_prefill and removal of is_torch_fx_available are separate code-path breaks." }, { - "left": "issue:44987", - "right": "issue:45459", + "left": "issue:43901", + "right": "issue:45003", "accept": false, - "reason": "Loading failure for a specific model and protobuf-error masking in tokenizers are different problem areas." + "reason": "Docs mismatch in TextClassificationPipeline is unrelated to modeling_utils sys.modules access." }, { - "left": "issue:44466", - "right": "issue:45335", + "left": "issue:44162", + "right": "issue:45200", "accept": false, - "reason": "Device-dependent tied-weight serialization and t5gemma resize_token_embeddings not updating decoder.embed_tokens are similar in theme only, not the same concrete bug." + "reason": "Different models and different bugs: ESM2 breakage vs Gemma 4 token-type defaults." }, { - "left": "issue:44995", - "right": "issue:45200", + "left": "issue:43352", + "right": "issue:45250", "accept": false, - "reason": "A stale indexer cache crash in GlmMoeDsa and Gemma 4 token-id defaults are unrelated." + "reason": "Generic flash-attention topic, but one is model support gap and the other is an unspecified flash-attention issue; not the same bug." }, { - "left": "issue:45137", - "right": "issue:45600", + "left": "issue:43257", + "right": "issue:45137", "accept": false, - "reason": "ZeRO3 empty-deque failure and CI breakage from removed Sam3LiteText configs are unrelated." + "reason": "Qwen3 MoE weight conversion on load is unrelated to a DeepSpeed ZeRO3 deque error." }, { - "left": "issue:43475", - "right": "issue:43540", + "left": "issue:44496", + "right": "issue:45003", "accept": false, - "reason": "Different Qwen/SAM multimodal paths: one missing an output attribute, the other a video-processing ValueError." + "reason": "Unrecognized model config and unsafe sys.modules access are distinct problems." }, { - "left": "issue:44060", - "right": "issue:44077", + "left": "issue:45292", + "right": "issue:45584", "accept": false, - "reason": "Qwen3-Next tied-weights warning and patchtsmixer post_init policy are unrelated." + "reason": "Resize-token-embeddings behavior is unrelated to Whisper empty-transcription failure." }, { - "left": "issue:45397", - "right": "issue:45656", + "left": "issue:45292", + "right": "issue:45701", "accept": false, - "reason": "Gemma-4 ZeRO3 loading and a double optimizer-step issue under DeepSpeed are different failure modes." + "reason": "Embedding resize bug and tokenizer behavior change across versions are not the same underlying issue." }, { - "left": "issue:43653", - "right": "issue:44162", + "left": "issue:43262", + "right": "issue:44246", "accept": false, - "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are separate issues." + "reason": "Audio processor sampling-rate default bug is unrelated to import-time slowness." }, { - "left": "issue:44479", - "right": "issue:46170", + "left": "issue:45375", + "right": "issue:45759", "accept": false, - "reason": "A Qwen video-input regression and a Windows/ROCm torch.distributed import problem are unrelated." + "reason": "Both touch Qwen multimodal config handling, but one drops a missing field and the other fails to unwrap text_config; different fixes." } ] }, @@ -12373,7 +12344,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 40, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -12381,294 +12352,287 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:28282", - "issue:30990", - "issue:42371", "issue:42886", - "issue:42994", "issue:43065", - "issue:43232", "issue:43408", "issue:43475", - "issue:43482", - "issue:43643", - "issue:43646", - "issue:43673", + "issue:43540", + "issue:43653", "issue:43906", "issue:43931", "issue:44038", + "issue:44060", + "issue:44077", "issue:44162", "issue:44188", "issue:44297", - "issue:44361", "issue:44368", + "issue:44466", + "issue:44479", "issue:44483", + "issue:44561", "issue:44623", - "issue:44749", - "issue:44869", - "issue:44871", "issue:44877", "issue:44977", + "issue:44987", "issue:44995", - "issue:45092", + "issue:45137", + "issue:45200", "issue:45229", - "issue:45230", - "issue:45310", - "issue:45356", - "issue:45381", + "issue:45335", + "issue:45375", "issue:45397", + "issue:45412", + "issue:45459", "issue:45563", "issue:45584", "issue:45588", + "issue:45600", "issue:45656", + "issue:45685", "issue:45987", "issue:46082", "issue:46123", - "issue:46132", "issue:46170", "issue:46202", - "issue:46421" + "issue:46399", + "issue:46421", + "issue:46431", + "issue:46506" ], "soft_pairs": [ + "issue:44877|issue:45375", + "issue:44038|issue:46506", + "issue:44561|issue:46399", + "issue:45412|issue:46123", + "issue:45685|issue:46431", + "issue:44987|issue:45459", + "issue:44466|issue:45335", + "issue:45137|issue:45600", + "issue:44995|issue:45200", + "issue:43475|issue:43540", + "issue:44060|issue:44077", + "issue:44479|issue:46170", + "issue:43653|issue:44162", + "issue:45397|issue:45656", + "issue:42886|issue:45229", "issue:44368|issue:44977", "issue:43906|issue:46202", - "issue:42886|issue:45229", - "issue:44995|issue:45588", "issue:44483|issue:46202", + "issue:44995|issue:45588", "issue:44368|issue:45563", "issue:43065|issue:43408", "issue:44188|issue:46202", "issue:43931|issue:44368", "issue:44877|issue:46170", "issue:45987|issue:46082", - "issue:44188|issue:46421", "issue:43065|issue:46123", + "issue:44188|issue:46421", "issue:45584|issue:46170", - "issue:44297|issue:44623", - "issue:44361|issue:44749", - "issue:43646|issue:44162", - "issue:44749|issue:45656", - "issue:44623|issue:45656", - "issue:43232|issue:43673", - "issue:44749|issue:45310", - "issue:44869|issue:45356", - "issue:44871|issue:46132", - "issue:42994|issue:43482", - "issue:45230|issue:45397", - "issue:43475|issue:43643", - "issue:44038|issue:44623", - "issue:28282|issue:30990", - "issue:42371|issue:45092", - "issue:43475|issue:45381" + "issue:44297|issue:44623" ], - "prepared_review_unit_hash": "936de4a8c621aedbe28d642a57598c2be6367bd8db35584a419f2ffb74968a19", + "prepared_review_unit_hash": "864a418d90e0909b033ad0f153b24da6c1fda5177756a47542e68ef71c15598b", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 30, - "serialized_chars": 35957, - "estimated_input_tokens": 8990, - "estimated_eval_tokens": 18236 + "soft_pair_count": 29, + "serialized_chars": 35808, + "estimated_input_tokens": 8952, + "estimated_eval_tokens": 18160 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly a loose collection of Transformers bugs and regressions rather than true duplicates. They share broad subsystem themes (tokenizers, generation, model loading, v5 regressions), but the concrete failure modes differ enough that none of the soft pairs should be merged.", - "confidence": 0.82, - "canonical_issue_reason": "issue:43646 is the broadest representative of the v5 initialization-regression theme and best serves as an umbrella issue among this set.", + "summary": "This is a loose, heterogeneous cluster of Transformers bug reports rather than a single duplicate family. The only near-duplicate-looking subthread is tokenizer/save_pretrained serialization, but even those reports describe different failure modes.", + "confidence": 0.83, + "canonical_issue_reason": "issue:44297 is the closest representative of the small tokenizer serialization subtheme: it is narrowly scoped, concrete, and the strongest semantic neighbor to 44623, even though the broader cluster is mixed.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45092 is the most concrete and actionable report, with a specific model family and a clear failure mode tied to meta initialization.", + "best_issue_reason": "issue:44297 is the most concrete and best-scoped issue here; it reads like a clear reproducible regression instead of a broad symptom report.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44368", - "right": "issue:44977", - "accept": false, - "reason": "Different problems: a config warning about tied embeddings vs a flash-attention generation failure." - }, - { - "left": "issue:43906", - "right": "issue:46202", + "left": "issue:44877", + "right": "issue:45375", "accept": false, - "reason": "Unrelated bugs; one is a reproduction of an older issue, the other is a Gemma pipeline roundtrip failure." + "reason": "Both are strict-config loading problems, but they hit different models and different missing/ignored fields; not the same bug." }, { - "left": "issue:42886", - "right": "issue:45229", + "left": "issue:44038", + "right": "issue:46506", "accept": false, - "reason": "Offline tokenizer cache loading and multi-GPU CUDA OOM are different failure modes." + "reason": "Unrelated failures: a Qwen3-VL-Moe bug versus CPU/CUDA output divergence in PPDocLayoutV3." }, { - "left": "issue:44995", - "right": "issue:45588", + "left": "issue:44561", + "right": "issue:46399", "accept": false, - "reason": "Different code paths: stale indexer cache on second forward vs flash_attention handling of s_aux=None." + "reason": "Both are v5 regressions, but the concrete code paths differ: removed helper breaking trust_remote_code models versus missing decorator on rotary registration." }, { - "left": "issue:44483", - "right": "issue:46202", + "left": "issue:45412", + "right": "issue:46123", "accept": false, - "reason": "API request rejection and tokenizer roundtrip failure are unrelated." + "reason": "RT-DETR memory retention and MaskGenerationPipeline partial-batch handling are different symptoms and subsystems." }, { - "left": "issue:44368", - "right": "issue:45563", + "left": "issue:45685", + "right": "issue:46431", "accept": false, - "reason": "A warning about tie_word_embeddings is not the same as stale num_return_sequences behavior." + "reason": "MPS histogram/int support failure and missing chat template support are unrelated processor/model issues." }, { - "left": "issue:43065", - "right": "issue:43408", + "left": "issue:44987", + "right": "issue:45459", "accept": false, - "reason": "Both touch SAM3, but one is a dummy Conv2d placeholder and the other is a config/model-type mismatch warning." + "reason": "Both touch loading/tokenizer behavior, but one is a model load failure on v5.1 and the other is protobuf-import error masking; different root causes." }, { - "left": "issue:44188", - "right": "issue:46202", + "left": "issue:44466", + "right": "issue:45335", "accept": false, - "reason": "Attention-kernel divergence under torch.compile is unrelated to Gemma tokenizer roundtripping." + "reason": "Both involve embeddings/tied weights, but one is serialization inconsistency and the other is resize_token_embeddings not propagating to decoder embeddings." }, { - "left": "issue:43931", - "right": "issue:44368", + "left": "issue:45137", + "right": "issue:45600", "accept": false, - "reason": "Weight-shape mismatch during model load is a different issue from the embedding warning." + "reason": "DeepSpeed deque exhaustion and stale auto_mappings references are unrelated failures." }, { - "left": "issue:44877", - "right": "issue:46170", + "left": "issue:44995", + "right": "issue:45200", "accept": false, - "reason": "Strict config loading and Windows ROCm/torch.distributed import problems are unrelated." + "reason": "A stale-cache second-forward crash is not the same as needing default mm_token_type_ids for text-only fine-tuning." }, { - "left": "issue:45987", - "right": "issue:46082", + "left": "issue:43475", + "right": "issue:43540", "accept": false, - "reason": "Cache length return type and LlamaConfig head_dim validation are separate bugs." + "reason": "Different model families and different failures: missing vision encoder output field versus video-input processing ValueError." }, { - "left": "issue:44188", - "right": "issue:46421", + "left": "issue:44060", + "right": "issue:44077", "accept": false, - "reason": "Both mention caching/compile, but one is kernel divergence and the other is lazy cache init causing recompilation." + "reason": "Tied-weights warning corruption and optional post_init policy are not the same issue." }, { - "left": "issue:43065", - "right": "issue:46123", + "left": "issue:44479", + "right": "issue:46170", "accept": false, - "reason": "Dummy layer definition in SAM3 is unrelated to a pipeline batch finalization bug." + "reason": "A multi-model video regression and a Windows/ROCm import dependency problem are unrelated." }, { - "left": "issue:45588", - "right": "issue:46170", + "left": "issue:43653", + "right": "issue:44162", "accept": false, - "reason": "Flash-attention sink handling and Windows ROCm import dependency issues do not match." + "reason": "Tokenizer special-token registration bug versus broad ESM2 breakage; too different to merge." }, { - "left": "issue:44297", - "right": "issue:44623", + "left": "issue:45397", + "right": "issue:45656", "accept": false, - "reason": "Both are save_pretrained-related, but one is tokenizer_class metadata and the other is missing processor files." + "reason": "Gemma-4 ZeRO3 loading failure and double optimizer-step behavior under deepspeed are different code-path problems." }, { - "left": "issue:44361", - "right": "issue:44749", + "left": "issue:42886", + "right": "issue:45229", "accept": false, - "reason": "Tokenizer AttributeError and a performance regression after upgrade are different bugs." + "reason": "Offline tokenizer cache loading and Gemma4 multi-GPU OOM are not the same underlying bug." }, { - "left": "issue:43646", - "right": "issue:44162", + "left": "issue:44368", + "right": "issue:44977", "accept": false, - "reason": "Custom model initialization regression and ESM2 breakage are not the same concrete failure." + "reason": "Both mention Qwen3.5, but one is a tied-weights warning and the other is a flash-attention generation regression." }, { - "left": "issue:44749", - "right": "issue:45656", + "left": "issue:43906", + "right": "issue:46202", "accept": false, - "reason": "Data filtering slowdown and deepspeed double optimizer step are unrelated." + "reason": "One is an isolated reproduction of an external issue; the other is a Gemma pipeline roundtrip failure on a specific token. Not enough evidence of a shared root cause." }, { - "left": "issue:44623", - "right": "issue:45656", + "left": "issue:44483", + "right": "issue:46202", "accept": false, - "reason": "Processor save_pretrained file emission and optimizer stepping behavior are different code paths." + "reason": "Chat-completions API rejection and Gemma token roundtrip failure are unrelated." }, { - "left": "issue:43232", - "right": "issue:43673", + "left": "issue:44995", + "right": "issue:45588", "accept": false, - "reason": "Both are generation/cache related, but they describe different cache/update bugs." + "reason": "Stale indexer cache on second forward is not the same as flash_attention crash on sink-less models." }, { - "left": "issue:44749", - "right": "issue:45310", + "left": "issue:44368", + "right": "issue:45563", "accept": false, - "reason": "Performance regression in filtering vs Qwen3.5 MoE from_pretrained failure are not duplicates." + "reason": "A tie_word_embeddings warning and a stale num_return_sequences warning are different warning sites with different triggers." }, { - "left": "issue:44869", - "right": "issue:45356", + "left": "issue:43065", + "right": "issue:43408", "accept": false, - "reason": "Whisper timestamp decoding crash and Kimi tokenizer regression are unrelated." + "reason": "A dummy Conv2d implementation and a model-type mismatch warning are not the same bug." }, { - "left": "issue:44871", - "right": "issue:46132", + "left": "issue:44188", + "right": "issue:46202", "accept": false, - "reason": "EOS token config inconsistency and AttentionInterface.register behavior change are different issues." + "reason": "torch.compile branch divergence in attention kernels is unrelated to Gemma tokenizer roundtrip failure." }, { - "left": "issue:42994", - "right": "issue:43482", + "left": "issue:43931", + "right": "issue:44368", "accept": false, - "reason": "Quantized model saving failure and GGUF loading failure are not the same bug." + "reason": "Weight-shape mismatch loading Qwen3-VL-30B is unrelated to the Qwen3.5 tie_word_embeddings warning." }, { - "left": "issue:45230", - "right": "issue:45397", + "left": "issue:44877", + "right": "issue:46170", "accept": false, - "reason": "Generic bug report and gemma-4 zero3 from_pretrained failure are too unspecific to be duplicates." + "reason": "Strict config loading for granite_speech and a Windows ROCm import problem do not share a code path." }, { - "left": "issue:43475", - "right": "issue:43643", + "left": "issue:45987", + "right": "issue:46082", "accept": false, - "reason": "SAM3 vision encoder attribute error and trust_remote_code config field loss are unrelated." + "reason": "StaticCache return-type contract and LlamaConfig head_dim validation are different API contract issues." }, { - "left": "issue:44038", - "right": "issue:44623", + "left": "issue:43065", + "right": "issue:46123", "accept": false, - "reason": "Qwen3-VL-Moe loading and processor save_pretrained file output are different failures." + "reason": "Sam3PixelDecoder dummy Conv2d and MaskGenerationPipeline partial-batch handling are unrelated." }, { - "left": "issue:28282", - "right": "issue:30990", + "left": "issue:44188", + "right": "issue:46421", "accept": false, - "reason": "PyTorch-missing ImportError and Sentence Transformers loading hang are distinct problems." + "reason": "Kernel divergence under torch.compile is not the same as cache lazy-init causing recompilation." }, { - "left": "issue:42371", - "right": "issue:45092", + "left": "issue:45584", + "right": "issue:46170", "accept": false, - "reason": "TF32 settings guidance and InternVL2 meta-initialization incompatibility are unrelated." + "reason": "Whisper empty-transcription failure after align_special_tokens and a Windows/ROCm dependency issue are unrelated." }, { - "left": "issue:43475", - "right": "issue:45381", + "left": "issue:44297", + "right": "issue:44623", "accept": false, - "reason": "SAM3 video encoder attribute error and Qwen2.5-VL vision_position_ids bug are different model-specific issues." + "reason": "Both are save_pretrained-related, but one is a tokenizer_config mismatch and the other is missing processor files; different failure modes." } ] }, @@ -12684,7 +12648,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 41, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -12692,307 +12656,288 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:28282", "issue:30990", - "issue:35141", - "issue:36010", "issue:41628", - "issue:42175", + "issue:42371", "issue:42673", - "issue:42831", - "issue:42898", "issue:42994", "issue:43232", - "issue:43278", - "issue:43377", - "issue:43504", + "issue:43475", + "issue:43482", "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43701", "issue:43704", "issue:43723", "issue:43784", - "issue:43824", "issue:43873", - "issue:43976", "issue:44038", + "issue:44162", "issue:44188", - "issue:44222", - "issue:44242", "issue:44351", - "issue:44496", - "issue:44568", - "issue:44792", + "issue:44361", + "issue:44623", + "issue:44749", + "issue:44869", + "issue:44871", "issue:44933", "issue:44945", "issue:44964", "issue:45081", + "issue:45092", "issue:45103", "issue:45137", + "issue:45230", "issue:45310", + "issue:45356", "issue:45375", + "issue:45381", + "issue:45397", "issue:45406", "issue:45478", - "issue:45561", - "issue:45600", - "issue:45712", + "issue:45656", "issue:45753", - "issue:46077", - "issue:46123", + "issue:46132", "issue:46202", "issue:46326", "issue:46399" ], "soft_pairs": [ + "issue:44361|issue:44749", + "issue:43646|issue:44162", + "issue:44749|issue:45656", + "issue:44623|issue:45656", + "issue:43232|issue:43673", + "issue:43701|issue:44749", + "issue:44869|issue:45356", + "issue:44871|issue:46132", + "issue:42994|issue:43482", + "issue:45230|issue:45397", + "issue:43475|issue:43643", + "issue:28282|issue:30990", + "issue:44038|issue:44623", + "issue:42371|issue:45092", + "issue:43475|issue:45381", "issue:41628|issue:44933", "issue:45375|issue:45478", "issue:45406|issue:46202", - "issue:45310|issue:45375", - "issue:43723|issue:46326", "issue:42994|issue:43525", "issue:43784|issue:46399", + "issue:45310|issue:45375", + "issue:43723|issue:46326", "issue:42673|issue:43704", "issue:43873|issue:44945", - "issue:43232|issue:44188", "issue:44964|issue:45137", + "issue:43232|issue:44188", "issue:45081|issue:45103", "issue:41628|issue:44351", - "issue:44945|issue:45753", - "issue:44222|issue:45310", - "issue:42994|issue:43824", - "issue:44038|issue:45310", - "issue:43232|issue:44792", - "issue:36010|issue:41628", - "issue:42831|issue:44222", - "issue:46077|issue:46123", - "issue:42175|issue:45712", - "issue:44496|issue:45561", - "issue:43504|issue:45561", - "issue:45561|issue:45600", - "issue:42175|issue:44933", - "issue:43976|issue:44933", - "issue:43377|issue:44242", - "issue:30990|issue:35141", - "issue:43278|issue:45137", - "issue:42898|issue:44568" + "issue:44945|issue:45753" ], - "prepared_review_unit_hash": "5b78ae1e185481f287e119ca32c9176666e70c49e68f61ff5b689eab3544d738", + "prepared_review_unit_hash": "30fa43a485a9b3ad6a723e37dd27f36f51d3474fb73932e83c6c8e3bf92a3db1", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 36334, - "estimated_input_tokens": 9084, - "estimated_eval_tokens": 18424 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36131, + "estimated_input_tokens": 9033, + "estimated_eval_tokens": 18322 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The packet contains two clear duplicate issue pairs overall: the surfaced soft-edge pair 42673/43704, and an additional exact-title pair 45310/45478 that also looks like a true duplicate even though it was not included in the soft edges. Most other candidates share a subsystem or model family but do not look like the same concrete bug.", + "summary": "Mostly conservative triage: the only clear duplicate in this packet is issue:42673 / issue:43704, which are the same VRAM-leak report for the same model and trigger. The other soft pairs look related by subsystem or version, but not identical enough to merge.", "confidence": 0.82, - "canonical_issue_reason": "issue:42673 is the best canonical issue because it is an exact duplicate anchor, older than 43704, and has the richer discussion record. It also represents one of the two clear duplicate pairs visible in the packet.", + "canonical_issue_reason": "issue:42673 is the strongest canonical issue because it is a specific report of the Qwen3ForCausalLM VRAM leak under multiple dataloader threads, and issue:43704 is the same bug with the same title and trigger.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the strongest overall issue candidate for triage because it is a clean duplicate representative with the most context. The packet also contains another exact-title duplicate pair, 45310/45478, but 42673 is still the better canonical anchor.", + "best_issue_reason": "issue:42673 is the best issue anchor for triage because it is the clearest, most specific instance of the duplicate VRAM-leak report and cleanly represents issue:43704.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:41628", - "right": "issue:44933", + "left": "issue:44361", + "right": "issue:44749", "accept": false, - "reason": "Both are import errors, but they target different missing symbols in different modules, so they do not look like the same bug." + "reason": "Different failures: tokenizer/task AttributeError vs a performance regression after upgrade." }, { - "left": "issue:45375", - "right": "issue:45478", + "left": "issue:43646", + "right": "issue:44162", "accept": false, - "reason": "Same model family, but one is about a missing config field and the other is a from_pretrained loading error; not the same concrete defect." + "reason": "Both are v5 breakages, but they affect different models and different failure modes." }, { - "left": "issue:45406", - "right": "issue:46202", + "left": "issue:44749", + "right": "issue:45656", "accept": false, - "reason": "Both involve Gemma-related loading/tokenization behavior, but the symptoms are different enough that they do not look like one bug." + "reason": "Unrelated bugs: data filtering slowdown vs optimizer step being called twice under deepspeed." }, { - "left": "issue:45310", - "right": "issue:45375", + "left": "issue:44623", + "right": "issue:45656", "accept": false, - "reason": "These mention the same general Qwen3.5 MoE area, but one is a from_pretrained error and the other is a missing vision config field." + "reason": "Saving processor files and duplicate optimizer stepping are different code paths." }, { - "left": "issue:43723", - "right": "issue:46326", + "left": "issue:43232", + "right": "issue:43673", "accept": false, - "reason": "Tokenizer loading in v5 and continue_final_message for mistral_common are different code paths and symptoms." + "reason": "Both are generation-related, but the reported crashes are different and not the same concrete bug." }, { - "left": "issue:42994", - "right": "issue:43525", + "left": "issue:43701", + "right": "issue:44749", "accept": false, - "reason": "One is a quantized model saving failure, the other is a missing config attribute on Llama4Config; they are not the same bug." + "reason": "Checkpoint resume key mismatch is unrelated to the slowdown report." }, { - "left": "issue:43784", - "right": "issue:46399", + "left": "issue:44869", + "right": "issue:45356", "accept": false, - "reason": "An import-time NameError in sentence-transformers and a Qwen3.5 kernel-registration crash are unrelated failure modes." + "reason": "Different tokenizer/decoding bugs with different models and triggers." }, { - "left": "issue:42673", - "right": "issue:43704", - "accept": true, - "reason": "Exact same VRAM-leak report for Qwen3ForCausalLM in multiple dataloader threads." + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "EOS config mismatch is unrelated to AttentionInterface.register behavior." }, { - "left": "issue:43873", - "right": "issue:44945", + "left": "issue:42994", + "right": "issue:43482", "accept": false, - "reason": "Quantization/offloading behavior and pipeline-parallel output corruption are different runtime problems." + "reason": "Quantized model saving and GGUF loading are separate problems." }, { - "left": "issue:43232", - "right": "issue:44188", - "accept": false, - "reason": "Generation kwargs handling after sync_gpus is not the same issue as attention-kernel divergence under torch.compile." - }, - { - "left": "issue:44964", - "right": "issue:45137", - "accept": false, - "reason": "Model loading failure for a multimodal checkpoint and a DeepSpeed ZeRO3 deque error are different defects." - }, - { - "left": "issue:45081", - "right": "issue:45103", - "accept": false, - "reason": "A tokenizer regex patch crash and an auto_docstring AttributeError involve different subsystems." - }, - { - "left": "issue:41628", - "right": "issue:44351", + "left": "issue:45230", + "right": "issue:45397", "accept": false, - "reason": "Both are import errors, but they concern different exported names and do not point to the same failure path." + "reason": "Generic placeholder issue vs a specific Gemma-4 ZeRO3 from_pretrained failure." }, { - "left": "issue:44945", - "right": "issue:45753", + "left": "issue:43475", + "right": "issue:43643", "accept": false, - "reason": "Pipeline-parallel output issues and an infinite loop on a specific image are not the same bug." + "reason": "Both mention missing fields, but they are different objects and different failure locations." }, { - "left": "issue:44222", - "right": "issue:45310", + "left": "issue:28282", + "right": "issue:30990", "accept": false, - "reason": "Both mention Qwen3.5/MoE/FP8-adjacent areas, but one is FP8 save_pretrained and the other is a from_pretrained loading error." + "reason": "ImportError about missing PyTorch is not the same as a loading hang." }, { - "left": "issue:42994", - "right": "issue:43824", + "left": "issue:44038", + "right": "issue:44623", "accept": false, - "reason": "Quantized model saving and a missing Qwen2.5-VL import are unrelated issues." + "reason": "Qwen3-VL-Moe v5 bug is unrelated to processor.save_pretrained missing files." }, { - "left": "issue:44038", - "right": "issue:45310", + "left": "issue:42371", + "right": "issue:45092", "accept": false, - "reason": "Same broader Qwen3 family, but the model/version and failure mode differ enough that this is not clearly one bug." + "reason": "TF32 API guidance is unrelated to InternVL2 meta-initialization incompatibility." }, { - "left": "issue:43232", - "right": "issue:44792", + "left": "issue:43475", + "right": "issue:45381", "accept": false, - "reason": "Generation kwargs handling and a janus image-generation test failure do not share the same concrete defect." + "reason": "Different multimodal model bugs with different symptoms and likely different fixes." }, { - "left": "issue:36010", - "right": "issue:41628", + "left": "issue:41628", + "right": "issue:44933", "accept": false, - "reason": "Both are import regressions, but they affect different APIs and different missing names." + "reason": "Both are import-related, but they concern different missing symbols and different modules." }, { - "left": "issue:42831", - "right": "issue:44222", + "left": "issue:45375", + "right": "issue:45478", "accept": false, - "reason": "Both involve FP8, but one is an accuracy issue and the other is a save_pretrained issue; not the same bug." + "reason": "Same model family, but one is a dropped config field and the other is a from_pretrained error; not the same bug." }, { - "left": "issue:46077", - "right": "issue:46123", + "left": "issue:45406", + "right": "issue:46202", "accept": false, - "reason": "Bidirectional attention mask construction and final-partial-batch result dropping are unrelated behaviors." + "reason": "Gemma processor _tokenizer crash and Gemma pipeline roundtrip failure are distinct issues." }, { - "left": "issue:42175", - "right": "issue:45712", + "left": "issue:42994", + "right": "issue:43525", "accept": false, - "reason": "A packaging/backend dependency issue and leftover dummy-class cleanup are different maintenance problems." + "reason": "Quantized saving failure is unrelated to missing pad_token_id on Llama4Config." }, { - "left": "issue:44496", - "right": "issue:45561", + "left": "issue:43784", + "right": "issue:46399", "accept": false, - "reason": "Model config loading failure and a pytest-xdist file race do not share the same code path." + "reason": "Different packages and failures: import-time nn error vs rotary-pos-emb registration error." }, { - "left": "issue:43504", - "right": "issue:45561", + "left": "issue:45310", + "right": "issue:45375", "accept": false, - "reason": "Legacy-field preset loading and xdist capture-file racing are unrelated issues." + "reason": "Related model family, but not the same concrete failure." }, { - "left": "issue:45561", - "right": "issue:45600", + "left": "issue:43723", + "right": "issue:46326", "accept": false, - "reason": "A testing-utils race and stale auto_mappings references are different bugs." + "reason": "Generic AutoTokenizer loading issue and mistral_common continue_final_message bug are separate." }, { - "left": "issue:42175", - "right": "issue:44933", - "accept": false, - "reason": "Backend dependency installation and an image_utils import exposure problem are unrelated." + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same title, same model, same VRAM leak under multiple dataloader threads; clearly duplicate reports." }, { - "left": "issue:43976", - "right": "issue:44933", + "left": "issue:43873", + "right": "issue:44945", "accept": false, - "reason": "Python-version compatibility and an import exposure error are different failure modes." + "reason": "Quantization offloading behavior is unrelated to pipeline parallelism output corruption." }, { - "left": "issue:43377", - "right": "issue:44242", + "left": "issue:44964", + "right": "issue:45137", "accept": false, - "reason": "Missing padding-mask support and omitted load-balancing loss are different model behaviors." + "reason": "Phi-4 multimodal loading failure is unrelated to the DeepSpeed deque error." }, { - "left": "issue:30990", - "right": "issue:35141", + "left": "issue:45081", + "right": "issue:45103", "accept": false, - "reason": "SentenceTransformers loading stall and embedding reinitialization after resize are unrelated." + "reason": "Tokenizer regex crash and auto-docstring annotation crash are unrelated." }, { - "left": "issue:43278", - "right": "issue:45137", + "left": "issue:41628", + "right": "issue:44351", "accept": false, - "reason": "Training/eval dtype mismatch and a ZeRO3 deque underflow are different failures." + "reason": "Both are missing imports, but they are different missing symbols with no clear shared code path." }, { - "left": "issue:42898", - "right": "issue:44568", + "left": "issue:44945", + "right": "issue:45753", "accept": false, - "reason": "Both are tokenizer behavior changes in v5, but one is cleanup-space semantics and the other is BOS/EOS insertion." + "reason": "Pipeline parallelism output corruption is not the same as an infinite loop on a specific image." } ] }, "evaluator_result": { "accept": false, - "feedback": "The summary overstates uniqueness: the packet contains another exact-title pair, issue:45310 and issue:45478, which also looks like a duplicate candidate. Because the analyst only accepted 42673/43704 and then claimed there was \u2018only one pair\u2019 that looked like a true duplicate, the triage conclusion is not fully grounded. Several soft-edge rejections are plausible, but a few reasons go beyond the packet\u2019s evidence by asserting concrete defect differences that are not visible here." + "feedback": "The duplicate-pair conclusion for issue:42673 / issue:43704 is grounded, but the canonical_issue_reason overstates the packet by claiming \u201cmore discussion/inbound references\u201d when both issues have inbound_references = 0. That unsupported detail is enough to reject. The rest of the soft-edge verdicts look conservative overall." }, "evaluator_used": true, "retried": true, @@ -13005,7 +12950,7 @@ "component_index": 1, "component_count": 20, "review_unit_index": 42, - "review_unit_count": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -13013,55 +12958,72 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:22355", - "issue:27957", - "issue:28282", "issue:30333", + "issue:30990", "issue:33290", - "issue:34689", "issue:35141", - "issue:36331", + "issue:36010", + "issue:41628", + "issue:42175", "issue:42222", - "issue:42796", "issue:42831", "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43377", "issue:43479", - "issue:43644", + "issue:43504", "issue:43723", "issue:43784", "issue:43824", "issue:43874", + "issue:43976", "issue:44016", - "issue:44060", - "issue:44075", + "issue:44038", "issue:44079", "issue:44222", - "issue:44295", - "issue:44297", - "issue:44448", + "issue:44242", + "issue:44496", "issue:44568", "issue:44623", "issue:44734", "issue:44779", + "issue:44792", "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:45561", + "issue:45600", "issue:45712", "issue:45812", - "issue:46132", - "issue:46209", - "issue:46231", - "issue:46431" + "issue:46077", + "issue:46123", + "issue:46231" ], "soft_pairs": [ + "issue:42994|issue:43824", + "issue:44222|issue:45310", + "issue:44038|issue:45310", + "issue:36010|issue:41628", + "issue:43232|issue:44792", + "issue:42831|issue:44222", + "issue:46077|issue:46123", + "issue:42175|issue:45712", + "issue:44496|issue:45561", + "issue:43504|issue:45561", + "issue:42175|issue:44933", + "issue:45561|issue:45600", + "issue:43976|issue:44933", + "issue:43278|issue:45137", + "issue:43377|issue:44242", + "issue:30990|issue:35141", + "issue:42898|issue:44568", "issue:30333|issue:33290", "issue:44623|issue:45310", "issue:43479|issue:44869", @@ -13072,250 +13034,536 @@ "issue:44016|issue:46231", "issue:43874|issue:45092", "issue:43784|issue:43824", - "issue:44734|issue:45103", "issue:42898|issue:44779", - "issue:42898|issue:45356", - "issue:44295|issue:45656", - "issue:42898|issue:44448", - "issue:44060|issue:45125", - "issue:44075|issue:46231", - "issue:45405|issue:45712", - "issue:44222|issue:44297", - "issue:44933|issue:45712", - "issue:44297|issue:45656", - "issue:27957|issue:30333", - "issue:30333|issue:34689", - "issue:44448|issue:46132", - "issue:42796|issue:43723", - "issue:44222|issue:44623", - "issue:42831|issue:44623", - "issue:43479|issue:43644", - "issue:30333|issue:36331", - "issue:30333|issue:35141", - "issue:22355|issue:28282", - "issue:46209|issue:46431", - "issue:45245|issue:46132" + "issue:44734|issue:45103", + "issue:42898|issue:45356" ], - "prepared_review_unit_hash": "811e1ffa6f9c6d4e5fde23a4257ae52c9c7ee6e6e404bec18362b5dc3dc96a94", + "prepared_review_unit_hash": "c4ebd2200a5161d469ac07be906636e650c2d6f8ca69df2c4f21d9eb9e8f4a09", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 37096, - "estimated_input_tokens": 9274, - "estimated_eval_tokens": 18804 + "soft_pair_count": 30, + "serialized_chars": 36211, + "estimated_input_tokens": 9053, + "estimated_eval_tokens": 18362 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a heterogeneous mix of unrelated Transformers issues: tokenizer regressions, model-loading/import errors, training/runtime bugs, docs/notebook problems, and FP8/deepspeed edge cases. The soft-similarity links do not indicate true duplicates, so all soft edges should be rejected.", - "confidence": 0.85, - "canonical_issue_reason": "issue:45812 is the clearest high-impact, well-scoped bug report in the set: a broad silent tokenizer regression affecting all Granite models, with a concrete symptom and likely actionable fix.", + "summary": "These are mostly isolated bug reports across different models, APIs, and failure modes. I did not find any soft-edge pair that clearly describes the same underlying defect closely enough to merge as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45812 is the strongest canonical issue candidate: it describes a broad, concrete, reproducible tokenizer regression with a clear v4\u2192v5 delta and wide user impact.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the strongest standalone issue for triage because it describes a clear, broadly affecting regression with strong user impact and a specific failure mode.", + "best_issue_reason": "issue:45812 is the best standalone issue to anchor triage because it is specific, high-signal, and clearly frames the regression for maintainers.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving failure and a Qwen2.5-VL import error are unrelated failure modes in different code paths." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "Both mention MoE/FP8, but one is save_pretrained on FP8 MoE and the other is from_pretrained loading for Qwen3.5; not the same bug." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen variants and different symptoms; this looks like separate model-specific regression reports." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Import errors for GenerationMixin and AutoImageProcessor are distinct missing-export issues." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation KV-sync bug and a Janus image-generation test failure do not point to the same code-path problem." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "FP8 accuracy degradation is not the same as FP8 MoE save_pretrained failure." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Bidirectional-attention masking and partial-batch result dropping are unrelated pipeline/model behaviors." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Backend dependency packaging and leftover dummy-class cleanup are unrelated maintenance issues." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Model config loading failure and a pytest-xdist file race are completely different bugs." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Legacy-field pretrained load failure and a parallel test race are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Missing TensorFlow backend dependency and a nonexistent image_utils import are separate packaging/import problems." + }, + { + "left": "issue:45561", + "right": "issue:45600", + "accept": false, + "reason": "A test-worker race on captured_info.txt is not the same as stale auto_mappings references breaking CI." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python version compatibility and an import-from-image_utils error are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift in evaluation and DeepSpeed ZeRO3 deque underflow are different runtime issues." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Missing padding-mask support in an encoder and missing load-balancing loss in MoE routing are different defects." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "A loading hang in Sentence Transformers and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Tokenization-space cleanup behavior and add_special_tokens not inserting BOS/EOS are distinct tokenizer behaviors." + }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLflow failure reporting vs deepspeed Adafactor OOM are unrelated training bugs with different symptoms and code paths." + "reason": "MLflow status reporting after failed training and an Adafactor OOM in DeepSpeed are unrelated." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "processor.save_pretrained missing files and Qwen3.5 Moe from_pretrained failures are different APIs and different regressions." + "reason": "Processor save_pretrained missing files and Qwen3.5 MoE from_pretrained loading failure are different save/load paths." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Phi4MultimodalConfig default-init bug and Whisper timestamp decode crash are unrelated model/config vs decoding issues." + "reason": "Default config initialization and Whisper word-timestamp decoding crashes are unrelated." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "VitPose model breakage and Gemma 4 mm_token_type_ids defaulting are different model-family problems." + "reason": "Broken ViTPose model paths and Gemma4 mm_token_type_ids defaults are separate model-specific issues." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Config initialization with None and ModelOutput key assignment are unrelated code paths." + "reason": "Config defaulting and ModelOutput key assignment are different mechanisms with no clear shared root cause." }, { "left": "issue:42898", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different behavior and different model families; not the same underlying bug." + "reason": "Both are tokenizer regressions, but one concerns whitespace cleanup while the other concerns wrong token IDs for Granite models; not the same bug." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "Tokenizer loading failure and missing BOS/EOS insertion are distinct tokenizer bugs with different fixes." + "reason": "General AutoTokenizer loading issues and a specific mDeBERTa special-tokens regression are not the same concrete defect." }, { "left": "issue:44016", "right": "issue:46231", "accept": false, - "reason": "Notebook syntax error and documentation parameter mismatch are not the same issue." + "reason": "Notebook syntax error and a docstring parameter-name mismatch are unrelated documentation issues." }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Missing image-processor method and remote-code checkpoint/meta-init incompatibility are different multimodal failures." + "reason": "Missing image-patch count method in GLM46V and old InternVL2 meta-init incompatibility are different multimodal bugs." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Sentence-transformers import NameError and missing Qwen2.5-VL import are separate import/dependency problems." + "reason": "Sentence-Transformers import NameError and Qwen2.5-VL import failure are unrelated import problems." }, { - "left": "issue:44734", - "right": "issue:45103", + "left": "issue:42898", + "right": "issue:44779", "accept": false, - "reason": "KV-cache continuation crash in serve and auto_docstring annotation crash are unrelated subsystems." + "reason": "Different tokenizer regressions affecting different behaviors and models; too broad to treat as one bug." }, { - "left": "issue:42898", - "right": "issue:44779", + "left": "issue:44734", + "right": "issue:45103", "accept": false, - "reason": "Both mention v5 tokenizer regressions, but the affected models and failure mechanisms differ; not mergeable as one fix." + "reason": "Serving KV-cache tensor indexing and auto_docstring future-annotations handling are unrelated crashes." }, { "left": "issue:42898", "right": "issue:45356", "accept": false, - "reason": "clean_up_tokenization_spaces behavior and Kimi codec-handling regression are distinct tokenizer issues." - }, + "reason": "Whitespace cleanup/tokenization behavior and Kimi-K2.5 codec handling are distinct tokenizer regressions, not one underlying 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": 20, + "review_unit_index": 43, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28282", + "issue:30333", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:41628", + "issue:42757", + "issue:42796", + "issue:42831", + "issue:42898", + "issue:43122", + "issue:43232", + "issue:43475", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:44060", + "issue:44075", + "issue:44206", + "issue:44222", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44448", + "issue:44484", + "issue:44623", + "issue:44933", + "issue:44977", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45245", + "issue:45310", + "issue:45381", + "issue:45405", + "issue:45656", + "issue:45712", + "issue:46082", + "issue:46132", + "issue:46209", + "issue:46231", + "issue:46305", + "issue:46424", + "issue:46431" + ], + "soft_pairs": [ + "issue:44295|issue:45656", + "issue:44075|issue:46231", + "issue:44060|issue:45125", + "issue:42898|issue:44448", + "issue:45405|issue:45712", + "issue:44222|issue:44297", + "issue:44933|issue:45712", + "issue:27957|issue:30333", + "issue:30333|issue:34689", + "issue:44297|issue:45656", + "issue:44448|issue:46132", + "issue:42796|issue:43723", + "issue:44222|issue:44623", + "issue:43479|issue:43644", + "issue:30333|issue:36331", + "issue:42831|issue:44623", + "issue:30333|issue:35141", + "issue:22355|issue:28282", + "issue:45245|issue:46132", + "issue:46209|issue:46431", + "issue:41628|issue:42757", + "issue:43475|issue:46305", + "issue:43122|issue:46209", + "issue:44297|issue:45310", + "issue:43723|issue:44977", + "issue:45381|issue:46305", + "issue:45071|issue:46082", + "issue:43232|issue:46424", + "issue:44206|issue:45103", + "issue:44263|issue:45092", + "issue:44222|issue:44484" + ], + "prepared_review_unit_hash": "8e09a453ea970d9f13e884328d85112f0f23afda798be6d19a028285fa32bd3a", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36223, + "estimated_input_tokens": 9056, + "estimated_eval_tokens": 18368 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a collection of unrelated Transformers issues with only superficial topical overlap. None of the soft candidate pairs look like the same underlying bug or change, so all soft edges are rejected. No PRs are present.", + "confidence": 0.93, + "canonical_issue_reason": "Issue #43723 is the broadest user-facing regression in the set (tokenizer loading in v5) and is the closest thing to a representative API-breakage report among otherwise disparate issues.", + "canonical_pr_reason": null, + "best_issue_reason": "#43723 is the best representative because it is concrete, widely applicable, and sits near several other tokenizer/import-related reports without being overly model-specific.", + "best_pr_reason": null, + "soft_edge_verdicts": [ { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "position_ids buffer read error and double optimizer-step in deepspeed are unrelated runtime bugs." + "reason": "Unrelated bugs: buffer access for position_ids vs a double optimizer-step complaint under deepspeed." }, { - "left": "issue:42898", - "right": "issue:44448", + "left": "issue:44075", + "right": "issue:46231", "accept": false, - "reason": "Tokenizer cleanup behavior and Pegasus v4/v5 output differences are different regression reports." + "reason": "Different problem classes: SGD args not used vs documentation naming mismatch." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Incorrect tied-weights warning and missing tensor-parallel plan are separate Qwen model-class issues." + "reason": "Both are Qwen-related, but one is a tied-weights warning bug and the other is missing tensor-parallel metadata; different code paths." }, { - "left": "issue:44075", - "right": "issue:46231", + "left": "issue:42898", + "right": "issue:44448", "accept": false, - "reason": "SGD args ignored is a behavior bug; the other item is only a docs parameter-name mismatch." + "reason": "Both mention v5 output changes, but one is about clean_up_tokenization_spaces semantics and the other is a broader model output regression; not the same bug." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "PEFT version pinning and leftover dummy classes are unrelated maintenance problems." + "reason": "A dependency version bump issue and leftover dummy classes/check_repo failure are unrelated." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "FP8 save_pretrained behavior and tokenizer_class metadata mismatch are different save-pretrained bugs." + "reason": "Both involve save_pretrained, but one is FP8 MoE serialization and the other is tokenizer class metadata; different artifacts and fixes." }, { "left": "issue:44933", "right": "issue:45712", "accept": false, - "reason": "Nonexistent image_utils import and dummy-class cleanup are unrelated repository hygiene issues." + "reason": "Import from image_utils vs dummy PT class cleanup are unrelated maintenance issues." }, { "left": "issue:27957", "right": "issue:30333", "accept": false, - "reason": "Flash Attention 2 with XLMRoberta and MLflow failure reporting are entirely different subsystems." + "reason": "Flash Attention 2 model support vs MLFlow reporting on failed training jobs are unrelated." }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "MLflow reporting on failure and Llama 3.2 Vision model loading break are unrelated issues." + "reason": "Training-job tracking bug vs model-loading regression for a specific Llama checkpoint; different subsystems." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save metadata bug vs optimizer double-step bug; no shared underlying code-path." }, { "left": "issue:44448", "right": "issue:46132", "accept": false, - "reason": "Pegasus output regression and AttentionInterface.register behavior change are separate code paths." + "reason": "Versioned output mismatch for a model vs behavior change in AttentionInterface.register; different features." }, { "left": "issue:42796", "right": "issue:43723", "accept": false, - "reason": "Tokenizer BPE-marker handling and AutoTokenizer load failure are both tokenizer-related but not the same bug." + "reason": "Tokenizer BPE marker handling vs tokenizer loading failure are adjacent topics but different bugs." }, { "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "FP8 moe save_pretrained and processor.save_pretrained missing files involve different object types and failure modes." - }, - { - "left": "issue:42831", - "right": "issue:44623", - "accept": false, - "reason": "FP8 accuracy loss and processor save_pretrained missing files are unrelated." + "reason": "FP8 MoE model serialization vs processor save_pretrained missing files are different save paths and failure modes." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Default config initialization and non-persistent buffer junk filling are different initialization bugs." + "reason": "Config default initialization bug vs non-persistent buffer junk in v5; unrelated internals." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "MLflow job-failure reporting and CustomTrainer.compute_loss signature break are unrelated trainer bugs." + "reason": "MLFlow reporting on failure vs CustomTrainer signature mismatch; unrelated training issues." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FP8 accuracy issue vs processor serialization missing files; no shared bug." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "MLflow failure reporting and embedding resize reinitialization are unrelated issues." + "reason": "Failed-job tracking and embedding resize/reinit are unrelated." }, { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Missing transformers.onnx module and missing PyTorch for AutoModel are different ImportErrors with different causes." + "reason": "Importing transformers.onnx vs AutoModel requiring torch; both import-related but different missing-module dependencies." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "Category cardinality runtime error vs attention interface behavior change; unrelated." }, { "left": "issue:46209", "right": "issue:46431", "accept": false, - "reason": "FP8 inference on B200 and missing chat template in processor are unrelated issues." + "reason": "FP8 inference on B200 vs chat template availability error; different features and failure causes." }, { - "left": "issue:45245", - "right": "issue:46132", + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Missing AutoImageProcessor export vs missing is_offline_mode export; both import errors, but not the same symbol or fix." + }, + { + "left": "issue:43475", + "right": "issue:46305", + "accept": false, + "reason": "SAM 3 Video output attribute bug vs MoonshineStreaming mask off-by-one; unrelated model-specific issues." + }, + { + "left": "issue:43122", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer-output regression across versions vs FP8 inference failure on B200; different subsystems." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer class metadata mismatch vs Qwen3.5 MoE from_pretrained error; not the same serialization/loading bug." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Generic tokenizer loading issue vs Qwen3.5 flash-attention generation bug; unrelated." + }, + { + "left": "issue:45381", + "right": "issue:46305", "accept": false, - "reason": "Category-count runtime error and AttentionInterface.register behavior change are unrelated." + "reason": "qwen2.5-vl vision_position_ids bug vs streaming mask off-by-one; different models and code paths." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "PretrainedConfig type-checking regression vs LlamaConfig head_dim validation; different config behaviors." + }, + { + "left": "issue:43232", + "right": "issue:46424", + "accept": false, + "reason": "Generation sync_gpus state update bug vs cache max_length only honored for quantized cache; related to generation, but not the same defect." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Feature extractor argument crash vs auto-docstring annotation parsing crash; unrelated tooling bugs." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "torch.split return handling in an indexer vs remote-code checkpoint meta-init incompatibility; different components." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 MoE save_pretrained bug vs a question about max_shard_size defaults; not the same issue." } ] }, @@ -13330,8 +13578,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 43, - "review_unit_count": 54, + "review_unit_index": 44, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -13339,289 +13587,267 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:22355", + "issue:27957", "issue:30990", "issue:38175", + "issue:39401", "issue:41628", "issue:42757", + "issue:42796", "issue:42831", + "issue:42886", "issue:42907", + "issue:43066", "issue:43122", "issue:43232", - "issue:43475", + "issue:43408", "issue:43479", "issue:43644", "issue:43723", + "issue:43824", "issue:44038", "issue:44190", - "issue:44206", - "issue:44222", "issue:44261", "issue:44263", "issue:44279", "issue:44295", - "issue:44297", "issue:44360", "issue:44373", "issue:44483", - "issue:44484", "issue:44485", + "issue:44568", "issue:44623", "issue:44734", "issue:44829", "issue:44871", "issue:44913", - "issue:44977", - "issue:45071", "issue:45092", - "issue:45103", + "issue:45278", "issue:45310", "issue:45357", - "issue:45381", "issue:45406", "issue:45834", "issue:46002", "issue:46050", "issue:46082", - "issue:46209", "issue:46305", "issue:46328", - "issue:46421", - "issue:46424" + "issue:46402", + "issue:46421" ], "soft_pairs": [ - "issue:41628|issue:42757", - "issue:44297|issue:45310", - "issue:43475|issue:46305", - "issue:43122|issue:46209", - "issue:43723|issue:44977", - "issue:45381|issue:46305", - "issue:45071|issue:46082", - "issue:43232|issue:46424", - "issue:44206|issue:45103", - "issue:44222|issue:44484", - "issue:44263|issue:45092", "issue:44734|issue:46421", "issue:43232|issue:45834", "issue:44373|issue:44485", "issue:44295|issue:44623", "issue:44190|issue:46082", "issue:43723|issue:44871", - "issue:44483|issue:45406", "issue:44279|issue:46328", - "issue:44263|issue:44360", + "issue:44483|issue:45406", "issue:42907|issue:46050", + "issue:44263|issue:44360", "issue:30990|issue:38175", - "issue:44261|issue:45092", "issue:45092|issue:45357", + "issue:44261|issue:45092", "issue:43644|issue:46305", "issue:42831|issue:45310", "issue:44829|issue:46002", "issue:44263|issue:44373", + "issue:41628|issue:43824", + "issue:42831|issue:44038", "issue:43479|issue:44913", - "issue:42831|issue:44038" + "issue:43408|issue:46402", + "issue:22355|issue:27957", + "issue:42796|issue:43122", + "issue:39401|issue:44568", + "issue:42757|issue:45278", + "issue:42886|issue:43066" ], - "prepared_review_unit_hash": "db72b15cc94c5fb6e6a2c74034de18ab0918d772b570ff3a15dc35e8c9e05c83", + "prepared_review_unit_hash": "a62b760a6751ea86f163806f7f72524e37e80c427508a4c98b6efed0fc8befb7", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36396, - "estimated_input_tokens": 9099, - "estimated_eval_tokens": 18454 + "soft_pair_count": 26, + "serialized_chars": 35118, + "estimated_input_tokens": 8780, + "estimated_eval_tokens": 17816 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a true duplicate cluster: the items span many unrelated Transformers bugs and regressions (imports, tokenization, quantization, cache, generation, save/load, and model-specific failures). The soft pairs are mostly superficial keyword matches, not the same underlying bug or change.", - "confidence": 0.95, - "canonical_issue_reason": "No single issue is a valid canonical representative because the cluster contains unrelated reports rather than one duplicated defect.", + "summary": "The set is highly heterogeneous: import regressions, tokenizer behavior changes, serving crashes, config/load-save bugs, and quantization issues. None of the soft pairs looks like the same concrete underlying bug or change, so they should stay separate.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical duplicate fits this cluster; the closest broad umbrella is issue 45278 for upgrade-related import regressions, but it does not cover most other items.", "canonical_pr_reason": null, - "best_issue_reason": "No issue is a strong global canonical choice for the same reason: the items do not converge on one concrete bug or code path.", + "best_issue_reason": "Issue 45278 is the broadest representative among the listed issues, but only for a narrow import-regression subset; it is not a true cluster canonical.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:41628", - "right": "issue:42757", - "accept": false, - "reason": "Different import failures: AutoImageProcessor vs is_offline_mode; unrelated APIs and code paths." - }, - { - "left": "issue:44297", - "right": "issue:45310", - "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch vs Qwen3.5 from_pretrained failure; different layers of the load/save stack." - }, - { - "left": "issue:43475", - "right": "issue:46305", + "left": "issue:44734", + "right": "issue:46421", "accept": false, - "reason": "SAM 3 vision encoder attribute error vs Moonshine streaming mask off-by-one; unrelated models and bugs." + "reason": "Serving KV-cache continuation crash vs chunked-prefill cache lazy-init recompilation; different code paths." }, { - "left": "issue:43122", - "right": "issue:46209", + "left": "issue:43232", + "right": "issue:45834", "accept": false, - "reason": "Tokenizer version behavior change vs FP8 inference failure on B200; no shared underlying defect." + "reason": "Generation kwargs sync_gpus bug vs long OCR input index error; unrelated failures." }, { - "left": "issue:43723", - "right": "issue:44977", + "left": "issue:44373", + "right": "issue:44485", "accept": false, - "reason": "AutoTokenizer loading issue vs Qwen3.5 flash-attention generation bug; different failure modes." + "reason": "Docstring typo for position_ids vs GLM-5 RoPE implementation bug; not the same issue." }, { - "left": "issue:45381", - "right": "issue:46305", + "left": "issue:44295", + "right": "issue:44623", "accept": false, - "reason": "Qwen2.5-VL vision_position_ids bug vs Moonshine streaming window off-by-one; distinct mechanisms." + "reason": "Reading position_ids after buffer registration vs processor.save_pretrained missing files; different subsystems." }, { - "left": "issue:45071", + "left": "issue:44190", "right": "issue:46082", "accept": false, - "reason": "PretrainedConfig type-checking regression vs LlamaConfig head_dim validation; separate config behaviors." - }, - { - "left": "issue:43232", - "right": "issue:46424", - "accept": false, - "reason": "Generation sync_gpus kwargs bug vs cache max_cache_length semantics; different cache/generation paths." + "reason": "Image-classification script dataset loading vs LlamaConfig head_dim validation; unrelated." }, { - "left": "issue:44206", - "right": "issue:45103", + "left": "issue:43723", + "right": "issue:44871", "accept": false, - "reason": "LaserFeatureExtractor center-arg crash vs auto_docstring AttributeError from annotations; unrelated code." + "reason": "AutoTokenizer loading regression vs Gemma eos_token_id mismatch; different concrete bugs." }, { - "left": "issue:44222", - "right": "issue:44484", + "left": "issue:44279", + "right": "issue:46328", "accept": false, - "reason": "FP8 save_pretrained bug vs max_shard_size default question; not the same issue." + "reason": "General dependency issue vs SequenceFeatureExtractor.pad inefficiency; not duplicates." }, { - "left": "issue:44263", - "right": "issue:45092", + "left": "issue:44483", + "right": "issue:45406", "accept": false, - "reason": "torch.split return-value bug in a DSA indexer vs remote-code checkpoint meta-init incompatibility; unrelated." + "reason": "Chat-completions request rejection vs Gemma4Processor missing _tokenizer crash; different failures." }, { - "left": "issue:44734", - "right": "issue:46421", + "left": "issue:42907", + "right": "issue:46050", "accept": false, - "reason": "serve KV-cache indexing crash vs cache lazy-init recompilation; different cache problems." + "reason": "Dequantized Ministral save/load failure vs sinq quantized round-trip breakage; different quantization paths." }, { - "left": "issue:43232", - "right": "issue:45834", + "left": "issue:44263", + "right": "issue:44360", "accept": false, - "reason": "Generation kwargs handling after sync_gpus vs Kosmos2.5 long-OCR index error; no shared code path." + "reason": "Both touch the DSA indexer, but one is torch.split return handling and the other is a missing ReLU; separate bugs." }, { - "left": "issue:44373", - "right": "issue:44485", + "left": "issue:30990", + "right": "issue:38175", "accept": false, - "reason": "Docstring error for position_ids vs GLM-5 RoPE implementation discussion; documentation vs implementation." + "reason": "SentenceTransformers loading hang vs siglip2 zero probabilities; unrelated." }, { - "left": "issue:44295", - "right": "issue:44623", + "left": "issue:45092", + "right": "issue:45357", "accept": false, - "reason": "position_ids buffer read error vs processor.save_pretrained missing files; separate serialization issues." + "reason": "Remote-code meta-init incompatibility vs Qwen3.5 save_pretrained visual-key regression; different issues." }, { - "left": "issue:44190", - "right": "issue:46082", + "left": "issue:44261", + "right": "issue:45092", "accept": false, - "reason": "Example script cannot load local dataset vs LlamaConfig head_dim validation; unrelated problems." + "reason": "MLA layernorm eps precision error vs InternVL2 meta-init incompatibility; not the same bug." }, { - "left": "issue:43723", - "right": "issue:44871", + "left": "issue:43644", + "right": "issue:46305", "accept": false, - "reason": "AutoTokenizer.from_pretrained loading issue vs Gemma-3 eos_token_id config mismatch; different tokenizer/config bugs." + "reason": "Non-persistent buffer junk fill vs sliding_window_mask off-by-one; unrelated." }, { - "left": "issue:44483", - "right": "issue:45406", + "left": "issue:42831", + "right": "issue:45310", "accept": false, - "reason": "Chat completions request rejection vs Gemma4Processor missing _tokenizer; unrelated server/processor failures." + "reason": "FineGrainedFP8 accuracy problem vs Qwen3.5 MoE from_pretrained error; different problems." }, { - "left": "issue:44279", - "right": "issue:46328", + "left": "issue:44829", + "right": "issue:46002", "accept": false, - "reason": "Generic dependency issue vs padding inefficiency in SequenceFeatureExtractor; not the same bug." + "reason": "Flash-attention-3 training degeneration vs NaNs after checkout/init; not the same concrete failure." }, { "left": "issue:44263", - "right": "issue:44360", + "right": "issue:44373", "accept": false, - "reason": "torch.split return values vs missing ReLU in DSA indexer; distinct implementations and symptoms." + "reason": "DSA indexer runtime bug vs wrong docstring for position_ids; one is documentation only." }, { - "left": "issue:42907", - "right": "issue:46050", + "left": "issue:41628", + "right": "issue:43824", "accept": false, - "reason": "Failed dequantized save on Ministrals/Devstrals vs sinq quantized model round-trip breakage; different quantization paths." + "reason": "Missing AutoImageProcessor import vs missing Qwen2_5_VLForConditionalGeneration import; distinct import errors." }, { - "left": "issue:30990", - "right": "issue:38175", + "left": "issue:42831", + "right": "issue:44038", "accept": false, - "reason": "Sentence Transformers loading hang vs zero probabilities in siglip2-base-patch16-224; unrelated model behaviors." + "reason": "FP8 accuracy regression vs Qwen3-VL-Moe bug; different model/features." }, { - "left": "issue:44261", - "right": "issue:45092", + "left": "issue:43479", + "right": "issue:44913", "accept": false, - "reason": "MLA layernorm eps precision issue vs InternVL2 meta-init incompatibility; different model families and fixes." + "reason": "Multimodal config default-init bug vs GPTNeoX rotary_pct reload regression; unrelated." }, { - "left": "issue:45092", - "right": "issue:45357", + "left": "issue:43408", + "right": "issue:46402", "accept": false, - "reason": "InternVL2 meta-init incompatibility vs Qwen3.5 save_pretrained wrong visual encoder keys; both model-specific but not the same defect." + "reason": "Wrong Sam3 model type warning vs text encoder weights not loaded; different Sam3 issues." }, { - "left": "issue:43644", - "right": "issue:46305", + "left": "issue:22355", + "right": "issue:27957", "accept": false, - "reason": "Non-persistent buffer junk filling vs Moonshine sliding-window off-by-one; unrelated bugs." + "reason": "transformers.onnx import failure vs XLMRoberta Flash Attention 2 support issue; unrelated." }, { - "left": "issue:42831", - "right": "issue:45310", + "left": "issue:42796", + "right": "issue:43122", "accept": false, - "reason": "FineGrainedFP8 accuracy degradation vs Qwen3.5 MoE loading error; different failure classes." + "reason": "Ministral BPE-marker tokenizer bug vs broad tokenization change across versions; not the same concrete bug." }, { - "left": "issue:44829", - "right": "issue:46002", + "left": "issue:39401", + "right": "issue:44568", "accept": false, - "reason": "flash_attention_3 training collapse vs NaNs after checkout/init; both training-related but not the same concrete bug." + "reason": "Wrong offset_mapping in Qwen3 tokenizer vs add_special_tokens BOS/EOS omission in mdeberta-v3-base; different tokenizer behaviors." }, { - "left": "issue:44263", - "right": "issue:44373", + "left": "issue:42757", + "right": "issue:45278", "accept": false, - "reason": "DSA indexer torch.split bug vs wrong position_ids docstring; one is runtime behavior, the other documentation." + "reason": "Specific is_offline_mode import error vs broad upgrade-time import regressions; too broad and not the same bug." }, { - "left": "issue:43479", - "right": "issue:44913", + "left": "issue:42886", + "right": "issue:43066", "accept": false, - "reason": "Phi4MultimodalConfig default-init bug vs GPTNeoXConfig rotary_pct not persisting on reload; separate config serialization issues." + "reason": "Offline cache load failure vs wrong tokenizer decoder type in v5; different tokenizer regressions." } ] }, @@ -13636,8 +13862,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 44, - "review_unit_count": 54, + "review_unit_index": 45, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -13653,13 +13879,10 @@ "issue:39401", "issue:41628", "issue:42371", - "issue:42757", - "issue:42796", "issue:42831", "issue:42886", "issue:42907", "issue:42994", - "issue:43066", "issue:43122", "issue:43232", "issue:43408", @@ -13668,7 +13891,6 @@ "issue:43646", "issue:43756", "issue:43784", - "issue:43824", "issue:43937", "issue:43976", "issue:44060", @@ -13678,262 +13900,259 @@ "issue:44263", "issue:44295", "issue:44297", + "issue:44360", "issue:44368", "issue:44373", + "issue:44485", "issue:44488", "issue:44521", - "issue:44568", "issue:44704", "issue:44945", + "issue:44964", "issue:45278", + "issue:45290", "issue:45593", + "issue:45600", "issue:45753", "issue:46002", "issue:46132", "issue:46305", - "issue:46369", - "issue:46402" + "issue:46326", + "issue:46369" ], "soft_pairs": [ - "issue:41628|issue:43824", - "issue:43408|issue:46402", - "issue:22355|issue:27957", - "issue:39401|issue:44568", - "issue:42796|issue:43122", - "issue:42757|issue:45278", "issue:42831|issue:45753", "issue:27957|issue:30990", - "issue:42886|issue:43066", "issue:22355|issue:28218", "issue:44368|issue:46305", - "issue:43232|issue:45593", "issue:42886|issue:42994", + "issue:43232|issue:45593", "issue:43408|issue:46369", "issue:43550|issue:46132", "issue:43646|issue:46002", "issue:43576|issue:43784", "issue:42371|issue:44263", - "issue:44117|issue:44704", - "issue:44295|issue:44297", "issue:44188|issue:44373", + "issue:44117|issue:44704", "issue:30990|issue:33357", - "issue:42907|issue:44945", + "issue:44295|issue:44297", "issue:43756|issue:44060", + "issue:42907|issue:44945", "issue:44261|issue:44373", "issue:42907|issue:43937", - "issue:43976|issue:44945", "issue:41628|issue:45278", + "issue:43976|issue:44945", "issue:39401|issue:44488", - "issue:43122|issue:44521" + "issue:43122|issue:44521", + "issue:44704|issue:45290", + "issue:42907|issue:44964", + "issue:43784|issue:45600", + "issue:43122|issue:46326", + "issue:39401|issue:44521", + "issue:44360|issue:44485" ], - "prepared_review_unit_hash": "258fb2d7c38b2351e3bb292e66a7c879c9de05943455b8ffd61eafcfa7142f84", + "prepared_review_unit_hash": "2db9545763907191a31ac18273a7f2c935b86fef2530f5d40dcd72e3cc3215c3", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36208, - "estimated_input_tokens": 9052, - "estimated_eval_tokens": 18360 + "soft_pair_count": 29, + "serialized_chars": 35872, + "estimated_input_tokens": 8968, + "estimated_eval_tokens": 18192 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly a loose semantic cluster of transformers issues, but the soft pairs are not true duplicates. They span distinct models, regressions, and code paths, so I rejected all soft merges and picked the broad import-regression issue as the best umbrella representative.", - "confidence": 0.41, - "canonical_issue_reason": "issue:45278 is the broadest and most general report here: a multi-import regression after upgrading Transformers, which makes it the best umbrella issue among otherwise unrelated, model-specific bugs.", + "summary": "These items are a mixed set of unrelated Transformers issues; none of the soft pairs look like the same underlying bug/change closely enough to merge. No PRs were provided.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45278 is the broadest and most representative regression report here: it captures multiple import failures after a version upgrade and could serve as an umbrella for related import-breakage reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most reusable representative for triage because it covers a broad class of import failures rather than a single model or narrow API quirk.", + "best_issue_reason": "issue:45278 is the best standalone issue because it is broad, clearly actionable, and likely to attract the right fix for the most general class of failures in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:41628", - "right": "issue:43824", - "accept": false, - "reason": "Both are import errors, but for different missing symbols and likely different regressions; not the same concrete failure." - }, - { - "left": "issue:43408", - "right": "issue:46402", - "accept": false, - "reason": "Both are Sam3-related, but one is a model-type warning and the other is missing text-encoder weights; different bugs." - }, - { - "left": "issue:22355", - "right": "issue:27957", - "accept": false, - "reason": "One is a missing package/module import, the other is a Flash Attention model issue; unrelated code paths." - }, - { - "left": "issue:39401", - "right": "issue:44568", - "accept": false, - "reason": "Both are tokenizer bugs, but they affect different models and behaviors: offset mapping vs special-token insertion." - }, - { - "left": "issue:42796", - "right": "issue:43122", - "accept": false, - "reason": "Similar tokenization theme, but one is a specific BPE-marker handling bug and the other is a version-to-version tokenization regression; not the same issue." - }, - { - "left": "issue:42757", - "right": "issue:45278", - "accept": false, - "reason": "The first is a specific missing import from huggingface_hub; the second is a broad upgrade-related import regression, not the same bug." - }, { "left": "issue:42831", "right": "issue:45753", "accept": false, - "reason": "Accuracy loss in FineGrainedFP8 is unrelated to a model-specific infinite loop on one image." + "reason": "Different problems: FP8 accuracy regression vs a multimodal generation infinite-loop on a specific image." }, { "left": "issue:27957", "right": "issue:30990", "accept": false, - "reason": "Flash Attention support for XLM-Roberta and Sentence-Transformers loading stalls are different failure modes." - }, - { - "left": "issue:42886", - "right": "issue:43066", - "accept": false, - "reason": "Offline cache loading failure and wrong tokenizer decoder type are different tokenizer/configuration problems." + "reason": "XLM-RoBERTa Flash Attention 2 behavior and Sentence-Transformers loading hang are unrelated code paths." }, { "left": "issue:22355", "right": "issue:28218", "accept": false, - "reason": "A missing transformers.onnx module and an added-token spacing bug are unrelated." + "reason": "Missing transformers.onnx module import and tokenizer spacing after added token are unrelated." }, { "left": "issue:44368", "right": "issue:46305", "accept": false, - "reason": "A tied-embeddings warning and a streaming mask off-by-one are different subsystems and symptoms." + "reason": "Weight-tying warning in Qwen3.5 fine-tuning is not the same as Moonshine streaming mask off-by-one." }, { - "left": "issue:43232", - "right": "issue:45593", + "left": "issue:42886", + "right": "issue:42994", "accept": false, - "reason": "Generation keyword update logic and D-FINE denoising losses are unrelated." + "reason": "Offline tokenizer cache loading and quantized model saving failure affect different subsystems and failure modes." }, { - "left": "issue:42886", - "right": "issue:42994", + "left": "issue:43232", + "right": "issue:45593", "accept": false, - "reason": "Offline tokenizer cache loading and quantized model saving are different code paths." + "reason": "Generation kwargs update after sync_gpus is unrelated to D-FINE auxiliary-loss behavior." }, { "left": "issue:43408", "right": "issue:46369", "accept": false, - "reason": "Both involve Sam3, but one is a model mismatch warning and the other is a memory overflow; not the same bug." + "reason": "Model-type mismatch warning for sam3_tracker vs memory overflow in sam3-litetext are different issues." }, { "left": "issue:43550", "right": "issue:46132", "accept": false, - "reason": "torch.compile SDPA failure in Bamba and AttentionInterface.register behavior changes are distinct attention-related issues." + "reason": "torch.compile/SDPA failure in Bamba is not the same as AttentionInterface.register changing registered-function behavior." }, { "left": "issue:43646", "right": "issue:46002", "accept": false, - "reason": "Custom model initialization breakage and NaNs in classification heads are different problems." + "reason": "Custom model initialization breakage and NaNs in classification heads after checkout/init are different bugs." }, { "left": "issue:43576", "right": "issue:43784", "accept": false, - "reason": "A broken CLI command and a NameError during sentence-transformers import are unrelated." + "reason": "Broken transformers env command and a NameError importing sentence-transformers are separate import/CLI problems." }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "TF32 API guidance and a torch.split return-value issue do not share the same underlying bug." - }, - { - "left": "issue:44117", - "right": "issue:44704", - "accept": false, - "reason": "Tokenizer mapping lookup returning None and AutoProcessor kwargs forwarding are separate from_pretrained issues." + "reason": "TF32 API settings warning is unrelated to the GlmMoeDsaIndexer torch.split issue." }, { - "left": "issue:44295", - "right": "issue:44297", + "left": "issue:44188", + "right": "issue:44373", "accept": false, - "reason": "Reading position_ids after buffer registration and tokenizer_class mismatch on save are unrelated." + "reason": "Attention-kernel divergence under torch.compile is unrelated to a docstring bug for position_ids." }, { - "left": "issue:44188", - "right": "issue:44373", + "left": "issue:44117", + "right": "issue:44704", "accept": false, - "reason": "A torch.compile attention-kernel divergence bug and a docstring typo are not the same issue." + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor not passing kwargs to cached_file are different loader bugs." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Loading hang for Sentence-Transformers and a Mac bus error on CLIP are different runtime failures." + "reason": "Loading hang for Sentence-Transformers and a MacOS bus error with community CLIP models are distinct runtime failures." }, { - "left": "issue:42907", - "right": "issue:44945", + "left": "issue:44295", + "right": "issue:44297", "accept": false, - "reason": "Quantized model save failure and incorrect output under pipeline parallelism are unrelated." + "reason": "Reading position_ids after buffer registration and tokenizer_class mismatch on save_pretrained are unrelated." }, { "left": "issue:43756", "right": "issue:44060", "accept": false, - "reason": "Different model configuration bugs: RoPE layer dropping versus incorrect tied-weights warning." + "reason": "Smollm3 RoPE-layer count mismatch and Qwen3-Next tied-weights warning concern different model configs." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Dequantized model saving failure and pipeline-parallelism output corruption are not the same code-path problem." }, { "left": "issue:44261", "right": "issue:44373", "accept": false, - "reason": "Layernorm epsilon precision bug and a docstring issue are unrelated." + "reason": "MLA q_a_layernorm epsilon precision issue is unrelated to the position_ids docstring." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Saving dequantized models and invalid GenerationConfig are separate issues." + "reason": "Model saving failure and invalid GenerationConfig for GLM-5 are separate failures." }, { - "left": "issue:43976", - "right": "issue:44945", + "left": "issue:41628", + "right": "issue:45278", "accept": false, - "reason": "Python-version incompatibility and pipeline-parallel output corruption are unrelated." + "reason": "Both are import-related, but one is a specific missing export while the other is a broad upgrade regression; not enough evidence they share the same fix." }, { - "left": "issue:41628", - "right": "issue:45278", + "left": "issue:43976", + "right": "issue:44945", "accept": false, - "reason": "Both are import problems, but one is a single missing class import while the other is a broad multi-import regression; too different to merge." + "reason": "Python-version compatibility failure is unrelated to pipeline-parallel generation correctness." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "A wrong offset-mapping bug and a model-loading failure are different tokenizer/model behaviors." + "reason": "Wrong offset_mapping in Qwen3 tokenizer and failure to load cjvt/sleng-bert are different tokenizer issues." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Version-to-version tokenization differences and zero assistant_masks for multimodal inputs are different behaviors." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "Passing kwargs to cached_file in AutoProcessor is unrelated to chat-template crashes on tool-call assistant messages." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Quantized/dequantized saving failure and inability to load Phi-4-multimodal-instruct are different model-loading/saving problems." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "sentence-transformers import NameError and stale auto_mappings references to removed Sam3LiteText configs are unrelated." }, { "left": "issue:43122", + "right": "issue:46326", + "accept": false, + "reason": "Tokenizer output changes across versions and continue_final_message behavior for mistral_common are different tokenization features." + }, + { + "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Version-dependent tokenization changes and zero assistant masks in multimodal chat templating are unrelated." + "reason": "Qwen3 offset_mapping bug and multimodal assistant_mask zeroing are different tokenizer/template issues." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer lacking ReLU and GLM-5 RoPE implementation are unrelated model internals." } ] }, @@ -13948,8 +14167,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 45, - "review_unit_count": 54, + "review_unit_index": 46, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -13962,8 +14181,8 @@ "issue:29159", "issue:36296", "issue:38617", - "issue:39401", "issue:42175", + "issue:42371", "issue:42673", "issue:42757", "issue:42907", @@ -13975,57 +14194,51 @@ "issue:43452", "issue:43475", "issue:43493", - "issue:43784", + "issue:43611", "issue:43824", + "issue:43874", "issue:43937", "issue:43992", "issue:44038", + "issue:44188", "issue:44222", "issue:44295", "issue:44297", "issue:44351", - "issue:44360", + "issue:44355", "issue:44485", "issue:44514", - "issue:44521", "issue:44625", "issue:44704", "issue:44779", "issue:44829", "issue:44938", "issue:44945", - "issue:44964", "issue:45072", "issue:45200", "issue:45216", "issue:45278", - "issue:45290", "issue:45478", - "issue:45600", "issue:45753", - "issue:46326" + "issue:46002", + "issue:46231", + "issue:46328" ], "soft_pairs": [ - "issue:44704|issue:45290", - "issue:42907|issue:44964", - "issue:43784|issue:45600", - "issue:43122|issue:46326", - "issue:39401|issue:44521", "issue:43116|issue:45200", - "issue:44360|issue:44485", "issue:44038|issue:44297", "issue:44625|issue:44829", - "issue:44038|issue:44295", "issue:43408|issue:43475", "issue:42907|issue:43452", + "issue:44038|issue:44295", "issue:43122|issue:44514", "issue:44038|issue:44222", "issue:45216|issue:45478", "issue:43824|issue:44351", "issue:22355|issue:29159", "issue:42757|issue:44351", - "issue:43278|issue:44945", "issue:36296|issue:43116", + "issue:43278|issue:44945", "issue:42175|issue:44938", "issue:43493|issue:45753", "issue:22355|issue:29127", @@ -14033,204 +14246,210 @@ "issue:43066|issue:44779", "issue:43937|issue:44485", "issue:42673|issue:45072", - "issue:38617|issue:45278" + "issue:38617|issue:45278", + "issue:44188|issue:46328", + "issue:43611|issue:46002", + "issue:44355|issue:46231", + "issue:42371|issue:43874", + "issue:43116|issue:43493", + "issue:44038|issue:45478" ], - "prepared_review_unit_hash": "45746de9ccbaf171ac8099715e20019936319d53c0b20916d979c9e5bce2c17a", + "prepared_review_unit_hash": "272c2052175ef61dd1803c20629e6caf5fd315ea891453e9c386512de9407b0e", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35170, - "estimated_input_tokens": 8793, - "estimated_eval_tokens": 17842 + "serialized_chars": 35237, + "estimated_input_tokens": 8810, + "estimated_eval_tokens": 17876 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated issue reports. A few share broad themes (imports, tokenization, Qwen3.5), but none are close enough in symptom and code path to merge as duplicates; I would reject all soft edges.", - "confidence": 0.62, - "canonical_issue_reason": "issue:45216 is the clearest, most actionable regression-style report with a specific model family, API, and version boundary, making it the best anchor candidate among this set.", + "summary": "These items are mostly unrelated issue reports across different models, APIs, and regressions. The soft-similarity links mostly reflect broad topical overlap (imports, tokenization, multimodal models, dtype/parallelism) rather than the same concrete bug, so none should be merged as duplicates.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 45278 is the broadest representative here because it summarizes a wide import-regression class, but even that is not a true umbrella for the whole set; there is no strong canonical duplicate among these issues.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45216 is the strongest standalone issue because it states a concrete regression, clear repro scope, and an explicit version threshold.", + "best_issue_reason": "45278 is the most general and cluster-like issue title, so it is the best single representative if one must be chosen, though the set does not form a real duplicate cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44704", - "right": "issue:45290", - "accept": false, - "reason": "Both involve helper APIs, but one is cached_file kwargs and the other is chat-template tokenization with tool-call messages; different code paths and failures." - }, - { - "left": "issue:42907", - "right": "issue:44964", - "accept": false, - "reason": "One is about saving dequantized Ministrals/Devstrals, the other loading Phi-4-multimodal-instruct; same general area, not the same bug." - }, - { - "left": "issue:43784", - "right": "issue:45600", - "accept": false, - "reason": "Importing sentence-transformers vs stale auto_mappings references to removed Sam3LiteText configs are unrelated import breakages." - }, - { - "left": "issue:43122", - "right": "issue:46326", - "accept": false, - "reason": "Both mention tokenizer behavior, but one is a v4-to-v5 tokenization regression and the other is continue_final_message for mistral_common tokenizers." - }, - { - "left": "issue:39401", - "right": "issue:44521", - "accept": false, - "reason": "Offset mappings and assistant masks are different tokenizer outputs; these are separate multimodal/tokenization bugs." - }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "Example-script multi-label classification returning empty results is unrelated to Gemma 4 mm_token_type_ids propagation." - }, - { - "left": "issue:44360", - "right": "issue:44485", - "accept": false, - "reason": "A DSA indexer ReLU discussion and GLM-5 RoPE implementation are not the same underlying defect." + "reason": "Different failures: classification script returning empty results vs Gemma 4 multimodal token defaults." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "A Qwen3-VL-Moe regression and tokenizer_class metadata mismatch are separate problems." + "reason": "Unrelated bugs: Qwen3-VL-Moe issue vs tokenizer save_pretrained metadata mismatch." }, { "left": "issue:44625", "right": "issue:44829", "accept": false, - "reason": "Config propagation for Qwen3.5 num_labels is unrelated to flash_attention_3 training degeneration." - }, - { - "left": "issue:44038", - "right": "issue:44295", - "accept": false, - "reason": "Qwen3-VL-Moe model loading and position_ids buffer access are different code paths." + "reason": "Different code paths: num_labels propagation vs flash_attention_3 training degeneration." }, { "left": "issue:43408", "right": "issue:43475", "accept": false, - "reason": "Both are SAM 3-related, but one is a model-type mismatch warning and the other is a missing attribute on vision encoder output." + "reason": "Same model family, but different concrete defects: warning mismatch vs missing attribute crash." }, { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals and gguf_file loading failures are different model-file handling bugs." + "reason": "Dequantized save failure and gguf_file loading failure are different problems." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Different bugs: Qwen3-VL-Moe regression vs position_ids buffer access error." }, { "left": "issue:43122", "right": "issue:44514", "accept": false, - "reason": "General tokenizer regression vs batched Qwen2_5_VLProcessor chat-template crash; same area, different failure." + "reason": "Both are tokenizer-related, but the concrete failures and affected paths differ." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Qwen3-VL-Moe version regression and FP8 moe save_pretrained behavior are not the same concrete bug." + "reason": "Different issues: Qwen3-VL-Moe bug vs FP8 save_pretrained moe bug." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both are Qwen3.5 on 5.4.0+, but one is save_pretrained checkpoint corruption and the other is from_pretrained loading error; related, but not clearly the same defect." + "reason": "Both concern Qwen3.5, but save_pretrained regression and from_pretrained error are distinct code paths." }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Both are missing-import errors from transformers, but they concern different symbols and model families." + "reason": "Both are import errors, but for different missing symbols and not the same underlying bug." }, { "left": "issue:22355", "right": "issue:29159", "accept": false, - "reason": "Missing transformers.onnx and slow-vs-fast tokenizer inconsistency are unrelated." + "reason": "No relation: ONNX import failure vs slow/fast tokenizer inconsistency." }, { "left": "issue:42757", "right": "issue:44351", "accept": false, - "reason": "Two separate missing imports from different packages/symbols; no shared bug mechanism." + "reason": "Both are import regressions, but they involve different missing APIs and are not the same defect." }, { - "left": "issue:43278", - "right": "issue:44945", + "left": "issue:36296", + "right": "issue:43116", "accept": false, - "reason": "Embedding dtype drift in eval and incorrect outputs under pipeline parallelism are different runtime issues." + "reason": "Tensor parallel training bug is unrelated to the classification script output issue." }, { - "left": "issue:36296", - "right": "issue:43116", + "left": "issue:43278", + "right": "issue:44945", "accept": false, - "reason": "Tensor parallel training instability and an example script returning empty classification results are unrelated." + "reason": "Different runtime problems: dtype change in evaluation vs incorrect output under pipeline parallelism." }, { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "A missing tensorflow extra in pip install versus Python 3.14 load failure are separate installation/runtime problems." + "reason": "Packaging/backend dependency issue vs Python 3.14 load failure; not the same bug." }, { "left": "issue:43493", "right": "issue:45753", "accept": false, - "reason": "SigLIP2 HF/JAX mismatch and Qwen3_5 infinite loop on a specific image are different model bugs." + "reason": "Different models and symptoms: SigLIP2 discrepancy vs Qwen3_5 infinite loop on an image." }, { "left": "issue:22355", "right": "issue:29127", "accept": false, - "reason": "Transformers.onnx import failure and a LayoutLMv3 error-message clarity issue are not the same issue." + "reason": "No underlying match: ONNX import error vs LayoutLMv3 error-message clarity." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Missing embed_tokens.weight on UMT5Encoder load is unrelated to AutoProcessor kwargs not reaching cached_file." + "reason": "Different areas: missing encoder weights vs AutoProcessor kwarg forwarding." }, { "left": "issue:43066", "right": "issue:44779", "accept": false, - "reason": "Wrong decoder type in v5 and DeepSeek tokenization output regressions may both be tokenizer-related, but they are distinct defects." + "reason": "Both are tokenizer regressions in v5, but they affect different models and likely different code paths." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "GLM-5 GenerationConfig validation and GLM-5 RoPE implementation are separate components and failure modes." + "reason": "Same model family, but invalid GenerationConfig and RoPE implementation are distinct problems." }, { "left": "issue:42673", - "right": "issue:45072", + "right": "issue:46328", + "accept": false, + "reason": "VRAM leak in threaded dataloading is unrelated to a padding performance issue." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "Broad import-error cluster overlap, but not the same concrete missing symbol or defect." + }, + { + "left": "issue:44188", + "right": "issue:46328", + "accept": false, + "reason": "Attention-kernel divergence under torch.compile is unrelated to feature-extractor padding overhead." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "Base-model-prefix loading regression and NaNs after checkout/init are different failures." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled-file import/runtime errors and a documentation parameter mismatch are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API guidance and a missing image-patch method are unrelated issues." + }, + { + "left": "issue:43116", + "right": "issue:43493", "accept": false, - "reason": "VRAM leakage with Qwen3ForCausalLM in dataloader threads and bfloat16 dtype mismatches are unrelated." + "reason": "Different models and behaviors: empty classification outputs vs SigLIP2 implementation discrepancy." }, { - "left": "issue:38617", - "right": "issue:45278", + "left": "issue:44038", + "right": "issue:45478", "accept": false, - "reason": "A specific layer_type_validation import error and a later broad wave of import errors after upgrade are not the same concrete bug." + "reason": "Both mention Qwen3-family MoE, but they are different models and different failure modes." } ] }, @@ -14245,8 +14464,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 46, - "review_unit_count": 54, + "review_unit_index": 47, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -14255,39 +14474,38 @@ "cluster_id": "cluster-10000-463", "nodes": [ "issue:33453", + "issue:38617", "issue:39401", + "issue:41628", "issue:42222", - "issue:42371", "issue:42673", "issue:42796", "issue:43064", "issue:43116", + "issue:43122", "issue:43441", "issue:43450", - "issue:43493", - "issue:43611", "issue:43673", "issue:43825", "issue:43827", - "issue:43874", "issue:43906", "issue:43937", "issue:43976", "issue:43986", - "issue:44038", "issue:44075", "issue:44117", - "issue:44188", "issue:44291", - "issue:44355", "issue:44380", "issue:44410", "issue:44466", "issue:44779", + "issue:44829", "issue:44936", "issue:44977", "issue:45071", + "issue:45081", "issue:45083", + "issue:45216", "issue:45310", "issue:45341", "issue:45357", @@ -14299,17 +14517,13 @@ "issue:45588", "issue:46002", "issue:46077", - "issue:46231", - "issue:46328" + "issue:46178", + "issue:46298", + "issue:46328", + "issue:46512" ], "soft_pairs": [ - "issue:44188|issue:46328", - "issue:44355|issue:46231", - "issue:43611|issue:46002", - "issue:43116|issue:43493", - "issue:42371|issue:43874", "issue:42222|issue:43441", - "issue:44038|issue:45478", "issue:43976|issue:45446", "issue:44075|issue:44936", "issue:45357|issue:45478", @@ -14321,8 +14535,8 @@ "issue:33453|issue:39401", "issue:44380|issue:46077", "issue:44117|issue:45588", - "issue:39401|issue:44779", "issue:43827|issue:45446", + "issue:39401|issue:44779", "issue:42673|issue:43906", "issue:43673|issue:46328", "issue:44291|issue:45071", @@ -14332,216 +14546,222 @@ "issue:45405|issue:45446", "issue:44410|issue:45310", "issue:45341|issue:45561", - "issue:45341|issue:46328" + "issue:45341|issue:46328", + "issue:43122|issue:46298", + "issue:38617|issue:41628", + "issue:44117|issue:45081", + "issue:45071|issue:46512", + "issue:42222|issue:46178", + "issue:44829|issue:45216" ], - "prepared_review_unit_hash": "2ded21166dddd09f498de1a11329b6228a1e3b577a1d848b80515ed6505be5b7", + "prepared_review_unit_hash": "4e50bd1d9d21900a1a1e713bd692f6d9b40b86345e5a3e699415d3b438b8f9ca", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 35917, - "estimated_input_tokens": 8980, - "estimated_eval_tokens": 18216 + "serialized_chars": 36668, + "estimated_input_tokens": 9167, + "estimated_eval_tokens": 18590 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is a set of mostly unrelated transformer bugs across tokenizers, model-specific loading/generation issues, training behavior, docs, and testing utilities. None of the soft-edge pairs look like true duplicates or the same underlying fix.", - "confidence": 0.87, - "canonical_issue_reason": "Issue 43611 is the clearest broad regression report in the set: it describes a concrete v5 model-loading failure tied to `base_model_prefix`, with a well-defined code path and wider potential impact than the more model-specific reports.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers v5 regressions, model-specific bugs, and a few docs/testing issues. I found no soft edge pair that clearly shares the same underlying bug/change closely enough to merge.", + "confidence": 0.62, + "canonical_issue_reason": "Issue 43122 is the best representative of the cluster\u2019s dominant theme: a concrete v4.57.3\u2192v5 tokenizer regression with user-visible impact and some follow-on references. The rest of the cluster is too heterogeneous for a single true duplicate canonical.", "canonical_pr_reason": null, - "best_issue_reason": "43611 is the best representative issue because it is a concrete, user-facing regression with a generic loading path and strong triage value.", + "best_issue_reason": "43122 is the strongest global representative because it is specific, regression-oriented, and broad enough to capture the tokenizer-loading/tokenization theme better than the more model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44188", - "right": "issue:46328", - "accept": false, - "reason": "Different subsystems and failure modes: attention-kernel divergence under `torch.compile` vs a performance issue in `SequenceFeatureExtractor.pad`." - }, - { - "left": "issue:44355", - "right": "issue:46231", - "accept": false, - "reason": "Running compiled Python files is unrelated to a docs/parameter-name mismatch in `set_training()`." - }, - { - "left": "issue:43611", - "right": "issue:46002", - "accept": false, - "reason": "Model loading with `base_model_prefix` is a different bug from NaNs in classification heads after checkpoint/init." - }, - { - "left": "issue:43116", - "right": "issue:43493", - "accept": false, - "reason": "Empty multi-label classification results in an example script is unrelated to a SigLIP2 implementation discrepancy." - }, - { - "left": "issue:42371", - "right": "issue:43874", - "accept": false, - "reason": "TF32 control settings and a missing image-patch helper on a video/image processor are unrelated bugs." - }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "Different model families and code paths: vitpose loading breakage vs Ministral flash-attention failure." - }, - { - "left": "issue:44038", - "right": "issue:45478", - "accept": false, - "reason": "Both mention Qwen3.5/MoE, but they point to different model variants and likely different `from_pretrained` failure paths; too broad to treat as one bug." + "reason": "Different models and failure modes: vitpose model breakage vs Ministral FlashAttention crash." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python version support regression is unrelated to a flex-attention import/version-check bug." + "reason": "Both are version-check issues, but they affect different imports/modules and do not describe the same bug." }, { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "Unused SGD arguments and `trainer.evaluate()` failing after training are different training bugs." + "reason": "Optimizer argument handling and trainer.evaluate failure are distinct code paths and symptoms." }, { "left": "issue:45357", "right": "issue:45478", "accept": false, - "reason": "Incorrect `save_pretrained` visual encoder keys is a save/serialization bug, not the same as a `from_pretrained` loading error." + "reason": "Same model family, but one is save_pretrained key serialization and the other is from_pretrained loading; not the same concrete bug." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "A multi-label example returning empty results is unrelated to an error message/doc pipeline task mismatch." + "reason": "Example-script multi-label behavior and pipeline docs/error-message wording are unrelated." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "A helper-length calculation bug in `qwen3_omni_moe` is unrelated to a PyTorch version check in flex attention." + "reason": "A qwen3_omni_moe helper bug and a flex_attention PyTorch version check are different problems." }, { "left": "issue:42673", "right": "issue:42796", "accept": false, - "reason": "VRAM leakage in dataloader threads and tokenizer BPE-marker handling are unrelated issues." + "reason": "VRAM leak in dataloader threads vs tokenizer BPE marker handling are unrelated." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Distributed optimizer-state corruption under FSDP/PEFT is a different code path from missing generation cache in chunked prefill." + "reason": "Distributed optimizer-state corruption and generation cache missing are different subsystems and bugs." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "Invalid `GenerationConfig` and a stale warning for `num_return_sequences` are different generation bugs." + "reason": "GenerationConfig validation failure and stale warning text are not the same issue." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Both are tokenizer-related, but one is a general loading regression and the other is a Qwen3 offset-mapping bug; not the same defect." + "reason": "Generic tokenizer-loading regression vs Qwen3 offset_mapping bug; no clear shared code-path shown." }, { "left": "issue:44380", "right": "issue:46077", "accept": false, - "reason": "GPT2 attention scaling under SDPA/FlashAttention is unrelated to Gemma4 building causal masks for bidirectional attention." + "reason": "Attention-scaling config being ignored and bidirectional mask construction are different model/back-end behaviors." }, { "left": "issue:44117", "right": "issue:45588", "accept": false, - "reason": "A tokenizer-mapping `None` handling bug is unrelated to a flash-attention attribute error on `s_aux=None`." + "reason": "Tokenizer mapping import assumptions and flash_attention sink-less-model crash are unrelated." }, { - "left": "issue:39401", - "right": "issue:44779", + "left": "issue:43827", + "right": "issue:45446", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different symptoms (`offset_mapping` vs incorrect tokenization output)." + "reason": "Docs still referencing pipeline() is a documentation issue, not the flex_attention version-check bug." }, { - "left": "issue:43827", - "right": "issue:45446", + "left": "issue:39401", + "right": "issue:44779", "accept": false, - "reason": "Docs still referencing `pipeline()` after removals is unrelated to the flex-attention PyTorch check." + "reason": "Both are tokenizer regressions, but they target different models and symptoms with no evidence of a shared defect." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "A VRAM leak in multi-threaded dataloader use is unrelated to an isolated reproduction issue." + "reason": "VRAM leak across dataloader threads is unrelated to an isolated reproduction of a different issue." }, { "left": "issue:43673", "right": "issue:46328", "accept": false, - "reason": "Missing generation cache during chunked prefill is unrelated to a feature-extractor padding performance issue." + "reason": "Generation cache behavior and a padding performance issue are distinct problems." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "`init_empty_weights` argument handling and `PretrainedConfig` type checking are distinct initialization/type-system problems." + "reason": "Unexpected _is_hf_initialized argument and PretrainedConfig type-checking regression are separate compatibility bugs." }, { "left": "issue:44466", "right": "issue:46002", "accept": false, - "reason": "Inconsistent serialization of tied weights is unrelated to NaNs in classification heads." + "reason": "Serialization of tied weights and NaNs in classification heads are different failure modes." }, { "left": "issue:43450", "right": "issue:43986", "accept": false, - "reason": "Incorrect batched video-processor output shape is not the same as crashing in `AutoProcessor` without torchvision." + "reason": "Batched video-processor shape handling and missing-torchvision AutoProcessor crash are not the same bug." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "GenerationConfig validity and flash-attention generation failures are different bugs." + "reason": "Invalid GenerationConfig vs flash-attention generation failure are unrelated." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "A release/version pin issue for PEFT is unrelated to a flex-attention import version check." + "reason": "Dependency version bump and a PyTorch version-check bug are different issues." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Missing Qwen3-next projections and Qwen3.5 MoE `from_pretrained` errors are different model-architecture issues." + "reason": "qwen3next missing projections and Qwen3.5 from_pretrained error are different model-specific defects." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "A small testing-utils bug is not the same as an xdist race on `captured_info.txt`." + "reason": "A minor testing_utils bug and a pytest-xdist captured_info race are distinct issues." }, { "left": "issue:45341", "right": "issue:46328", "accept": false, - "reason": "Testing-utils behavior and feature-extractor padding performance are unrelated." + "reason": "testing_utils bug and SequenceFeatureExtractor.pad performance regression are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:46298", + "accept": false, + "reason": "Tokenizer behavior regression and a noisy logging message do not share the same underlying bug." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but for different symbols and modules; no duplicate evidence." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and _patch_mistral_regex backend_tokenizer AttributeError are different tokenizer-loading failures." + }, + { + "left": "issue:45071", + "right": "issue:46512", + "accept": false, + "reason": "PretrainedConfig type checking and stale config.dtype after from_config are separate config regressions." + }, + { + "left": "issue:42222", + "right": "issue:46178", + "accept": false, + "reason": "vitpose model breakage and tensorflow_text backend mapping errors are unrelated." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 causing degenerate training and incorrect save_pretrained checkpoint serialization are different bugs." } ] }, @@ -14556,8 +14776,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 47, - "review_unit_count": 54, + "review_unit_index": 48, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -14565,6 +14785,7 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:27957", "issue:38617", "issue:39401", "issue:41628", @@ -14578,27 +14799,28 @@ "issue:43066", "issue:43097", "issue:43122", + "issue:43352", "issue:43408", "issue:43577", "issue:43582", "issue:43673", "issue:43742", + "issue:43874", "issue:43937", "issue:44075", - "issue:44117", "issue:44220", "issue:44261", + "issue:44263", + "issue:44360", "issue:44410", "issue:44483", "issue:44484", + "issue:44485", "issue:44493", "issue:44496", - "issue:44829", "issue:44877", "issue:44933", "issue:45071", - "issue:45081", - "issue:45216", "issue:45278", "issue:45310", "issue:45341", @@ -14607,28 +14829,20 @@ "issue:45478", "issue:45542", "issue:45656", - "issue:46178", + "issue:46172", "issue:46209", "issue:46231", - "issue:46298", - "issue:46328", - "issue:46512" + "issue:46328" ], "soft_pairs": [ - "issue:43122|issue:46298", - "issue:38617|issue:41628", - "issue:44117|issue:45081", - "issue:45071|issue:46512", - "issue:44829|issue:45216", - "issue:42222|issue:46178", "issue:45362|issue:46328", "issue:43408|issue:43742", "issue:42617|issue:45341", "issue:42673|issue:43582", "issue:42371|issue:44261", "issue:44877|issue:46231", - "issue:39401|issue:46209", "issue:44484|issue:45310", + "issue:39401|issue:46209", "issue:38617|issue:44220", "issue:43937|issue:44496", "issue:44933|issue:45278", @@ -14641,198 +14855,218 @@ "issue:43012|issue:43577", "issue:42175|issue:42222", "issue:43097|issue:44493", - "issue:43673|issue:45341" + "issue:43673|issue:45341", + "issue:43874|issue:44261", + "issue:41628|issue:45071", + "issue:44263|issue:44485", + "issue:45071|issue:45542", + "issue:27957|issue:43352", + "issue:44484|issue:45478", + "issue:44263|issue:46172", + "issue:44360|issue:46172" ], - "prepared_review_unit_hash": "70aa8c07870e8982f81cf57d4ec2fab65e5a06311cacc75c2c7f369b57243188", + "prepared_review_unit_hash": "64a1b7635f9c80d7d7ea6c7cdd8465415602ab5ebdf14c583733902d598c19ed", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 35432, - "estimated_input_tokens": 8858, - "estimated_eval_tokens": 17972 + "soft_pair_count": 29, + "serialized_chars": 35689, + "estimated_input_tokens": 8923, + "estimated_eval_tokens": 18102 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a heterogeneous cluster of Transformers issues spanning import errors, tokenizer regressions, config/dtype mismatches, backend/package gaps, and model-specific bugs. The soft pairs are only superficially similar, so none look like the same underlying bug or change.", - "confidence": 0.44, - "canonical_issue_reason": "issue:45278 is the best umbrella representative because it captures the broad version-upgrade/import-regression theme that recurs across several items, even though the cluster overall is still mixed.", + "summary": "This cluster is heterogeneous: the issues span unrelated model-loading regressions, tokenizer bugs, optimizer/training behavior, import errors, and packaging/docs problems. None of the soft pairs look like the same underlying defect, so there is no convincing duplicate set here.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most general and cluster-like issue here; it best represents the repeated v5 upgrade breakage pattern without being overly specific to one model or API path.", + "best_issue_reason": "issue:45278 is the broadest representative ('many import errors after update'), but it still only loosely anchors one subtheme and does not explain the rest of the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43122", - "right": "issue:46298", - "accept": false, - "reason": "Tokenization behavior regression vs a noisy logging issue; unrelated bug surfaces." - }, - { - "left": "issue:38617", - "right": "issue:41628", - "accept": false, - "reason": "Both are import errors, but for different missing symbols and different code paths." - }, - { - "left": "issue:44117", - "right": "issue:45081", - "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES None handling is distinct from the Mistral regex/backend_tokenizer AttributeError." - }, - { - "left": "issue:45071", - "right": "issue:46512", - "accept": false, - "reason": "Type-checking regression and stale config.dtype are separate dtype/config bugs." - }, - { - "left": "issue:44829", - "right": "issue:45216", - "accept": false, - "reason": "FlashAttention-3 training degeneration is unrelated to the Qwen3.5 save_pretrained checkpoint regression." - }, - { - "left": "issue:42222", - "right": "issue:46178", - "accept": false, - "reason": "vitpose model breakage and tensorflow_text backend registration are different failures." - }, { "left": "issue:45362", "right": "issue:46328", "accept": false, - "reason": "Chat runtime crash for Qwen3.5 is unrelated to a SequenceFeatureExtractor.pad performance issue." + "reason": "Different areas: Qwen chat/runtime failure vs SequenceFeatureExtractor padding performance." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "A model-type warning for sam3 tracking is not the same as the MobileLLM key error." + "reason": "Unrelated model families and failure modes." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel.py execution failure and a testing_utils bug are unrelated." + "reason": "Different utilities; a runtime script failure vs a testing helper bug." }, { "left": "issue:42673", "right": "issue:43582", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM and Apple Silicon warmup TypeError are different problems." + "reason": "Different platforms and code paths: VRAM leak vs Apple Silicon TypeError." }, { "left": "issue:42371", "right": "issue:44261", "accept": false, - "reason": "TF32 settings warning is unrelated to MLA layernorm epsilon/config precision error." + "reason": "TF32 settings issue is not the same as the q_a_layernorm epsilon bug." }, { "left": "issue:44877", "right": "issue:46231", "accept": false, - "reason": "Strict config loading for granite_speech and a documentation parameter-name mismatch are not the same bug." + "reason": "Config parsing regression vs documentation typo." }, { - "left": "issue:39401", - "right": "issue:46209", + "left": "issue:44484", + "right": "issue:45310", "accept": false, - "reason": "Tokenizer offset_mapping bug and FP8 inference failure on B200 are unrelated." + "reason": "save_pretrained shard-size question vs Qwen3.5 MoE loading error." }, { - "left": "issue:44484", - "right": "issue:45310", + "left": "issue:39401", + "right": "issue:46209", "accept": false, - "reason": "max_shard_size question and Qwen3.5 Moe from_pretrained error are different issues." + "reason": "Tokenizer offset_mapping bug vs FP8 inference failure on B200." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "ImportError for layer_type_validation is unrelated to _torch_extract_fbank_features()." + "reason": "ImportError in configuration_utils vs audio feature extraction issue." }, { "left": "issue:43937", "right": "issue:44496", "accept": false, - "reason": "Invalid GenerationConfig for GLM-5 and unrecognized model/config.json failure are separate loader issues." + "reason": "GenerationConfig validation failure vs unrecognized model/config error." }, { "left": "issue:44933", "right": "issue:45278", "accept": false, - "reason": "A missing import from image_utils is narrower than the broad post-upgrade import-error report; not the same bug." + "reason": "Shared import theme only; one is a specific missing symbol, the other is broad post-upgrade import failures." }, { "left": "issue:43066", "right": "issue:43122", "accept": false, - "reason": "Wrong tokenizer decoder type may affect tokenization, but it is not clearly the same issue as differing tokenization output across versions." + "reason": "Both are tokenizer regressions, but one is decoder-type metadata and the other is tokenization drift." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 behavior control and FineGrainedFP8 accuracy are different numerical-behavior concerns." + "reason": "TF32 control bug vs FineGrainedFP8 accuracy bug." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Missing qwen3next projections and Qwen3.5 Moe from_pretrained error are unrelated model issues." + "reason": "Different model code paths: missing projections vs from_pretrained loading error." }, { "left": "issue:42222", "right": "issue:45542", "accept": false, - "reason": "Broken vitpose models and an undefined tf backend from tensorboard-only install are different backend/package issues." + "reason": "VitPose breakage vs TensorFlow backend packaging issue." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "SGD args ignored and optimizer.step being called twice under deepspeed are distinct optimizer bugs." + "reason": "Both optimizer-related, but SGD args being ignored is not the same as double-stepping with deepspeed." }, { "left": "issue:44483", "right": "issue:45464", "accept": false, - "reason": "Generic /v1/chat/completions rejection and Qwen3.5-0.8B streaming failure are not the same concrete API bug." + "reason": "Different request modes/endpoints; not the same chat/completions defect." }, { "left": "issue:43012", "right": "issue:43577", "accept": false, - "reason": "Float32 warning during compile and Blip2 dtype fields staying float32 are separate dtype issues." + "reason": "Both involve dtype, but one is a compile-time warning and the other is incorrect model dtype on load." }, { "left": "issue:42175", "right": "issue:42222", "accept": false, - "reason": "Missing TensorFlow backend in a torch extra install is unrelated to vitpose breakage." + "reason": "Packaging/backend dependency issue vs model-specific VitPose breakage." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Removed tie_embeddings_and_encoder_decoder and unexpected position-id keys are different v5 regressions." + "reason": "API removal in config semantics vs unexpected key warnings from position IDs." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "chunked_prefill cache regression and a testing_utils bug are unrelated." + "reason": "Generation cache regression vs a testing utility bug." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Image patch counting bug vs rms_norm_eps precision bug." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "Missing import symbol vs PreTrainedConfig type-checking regression." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "Different GLM submodules and different bugs." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Type-checking regression vs TensorFlow backend error." + }, + { + "left": "issue:27957", + "right": "issue:43352", + "accept": false, + "reason": "Different models and different Flash Attention 2 support states." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "save_pretrained shard sizing is unrelated to Qwen3.5 MoE loading." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "Different models and code paths: GlmMoeDsaIndexer vs deepseek-v4 dtype/transpose issue." + }, + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "DSA indexer ReLU bug is unrelated to the deepseek-v4 tensor transform issue." } ] }, @@ -14847,8 +15081,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 48, - "review_unit_count": 54, + "review_unit_index": 49, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -14856,19 +15090,17 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:27957", "issue:38617", - "issue:41628", "issue:42757", "issue:42890", "issue:43064", "issue:43097", - "issue:43352", "issue:43441", "issue:43452", "issue:43502", "issue:43531", "issue:43673", + "issue:43824", "issue:43873", "issue:43874", "issue:43931", @@ -14881,21 +15113,21 @@ "issue:44263", "issue:44315", "issue:44351", - "issue:44360", + "issue:44380", "issue:44410", "issue:44484", "issue:44485", + "issue:44704", "issue:44829", "issue:44933", + "issue:44936", "issue:44987", "issue:45003", - "issue:45071", + "issue:45103", "issue:45341", "issue:45372", "issue:45375", "issue:45440", - "issue:45478", - "issue:45542", "issue:45588", "issue:45753", "issue:46050", @@ -14903,283 +15135,264 @@ "issue:46172", "issue:46231", "issue:46298", - "issue:46369" + "issue:46365", + "issue:46369", + "issue:46421" ], "soft_pairs": [ - "issue:43874|issue:44261", - "issue:44263|issue:44485", - "issue:41628|issue:45071", - "issue:45071|issue:45542", - "issue:27957|issue:43352", - "issue:44484|issue:45478", - "issue:44263|issue:46172", - "issue:44360|issue:46172", "issue:43931|issue:44410", "issue:43441|issue:44829", "issue:43931|issue:44038", "issue:43097|issue:43452", - "issue:44933|issue:46231", - "issue:45753|issue:46369", "issue:45003|issue:45341", + "issue:45753|issue:46369", "issue:43986|issue:45341", - "issue:43673|issue:46123", + "issue:44933|issue:46231", "issue:43874|issue:44263", + "issue:43673|issue:46123", "issue:43986|issue:44484", "issue:44117|issue:44987", - "issue:45440|issue:46172", "issue:44261|issue:44485", + "issue:45440|issue:46172", "issue:43986|issue:45372", "issue:43531|issue:45341", "issue:43064|issue:44315", - "issue:38617|issue:42757", "issue:43097|issue:45588", + "issue:38617|issue:42757", "issue:43097|issue:46050", "issue:44410|issue:45375", "issue:43502|issue:43992", "issue:44038|issue:44410", "issue:42890|issue:46298", - "issue:44062|issue:44117", "issue:38617|issue:44351", - "issue:43097|issue:43873" + "issue:44062|issue:44117", + "issue:43097|issue:43873", + "issue:44936|issue:46365", + "issue:44704|issue:45103", + "issue:43986|issue:44117", + "issue:44380|issue:46421", + "issue:43824|issue:43986" ], - "prepared_review_unit_hash": "92d74453a87a84b733721cba7ecc0e1e88b4e5e82200e09cfd0dfc55c52fb75b", + "prepared_review_unit_hash": "c8a3a0704c3b2594c779c281c6d41cf309eb076c84f64da72efaf0d887b021ab", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 35, - "serialized_chars": 37902, - "estimated_input_tokens": 9476, - "estimated_eval_tokens": 19208 + "soft_pair_count": 32, + "serialized_chars": 37186, + "estimated_input_tokens": 9297, + "estimated_eval_tokens": 18850 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous single-issue reports; the soft edges are false positives driven by broad keyword similarity (imports, Flash Attention, Qwen/GLM/DeepSeek names, etc.) rather than shared root causes.", - "confidence": 0.95, - "canonical_issue_reason": "No single duplicate anchor stands out because the items cover unrelated bugs. If forced to pick one representative issue, issue:43873 is the most engaged and broadly impactful standalone bug report, but it is not a true canonical duplicate for the rest.", + "summary": "The items are mostly unrelated Transformer issues that only share broad keywords; none of the proposed soft pairs look like true duplicates or mergeable PR equivalents.", + "confidence": 0.84, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue:43873 is the strongest standalone issue in the set: open, high discussion/inbound references, and clearly framed as a real product bug. It is the best representative issue, though not a cluster-wide duplicate target.", + "best_issue_reason": "issue:43097 is the strongest standalone representative: it describes a concrete v5 regression, has discussion/inbound references, and is more actionable than the many one-off import/load errors. Still, this is not a true duplicate center for the full set.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43874", - "right": "issue:44261", - "accept": false, - "reason": "Different bugs: a missing image-processor method vs an MLA layernorm precision/config issue." - }, - { - "left": "issue:44263", - "right": "issue:44485", - "accept": false, - "reason": "Unrelated code paths: a torch.split indexer issue vs a GLM-5 RoPE implementation question." - }, - { - "left": "issue:41628", - "right": "issue:45071", - "accept": false, - "reason": "Both are import/type issues, but for different symbols and subsystems; not the same underlying defect." - }, - { - "left": "issue:45071", - "right": "issue:45542", - "accept": false, - "reason": "PretrainedConfig type checking regression is unrelated to the TensorFlow/tensorboard backend import problem." - }, - { - "left": "issue:27957", - "right": "issue:43352", - "accept": false, - "reason": "Both mention Flash Attention 2, but they affect different models and failure modes." - }, - { - "left": "issue:44484", - "right": "issue:45478", - "accept": false, - "reason": "A doc/question about shard size is unrelated to a Qwen3.5 MoE from_pretrained loading error." - }, - { - "left": "issue:44263", - "right": "issue:46172", - "accept": false, - "reason": "Different failures: torch.split return semantics vs a DeepSeek tensor transpose bug." - }, - { - "left": "issue:44360", - "right": "issue:46172", - "accept": false, - "reason": "A missing ReLU in the DSA indexer is unrelated to DeepSeek v4 tensor shaping." - }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Both are Qwen-family, but one is a model-loading weight mismatch and the other is missing projections in layer 0." + "reason": "Both are model-shape/config complaints for different Qwen/vision models; different symptoms and code paths." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "Both involve FlashAttention, but one is model compatibility and the other is degenerate training behavior with FA3." + "reason": "Both mention attention backends, but one is a Ministral FlashAttention failure and the other is a sequence-classification training degeneration; not the same bug." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Similar model family, but one is a shape-mismatch load failure and the other is a distinct Transformers 5/Qwen3-VL-Moe bug." + "reason": "Different Qwen-VL model families and different loading failures; only broad VL overlap." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "API removal for tie_embeddings is unrelated to gguf_file breaking tokenizer/model loading." + "reason": "One is a removed trainer/model API in v5, the other is gguf_file tokenizer/model loading; unrelated code paths." }, { - "left": "issue:44933", - "right": "issue:46231", + "left": "issue:45003", + "right": "issue:45341", "accept": false, - "reason": "A missing import and a docs parameter-name mismatch are not the same bug." + "reason": "Both involve code bugs, but one is unsafe sys.modules access and the other is a small testing_utils issue; no shared underlying defect." }, { "left": "issue:45753", "right": "issue:46369", "accept": false, - "reason": "Qwen3_5 infinite-loop generation and SAM memory overflow are different model/runtime failures." - }, - { - "left": "issue:45003", - "right": "issue:45341", - "accept": false, - "reason": "Unsafe sys.modules access is a runtime import issue; testing_utils has a separate small bug." + "reason": "Both are model-specific runtime problems, but one is a Qwen image infinite loop and the other is SAM memory overflow; different causes." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "Video model loading without torchvision is unrelated to the small testing_utils bug." + "reason": "Video-model processor import crash vs a minor testing utility bug; no overlap beyond general bugs." }, { - "left": "issue:43673", - "right": "issue:46123", + "left": "issue:44933", + "right": "issue:46231", "accept": false, - "reason": "Generation cache missing during chunked prefill is not the same as MaskGenerationPipeline dropping the last partial batch." + "reason": "One is an import failure from image_utils, the other is a documentation parameter-name mismatch." }, { "left": "issue:43874", "right": "issue:44263", "accept": false, - "reason": "Different models and symptoms: missing GLM46V processor method vs torch.split return values." + "reason": "Both are GLM-related, but one is a missing image-patch method and the other is torch.split return handling; distinct failures." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Both mention cache/prefill behavior, but one is missing GenerationMixin cache in v5 and the other is mask pipeline batch handling." }, { "left": "issue:43986", "right": "issue:44484", "accept": false, - "reason": "A video-loading import failure is unrelated to a save_pretrained shard-size question." + "reason": "Processor import-time crash vs save_pretrained shard-size question; unrelated." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Both touch loading, but one is a tokenizer-mapping None bug and the other is a specific model load failure in v5.1.0." + "reason": "Tokenizer-mapping None handling vs model load failure on a specific package/version; different layers of the stack." }, { - "left": "issue:45440", - "right": "issue:46172", + "left": "issue:44261", + "right": "issue:44485", "accept": false, - "reason": "DeepseekV3MoE divergence and deepseek-v4 tensor transpose are different model-specific issues." + "reason": "Both touch precision/config in attention/rope-style math, but one is a missing rms_norm_eps in MLA and the other is GLM-5 RoPE implementation discussion; not the same bug." }, { - "left": "issue:44261", - "right": "issue:44485", + "left": "issue:45440", + "right": "issue:46172", "accept": false, - "reason": "MLA q_a_layernorm precision/config issues do not match GLM-5 RoPE implementation concerns." + "reason": "Different DeepSeek issues: one is native-vs-remote implementation divergence, the other is a specific dtype/transpose bug." }, { "left": "issue:43986", "right": "issue:45372", "accept": false, - "reason": "Loading a video model without torchvision and a mistral_common import error breaking Gemma 4 are unrelated." + "reason": "Both are loading-time import problems, but one concerns missing torchvision for video processors and the other missing ReasoningEffort from mistral_common." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to the testing_utils bug." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to a small testing utility bug." }, { "left": "issue:43064", "right": "issue:44315", "accept": false, - "reason": "Trainer/FSDP optimizer-state corruption is a different failure mode from Liger Kernel not being applied with model_init." + "reason": "Trainer/FSDP2 optimizer state corruption is unrelated to Liger Kernel not being applied with model_init." }, { - "left": "issue:38617", - "right": "issue:42757", + "left": "issue:43097", + "right": "issue:45588", "accept": false, - "reason": "Both are import errors, but they involve different missing names from different packages and releases." + "reason": "Trainer API removal regression vs flash_attention integration crash on s_aux=None; different subsystems and fixes." }, { - "left": "issue:43097", - "right": "issue:45588", + "left": "issue:38617", + "right": "issue:42757", "accept": false, - "reason": "A removed config helper and a flash_attention.py s_aux AttributeError are unrelated code paths." + "reason": "Two distinct import errors for different symbols from different packages; only superficial similarity." }, { "left": "issue:43097", "right": "issue:46050", "accept": false, - "reason": "The tie_embeddings removal regression is unrelated to quantized save/load breakage." + "reason": "Trainer v5 API regression vs quantized model save/load corruption after sinq; not the same underlying defect." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Missing attention projections in qwen3next and a silently dropped config field in Qwen3_5MoeVisionConfig are different bugs." + "reason": "Both are Qwen vision configs, but one is missing projections at layer 0 and the other is a missing config field silently dropped by strict validation." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "local_files_only network leakage and a missing embed_tokens.weight on UMT5Encoder are unrelated." + "reason": "Local-files-only network leak vs missing embed_tokens.weight on UMT5Encoder; different loading bugs." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "Qwen3-VL-Moe loading issues are distinct from qwen3next missing attn_qkv/attn_gate projections." + "reason": "Both are Qwen-related, but one is a Transformers v5 bug report and the other is a specific config/projection omission." }, { "left": "issue:42890", "right": "issue:46298", "accept": false, - "reason": "A nondeterministic test due to missing set_seed is unrelated to a noisy logging message." + "reason": "Integration-test flakiness from missing set_seed is unrelated to a noisy logging message." }, { - "left": "issue:44062", - "right": "issue:44117", + "left": "issue:38617", + "right": "issue:44351", "accept": false, - "reason": "Duplicate AddedToken kwargs and TOKENIZER_MAPPING_NAMES returning None are different tokenizer-loading defects." + "reason": "Different import failures: layer_type_validation vs HybridCache; same symptom class, different symbol and likely different fixes." }, { - "left": "issue:38617", - "right": "issue:44351", + "left": "issue:44062", + "right": "issue:44117", "accept": false, - "reason": "Both are import errors, but they concern different missing exports and likely different fix paths." + "reason": "Tokenizer AddedToken duplicate kwargs bug is unrelated to TOKENIZER_MAPPING_NAMES returning None." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "A v5 API removal issue is unrelated to quantization/offloading behavior." + "reason": "Trainer state regression vs offloading/quantization behavior; not the same code-path problem." + }, + { + "left": "issue:44936", + "right": "issue:46365", + "accept": false, + "reason": "Trainer.evaluate after train is a runtime state issue; the other is a Trainer import problem influenced by datasets installation." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "AutoProcessor cached_file kwargs forwarding vs auto_docstring annotation introspection crash; different functionality." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Video processor import crash vs tokenizer mapping None handling; unrelated." + }, + { + "left": "issue:44380", + "right": "issue:46421", + "accept": false, + "reason": "GPT2 attention scaling ignored in backends vs cache lazy initialization causing recompilation; different bugs." + }, + { + "left": "issue:43824", + "right": "issue:43986", + "accept": false, + "reason": "Both are import/loading issues, but one is a missing Qwen2.5-VL class export and the other is a torchvision-dependent video processor crash." } ] }, @@ -15194,8 +15407,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 49, - "review_unit_count": 54, + "review_unit_index": 50, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -15207,6 +15420,7 @@ "issue:38617", "issue:42175", "issue:42490", + "issue:42757", "issue:42907", "issue:43012", "issue:43064", @@ -15218,16 +15432,18 @@ "issue:43526", "issue:43550", "issue:43582", + "issue:43618", + "issue:43761", + "issue:43824", "issue:43825", "issue:43906", "issue:43976", - "issue:43986", "issue:43992", - "issue:44016", + "issue:44038", "issue:44077", - "issue:44117", "issue:44162", "issue:44222", + "issue:44230", "issue:44368", "issue:44380", "issue:44483", @@ -15235,37 +15451,30 @@ "issue:44514", "issue:44655", "issue:44683", - "issue:44704", - "issue:44936", "issue:44938", - "issue:45103", "issue:45478", "issue:45479", "issue:45584", + "issue:45600", + "issue:45698", "issue:45712", "issue:45753", "issue:46132", "issue:46172", "issue:46178", "issue:46298", - "issue:46365", - "issue:46421" + "issue:46365" ], "soft_pairs": [ - "issue:44936|issue:46365", - "issue:44704|issue:45103", - "issue:44380|issue:46421", - "issue:43986|issue:44117", - "issue:43986|issue:44016", "issue:44938|issue:46178", "issue:43992|issue:44484", "issue:43066|issue:43906", "issue:43208|issue:43550", - "issue:43012|issue:43550", "issue:42907|issue:44655", + "issue:43012|issue:43550", "issue:43012|issue:44368", - "issue:43825|issue:44655", "issue:45712|issue:46298", + "issue:43825|issue:44655", "issue:42907|issue:45753", "issue:43352|issue:44380", "issue:43097|issue:46298", @@ -15273,238 +15482,250 @@ "issue:44162|issue:46172", "issue:45584|issue:46365", "issue:44380|issue:46132", - "issue:44380|issue:45479", "issue:44077|issue:46365", + "issue:44380|issue:45479", "issue:43064|issue:44380", - "issue:44222|issue:45478", "issue:36683|issue:43582", + "issue:44222|issue:45478", "issue:42175|issue:46365", "issue:42490|issue:43012", "issue:44483|issue:44514", "issue:43012|issue:44683", "issue:38617|issue:43502", - "issue:43208|issue:43526" + "issue:43208|issue:43526", + "issue:42757|issue:43824", + "issue:45600|issue:46298", + "issue:45698|issue:46172", + "issue:43618|issue:44380", + "issue:43618|issue:43761", + "issue:44038|issue:44230" ], - "prepared_review_unit_hash": "275fa364392e8a46f0bd917f6217b97043f703725bc7523e75b074c66335af8b", + "prepared_review_unit_hash": "ed1c8efd9d8089edde8cd53af94b7c77bd192d6006f87c733dc0653e24260705", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36579, - "estimated_input_tokens": 9145, - "estimated_eval_tokens": 18546 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37301, + "estimated_input_tokens": 9326, + "estimated_eval_tokens": 18908 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated standalone issues across different subsystems; none of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates.", - "confidence": 0.97, - "canonical_issue_reason": "No strong duplicate cluster is evident. If one issue must serve as an anchor, issue 43502 is the most concrete and broadly actionable standalone bug report among this set.", + "summary": "This cluster is a mix of distinct Transformers issues: import/API breakages, save/load regressions, v5 behavioral changes, and model-specific bugs. None of the soft-similarity pairs look like the same underlying defect, so they should stay separate.", + "confidence": 0.64, + "canonical_issue_reason": "issue:43097 is the broadest and most central regression report: a concrete v5 backward-compatibility break with clear user impact and inbound references.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43502 is a good representative standalone issue because it has a clear reproducible behavior, broader user impact, and substantial discussion compared with the others.", + "best_issue_reason": "issue:43097 is the strongest representative issue here because it describes a specific, high-impact v5 API regression rather than a narrow model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44936", - "right": "issue:46365", - "accept": false, - "reason": "Different failures: Trainer evaluation after training vs import failure tied to datasets installation." - }, - { - "left": "issue:44704", - "right": "issue:45103", - "accept": false, - "reason": "Unrelated code paths: cached_file kwargs passthrough vs auto-docstring crash under future annotations." - }, - { - "left": "issue:44380", - "right": "issue:46421", - "accept": false, - "reason": "Different bugs in different areas: GPT2 attention scaling vs cache lazy initialization recompilation." - }, - { - "left": "issue:43986", - "right": "issue:44117", - "accept": false, - "reason": "Video AutoProcessor/torchvision crash is unrelated to tokenizer mapping None handling." - }, - { - "left": "issue:43986", - "right": "issue:44016", - "accept": false, - "reason": "Different subsystems: AutoProcessor video loading vs notebook syntax error." - }, { "left": "issue:44938", "right": "issue:46178", "accept": false, - "reason": "Python 3.14 import/load issue is unrelated to undefined backend tensorflow_text." + "reason": "Python 3.14 load failure vs undefined backend mapping for tensorflow_text; different failure modes and code paths." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing embed_tokens.weight in UMT5Encoder is unrelated to save_pretrained shard size behavior." + "reason": "Missing embed_tokens.weight during UMT5 loading is unrelated to the 50GB shard-size question in save_pretrained()." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type is not the same as an isolated reproduction issue without matching details." + "reason": "Wrong tokenizer decoder type is a separate tokenizer regression from the isolated reproduction of another issue." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training bugs and Bamba SDPA compile failure are different model-specific problems." - }, - { - "left": "issue:43012", - "right": "issue:43550", - "accept": false, - "reason": "PyTorch bfloat16 warning is unrelated to Bamba SDPA compile failure." + "reason": "xLSTM training bugs and Bamba torch.compile+SDPA failure affect different models and different mechanics." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals is a different failure from saving Pipeline objects." + "reason": "Dequantized model saving and pipeline save_pretrained() are both save-related, but the concrete bugs differ." }, { "left": "issue:43012", - "right": "issue:44368", + "right": "issue:43550", "accept": false, - "reason": "Precision warning on compile is unrelated to tie_word_embeddings warning during Qwen3.5 LoRA fine-tuning." + "reason": "bfloat16 compile warning is just a warning issue; Bamba compile with SDPA is a functional model/backend failure." }, { - "left": "issue:43825", - "right": "issue:44655", + "left": "issue:43012", + "right": "issue:44368", "accept": false, - "reason": "Pipeline translation-task messaging and pipeline save_pretrained are unrelated issues." + "reason": "PyTorch precision warning and tie_word_embeddings warning are separate warning paths with different causes." }, { "left": "issue:45712", "right": "issue:46298", "accept": false, - "reason": "Leftover dummy classes in dir(transformers) is not the same as a noisy logging message." + "reason": "Left is leftover dummy-class cleanup; right is a noisy logging message. Different scope and fix." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "pipeline translation-task messaging is unrelated to pipeline save_pretrained() serialization behavior." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "Dequantized save failure and Qwen3_5 infinite loop are different bugs." + "reason": "Saving dequantized models and Qwen3_5 infinite-loop generation are unrelated model lifecycle vs inference bugs." }, { "left": "issue:43352", "right": "issue:44380", "accept": false, - "reason": "Nemotron FlashAttention support and GPT2 attention scaling under SDPA/FlashAttention are different code paths." + "reason": "Flash Attention support error for NemotronH vs GPT2 attention scaling ignored under SDPA/FlashAttention are different code paths." }, { "left": "issue:43097", "right": "issue:46298", "accept": false, - "reason": "Removed tie_embeddings_and_encoder_decoder API and a noisy logging message are unrelated." + "reason": "A v5 API removal regression is not the same as an unspecified noisy logging message." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "FlashAttention support bug is unrelated to the Python 3.9/3.10 compatibility report." + "reason": "Nemotron Flash Attention support and Python 3.9/3.10 compatibility are distinct issues." }, { "left": "issue:44162", "right": "issue:46172", "accept": false, - "reason": "ESM2 breakage is unrelated to deepseek-v4 comb.to(dtype).transpose behavior." + "reason": "ESM2 training/model breakage and a DeepSeek tensor transpose bug are unrelated model-specific failures." }, { "left": "issue:45584", "right": "issue:46365", "accept": false, - "reason": "Whisper empty transcription failure is unrelated to importing Trainer with datasets installed." + "reason": "Whisper empty transcription after token alignment is unrelated to importing Trainer failing because of datasets." }, { "left": "issue:44380", "right": "issue:46132", "accept": false, - "reason": "GPT2 attention scaling ignored in backends is different from AttentionInterface.register changing function behavior." + "reason": "Ignoring GPT2 attention scaling and AttentionInterface.register behavior changes are different attention APIs and defects." }, { - "left": "issue:44380", - "right": "issue:45479", + "left": "issue:44077", + "right": "issue:46365", "accept": false, - "reason": "GPT2 SDPA/FlashAttention scaling bug is unrelated to degenerate zero-loss sequence classification." + "reason": "patchtsmixer post_init validation and Trainer import failure are unrelated subsystems." }, { - "left": "issue:44077", - "right": "issue:46365", + "left": "issue:44380", + "right": "issue:45479", "accept": false, - "reason": "patchtsmixer post_init policy is unrelated to the Trainer import issue." + "reason": "Attention scaling backend bug and degenerate zero-loss classification bug are different model behaviors." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption is a different training bug from GPT2 attention scaling." + "reason": "FSDP2/PEFT optimizer state mismatch and GPT2 attention scaling under SDPA are unrelated training vs inference bugs." }, { - "left": "issue:44222", - "right": "issue:45478", + "left": "issue:36683", + "right": "issue:43582", "accept": false, - "reason": "FP8 save_pretrained on MoE models and Qwen3.5 MoE from_pretrained errors are related by model family only, not the same concrete bug." + "reason": "Gemma3Config missing vocab_size and Apple Silicon caching_allocator_warmup TypeError are different model/runtime failures." }, { - "left": "issue:36683", - "right": "issue:43582", + "left": "issue:44222", + "right": "issue:45478", "accept": false, - "reason": "Gemma3Config vocab_size AttributeError is unrelated to Apple Silicon caching_allocator_warmup TypeError." + "reason": "FP8 save_pretrained() for moe and Qwen3.5 moe from_pretrained() error are related to the model family but not the same bug." }, { "left": "issue:42175", "right": "issue:46365", "accept": false, - "reason": "TensorFlow backend packaging mismatch is unrelated to the Trainer import/uninstall-datasets report." + "reason": "Tensorflow absent from a torch extra install is a packaging issue, not the same as Trainer import failing due to datasets." }, { "left": "issue:42490", "right": "issue:43012", "accept": false, - "reason": "Local path trailing-slash import warning is unrelated to the bfloat16 compile warning." + "reason": "Trailing-slash local model import warning and bfloat16 compile warning are unrelated warning paths." }, { "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "Chat completions request rejection and Qwen2_5_VLProcessor batched padding=False crash are different issues." + "reason": "Chat completions API acceptance regression and Qwen2_5_VLProcessor batching crash are different request/processor bugs." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "A precision warning during compilation is not the same as flex_attention failing on torch >= 2.9." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "Missing layer_type_validation import and unwanted API requests despite local_files_only are separate import vs network issues." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "xLSTM training blockers and BeitImageProcessorFast reduce_labels returning one label are unrelated." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Missing is_offline_mode import and missing Qwen2_5_VLForConditionalGeneration export are both import errors but from different symbols and fixes." + }, + { + "left": "issue:45600", + "right": "issue:46298", + "accept": false, + "reason": "auto_mappings.py stale config references and a noisy logging message are not the same defect." + }, + { + "left": "issue:45698", + "right": "issue:46172", + "accept": false, + "reason": "Wrong custom module loading after save_pretrained() and a DeepSeek tensor transpose bug are unrelated save/load vs tensor-shape issues." }, { - "left": "issue:43012", - "right": "issue:44683", + "left": "issue:43618", + "right": "issue:44380", "accept": false, - "reason": "bfloat16 precision warning and compiled flex_attention failure on torch>=2.9 are distinct backend problems." + "reason": "CLIPOutput attentions assignment regression is separate from GPT2 attention scaling being ignored under backend kernels." }, { - "left": "issue:38617", - "right": "issue:43502", + "left": "issue:43618", + "right": "issue:43761", "accept": false, - "reason": "ImportError for layer_type_validation is unrelated to local_files_only still making API requests." + "reason": "Missing CLIP attentions output and CLIPVisionModel hidden_states regression are adjacent but distinct output-field bugs." }, { - "left": "issue:43208", - "right": "issue:43526", + "left": "issue:44038", + "right": "issue:44230", "accept": false, - "reason": "xLSTM training bugs and BeitImageProcessorFast reduce_labels returning one label are unrelated." + "reason": "Qwen3-VL-Moe version bug and fp8 Qwen3.5 moe inference support are different model/version problems." } ] }, @@ -15519,8 +15740,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 50, - "review_unit_count": 54, + "review_unit_index": 51, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -15530,7 +15751,9 @@ "nodes": [ "issue:36010", "issue:36683", + "issue:41669", "issue:42757", + "issue:42890", "issue:42898", "issue:42994", "issue:43010", @@ -15543,19 +15766,15 @@ "issue:43611", "issue:43618", "issue:43756", - "issue:43761", - "issue:43824", "issue:43867", "issue:43901", "issue:43937", "issue:43976", "issue:43992", - "issue:44038", "issue:44075", "issue:44077", - "issue:44230", + "issue:44246", "issue:44355", - "issue:44380", "issue:44483", "issue:44485", "issue:44492", @@ -15563,44 +15782,40 @@ "issue:44655", "issue:44855", "issue:44938", + "issue:45071", "issue:45216", "issue:45276", "issue:45278", "issue:45290", "issue:45310", + "issue:45357", + "issue:45375", + "issue:45561", "issue:45593", - "issue:45600", - "issue:45698", + "issue:45850", "issue:46077", "issue:46082", - "issue:46172", - "issue:46298", + "issue:46202", "issue:46365" ], "soft_pairs": [ - "issue:42757|issue:43824", - "issue:45600|issue:46298", - "issue:45698|issue:46172", - "issue:43618|issue:44380", - "issue:43618|issue:43761", - "issue:44038|issue:44230", "issue:44483|issue:45290", "issue:43299|issue:43867", "issue:42994|issue:44655", - "issue:42898|issue:43066", "issue:36683|issue:43475", + "issue:42898|issue:43066", "issue:44625|issue:46365", "issue:43756|issue:44485", - "issue:43352|issue:43937", "issue:45276|issue:46077", - "issue:43901|issue:44655", - "issue:43611|issue:44655", + "issue:43352|issue:43937", "issue:36010|issue:42757", + "issue:43611|issue:44655", + "issue:43901|issue:44655", "issue:36683|issue:44077", "issue:43867|issue:45216", "issue:43976|issue:44938", - "issue:43352|issue:45593", "issue:44625|issue:46077", + "issue:43352|issue:45593", "issue:46077|issue:46082", "issue:36683|issue:43525", "issue:43618|issue:45216", @@ -15609,235 +15824,252 @@ "issue:43064|issue:44492", "issue:43010|issue:45278", "issue:44075|issue:46365", - "issue:44355|issue:44855" + "issue:44355|issue:44855", + "issue:43867|issue:45375", + "issue:43867|issue:46202", + "issue:44938|issue:45071", + "issue:36683|issue:43064", + "issue:41669|issue:44246", + "issue:42890|issue:45561", + "issue:43010|issue:45850", + "issue:43867|issue:45357" ], - "prepared_review_unit_hash": "2690d11b039c7435a0e69758a961972a9ff78fac065afda86a3ccd4b83f3cd0b", + "prepared_review_unit_hash": "9e8a036bad02e39284828bb4a3c20e9d7655788f526cf74d6a5b0eecf421c81e", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 36800, - "estimated_input_tokens": 9200, - "estimated_eval_tokens": 18656 + "soft_pair_count": 34, + "serialized_chars": 37463, + "estimated_input_tokens": 9366, + "estimated_eval_tokens": 18988 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a grab bag of Transformers issues with only superficial overlap around imports, save/load, and v5 regressions. The soft pairs mostly share broad wording but do not look like the same underlying bug or change.", - "confidence": 0.42, - "canonical_issue_reason": "issue:45698 is the broadest and most central save/load regression in the set: it describes a concrete from_pretrained/save_pretrained failure, has strong discussion activity and references, and is more representative than the many model-specific or unrelated issues.", + "summary": "Most pairs are related only by broad theme (imports, saving, loading, config regressions) and are not the same concrete bug. I accepted only the import-time slowdown pair; the rest look like distinct issues.", + "confidence": 0.62, + "canonical_issue_reason": "issue:45278 is the broadest representative here: it explicitly summarizes widespread import failures after upgrading to 5.5.0, which is the closest umbrella for the import-regression slice of this set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45698 is the best single issue to anchor this cluster because it is actionable, user-facing, and covers a core serialization path that is closer to the cluster's recurring load/save regressions.", + "best_issue_reason": "issue:45278 is the best overall issue representative because it captures the most general shared symptom in the cluster (post-upgrade import breakage), even though it is broader than most concrete reports.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:42757", - "right": "issue:43824", - "accept": false, - "reason": "Both are import errors, but for different symbols, different package boundaries, and likely different root causes." - }, - { - "left": "issue:45600", - "right": "issue:46298", - "accept": false, - "reason": "One is a CI break from stale auto_mappings references; the other is a noisy logging message. Not the same bug." - }, - { - "left": "issue:45698", - "right": "issue:46172", - "accept": false, - "reason": "Save/load custom-module regression vs a DeepSeek tensor transpose bug; unrelated code paths." - }, - { - "left": "issue:43618", - "right": "issue:44380", - "accept": false, - "reason": "Both involve attention-related outputs, but CLIP attentions missing and GPT-2 attention scaling being ignored are different model/backend bugs." - }, - { - "left": "issue:43618", - "right": "issue:43761", - "accept": false, - "reason": "Same model family, but one loses attentions and the other loses hidden_states; likely separate output-assembly regressions." - }, - { - "left": "issue:44038", - "right": "issue:44230", - "accept": false, - "reason": "Regression loading Qwen3-VL/MoE models vs a feature request for FP8 inference support; not the same issue." - }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "API request validation failure and chat-template tool-call crash are different layers and different failure modes." + "reason": "Both are chat-related crashes, but one is /v1/chat/completions request handling and the other is apply_chat_template with tool-call messages; different code paths." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Both are model loading problems, but one is Qwen3-VL-MoE loading and the other is a sorted state_dict load error. Not the same bug." + "reason": "Both are model-loading problems, but one is a Qwen3VL MoE loading regression and the other is a state_dict-sorting load error; not the same bug." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving failure and pipeline save_pretrained failure are different serialization paths." + "reason": "Quantized model saving and Pipeline.save_pretrained are different serialization paths with different failure modes." }, { - "left": "issue:42898", - "right": "issue:43066", + "left": "issue:36683", + "right": "issue:43475", "accept": false, - "reason": "Tokenizer cleanup behavior change vs wrong decoder type; related area, but not the same defect." + "reason": "Missing Gemma3Config.vocab_size and missing Sam3VisionEncoderOutput.fpn_position_embeddings are unrelated attribute bugs." }, { - "left": "issue:36683", - "right": "issue:43475", + "left": "issue:42898", + "right": "issue:43066", "accept": false, - "reason": "Two AttributeErrors on different model objects and different missing fields; no shared underlying change." + "reason": "Tokenizer behavior change and wrong decoder type are both v5 tokenizer regressions, but they are distinct concrete issues." }, { "left": "issue:44625", "right": "issue:46365", "accept": false, - "reason": "A config propagation bug and an import-order issue with Trainer/datasets are unrelated." + "reason": "num_labels propagation and Trainer import being blocked by datasets are unrelated problems." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "RoPE-layer dropping in Smollm3 vs a GLM-5 RoPE implementation discussion; different models and concrete problems." + "reason": "Both mention RoPE, but one is Smollm3 dropping layers and the other is a GLM-5 RoPE discussion; different bugs." }, { - "left": "issue:43352", - "right": "issue:43937", + "left": "issue:45276", + "right": "issue:46077", "accept": false, - "reason": "Flash-Attention support gating for Nemotron vs invalid GenerationConfig for GLM-5 are unrelated." + "reason": "Both are Gemma4 regressions, but one is resize_token_embeddings behavior and the other is attention-mask construction." }, { - "left": "issue:45276", - "right": "issue:46077", + "left": "issue:43352", + "right": "issue:43937", "accept": false, - "reason": "Both are Gemma4 issues, but embedding resize propagation and attention-mask construction are different code paths." + "reason": "Flash-Attention support and invalid GenerationConfig are unrelated failures." }, { - "left": "issue:43901", - "right": "issue:44655", + "left": "issue:36010", + "right": "issue:42757", "accept": false, - "reason": "Documentation mismatch for a pipeline argument vs pipeline save_pretrained failure; not the same underlying issue." + "reason": "Both are import errors, but they concern different missing symbols from different dependencies and do not share a concrete root cause." }, { "left": "issue:43611", "right": "issue:44655", "accept": false, - "reason": "Base-model-prefix loading regression and pipeline saving failure are separate serialization bugs." + "reason": "Loading models with base_model_prefix and saving Pipeline objects are different code paths and different regressions." }, { - "left": "issue:36010", - "right": "issue:42757", + "left": "issue:43901", + "right": "issue:44655", "accept": false, - "reason": "Both are import errors, but they concern different missing exports and different dependency changes." + "reason": "This is a docs mismatch versus a save_pretrained bug; not the same underlying issue." }, { "left": "issue:36683", "right": "issue:44077", "accept": false, - "reason": "Gemma3 config attribute loss vs patchtsmixer post_init policy change; unrelated." + "reason": "A missing config attribute and ignored optimizer SGD args are unrelated." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "Both are load/save problems, but one is state_dict ordering and the other is a Qwen3.5 checkpoint regression." + "reason": "Generic load_model failure from sorted state_dict and Qwen3.5 save_pretrained corruption are distinct bugs." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Different Python-version compatibility failures with no shared code path." + "reason": "Both are Python-version compatibility reports, but they target different versions and different failure modes." }, { - "left": "issue:43352", - "right": "issue:45593", + "left": "issue:44625", + "right": "issue:46077", "accept": false, - "reason": "Flash-attention support issue and D-FINE auxiliary-loss behavior are unrelated." + "reason": "num_labels propagation and Gemma4 attention-mask logic are unrelated." }, { - "left": "issue:44625", - "right": "issue:46077", + "left": "issue:43352", + "right": "issue:45593", "accept": false, - "reason": "Num-labels propagation and Gemma4 bidirectional attention are separate model-config bugs." + "reason": "Flash-Attention support and D-FINE auxiliary-loss handling are unrelated model issues." }, { "left": "issue:46077", "right": "issue:46082", "accept": false, - "reason": "Gemma4 attention-mask construction and LlamaConfig head_dim validation are different config/runtime checks." + "reason": "Gemma4 attention masks and LlamaConfig head_dim validation are different code paths." }, { "left": "issue:36683", "right": "issue:43525", "accept": false, - "reason": "Different config attributes missing on different models; too generic to be the same bug." + "reason": "Gemma3Config.vocab_size and Llama4Config.pad_token_id are different missing fields in different models." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "CLIP attentions regression and Qwen3.5 checkpoint save regression are unrelated." + "reason": "CLIPOutput attentions regression and Qwen3.5 checkpoint saving regression are unrelated." }, { "left": "issue:36683", "right": "issue:44625", "accept": false, - "reason": "Different model config fields with no shared failure mechanism." + "reason": "Both involve configs, but one is a missing attribute and the other is num_labels propagation; not the same bug." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "UMT5 weight-loading miss and Qwen3.5 MoE from_pretrained error are separate loading bugs." + "reason": "UMT5 weight loading and Qwen3.5 MoE from_pretrained failure are different model-loading issues." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Distributed optimizer-state corruption vs a typo in cache strategy docs/code; unrelated." + "reason": "Trainer optimizer-state corruption and a cache-strategy typo are unrelated." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "A no_grad decorator request for cache updates and a broad import-error report are not the same defect." + "reason": "@torch.no_grad on cache updates and broad import errors after upgrade are different problems." }, { "left": "issue:44075", "right": "issue:46365", "accept": false, - "reason": "Optimizer argument handling vs Trainer import breakage from datasets are different issues." + "reason": "Ignored SGD args and Trainer import blocked by datasets are unrelated." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Both mention import/runtime errors, but one is compiled Python files and the other is a Python 3.13 torch.jit parsing problem; not the same root cause." + "reason": "Compiled Python file errors and the Python 3.13 @torch.jit.script parsing/IndentationError regression are different causes." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "Generic load_model failure and a missing Qwen3.5 vision config field are unrelated." + }, + { + "left": "issue:43867", + "right": "issue:46202", + "accept": false, + "reason": "Load error from state_dict sorting and Gemma pipeline roundtrip failure on \u2581 are unrelated." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python 3.14 load failure and PretrainedConfig type-checking regression are distinct issues." + }, + { + "left": "issue:36683", + "right": "issue:43064", + "accept": false, + "reason": "Missing config attribute and distributed optimizer-state bug are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "Both describe the same import-performance regression: importing transformers is slow, and 41669 identifies import-star usage in models as the cause." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "A missing set_seed in integration tests and an xdist race on captured_info.txt are different testing issues." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "Cache update decoration and a broad 'v5 breaks too many models' complaint are not the same concrete bug." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: issue:45278 is indeed the broad import-regression umbrella, and the one accepted pair (issue:41669 vs issue:44246) is well supported by the issue titles. The remaining soft-edge verdicts are conservative and do not overstate shared root cause." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -15845,8 +16077,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 51, - "review_unit_count": 54, + "review_unit_index": 52, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -15854,64 +16086,57 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:27957", "issue:28282", "issue:30333", + "issue:30990", + "issue:31795", "issue:33453", + "issue:34689", "issue:36296", - "issue:36683", - "issue:41669", "issue:42548", + "issue:42831", "issue:42890", "issue:43010", - "issue:43064", "issue:43065", "issue:43232", + "issue:43299", "issue:43502", "issue:43575", "issue:43576", "issue:43618", "issue:43723", - "issue:43867", + "issue:43827", + "issue:43874", "issue:43976", "issue:43992", "issue:44062", "issue:44077", "issue:44112", "issue:44222", - "issue:44246", + "issue:44315", "issue:44351", "issue:44355", "issue:44360", "issue:44387", "issue:44485", "issue:44556", + "issue:44655", "issue:44749", "issue:44857", - "issue:44938", + "issue:44961", "issue:44964", "issue:45071", "issue:45125", - "issue:45357", "issue:45375", "issue:45440", "issue:45478", - "issue:45561", "issue:45600", - "issue:45850", "issue:46132", - "issue:46202", "issue:46285" ], "soft_pairs": [ - "issue:43867|issue:45375", - "issue:43867|issue:46202", - "issue:44938|issue:45071", - "issue:36683|issue:43064", - "issue:41669|issue:44246", - "issue:42890|issue:45561", - "issue:43010|issue:45850", "issue:43992|issue:45478", - "issue:43867|issue:45357", "issue:43010|issue:43976", "issue:36296|issue:43575", "issue:43618|issue:44077", @@ -15935,243 +16160,269 @@ "issue:30333|issue:33453", "issue:45125|issue:45375", "issue:43010|issue:43232", - "issue:43502|issue:44222" - ], - "prepared_review_unit_hash": "bb5f9daee5b7dff778e1fbe04e37c0a206ee230298072fa11c84a6bee2b639e4", - "component_budget": { - "node_count": 463, - "item_count": 463, - "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 - }, - "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 36543, - "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 still mostly heterogeneous. 45125 remains the best anchor because it names a concrete Qwen3.5 MoE tensor-parallel bug, but 45478 is only a plausible same-family loading regression from the title, not enough to accept as the same underlying issue.", - "confidence": 0.62, - "canonical_issue_reason": "Issue 45125 is the strongest canonical anchor because it identifies a specific failure mode (`_tp_plan` missing for Qwen3_5MoeForConditionalGeneration) and a concrete code path. It is also the closest topical neighbor to 45478, but the evidence is still too thin to treat that pair as a confirmed duplicate.", - "canonical_pr_reason": null, - "best_issue_reason": "45125 is the best global issue candidate because it is the most specific, actionable bug report in the set and the likeliest place a fix would land. The rest are either broader regressions, unrelated subsystems, or weaker title-only similarities.", - "best_pr_reason": null, - "soft_edge_verdicts": [ - { - "left": "issue:43867", - "right": "issue:45375", - "accept": false, - "reason": "Both are loading-related, but one is a generic state_dict/load error and the other is a missing vision config field. Different bugs." - }, - { - "left": "issue:43867", - "right": "issue:46202", - "accept": false, - "reason": "Different symptoms and code paths: model load/state_dict handling vs Gemma pipeline roundtrip tokenization." - }, - { - "left": "issue:44938", - "right": "issue:45071", - "accept": false, - "reason": "Python-version load failure and PretrainedConfig type-checking breakage are not the same bug." - }, - { - "left": "issue:36683", - "right": "issue:43064", - "accept": false, - "reason": "Missing `vocab_size` in Gemma3Config is unrelated to FSDP2/PEFT optimizer-state corruption." - }, - { - "left": "issue:41669", - "right": "issue:44246", - "accept": false, - "reason": "Both mention import cost, but one is a specific import-* regression and the other is an intermittent slowdown report." - }, - { - "left": "issue:42890", - "right": "issue:45561", - "accept": false, - "reason": "Test flakiness from missing `set_seed()` is different from a pytest-xdist race on `captured_info.txt`." - }, - { - "left": "issue:43010", - "right": "issue:45850", - "accept": false, - "reason": "A cache/layer `no_grad` change is unrelated to a broad complaint that v5 breaks too many models." - }, + "issue:30990|issue:31795", + "issue:43502|issue:44222", + "issue:43299|issue:44556", + "issue:43502|issue:44961", + "issue:42831|issue:43874", + "issue:30333|issue:36296", + "issue:44355|issue:45071", + "issue:43827|issue:44961", + "issue:44355|issue:44655", + "issue:33453|issue:34689", + "issue:44315|issue:44360", + "issue:27957|issue:33453" + ], + "prepared_review_unit_hash": "4c368fe1523b16e11ba2fa9285d455a872e75d92b218527766d6b30d2f44d4fc", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 37972, + "estimated_input_tokens": 9493, + "estimated_eval_tokens": 19242 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the issues cover unrelated loading, configuration, CI, quantization, and documentation problems. I found no soft-edge pair that looks like the same underlying bug/change, so none are accepted.", + "confidence": 0.97, + "canonical_issue_reason": "issue:33453 is the broadest and most discussion-heavy user-facing regression here, so it is the best anchor if one must be chosen; however, the set is too mixed to form a true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:33453 is the most representative standalone issue in the group because it describes a generic regression and has the highest discussion activity among the clearly bug-focused reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Both are Qwen/loader failures, but the titles point to different missing pieces: `embed_tokens.weight` vs a generic Qwen3.5 MoE `from_pretrained` error." - }, - { - "left": "issue:43867", - "right": "issue:45357", - "accept": false, - "reason": "Both are model-loading regressions, but one is a sorted-state_dict issue while the other is incorrect visual encoder save keys." + "reason": "Both are model-loading problems, but one is UMT5 encoder state loading and the other is Qwen3.5 MoE from_pretrained failure; different code paths and symptoms." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "Generation kwargs handling and Python compatibility are unrelated issues." + "reason": "Cache no_grad decoration vs a Python version compatibility regression; unrelated problems." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Tensor-parallel training bug and TP-related OOM on a specific model are not clearly the same failure." + "reason": "Tensor-parallel training bug and an OOM while loading a specific model are not the same failure mode." }, { "left": "issue:43618", "right": "issue:44077", "accept": false, - "reason": "CLIP attentions regression and `patchtsmixer` post_init validation are unrelated." + "reason": "CLIP attentions regression and patchtsmixer post_init restriction are unrelated model behaviors." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "PyTorch import missing vs tokenizer-loading regression are different root causes." + "reason": "Missing PyTorch import error vs tokenizer loading regression; different underlying causes." }, { "left": "issue:43618", "right": "issue:46285", "accept": false, - "reason": "Both touch CLIP, but they describe different structural regressions and different model variants." + "reason": "Both mention CLIP internals, but one is about attentions being dropped and the other about text model structure divergence; not the same bug." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "CLIP attentions assignment and LwDetrImageLoss AMP/CUDA crashes are unrelated." + "reason": "Attention output assignment and LwDetrImageLoss AMP crashes are unrelated." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "DSA/ReLU discussion is unrelated to DeepseekV3MoE implementation divergence." + "reason": "A missing ReLU in an indexer is a different issue from DeepseekV3MoE diverging from remote behavior." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "Local-files-only network leakage and `AddedToken` keyword collision are separate bugs." + "reason": "Local-files-only network requests and AddedToken argument duplication are separate bugs." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "A flaky integration test and a model-load failure for Phi-4 multimodal are different issues." + "reason": "A flaky SamHQ test and a Phi-4 multimodal loading failure are unrelated." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "Seed-related test instability is not the same as `auto_mappings.py` referencing removed configs." + "reason": "Test seeding flakiness and removed Sam3LiteText configs are different CI issues." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Test seed flakiness and CI device-override failure are different failures." + "reason": "Both involve tests, but they target different models and failure mechanisms." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Tokenizer loading regression and checkpoint reload after upgrade are related broadly, but not the same concrete bug." + "reason": "Tokenizer loading in v5 and checkpoint reload compatibility after upgrading versions are similar only at a high level." }, { "left": "issue:44222", "right": "issue:45125", "accept": false, - "reason": "FP8 save_pretrained on MoE and missing `_tp_plan` for tensor parallelism are different code paths." + "reason": "FP8 save_pretrained behavior and missing tensor-parallel plan are distinct MoE issues." }, { "left": "issue:44387", "right": "issue:44556", "accept": false, - "reason": "Int4 quantization memory regression is unrelated to checkpoint reload incompatibility." + "reason": "CUDA reserved-memory/OOM under int4 quantization is not the same as reload incompatibility after version upgrade." }, { "left": "issue:43010", "right": "issue:43065", "accept": false, - "reason": "Cache/layer no_grad proposal and dummy Conv2d in Sam3PixelDecoder are unrelated." + "reason": "Generation cache no_grad decoration and a dummy Conv2d in Sam3PixelDecoder are unrelated." }, { "left": "issue:43576", "right": "issue:46132", "accept": false, - "reason": "Broken `transformers env` command and AttentionInterface.register behavior change are distinct issues." + "reason": "Broken env command and AttentionInterface.register behavior change do not share a concrete code-path bug." }, { "left": "issue:43502", - "right": "issue:44485", + "right": "issue:44222", "accept": false, - "reason": "Network requests despite `local_files_only` and GLM-5 RoPE implementation are unrelated." + "reason": "Network access despite local_files_only and FP8 MoE save_pretrained are different concerns." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "The titles are plausibly related, but the packet does not show enough evidence to confirm the same tensor-parallel loading failure; treat as possible duplicate only." + "reason": "Both concern Qwen3.5 MoE loading, but one is missing _tp_plan for tensor parallelism while the other is a broader from_pretrained error; not enough evidence of the same fix." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Compiled Python file loading errors and post-upgrade slowdown are different problems." + "reason": "Compiled-file runtime errors and Chinese-described slow filtering after upgrade are unrelated." }, { "left": "issue:42548", "right": "issue:44351", "accept": false, - "reason": "Both are import-name errors, but they involve different missing exports and likely different fixes." + "reason": "Missing PreTrainedModel import and missing HybridCache import are different export regressions." }, { "left": "issue:30333", "right": "issue:33453", "accept": false, - "reason": "MLflow reporting and tokenizer loading regression are unrelated." + "reason": "MLFlow reporting on failed training and tokenizer loading regression are unrelated." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Missing tensor-parallel plan for Qwen3.5 MoE is unrelated to a missing vision-config field." + "reason": "Missing tensor-parallel plan and a missing vision config field are separate config/export issues." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Cache/layer `no_grad` decoration and generation sync_gpus handling are different bugs." + "reason": "Different generation-related bugs: cache update decoration vs sync_gpus update_model_kwargs handling." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Sentence Transformers loading hangs and forward-doc confusion are not the same issue." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Local-files-only network leakage and FP8 MoE save_pretrained are unrelated failures." + "reason": "Repeated pair in the data; still unrelated network-access and FP8 save_pretrained problems." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Qwen3VL MoE loading failure in dev0 and v4.57-to-v5.2/v5.3 reload breakage are different version-specific regressions." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "local_files_only network requests and a one-word issue title do not indicate the same bug." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FineGrainedFP8 accuracy and a missing image-patch count method are unrelated multimodal/quantization issues." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLFlow failure reporting and tensor-parallel training bugs are different subsystems and symptoms." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled-file errors and PretrainedConfig type checking regression are separate breakages." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs still referencing pipeline() and an unrelated one-word issue are not duplicates." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled Python file errors and inability to save Pipeline objects are distinct problems." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer loading regression and Llama 3.2 90B Vision model loading regression are both load-related, but they affect different models and failure points." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger kernel application and DSA indexer activation are unrelated." + }, + { + "left": "issue:27957", + "right": "issue:33453", + "accept": false, + "reason": "Flash Attention 2 support for XLMRoberta and tokenizer-loading regression are different issues." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The cluster-level summary is mostly grounded, but the one accepted soft-edge verdict overstates the evidence available in the packet. From the titles alone, 45125 and 45478 look plausibly related, but not enough to confidently call them the same tensor-parallel loading failure. The wording should stay more conservative (e.g. \u2018possible duplicate / likely same model family issue\u2019)." - }, - "evaluator_used": true, - "retried": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -16180,8 +16431,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 52, - "review_unit_count": 54, + "review_unit_index": 53, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -16189,80 +16440,69 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ - "issue:27957", - "issue:30333", - "issue:30990", "issue:31795", - "issue:33453", - "issue:34689", "issue:36296", "issue:38617", "issue:41628", + "issue:41669", + "issue:42371", "issue:42548", "issue:42757", - "issue:42831", - "issue:43299", - "issue:43502", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43525", "issue:43526", + "issue:43572", "issue:43576", "issue:43673", "issue:43824", - "issue:43827", - "issue:43874", "issue:43901", "issue:44038", "issue:44222", "issue:44230", + "issue:44246", "issue:44273", "issue:44315", - "issue:44355", - "issue:44360", "issue:44393", "issue:44485", "issue:44492", "issue:44556", "issue:44623", - "issue:44655", "issue:44908", "issue:44938", "issue:44961", "issue:45003", "issue:45071", + "issue:45092", "issue:45125", "issue:45230", + "issue:45278", "issue:45362", "issue:45468", "issue:45542", + "issue:45753", "issue:45850", + "issue:46095", "issue:46143", "issue:46144", - "issue:46209" + "issue:46209", + "issue:46218" ], "soft_pairs": [ - "issue:30990|issue:31795", - "issue:43299|issue:44556", - "issue:43502|issue:44961", - "issue:42831|issue:43874", - "issue:30333|issue:36296", - "issue:44355|issue:45071", - "issue:43827|issue:44961", - "issue:44355|issue:44655", - "issue:33453|issue:34689", - "issue:44315|issue:44360", - "issue:27957|issue:33453", "issue:42757|issue:43576", "issue:43576|issue:45071", "issue:43576|issue:43824", "issue:31795|issue:38617", "issue:44961|issue:45468", "issue:43576|issue:44938", - "issue:44908|issue:44961", "issue:44485|issue:44961", + "issue:44908|issue:44961", "issue:46143|issue:46144", "issue:44222|issue:44230", "issue:36296|issue:43901", - "issue:44315|issue:46209", "issue:41628|issue:42548", + "issue:44315|issue:46209", "issue:36296|issue:43526", "issue:43673|issue:44492", "issue:45362|issue:45850", @@ -16270,228 +16510,274 @@ "issue:44273|issue:45542", "issue:44038|issue:45125", "issue:44393|issue:45071", - "issue:44273|issue:45003" + "issue:44273|issue:45003", + "issue:44393|issue:45753", + "issue:43576|issue:44556", + "issue:43525|issue:45850", + "issue:41669|issue:43673", + "issue:44230|issue:46095", + "issue:44556|issue:45092", + "issue:43572|issue:45850", + "issue:44492|issue:45850", + "issue:45278|issue:45850", + "issue:42371|issue:44492", + "issue:41669|issue:46218", + "issue:43408|issue:44315", + "issue:42890|issue:44908", + "issue:43352|issue:44393", + "issue:44246|issue:44273", + "issue:43576|issue:44908" ], - "prepared_review_unit_hash": "1ff8e9e3d85648c221371dd2811c5173f9db721be767467ef465db57e897944f", + "prepared_review_unit_hash": "b23039210d877bcec8485f7e6176c6ccf1dc0763fdc51d68af328504a8ed692f", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 36297, - "estimated_input_tokens": 9075, - "estimated_eval_tokens": 18406 + "soft_pair_count": 37, + "serialized_chars": 37987, + "estimated_input_tokens": 9497, + "estimated_eval_tokens": 19250 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a heterogeneous Transformers issue cluster with a loose theme of v5 regressions, loading/import breakages, and a few model-specific bugs, but the pairwise candidates are not the same underlying defects. No PRs are present.", - "confidence": 0.82, - "canonical_issue_reason": "Issue 44556 is the clearest concrete version-regression report in the cluster: reloading a checkpoint trained on v4.57 fails after upgrading to v5.2/v5.3. It best represents the recurring \"v5 broke loading\" theme without being overly vague.", + "summary": "This cluster mixes several broad v5 migration/import regressions with many unrelated model-specific bugs. None of the soft pairs appear to be the same concrete underlying issue, so I would not merge any of them.", + "confidence": 0.64, + "canonical_issue_reason": "issue:45278 is the best umbrella issue here: it explicitly captures the v5-era import breakage pattern that most closely matches the cluster\u2019s import/lazy-loading failures.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44556 is the most representative of the cluster\u2019s broad regression theme while still being specific and actionable. The rest of the items are mostly separate model-, API-, or docs-level problems.", + "best_issue_reason": "issue:45278 is the strongest representative because it is broad, user-facing, and directly describes the dominant failure mode seen across the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:30990", - "right": "issue:31795", + "left": "issue:42757", + "right": "issue:43576", "accept": false, - "reason": "Different bugs: loading hangs for Sentence Transformers vs confusing model.forward docs." + "reason": "Different failures: huggingface_hub import vs transformers env CLI." }, { - "left": "issue:43299", - "right": "issue:44556", + "left": "issue:43576", + "right": "issue:45071", "accept": false, - "reason": "Both are v5-related loading regressions, but they affect different models and failure modes." + "reason": "Unrelated surfaces: env command vs PretrainedConfig type checking." }, { - "left": "issue:43502", - "right": "issue:44961", + "left": "issue:43576", + "right": "issue:43824", "accept": false, - "reason": "local_files_only network leakage vs an unrelated racoon issue." + "reason": "CLI regression is unrelated to a specific Qwen2.5-VL import export." }, { - "left": "issue:42831", - "right": "issue:43874", + "left": "issue:31795", + "right": "issue:38617", "accept": false, - "reason": "FP8 accuracy regression vs missing image-patch method in a specific processor." + "reason": "Docs confusion is unrelated to an ImportError for layer_type_validation." }, { - "left": "issue:30333", - "right": "issue:36296", + "left": "issue:44961", + "right": "issue:45468", "accept": false, - "reason": "MLflow reporting on training failure vs tensor-parallel training bug; unrelated code paths." + "reason": "No overlap: a generic titled issue vs Gemma audio positional encoding." }, { - "left": "issue:44355", - "right": "issue:45071", + "left": "issue:43576", + "right": "issue:44938", "accept": false, - "reason": "Compiled-file loading errors vs PretrainedConfig type-checking regression." + "reason": "env command failure is not the same as Python 3.14 load failure." }, { - "left": "issue:43827", + "left": "issue:44485", "right": "issue:44961", "accept": false, - "reason": "Docs still mention pipeline() vs unrelated racoon report." + "reason": "GLM RoPE discussion is unrelated to the placeholder racoon report." }, { - "left": "issue:44355", - "right": "issue:44655", + "left": "issue:44908", + "right": "issue:44961", "accept": false, - "reason": "Compiled Python file failures vs save_pretrained on Pipeline objects." + "reason": "Scheduler kwargs bug is unrelated to the racoon report." }, { - "left": "issue:33453", - "right": "issue:34689", + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "Both RoFormer-related, but one is kwargs forwarding and the other is attention interface usage." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "FP8 save_pretrained and FP8 inference support are different code paths." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor parallel training bug is unrelated to docs for return_all_scores." + }, + { + "left": "issue:41628", + "right": "issue:42548", "accept": false, - "reason": "Tokenizer-loading regression vs Llama 3.2 Vision model-loading breakage; different failures." + "reason": "Both are import errors, but for different missing symbols with no concrete shared bug shown." }, { "left": "issue:44315", - "right": "issue:44360", + "right": "issue:46209", "accept": false, - "reason": "Liger Kernel not applied in model_init vs DSA indexer missing ReLU." + "reason": "Liger Kernel application and B200 FP8 inference are unrelated issues." }, { - "left": "issue:27957", - "right": "issue:33453", + "left": "issue:36296", + "right": "issue:43526", "accept": false, - "reason": "Flash Attention 2 support for XLMRoberta vs tokenizer-loading regression." + "reason": "Tensor parallel training bug is unrelated to BeitImageProcessorFast label reduction." }, { - "left": "issue:42757", - "right": "issue:43576", + "left": "issue:43673", + "right": "issue:44492", "accept": false, - "reason": "Missing is_offline_mode import vs broken transformers env CLI; both import-ish, but not the same bug." + "reason": "Generation cache missing and cache-strategy typo are not the same bug." }, { - "left": "issue:43576", - "right": "issue:45071", + "left": "issue:45362", + "right": "issue:45850", "accept": false, - "reason": "Broken env command vs PretrainedConfig type-checking regression; same release family, different symptom." + "reason": "Specific Qwen chat crash is not the same as a broad v5 breakage complaint." }, { - "left": "issue:43576", - "right": "issue:43824", + "left": "issue:44623", + "right": "issue:45230", "accept": false, - "reason": "CLI env breakage vs missing Qwen2_5_VL export; different components." + "reason": "Missing processor.save_pretrained files is unrelated to a generic bug report." }, { - "left": "issue:31795", - "right": "issue:38617", + "left": "issue:44273", + "right": "issue:45542", "accept": false, - "reason": "Documentation confusion vs ImportError for layer_type_validation; unrelated." + "reason": "Lazy loading regression is unrelated to the undefined tf backend error." }, { - "left": "issue:44961", - "right": "issue:45468", + "left": "issue:44038", + "right": "issue:45125", "accept": false, - "reason": "racoon is unrelated to Gemma-4 audio positional encoding." + "reason": "Qwen3-VL-Moe support issue is unrelated to missing _tp_plan for tensor parallelism." }, { - "left": "issue:43576", - "right": "issue:44938", + "left": "issue:44393", + "right": "issue:45071", "accept": false, - "reason": "env CLI breakage vs Python 3.14 load failure; different root causes." + "reason": "2D bounding-box output bug is unrelated to PretrainedConfig type checking." }, { - "left": "issue:44908", - "right": "issue:44961", + "left": "issue:44273", + "right": "issue:45003", "accept": false, - "reason": "Scheduler kwargs bug vs unrelated racoon issue." + "reason": "Lazy loading problems may be related in theme, but sys.modules access is not the same concrete bug." }, { - "left": "issue:44485", - "right": "issue:44961", + "left": "issue:44393", + "right": "issue:45753", "accept": false, - "reason": "GLM-5 RoPE implementation discussion vs unrelated racoon issue." + "reason": "Different Qwen3-VL behaviors: bbox output error vs infinite loop on a specific image." }, { - "left": "issue:46143", - "right": "issue:46144", + "left": "issue:43576", + "right": "issue:44556", "accept": false, - "reason": "Both are RoFormer-related, but one is kwargs propagation and the other is attention-interface usage; not the same defect." + "reason": "Broken env command and checkpoint reload compatibility are different migration regressions." }, { - "left": "issue:44222", - "right": "issue:44230", + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "Llama4Config pad_token_id error is a specific config bug, not the broad v5 breakage report." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import * performance/refactor issue is unrelated to missing GenerationMixin cache." + }, + { + "left": "issue:44230", + "right": "issue:46095", "accept": false, - "reason": "FP8 save_pretrained bug vs FP8 inference support request; related area, different change." + "reason": "FP8 inference support and a weight-conversion-script question are unrelated." }, { - "left": "issue:36296", - "right": "issue:43901", + "left": "issue:44556", + "right": "issue:45092", "accept": false, - "reason": "Tensor-parallel training bug vs TextClassificationPipeline docs mismatch." + "reason": "Checkpoint reload incompatibility and remote-code meta-init incompatibility are different failure modes." }, { - "left": "issue:44315", - "right": "issue:46209", + "left": "issue:43572", + "right": "issue:45850", "accept": false, - "reason": "Liger Kernel/model_init bug vs FP8 inference failure on B200; unrelated." + "reason": "StableLm pad_token_idx regression is specific, while 45850 is only a broad complaint." }, { - "left": "issue:41628", - "right": "issue:42548", + "left": "issue:44492", + "right": "issue:45850", "accept": false, - "reason": "Different import errors: AutoImageProcessor vs PreTrainedModel." + "reason": "Cache-strategy typo is not the same as a generic v5 breakage report." }, { - "left": "issue:36296", - "right": "issue:43526", + "left": "issue:45278", + "right": "issue:45850", "accept": false, - "reason": "Tensor-parallel training bug vs BeitImageProcessorFast reduce_labels bug." + "reason": "Both are broad v5 regressions, but one is specifically import errors and the other is a generic complaint." }, { - "left": "issue:43673", + "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "Generation cache missing vs typo in cache strategies; cache-related but not the same bug." + "reason": "TF32 settings warning is unrelated to cache strategies typos." }, { - "left": "issue:45362", - "right": "issue:45850", + "left": "issue:41669", + "right": "issue:46218", "accept": false, - "reason": "Specific Qwen3.5 chat crash vs broad complaint that v5 breaks too many models." + "reason": "Import * cleanup/perf issue is unrelated to the Qwen3_5ForCausalLM import failure." }, { - "left": "issue:44623", - "right": "issue:45230", + "left": "issue:43408", + "right": "issue:44315", "accept": false, - "reason": "Missing processor.save_pretrained files vs generic bug report." + "reason": "sam3 tracker/model-type warning is unrelated to Liger Kernel application." }, { - "left": "issue:44273", - "right": "issue:45542", + "left": "issue:42890", + "right": "issue:44908", "accept": false, - "reason": "Lazy loading issue vs tensorboard-only installation causing tf backend error." + "reason": "Seed-related test flakiness is unrelated to inverse_sqrt scheduler kwargs." }, { - "left": "issue:44038", - "right": "issue:45125", + "left": "issue:43352", + "right": "issue:44393", "accept": false, - "reason": "Qwen3-VL-Moe breakage vs missing _tp_plan for Qwen3_5Moe; same model family, different code path." + "reason": "Flash Attention 2 support and 2D bounding-box output are distinct model issues." }, { - "left": "issue:44393", - "right": "issue:45071", + "left": "issue:44246", + "right": "issue:44273", "accept": false, - "reason": "Qwen3-VL bbox output problem vs PretrainedConfig type-checking regression." + "reason": "Slow import timing and lazy loading failure are related broadly, but not the same concrete bug." }, { - "left": "issue:44273", - "right": "issue:45003", + "left": "issue:43576", + "right": "issue:44908", "accept": false, - "reason": "Lazy loading bug vs unsafe sys.modules access in modeling_utils." + "reason": "env CLI regression is unrelated to scheduler kwargs handling." } ] }, @@ -16506,8 +16792,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 53, - "review_unit_count": 54, + "review_unit_index": 54, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -16515,24 +16801,26 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-463", "nodes": [ + "issue:31515", "issue:31795", "issue:33453", + "issue:34689", "issue:41669", "issue:42371", - "issue:42890", + "issue:42913", + "issue:43066", + "issue:43298", "issue:43352", - "issue:43408", + "issue:43502", "issue:43519", - "issue:43525", - "issue:43572", "issue:43576", - "issue:43673", + "issue:43606", "issue:43704", "issue:43827", "issue:43976", "issue:44077", - "issue:44230", - "issue:44246", + "issue:44261", + "issue:44263", "issue:44273", "issue:44295", "issue:44297", @@ -16542,21 +16830,19 @@ "issue:44393", "issue:44485", "issue:44492", - "issue:44556", + "issue:44623", "issue:44655", "issue:44704", "issue:44829", "issue:44861", "issue:44908", + "issue:44998", "issue:45071", - "issue:45092", "issue:45230", - "issue:45278", "issue:45310", "issue:45341", "issue:45468", "issue:45478", - "issue:45753", "issue:45850", "issue:46018", "issue:46095", @@ -16565,292 +16851,336 @@ "issue:46328" ], "soft_pairs": [ - "issue:44393|issue:45753", - "issue:43576|issue:44556", - "issue:44230|issue:46095", - "issue:43525|issue:45850", - "issue:41669|issue:43673", - "issue:44556|issue:45092", - "issue:43572|issue:45850", - "issue:44492|issue:45850", - "issue:45278|issue:45850", - "issue:42371|issue:44492", - "issue:41669|issue:46218", - "issue:43408|issue:44315", - "issue:42890|issue:44908", - "issue:43352|issue:44393", - "issue:43576|issue:44908", - "issue:44246|issue:44273", "issue:43976|issue:44393", "issue:44908|issue:45071", - "issue:44297|issue:45230", "issue:44295|issue:45230", + "issue:44297|issue:45230", "issue:44704|issue:44908", "issue:41669|issue:44351", "issue:44492|issue:46328", "issue:41669|issue:45341", - "issue:44393|issue:45478", "issue:43976|issue:45850", + "issue:44393|issue:45478", "issue:43976|issue:46218", "issue:44077|issue:44861", "issue:43827|issue:44908", "issue:31795|issue:33453", "issue:44315|issue:44829", "issue:43576|issue:44273", - "issue:44360|issue:45468", "issue:44655|issue:44861", + "issue:44360|issue:45468", "issue:45230|issue:45310", "issue:44273|issue:44908", "issue:43576|issue:46218", "issue:43519|issue:44485", "issue:43519|issue:44908", "issue:46018|issue:46172", - "issue:41669|issue:43704" + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:31515|issue:31795", + "issue:43704|issue:44485", + "issue:44485|issue:45468", + "issue:45850|issue:46218", + "issue:42371|issue:43704", + "issue:43502|issue:43519", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:43298|issue:46095", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:44261|issue:45468", + "issue:44351|issue:46218", + "issue:43352|issue:46218", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44492|issue:44998" ], - "prepared_review_unit_hash": "ae8d46b9de7b3f08b603be46fdb0d98cd6ffe2bd39db723101156bdce74a04b1", + "prepared_review_unit_hash": "360dc2b5de1027460ac5cc1603d312c50857f79b995ca013afb900c4d4850a06", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 41, - "serialized_chars": 39338, - "estimated_input_tokens": 9835, - "estimated_eval_tokens": 19926 + "soft_pair_count": 43, + "serialized_chars": 39766, + "estimated_input_tokens": 9942, + "estimated_eval_tokens": 20140 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is heterogeneous, with a loose theme of Transformers v5 regressions and model/config compatibility issues, but no clear duplicate cluster. I\u2019d use the broad v5 upgrade regression report as the best representative issue.", - "confidence": 0.25, - "canonical_issue_reason": "issue:45278 is the broadest concrete user-facing regression report in the set (\u201cmany import errors after update from 4.57.0 to 5.5.0\u201d), and it best summarizes the v5 breakage theme without being as vague as the umbrella complaints.", + "summary": "Mostly unrelated closed issues with a few superficial overlaps around tokenizer/model-loading regressions in v5. I did not find any soft pair that safely looks like the same underlying bug/change, so I rejected all soft edges.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45310 is the most concrete, actionable loading regression in the set and the best anchor for the small v5/from_pretrained-related subgroup.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most representative issue because it captures the widest class of compatibility failures among these reports; the others are more specific to one model, API, or bug.", + "best_issue_reason": "issue:45310 has the clearest bug description, version scope, and likely reproducible model-loading failure, making it the best representative issue here.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44393", - "right": "issue:45753", + "left": "issue:43976", + "right": "issue:44393", "accept": false, - "reason": "Both are Qwen3-VL problems, but one is 2D bbox output hallucination/error and the other is an infinite loop on a specific image; different failure modes." + "reason": "Different problems: Python version compatibility vs Qwen3-VL output/vision bug." }, { - "left": "issue:43576", - "right": "issue:44556", + "left": "issue:44908", + "right": "issue:45071", "accept": false, - "reason": "Both relate to v5 breakage, but one is the `transformers env` command and the other is checkpoint reload compatibility; different code paths." + "reason": "Unrelated subsystems: LR scheduler kwarg handling vs PretrainedConfig type checking." }, { - "left": "issue:44230", - "right": "issue:46095", + "left": "issue:44295", + "right": "issue:45230", "accept": false, - "reason": "One asks for FP8 inference support for Qwen3-VL/Qwen3.5 MoE, the other asks for a DeepSeek-V4 weight conversion script; not the same bug/change." + "reason": "One is a buffer access error, the other is a generic bug report with no clear matching failure mode." }, { - "left": "issue:43525", - "right": "issue:45850", + "left": "issue:44297", + "right": "issue:45230", "accept": false, - "reason": "`Llama4Config` missing `pad_token_id` is a specific config bug; \u201cSince 5.0 version it breaks too many models\u201d is a broad complaint, not the same issue." + "reason": "Tokenizer save_pretrained metadata mismatch does not match the generic bug report enough to treat as the same bug." }, { - "left": "issue:41669", - "right": "issue:43673", + "left": "issue:44704", + "right": "issue:44908", "accept": false, - "reason": "Import-star usage slowing imports is unrelated to `GenerationMixin` cache missing during chunked prefill." + "reason": "AutoProcessor cached_file kwargs forwarding and inverse_sqrt scheduler kwargs are unrelated." }, { - "left": "issue:44556", - "right": "issue:45092", + "left": "issue:41669", + "right": "issue:44351", "accept": false, - "reason": "Both are upgrade incompatibilities, but one is checkpoint reload after v5.2/v5.3 and the other is meta-init incompatibility for old InternVL2 remote-code checkpoints; distinct failures." + "reason": "Import-star cleanup/perf issue is not the same as the HybridCache import failure." }, { - "left": "issue:43572", - "right": "issue:45850", + "left": "issue:44492", + "right": "issue:46328", "accept": false, - "reason": "StableLmConfig missing `pad_token_idx` is a specific config regression, not the broad \u201ctoo many models\u201d complaint." + "reason": "A cache-strategy typo and a padding performance bug are different issues." }, { - "left": "issue:44492", - "right": "issue:45850", + "left": "issue:41669", + "right": "issue:45341", "accept": false, - "reason": "A typo in cache strategy docs is not the same as a generic v5 model breakage report." + "reason": "Import-star usage cleanup is unrelated to a testing_utils bug." }, { - "left": "issue:45278", + "left": "issue:43976", "right": "issue:45850", "accept": false, - "reason": "Both are broad v5 complaints, but one is specifically about import errors after upgrade and the other is a generic statement that many models break; not the same concrete defect." + "reason": "Both mention broad v5 breakage, but the concrete failures are different." }, { - "left": "issue:42371", - "right": "issue:44492", + "left": "issue:44393", + "right": "issue:45478", "accept": false, - "reason": "TF32 API guidance and a cache-strategy typo are unrelated." + "reason": "Different model families and failure modes; both are vision/model-specific bugs, not the same one." }, { - "left": "issue:41669", + "left": "issue:43976", "right": "issue:46218", "accept": false, - "reason": "Import-star slowdown in model packages is unrelated to failing to import `Qwen3_5ForCausalLM`." + "reason": "Python version support and Qwen3_5 import failure are not the same underlying bug." }, { - "left": "issue:43408", - "right": "issue:44315", + "left": "issue:44077", + "right": "issue:44861", "accept": false, - "reason": "Wrong model type warning for sam3_tracker is unrelated to Liger Kernel not applying with `model_init`." + "reason": "patchtsmixer post_init policy and tied-weight key handling are separate code paths." }, { - "left": "issue:42890", + "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Missing `set_seed()` in SAM HQ tests is unrelated to inverse_sqrt scheduler kwargs being ignored." + "reason": "Docs referencing removed pipeline APIs is unrelated to scheduler kwargs behavior." }, { - "left": "issue:43352", - "right": "issue:44393", + "left": "issue:31795", + "right": "issue:33453", "accept": false, - "reason": "Flash Attention 2.0 unsupported for NemotronH is unrelated to Qwen3-VL bbox output errors." + "reason": "Documentation confusion in model.forward is unrelated to tokenizer loading regression." }, { - "left": "issue:43576", - "right": "issue:44908", + "left": "issue:44315", + "right": "issue:44829", "accept": false, - "reason": "`transformers env` command breakage is unrelated to scheduler kwargs handling." + "reason": "Different training/runtime behaviors: model_init Liger application vs flash_attention_3 degenerate training." }, { - "left": "issue:44246", + "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Import slowdown is about startup/perf, while lazy loading not working properly is a different mechanism-level bug." + "reason": "Broken transformers env command and lazy loading are different tools/features." }, { - "left": "issue:43976", - "right": "issue:44393", + "left": "issue:44655", + "right": "issue:44861", "accept": false, - "reason": "Python 3.9/3.10 compatibility is unrelated to Qwen3-VL output hallucination." + "reason": "Pipeline save_pretrained support and tied-weight-key AttributeError are not the same bug." }, { - "left": "issue:44908", - "right": "issue:45071", + "left": "issue:44360", + "right": "issue:45468", "accept": false, - "reason": "Scheduler kwargs being ignored is unrelated to `PretrainedConfig` type checking regressions." + "reason": "DSA indexer math and Gemma4AudioRelPositionalEncoding are unrelated." }, { - "left": "issue:44297", - "right": "issue:45230", + "left": "issue:45230", + "right": "issue:45310", "accept": false, - "reason": "Tokenizer class mismatch on save is unrelated to a generic, underspecified bug report." + "reason": "The first is too generic to establish that it is the same concrete Qwen3.5 MoE from_pretrained regression." }, { - "left": "issue:44295", - "right": "issue:45230", + "left": "issue:44273", + "right": "issue:44908", "accept": false, - "reason": "`position_ids` buffer access failure is a specific tensor/buffer bug, not the same as an unspecified bug report." + "reason": "Lazy loading and inverse_sqrt scheduler kwargs are unrelated." }, { - "left": "issue:44704", + "left": "issue:43576", + "right": "issue:46218", + "accept": false, + "reason": "Broken env subcommand and Qwen3_5 import failure are different failures." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Timestamp calculation in Qwen3VL Processor is not the same as GLM-5 RoPE implementation." + }, + { + "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "`AutoProcessor.from_pretrained` kwargs forwarding is unrelated to inverse_sqrt scheduler kwargs." + "reason": "Processor timestamp math and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeek-V4 gating logic and a dtype/transpose expression bug are different concrete issues." }, { "left": "issue:41669", - "right": "issue:44351", + "right": "issue:43704", "accept": false, - "reason": "Import-star slowdown is unrelated to missing `HybridCache` export/import errors." + "reason": "Import-star cleanup is not the same as VRAM leakage in multi-threaded dataloading." }, { - "left": "issue:44492", - "right": "issue:46328", + "left": "issue:31795", + "right": "issue:34689", "accept": false, - "reason": "A cache-strategy typo is unrelated to `SequenceFeatureExtractor.pad` wasting time on numpy-to-list conversion." + "reason": "Model.forward documentation confusion and Llama 3.2 Vision model loading breakage are unrelated." }, { - "left": "issue:41669", - "right": "issue:45341", + "left": "issue:31515", + "right": "issue:31795", "accept": false, - "reason": "Model import-star overhead is unrelated to a bug in `testing_utils.py`." + "reason": "Slow from_pretrained checkpoint loading and forward API documentation confusion are different issues." }, { - "left": "issue:44393", - "right": "issue:45478", + "left": "issue:43704", + "right": "issue:44485", "accept": false, - "reason": "Both are Qwen3.5/Qwen3-VL-related, but one is bbox output hallucination and the other is a from_pretrained import/error regression; different bugs." + "reason": "VRAM leak in dataloader threads and GLM-5 RoPE implementation are unrelated." }, { - "left": "issue:43976", - "right": "issue:45850", + "left": "issue:44485", + "right": "issue:45468", "accept": false, - "reason": "Python-version support is unrelated to the broad \u201cmany models break\u201d complaint." + "reason": "Different model families and different numerical/positional-encoding problems." }, { - "left": "issue:43976", + "left": "issue:45850", "right": "issue:46218", "accept": false, - "reason": "Python compatibility does not match a missing Qwen3_5 model import." + "reason": "Generic v5 breakage and Qwen3_5 import failure are too broad to treat as the same bug." }, { - "left": "issue:44077", - "right": "issue:44861", + "left": "issue:42371", + "right": "issue:43704", "accept": false, - "reason": "Optional `post_init` being disallowed and `_tied_weights_keys` crashing on list input are different model config/utility issues." + "reason": "TF32 settings guidance and VRAM leakage are unrelated." }, { - "left": "issue:43827", - "right": "issue:44908", + "left": "issue:43502", + "right": "issue:43519", "accept": false, - "reason": "Docs still referencing `pipeline()` is unrelated to scheduler kwargs handling." + "reason": "local_files_only network requests and timestamp calculation are different code paths." }, { - "left": "issue:31795", - "right": "issue:33453", + "left": "issue:43519", + "right": "issue:44623", "accept": false, - "reason": "Documentation confusion around `input_ids`/`past_key_values` is unrelated to tokenizer loading regression." + "reason": "Processor timestamp math and processor.save_pretrained missing files are different bugs." }, { - "left": "issue:44315", - "right": "issue:44829", + "left": "issue:43519", + "right": "issue:43704", "accept": false, - "reason": "Liger Kernel not applied with `model_init` is unrelated to degenerate training under flash_attention_3." + "reason": "Timestamp calculation and VRAM leakage are unrelated." }, { - "left": "issue:43576", - "right": "issue:44273", + "left": "issue:43298", + "right": "issue:46095", "accept": false, - "reason": "Broken `transformers env` command and lazy loading failure are different issues." + "reason": "Audio-model continuous batching support and DeepSeek-V4 weight conversion are different concerns." }, { - "left": "issue:44360", + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "GlmMoeDsaIndexer split return values and Gemma4AudioRelPositionalEncoding are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch in bark-small does not match Gemma-4 positional encoding." + }, + { + "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "DSA indexer discussion and Gemma-4 audio positional encoding are unrelated model/subsystem problems." + "reason": "MLA q_a_layernorm epsilon precision issue is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "HybridCache import failure and Qwen3_5 import failure are different missing-symbol problems." + }, + { + "left": "issue:43352", + "right": "issue:46218", + "accept": false, + "reason": "Flash Attention 2.0 unsupported for NemotronHForCausalLM is unrelated to Qwen3_5 import errors." }, { - "left": "issue:44655", - "right": "issue:44861", + "left": "issue:42913", + "right": "issue:43066", "accept": false, - "reason": "Pipeline save/load failure is unrelated to tied-weight-key handling." + "reason": "Both are tokenizer v5 complaints, but one is general behavior drift and the other is a specific decoder-type mismatch." }, { - "left": "issue:45230", - "right": "issue:45310", + "left": "issue:44908", + "right": "issue:44998", "accept": false, - "reason": "A generic bug report is not enough evidence to merge with a specific Qwen3.5 MoE from_pretrained error." + "reason": "Scheduler kwarg handling is unrelated to the placeholder 'Unemployment' issue." }, { - "left": "issue:44273", - "right": "issue:44908", + "left": "issue:44492", + "right": "issue:44998", "accept": false, - "reason": "Lazy loading malfunction is unrelated to inverse_sqrt scheduler kwargs being ignored." + "reason": "Cache-strategy typo is unrelated to the placeholder issue." } ] }, @@ -16865,8 +17195,8 @@ "label": "issue", "component_index": 1, "component_count": 20, - "review_unit_index": 54, - "review_unit_count": 54, + "review_unit_index": 55, + "review_unit_count": 55, "status": "reviewed", "reason": null, "source": "llm", @@ -16876,69 +17206,32 @@ "nodes": [ "issue:30333", "issue:31515", - "issue:31795", "issue:33453", - "issue:34689", "issue:36296", "issue:37934", "issue:38617", - "issue:42371", - "issue:42913", "issue:42936", - "issue:43066", "issue:43240", "issue:43298", - "issue:43352", - "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:44351", - "issue:44485", - "issue:44492", "issue:44623", "issue:44861", - "issue:44908", "issue:44938", "issue:44998", "issue:45290", "issue:45310", - "issue:45468", "issue:45850", "issue:46018", - "issue:46095", - "issue:46144", - "issue:46218" + "issue:46144" ], "soft_pairs": [ - "issue:31795|issue:34689", - "issue:31515|issue:31795", - "issue:43704|issue:44485", - "issue:44485|issue:45468", - "issue:45850|issue:46218", - "issue:42371|issue:43704", - "issue:43502|issue:43519", - "issue:43298|issue:46095", - "issue:43519|issue:44623", - "issue:43519|issue:43704", - "issue:44263|issue:45468", - "issue:43606|issue:45468", - "issue:44261|issue:45468", - "issue:44351|issue:46218", - "issue:43352|issue:46218", - "issue:42913|issue:43066", - "issue:44908|issue:44998", - "issue:44492|issue:44998", "issue:43618|issue:44998", "issue:44998|issue:45850", "issue:44938|issue:44998", @@ -16965,292 +17258,184 @@ "issue:37934|issue:43668", "issue:42936|issue:43668" ], - "prepared_review_unit_hash": "638c407292fd2022c55a29bb3a254819306949e034c3bfa801f4f4d5e87dd69c", + "prepared_review_unit_hash": "31655ebedf486722338388cb7c7f83b4d557a8c49aac22970a530f463811996a", "component_budget": { "node_count": 463, "item_count": 463, "soft_pair_count": 1704, - "serialized_chars": 883317, - "estimated_input_tokens": 220830, - "estimated_eval_tokens": 441916 + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 }, "budget": { - "node_count": 45, - "item_count": 45, - "soft_pair_count": 43, - "serialized_chars": 38002, - "estimated_input_tokens": 9501, - "estimated_eval_tokens": 19258 + "node_count": 26, + "item_count": 26, + "soft_pair_count": 25, + "serialized_chars": 21932, + "estimated_input_tokens": 5483, + "estimated_eval_tokens": 11222 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous; the soft pairs share superficial release/model/tokenizer language but not a single fixable bug, so none should be merged as duplicates.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45850 is the only broad umbrella describing a widespread 5.0 regression pattern, so it is the closest cluster-level representative.", + "summary": "This is a noisy cluster of mostly unrelated issues that only share vague transformer-regression language. The soft similarity edges are not true duplicates; they span different models, subsystems, and failure modes. No pull requests are present.", + "confidence": 0.22, + "canonical_issue_reason": "Issue 45850 is the broadest umbrella complaint about post-5.0 breakage and is the closest thing to a cluster anchor, though it is still too generic to represent a real duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "Same as canonical: it is the broadest issue in the set and the best available representative, though still too generic for a precise duplicate target.", + "best_issue_reason": "Issue 45850 is the most central/high-level issue in this set, but it should be treated as a coarse regression report rather than a duplicate target for the other concrete bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:31795", - "right": "issue:34689", - "accept": false, - "reason": "Both concern loading/docs around models, but one is documentation confusion and the other is a concrete model-loading regression." - }, - { - "left": "issue:31515", - "right": "issue:31795", - "accept": false, - "reason": "Slow checkpoint loading and forward-API documentation confusion are unrelated problems." - }, - { - "left": "issue:43704", - "right": "issue:44485", - "accept": false, - "reason": "VRAM leakage in multithreaded dataloading and GLM-5 RoPE implementation are different code paths." - }, - { - "left": "issue:44485", - "right": "issue:45468", - "accept": false, - "reason": "RoPE math for GLM-5 and Gemma-4 audio positional encoding are separate model-specific bugs." - }, - { - "left": "issue:45850", - "right": "issue:46218", - "accept": false, - "reason": "A broad complaint about many v5 regressions is not the same as a specific Qwen3_5 import failure." - }, - { - "left": "issue:42371", - "right": "issue:43704", - "accept": false, - "reason": "TF32 behavior settings and a VRAM leak in dataloader threads do not overlap concretely." - }, - { - "left": "issue:43502", - "right": "issue:43519", - "accept": false, - "reason": "local_files_only network leakage and Qwen3VL timestamp calculation are unrelated." - }, - { - "left": "issue:43298", - "right": "issue:46095", - "accept": false, - "reason": "Audio continuous batching support and DeepSeek-V4 weight conversion are different requests." - }, - { - "left": "issue:43519", - "right": "issue:44623", - "accept": false, - "reason": "Timestamp calculation in a processor and missing files from save_pretrained are distinct bugs." - }, - { - "left": "issue:43519", - "right": "issue:43704", - "accept": false, - "reason": "A processor timestamp bug and a VRAM leak in training threads are unrelated." - }, - { - "left": "issue:44263", - "right": "issue:45468", - "accept": false, - "reason": "torch.split return handling and Gemma-4 audio positional encoding are not the same failure." - }, - { - "left": "issue:43606", - "right": "issue:45468", - "accept": false, - "reason": "CPU offload device mismatch is a different issue from audio positional encoding logic." - }, - { - "left": "issue:44261", - "right": "issue:45468", - "accept": false, - "reason": "Missing rms_norm_eps in MLA q_a_layernorm and Gemma-4 audio positional encoding are separate model bugs." - }, - { - "left": "issue:44351", - "right": "issue:46218", - "accept": false, - "reason": "HybridCache import failure and Qwen3_5 module import failure are related to imports, but not the same missing symbol/module." - }, - { - "left": "issue:43352", - "right": "issue:46218", - "accept": false, - "reason": "Flash Attention 2 support for Nemotron and a Qwen3_5 import error are different problems." - }, - { - "left": "issue:42913", - "right": "issue:43066", - "accept": false, - "reason": "Both mention tokenizer changes, but one is a broad behavior difference while the other is a specific decoder-type regression." - }, - { - "left": "issue:44908", - "right": "issue:44998", - "accept": false, - "reason": "Scheduler kwargs handling and an unrelated social/placeholder issue share no bug or code path." - }, - { - "left": "issue:44492", - "right": "issue:44998", - "accept": false, - "reason": "A cache-strategy typo and an unrelated placeholder issue are not duplicates." - }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions regression and an unrelated placeholder issue are clearly different." + "reason": "Different topics entirely: CLIPOutput attentions vs an unrelated one-word issue." }, { "left": "issue:44998", "right": "issue:45850", "accept": false, - "reason": "The placeholder issue is unrelated to the broad v5 regression umbrella." + "reason": "Unrelated issue titles; no shared bug or code path." }, { "left": "issue:44938", "right": "issue:44998", "accept": false, - "reason": "Python 3.14 loading failure and the unrelated placeholder issue do not match." + "reason": "Python 3.14 import/load failure is unrelated to the other issue." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "_get_tied_weight_keys list handling and apply_chat_template tool-call crashes are different code paths." + "reason": "Different code paths: tied-weight key handling vs chat template/tool-call crash." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Missing pad_token_id in Llama4Config and tied-weight-key list handling are unrelated." + "reason": "Llama4Config pad token attribute regression is unrelated to tied-weight key crashes." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "Slow from_pretrained checkpoint loading and an ImportError for layer_type_validation are unrelated." + "reason": "Checkpoint loading slowness and an ImportError are separate problems." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "StableLmConfig pad_token_idx regression and tied-weight-key attribute handling are distinct bugs." + "reason": "StableLmConfig pad token attribute issue is not the same as tied-weight key handling." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "FP8 support for qwen3-vl/qwen3.5 and missing processor save files are different feature/bug areas." + "reason": "FP8 model support and missing processor save files are different issues." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLFlow reporting on failed training jobs and slow checkpoint loading are unrelated." + "reason": "MLFlow reporting on failed training jobs is unrelated to from_pretrained performance." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "Checkpoint load slowness and tensor-parallel training failure are different issues." + "reason": "Checkpoint load slowdown and tensor-parallel training bugs do not match." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "FP8 inference support and a from_pretrained error on Qwen3.5 MoE are separate concerns." + "reason": "Both mention Qwen3.5, but one is FP8 support and the other is a from_pretrained regression; not the same fix." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "Both involve loading, but checkpoint slowness and tokenizer-loading regression are not the same bug." + "reason": "Tokenizer loading regression and checkpoint loading slowdown are distinct." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching for audio models and FP8 qwen3-vl support are different requests." + "reason": "Continuous batching for audio models is unrelated to FP8 Qwen support." }, { "left": "issue:43717", "right": "issue:46144", "accept": false, - "reason": "Mamba init_weights behavior and RoFormer attention-interface usage are unrelated model implementation issues." + "reason": "Mamba init-weights behavior and RoFormer attention-interface usage are different model implementations." }, { "left": "issue:42936", "right": "issue:46018", "accept": false, - "reason": "Mask2Former ignore_value handling and DeepSeek-V4 shared expert gating are different model bugs." + "reason": "Mask2former ignore_value bug and DeepSeek-V4 gating issue are unrelated model-specific defects." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "A config type-hint issue and tokenizer regex warnings after 4.57.3 are unrelated." + "reason": "Type-hint metadata and tokenizer regex warnings do not describe the same bug." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Mask2Former ignore_value and missing kwargs in loss calculation do not describe the same defect." + "reason": "Loss kwargs propagation and Mask2former ignore_value handling are separate bugs." }, { "left": "issue:43240", "right": "issue:46144", "accept": false, - "reason": "Loss-function kwargs propagation and RoFormer attention interface usage are unrelated." + "reason": "Loss kwargs propagation is unrelated to RoFormer attention dispatch." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2Former ignore_value and tokenizer regex warnings are not duplicates." + "reason": "No shared underlying issue; one is segmentation-loss logic, the other a tokenizer warning." }, { "left": "issue:42936", "right": "issue:46144", "accept": false, - "reason": "Mask2Former ignore_value and RoFormer attention interface usage are separate bugs." + "reason": "Different model subsystems and failure modes." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "Mask2Former ignore_value and a typo in quantization operation naming are unrelated." + "reason": "Mask2former loss behavior and a quantization typo are unrelated." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Missing loss kwargs and a ModernBERT config type-hint error are different issues." + "reason": "Loss kwarg forwarding and a config type-hint issue are not duplicates." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "Incorrect type hints and a typo in a quantization symbol are not the same bug." + "reason": "Incorrect type hint and typo in a quantization attribute are unrelated." }, { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "A question about RMS norm in Llama4TextL2Norm and a ModernBERT type-hint correction are unrelated." + "reason": "RMS norm question for Llama4 is unrelated to ModernBERT type hints." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Mask2Former ignore_value handling and ModernBERT type hints are different problems." + "reason": "No common bug; model loss logic vs config typing." } ] }, @@ -17283,30 +17468,30 @@ "issue:10519|issue:11005", "issue:11005|issue:11307" ], - "prepared_review_unit_hash": "67b6e12a78bbf0598a2a79eebad12e2533612e73ff197f6c397f1ea81459bcc6", + "prepared_review_unit_hash": "6d70b3d2beccacc6eaf668728b81e99f027b2bfba7fb5aac82e3dc37f2e70f19", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2874, - "estimated_input_tokens": 719, - "estimated_eval_tokens": 1694 + "serialized_chars": 2878, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2874, - "estimated_input_tokens": 719, - "estimated_eval_tokens": 1694 + "serialized_chars": 2878, + "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: one is about tokenizer truncation direction, one asks for ReduceLROnPlateau-like scheduler behavior, and one concerns Wav2Vec2 word boundary timestamps. They describe unrelated features/problems, so there is no clear canonical issue in this cluster.", - "confidence": 0.96, + "summary": "The three issues are unrelated feature requests: truncation direction, ReduceLROnPlateau-like behavior, and Wav2Vec2 word time offsets. This cluster does not look like a true duplicate set.", + "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, @@ -17316,19 +17501,19 @@ "left": "issue:10519", "right": "issue:11307", "accept": false, - "reason": "Different domains and code paths: truncation strategy vs Wav2Vec2 timestamp extraction." + "reason": "Different topics and code paths: token truncation behavior vs Wav2Vec2 word boundary timestamps." }, { "left": "issue:10519", "right": "issue:11005", "accept": false, - "reason": "Unrelated requests: tokenizer truncation option vs learning-rate scheduler functionality." + "reason": "One is about truncation direction, the other about scheduler functionality; no shared underlying bug or change." }, { "left": "issue:11005", "right": "issue:11307", "accept": false, - "reason": "Separate topics with no shared underlying bug or change: scheduler behavior vs audio model word offsets." + "reason": "ReduceLROnPlateau-like scheduler behavior is unrelated to extracting time offsets in Wav2Vec2." } ] }, @@ -17360,12 +17545,12 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "3c993acaf5a65bec9bc9b17c4cfcf00abfef799b2e5b884bd4b64ef31a6cbc72", + "prepared_review_unit_hash": "279a3cab8c24c8e0fb2375c4fa5c28131dcf9784b77a37d573b4b89feb0f734e", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2479, + "serialized_chars": 2478, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -17373,7 +17558,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2479, + "serialized_chars": 2478, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -17382,24 +17567,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These issues are not duplicates of each other: one is a tokenization discrepancy, one is an auth/error-message problem for private model downloads, and one is a generic model-loading failure for bert-base-uncased.", - "confidence": 0.89, - "canonical_issue_reason": "Issue 16618 is the broadest and most central model-loading report in the set, with the highest discussion activity, while the other two describe different problems.", + "summary": "No duplicate-worthy pair is evident here. The three issues cover different failure modes: tokenization mismatch, private-model auth/error messaging, and a generic model-loading failure. The soft similarities are superficial.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 16618 is the best representative of the cluster only in the loose sense that it is the most general loading-related issue; it is still not the same underlying bug as the others.", + "best_issue_reason": "issue:16618 is the broadest user-facing loading problem, but the cluster is too heterogeneous to treat it as a true canonical issue.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Tokenization output differences and a model-loading failure are different code paths and different bugs; nothing here suggests one fix would resolve both." + "reason": "Tokenization output mismatch with TensorFlow code is a different problem from failing to load bert-base-uncased; no shared concrete code-path bug is evident." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "Private-model auth/download error messaging is a different problem from failing to load bert-base-uncased; they are not clearly the same underlying bug." + "reason": "Bad error messaging for private-model download without login is an authentication/error-handling issue, not the same as a generic model-load failure." } ] }, @@ -17429,40 +17614,40 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "779867657a1ee029b83acc129ba74596663776cdf613bff7998f8cdb6b7d3f8a", + "prepared_review_unit_hash": "38e586304b66c31765c0ff4020f9734724fd0d4b67333806669638e0dcc2fa82", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1590, - "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": 1590, - "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 both community-contribution feature requests, but for different capabilities: Flash Attention 2 architecture support vs. GGUF architecture support. They are not duplicates.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "These are not duplicates: one requests Flash Attention 2 support for more architectures, the other requests GGUF support for more architectures. They share the same \u201ccommunity contribution\u201d framing but target different code paths and feature areas.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue applies here; the two issues are separate feature requests (Flash Attention 2 vs GGUF support).", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Neither is a duplicate of the other. If a representative issue is needed for the Flash Attention 2 topic, issue 26350 is the more established thread and has substantially more discussion.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different underlying requests affecting different model-format/backend support paths; they are only loosely related as community contribution proposals." + "reason": "Different underlying requests: Flash Attention 2 architecture support vs GGUF architecture support. Similar wording and subsystem scope are not enough to treat them as the same bug/change." } ] }, @@ -17521,185 +17706,188 @@ "issue:44370|issue:46226", "issue:46222|issue:46226" ], - "prepared_review_unit_hash": "2e55cafba96425bae06e1d5ec826f55e5e19fde15e5a1879a7c21189c75859de", + "prepared_review_unit_hash": "db5cb3e904618b82bdf7735a502e2b72577929c2762540d4e6e60e861d9b2272", "component_budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 24, - "serialized_chars": 13532, - "estimated_input_tokens": 3383, - "estimated_eval_tokens": 7022 + "serialized_chars": 13528, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 24, - "serialized_chars": 13532, - "estimated_input_tokens": 3383, - "estimated_eval_tokens": 7022 + "serialized_chars": 13528, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is dominated by similarly templated i18n translation-request issues, but they are not the same underlying request. The language-specific issues (Turkish, Spanish, Romanian) are distinct, and the placeholder/nonsense issues look like template or spam artifacts rather than true duplicates.", - "confidence": 0.86, - "canonical_issue_reason": "If a representative must be chosen, issue:43834 is the most generic placeholder-style i18n issue and sits near the center of the similarity clique. It is only a stand-in representative, not a true duplicate of the language-specific translation requests.", + "summary": "This cluster is mixed: the Turkish, Spanish, and Romanian issues are distinct language-specific translation requests, while the placeholder-title issues form one duplicate subgroup and the two Thai-titled issues form another exact duplicate pair.", + "confidence": 0.76, + "canonical_issue_reason": "issue:43834 is the best anchor for the largest clear duplicate subgroup: the three generic placeholder-title translation issues are near-identical, and this one is the earliest of that set. The language-specific requests are separate issues, so the full cluster is not one single duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43834 is the best cluster representative because it uses the canonical template title and is the clearest example of the repeated boilerplate pattern.", + "best_issue_reason": "issue:43834 is the most useful representative of the duplicate placeholder subgroup; the only other clear duplicate is the exact Thai pair. No single issue cleanly represents the entire mixed cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Different translation requests; soft title/body similarity is just the shared i18n template, not the same language issue." + "reason": "Different language-specific translation request; boilerplate similarity alone is not enough." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "Turkish vs generic placeholder request; no evidence they refer to the same concrete translation task." + "reason": "Different language-specific translation request; not the same underlying issue." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Turkish request is distinct from the generic template issue; similarity is only boilerplate." + "reason": "Turkish request vs generic placeholder-title issue; too little to treat as the same issue." }, { "left": "issue:27088", "right": "issue:38435", "accept": false, - "reason": "Turkish and Romanian translation requests are separate issues, not duplicates." + "reason": "Different target languages and request contexts." }, { "left": "issue:27088", "right": "issue:28936", "accept": false, - "reason": "Turkish and Spanish translation requests are different localized docs efforts." + "reason": "Different target language; these are separate translation requests." }, { "left": "issue:28936", "right": "issue:43834", "accept": false, - "reason": "Spanish-specific request does not match the generic placeholder issue beyond shared template text." + "reason": "Spanish request is distinct from the generic placeholder-titled issues." }, { "left": "issue:28936", "right": "issue:44365", "accept": false, - "reason": "Spanish translation request is distinct from the generic template issue." + "reason": "Different target language/request, despite similar template text." }, { "left": "issue:28936", "right": "issue:38435", "accept": false, - "reason": "Spanish and Romanian translation requests are separate, not the same underlying bug/change." + "reason": "Spanish vs Romanian translation request; not the same issue." }, { "left": "issue:28936", "right": "issue:44370", "accept": false, - "reason": "Spanish request vs generic placeholder; no concrete overlap beyond template wording." + "reason": "Different language-specific request; similarity is only template-level." }, { "left": "issue:38435", "right": "issue:44370", "accept": false, - "reason": "Romanian request and generic placeholder issue are not the same concrete task." + "reason": "Romanian request is distinct from the placeholder-title issues." }, { "left": "issue:38435", "right": "issue:44365", "accept": false, - "reason": "Romanian request is distinct from the generic placeholder issue." + "reason": "Romanian request vs generic placeholder-titled issue; not a duplicate." }, { "left": "issue:38435", "right": "issue:43834", "accept": false, - "reason": "Romanian-specific translation request does not duplicate the generic template issue." + "reason": "Different language-specific request; only the boilerplate overlaps." }, { "left": "issue:38435", "right": "issue:46222", "accept": false, - "reason": "Romanian translation request is unrelated to the Thai-text issue; only surface similarity exists." + "reason": "Romanian translation request and Thai-titled issue do not look like the same underlying request." }, { "left": "issue:38435", "right": "issue:46226", "accept": false, - "reason": "Romanian translation request is unrelated to the Thai-text issue; only surface similarity exists." + "reason": "Romanian translation request and Thai-titled issue do not look like the same underlying request." }, { "left": "issue:43834", "right": "issue:44370", - "accept": false, - "reason": "Both are generic template-style issues, but there is no evidence they are the same concrete request or change." + "accept": true, + "reason": "Same generic placeholder-title translation request with near-identical body; looks like a duplicate." }, { "left": "issue:43834", "right": "issue:44365", - "accept": false, - "reason": "Generic placeholder titles alone are insufficient to treat these as duplicates of the same underlying issue." + "accept": true, + "reason": "Same generic placeholder-title translation request with near-identical body; looks like a duplicate." }, { "left": "issue:43834", "right": "issue:46222", "accept": false, - "reason": "Template issue vs nonsensical Thai text; no shared concrete bug/change." + "reason": "Different title/script and likely different request context; boilerplate similarity alone is insufficient." }, { "left": "issue:43834", "right": "issue:46226", "accept": false, - "reason": "Template issue vs nonsensical Thai text; similarity is not enough to merge." + "reason": "Different title/script and likely different request context; not enough evidence for a duplicate." }, { "left": "issue:44365", "right": "issue:44370", - "accept": false, - "reason": "Both are placeholder-style issues, but the available evidence does not show the same underlying request." + "accept": true, + "reason": "Same generic placeholder-title translation request with near-identical body; looks like a duplicate." }, { "left": "issue:44365", "right": "issue:46222", "accept": false, - "reason": "Generic template issue and Thai-text issue do not indicate the same concrete problem." + "reason": "Different title/script and likely different request context." }, { "left": "issue:44365", "right": "issue:46226", "accept": false, - "reason": "Generic template issue and Thai-text issue do not indicate the same concrete problem." + "reason": "Different title/script and likely different request context." }, { "left": "issue:44370", "right": "issue:46222", "accept": false, - "reason": "Generic template issue and Thai-text issue do not indicate the same concrete problem." + "reason": "Different title/script and likely different request context." }, { "left": "issue:44370", "right": "issue:46226", "accept": false, - "reason": "Generic template issue and Thai-text issue do not indicate the same concrete problem." + "reason": "Different title/script and likely different request context." }, { "left": "issue:46222", "right": "issue:46226", - "accept": false, - "reason": "Identical-looking Thai-text titles alone are not enough to confirm a true duplicate underlying bug/change; looks more like repeated spam or malformed submissions." + "accept": true, + "reason": "Exact same Thai-titled issue with identical body shape; this looks like a straightforward duplicate." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The mixed-cluster summary matches the packet: Turkish/Spanish/Romanian are clearly separate language-tagged requests, the three placeholder-title issues form a tight duplicate subgroup, and the two Thai-titled issues are an exact duplicate pair by title/body-length pattern. The soft-edge decisions stay cautious and do not overclaim beyond the available evidence." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -17771,250 +17959,250 @@ "issue:43837|issue:45177", "issue:45322|issue:46133" ], - "prepared_review_unit_hash": "a293327442a37812e3b6fb7aa2526ceeeab69010be6e92cc9abe8d4ae036fa3a", + "prepared_review_unit_hash": "1809bfe7d2db5b243266308a25dd9e89f78249501be2f03d1324a3816789295a", "component_budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21659, - "estimated_input_tokens": 5415, - "estimated_eval_tokens": 11086 + "serialized_chars": 21661, + "estimated_input_tokens": 5416, + "estimated_eval_tokens": 11088 }, "budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21659, - "estimated_input_tokens": 5415, - "estimated_eval_tokens": 11086 + "serialized_chars": 21661, + "estimated_input_tokens": 5416, + "estimated_eval_tokens": 11088 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a loose thematic group of unrelated model-add requests, not a true duplicate set. The items span different architectures and modalities (text, vision, audio, OCR, sparse transformers), so none of the soft edges look like the same underlying change.", - "confidence": 0.96, - "canonical_issue_reason": "issue:36879 is the oldest and clearest single model-request thread, but the cluster is too heterogeneous to treat the other issues as duplicates of it.", + "summary": "This cluster is a loose collection of model-request issues with overlapping wording, but they target different models, families, and modalities. I found no true duplicate pair and no PRs in the set.", + "confidence": 0.98, + "canonical_issue_reason": "issue:42738 is the most umbrella-like representative in the group; it\u2019s generic and broad compared with the many specific model-request tickets, so it works best as a cluster anchor.", "canonical_pr_reason": null, - "best_issue_reason": "issue:36879 is the best available representative anchor because it is concise and closed, but it should only be used as a loose cluster exemplar, not as a duplicate target.", + "best_issue_reason": "issue:42738 is the best issue candidate because it is the least model-specific and most likely to summarize the broader theme of the cluster, even though the items are not duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "Different requested models (EuroBert vs DEIMv2); only broad 'add a model' similarity." + "reason": "Different model requests: EuroBert vs DEIMv2; same template, not same underlying change." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "Text model request vs TTS support request; different code paths and deliverables." + "reason": "EuroBert config addition is unrelated to Qwen3-TTS support." }, { "left": "issue:36817", "right": "issue:36879", "accept": false, - "reason": "EuroBert and RF-DETR are unrelated architectures; not the same bug/change." + "reason": "EuroBert model config and RF-DETR model support are different models and code paths." }, { "left": "issue:36879", "right": "issue:41211", "accept": false, - "reason": "RF-DETR vs DEIMv2: different model families and implementation work." + "reason": "RF-DETR and DEIMv2 are separate model additions." }, { "left": "issue:36879", "right": "issue:43909", "accept": false, - "reason": "Vision model request vs audio model request; not mergeable into one concrete PR." + "reason": "Different audio/vision model requests; no shared concrete bug or change." }, { "left": "issue:36879", "right": "issue:44121", "accept": false, - "reason": "RF-DETR and sparse transformer requests address different model implementations." + "reason": "RF-DETR and a weight-sparse transformer request are unrelated model implementations." }, { "left": "issue:36879", "right": "issue:43837", "accept": false, - "reason": "RF-DETR vs ASR support; distinct model support areas." + "reason": "RF-DETR vs Qwen3-ASR are different models and modalities." }, { "left": "issue:36879", "right": "issue:43671", "accept": false, - "reason": "RF-DETR and TTS are different modalities and code changes." + "reason": "RF-DETR model support is unrelated to Qwen3-TTS support." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are unrelated requests." + "reason": "NeoBERT and Qwen3-ASR are different model families and changes." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "Different model targets: NeoBERT vs weight-sparse transformer." + "reason": "NeoBERT addition is not the same as adding the weight-sparse transformer model." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "Text-model request vs audio-model request; not the same underlying work." + "reason": "NeoBERT and LFM2.5 Audio are unrelated model requests." }, { "left": "issue:36879", "right": "issue:37015", "accept": false, - "reason": "RF-DETR and NeoBERT are different model additions, not duplicates." + "reason": "RF-DETR and NeoBERT are distinct model additions." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT vs ModernVBERT: separate model support requests." + "reason": "NeoBERT and ModernVBERT are separate model requests despite similar naming." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 support and Qwen3-ASR support are different model integrations." + "reason": "MiniCPM3 architecture support is unrelated to Qwen3-ASR support." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 architecture request vs sparse-transformer model request." + "reason": "MiniCPM3 and the weight-sparse transformer request target different code paths." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 and ModernVBERT are distinct model requests." + "reason": "MiniCPM3 architecture and ModernVBERT are different model implementations." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 vs Voxtral-TTS: different models and modalities." + "reason": "MiniCPM3 and Voxtral TTS support are different model families." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 vision model request vs Qwen3 TTS request; unrelated." + "reason": "DEIMv2 and Qwen3-TTS are unrelated additions." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT and Qwen3-ASR are different support asks." + "reason": "ModernVBERT models and Qwen3-ASR are different model targets." }, { "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT vs TTS support; not the same concrete change." + "reason": "ModernVBERT and Qwen3-TTS are not the same change." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "Related BERT family wording, but one is a specific model request and the other is a broad RoPE-family proposal." + "reason": "ModernVBERT models are more specific than the broad BERT-like RoPE request; not duplicates." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "BERT-like RoPE proposal vs Voxtral-TTS; no shared code path." + "reason": "BERT-like RoPE support and Voxtral TTS support are unrelated model requests." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "TTS and ASR are related audio tasks but separate implementation requests." + "reason": "TTS and ASR requests for Qwen3 are related by family but still distinct support work." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Different model families and request scopes within audio support." + "reason": "Qwen3-TTS support and Voxtral TTS support are different model-specific changes." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "TTS support vs OCR model support; different functionality." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and implementations." }, { "left": "issue:43671", "right": "issue:44121", "accept": false, - "reason": "TTS request and sparse-transformer request are unrelated." + "reason": "Qwen3-TTS support is unrelated to the weight-sparse transformer request." }, { "left": "issue:43671", "right": "issue:43909", "accept": false, - "reason": "Both audio-adjacent, but TTS and LFM2.5 Audio are distinct model additions." + "reason": "Qwen3-TTS and LFM2.5 Audio are separate audio model requests." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "ASR support vs audio model support; not the same change." + "reason": "Qwen3-ASR and LFM2.5 Audio are distinct speech/audio model requests." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and EUPE target different model architectures and tasks." + "reason": "Qwen3-ASR and EUPE are different model families and tasks." }, { "left": "issue:43837", "right": "issue:44121", "accept": false, - "reason": "ASR support vs weight-sparse transformer support; different implementations." + "reason": "Qwen3-ASR support is unrelated to the sparse transformer model request." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "Audio model request vs SAM3-LiteText request; unrelated." + "reason": "LFM2.5 Audio and SAM3-LiteText are different model additions." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Sparse transformer request vs EUPE model request; not the same underlying bug/change." + "reason": "OpenAI weight-sparse transformer and EUPE are unrelated architectures." }, { "left": "issue:43837", "right": "issue:45027", "accept": false, - "reason": "ASR and Voxtral-TTS are different audio model requests." + "reason": "Qwen3-ASR and Voxtral TTS are distinct speech model support requests." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT vs Voxtral-TTS: unrelated model integrations." + "reason": "NeoBERT and Voxtral-4B-TTS are unrelated model additions." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "ASR support vs DeepSeek-OCR2 support; different tasks." + "reason": "Qwen3-ASR and DeepSeek-OCR2 are different modality/model requests." }, { "left": "issue:45322", "right": "issue:46133", "accept": false, - "reason": "EUPE and TIPSv2 are separate model requests from different authors and architectures." + "reason": "EUPE and TIPSv2 are separate vision-language model requests; same broad area, not the same change." } ] }, @@ -18044,12 +18232,12 @@ "soft_pairs": [ "issue:38740|issue:46443" ], - "prepared_review_unit_hash": "2c0adbc07dce7c35314ed8caacf08beb3ec67d09d3bc4fd89158e2d956fe81d6", + "prepared_review_unit_hash": "abd9eec779977db39f047fee02065cfee0208a4393b2c31c9341c1d96525be8a", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1506, + "serialized_chars": 1505, "estimated_input_tokens": 377, "estimated_eval_tokens": 1010 }, @@ -18057,7 +18245,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1506, + "serialized_chars": 1505, "estimated_input_tokens": 377, "estimated_eval_tokens": 1010 }, @@ -18066,18 +18254,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are unrelated: one is a docs request about adding Pruna as an optimization framework, and the other is a Vision RoPE refactor. They should not be clustered or treated as duplicates.", - "confidence": 0.99, - "canonical_issue_reason": "No canonical issue: the issues cover different topics and different kinds of work, with no overlap in target area or problem statement.", + "summary": "The two issues are unrelated: one is a closed docs request about adding Pruna as an optimization framework, and the other is an open technical issue about Vision RoPE refactoring. No duplicate or same-change relationship is evident.", + "confidence": 0.98, + "canonical_issue_reason": "If a representative issue is needed, 46443 is the better anchor because it is the open technical bug/feature discussion; 38740 is a separate closed documentation request.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good global representative for the other; they are unrelated and not duplicates.", + "best_issue_reason": "46443 is the best issue to keep as the cluster representative because it is the substantive open code-related issue, whereas 38740 is a closed docs-only item.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:38740", "right": "issue:46443", "accept": false, - "reason": "Different subject matter entirely: docs/framework integration request vs. vision positional encoding refactor. No evidence they refer to the same underlying bug or change." + "reason": "These describe different topics and different artifact types of work: a docs request for Pruna versus a Vision RoPE refactor. They do not look like the same underlying bug or change." } ] }, @@ -18107,7 +18295,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "af3d6534083f3f87fd12ee9bc3767b406d9ac6a1d87825af4840a439c123c8a7", + "prepared_review_unit_hash": "27e0706e1d263b1917a29471235194bc690c815049d97956cbcd760a7db7bbfb", "component_budget": { "node_count": 2, "item_count": 2, @@ -18129,18 +18317,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one requests block decoding support, the other requests DeepSeek v3.2 support. They describe different functionality and likely different code paths.", - "confidence": 0.94, - "canonical_issue_reason": "41084 is the broader technical feature request and is the better cluster representative, but it is not the same ask as 43037.", + "summary": "These two issues are not duplicates: one is about block decoding behavior, the other is about adding DeepSeek v3.2 model support. The only link is weak semantic similarity, with no shared targets, files, or explicit references.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "41084 is the more general infrastructure issue; 43037 is a model-specific support request and is less suitable as the cluster representative.", - "best_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical representative for the other; they describe different features/problems and should remain separate.", + "best_pr_reason": "No PRs are present.", "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different requests with no clear overlap: block decoding behavior vs support for a specific model. They do not look like the same underlying bug or change." + "reason": "Different underlying requests: block decoding mechanics vs model/version support. No shared code path, filenames, targets, or explicit linkage." } ] }, @@ -18170,12 +18358,12 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "28a5adca3a2e6cf7e04aa200c4e366735119503d215480e04b5bfa23c64c9115", + "prepared_review_unit_hash": "4fe11e143217722151147f55411568f05ca25c06ea378d7b6a68547da6db9992", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1584, + "serialized_chars": 1583, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -18183,7 +18371,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1584, + "serialized_chars": 1583, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -18192,18 +18380,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are related to Trainer customization, but they describe different problems: one is about `remove_unused_column` behavior, the other about FLOPs estimation for custom trainers. They do not look like duplicates.", - "confidence": 0.9, - "canonical_issue_reason": "No true canonical duplicate emerges; if forced, #43598 is the more concrete issue about a specific Trainer behavior change.", + "summary": "These two issues are both about custom trainers, but they describe different problems: one is about `remove_unused_column` customizability in `Trainer`, the other is about FLOPS estimation handling for custom trainers. They should not be treated as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 43598 is the cleaner representative of the cluster because it has the more concrete Trainer customization focus and a clearer code-path target.", "canonical_pr_reason": null, - "best_issue_reason": "#43598 is the better representative because it targets a clearly scoped Trainer customization problem with a specific mechanism (`remove_unused_column`).", + "best_issue_reason": "Issue 43598 is the best issue to keep as a representative example of the cluster, but not because it duplicates 43600; it is simply the more direct and actionable Trainer customization report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Both mention custom trainers, but one concerns unused-column removal and the other concerns FLOPs counting; they affect different code paths and would not be merged as the same bug." + "reason": "Related theme, but different underlying behavior: `remove_unused_column` customization versus FLOPS estimation for custom trainers. Not the same bug or change." } ] }, @@ -18233,12 +18421,12 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "b0d7db918507534b8c0a134d3cbc65dcd0b712ff1d7042b6c1c094b78348c74c", + "prepared_review_unit_hash": "ec57effae4bd49b7794d11f1798968ebe291576c5cb9afe2c7acbc01ef550920", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1611, + "serialized_chars": 1612, "estimated_input_tokens": 403, "estimated_eval_tokens": 1062 }, @@ -18246,7 +18434,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1611, + "serialized_chars": 1612, "estimated_input_tokens": 403, "estimated_eval_tokens": 1062 }, @@ -18255,18 +18443,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are about different model families and different AutoModel entry points, so they should not be deduplicated.", - "confidence": 0.95, - "canonical_issue_reason": "Issue 43684 is the broader and more central request: adding Qwen3-Omni registration to generic auto classes. Issue 44405 is a separate request for sequence-classification support on Qwen3.5.", + "summary": "These are related only at a very high level (Qwen model support), but they describe different gaps: one is about registering Qwen3-Omni in AutoModel/AutoModelForConditionalGeneration, the other is about adding AutoModelForSequenceClassification support for Qwen3.5.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #43684 is the cleaner representative because it targets a concrete registration gap in the core auto-model mappings; #44405 is a separate task for a different config and a different task head.", "canonical_pr_reason": null, - "best_issue_reason": "43684 is the better representative issue because it targets core model registration and is not limited to one task head.", + "best_issue_reason": "#43684 is the best issue to keep as the representative item since it is broader and more directly about model registration, whereas #44405 is a distinct sequence-classification support request.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different underlying requests: Qwen3-Omni auto-registration for conditional generation vs Qwen3.5 sequence-classification support. They affect different configs, tasks, and code paths." + "reason": "Different underlying requests: Qwen3-Omni auto-model registration vs Qwen3.5 sequence-classification support. They affect different model families and API paths, so they should not be merged as duplicates." } ] }, @@ -18303,80 +18491,77 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "b3ffcb8cad40fca7773ca1ee71030dab9d62eaf9ce9005e8b556ca669ff9bc65", + "prepared_review_unit_hash": "e82584b54ea234ccfd57dc23605fb5c26b37d6c2fd3a528d185477da47dca653", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4264, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4264, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "43835 appears separate ('Fett'). The three 'huggingface ' issues look like the same duplicate report cluster, with 43845 as the oldest representative.", - "confidence": 0.87, - "canonical_issue_reason": "issue:43845 is a good canonical issue because it is the oldest of the similar 'huggingface ' reports and can stand in for that duplicate cluster; 43846 has the same body length, so the choice is based on age/representativeness rather than uniqueness of body size.", + "summary": "All four issues are extremely sparse and mostly consist of generic or identical titles with no meaningful body evidence. The cluster looks noisy rather than clearly duplicate-backed, so I would not merge any pair on the current data.", + "confidence": 0.36, + "canonical_issue_reason": "Issue 43859 is the only one with any discussion activity, so it is the least empty anchor in an otherwise low-signal cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43845 is the best overall issue candidate because it is the earliest of the near-identical 'huggingface ' reports and best represents that repeated cluster.", + "best_issue_reason": "43859 is the best available representative because it has the most activity and shares the same generic title pattern as the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Different title and no concrete evidence that 'Fett' is the same underlying issue as the 'huggingface ' reports." + "reason": "No shared bug description or target is available; the similarity is only superficial from short titles/body snippets." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Different title and no shared details beyond weak generic similarity." + "reason": "Insufficient evidence that these describe the same underlying problem; only weak title similarity is present." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "Different title and no concrete sign they describe the same bug." + "reason": "The available metadata does not show a common concrete bug, just generic title overlap." }, { "left": "issue:43845", "right": "issue:43846", - "accept": true, - "reason": "Same title, same body length, and created one minute apart; they look like duplicate reports of the same issue." + "accept": false, + "reason": "Identical generic titles alone are not enough to confirm a duplicate without body or target evidence." }, { "left": "issue:43845", "right": "issue:43859", - "accept": true, - "reason": "Same repeated 'huggingface ' report with no distinguishing evidence; likely a duplicate of the same underlying issue." + "accept": false, + "reason": "No concrete issue details are present to justify treating these as the same report." }, { "left": "issue:43846", "right": "issue:43859", - "accept": true, - "reason": "Same title and no distinguishing evidence; these look like duplicate reports of the same issue." + "accept": false, + "reason": "The pair lacks any substantive evidence beyond broad title similarity, so duplicate status is not supported." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The clustering is mostly grounded, but the canonical-issue rationale overstates the evidence: issue:43845 is not the sole 'fullest body length' match because issue:43846 has the same body length (84). More conservatively, 43845 can be chosen as the oldest representative of the similar 'huggingface ' cluster, but the current wording goes beyond the packet." - }, - "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -18410,88 +18595,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "5cd51429b0076f67e3d4dc60d0dc3f2b83facddf15ef66789172d572824333a1", + "prepared_review_unit_hash": "793e4744f25e0f8627164b3db25be0139b7253e09f8e5de68abd3ad7a5954f2e", "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": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5791, - "estimated_input_tokens": 1448, - "estimated_eval_tokens": 3152 + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are low-signal, likely spam or placeholder issues with no concrete shared bug report, target, or code path. They should not be merged as duplicates; the cluster only shares generic similarity in wording.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 44418 is the most interpretable of the set because it has a generic but at least descriptive title, making it the least arbitrary anchor for the cluster. That said, it is still not clearly related to the others as a duplicate.", + "summary": "These are not genuine duplicates of one bug/change; they look like unrelated spam or low-signal issues with only superficial title similarity. No PRs are present.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 44418 is the best issue candidate only in the sense that it is the most readable and least obviously placeholder-like. None of the issues provide enough substance to justify stronger canonical status.", + "best_issue_reason": "issue:44371 is the most generic placeholder-style item, but the cluster does not form a true duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both look like generic spam placeholders with no shared defect, target, or actionable description." + "reason": "Both are spam-like placeholders, but there is no evidence they describe the same underlying issue." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "The titles are identical placeholders, but there is no evidence of the same underlying bug or request." + "reason": "Shared spammy format only; no concrete bug or change matches." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "One is a spam placeholder and the other is a vague documentation suggestion; they do not describe the same issue." + "reason": "One is a generic spam placeholder and the other is a documentation-suggestion title; unrelated." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "A spam placeholder and a generic name-related title are not enough to establish the same underlying problem." + "reason": "Generic spam vs. a specific-looking phrase about Google LLC; no same underlying problem." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "A vague documentation suggestion and a placeholder spam issue are unrelated at the level needed for duplicate merging." + "reason": "Documentation enhancement and spam placeholder are not the same issue." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "No concrete shared bug or request is described; similarity is only superficial wording." + "reason": "Only superficial similarity in low-information titles; no shared problem statement." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "These are identical spam placeholders, but there is no substantive evidence they report the same problem rather than separate spam submissions." + "reason": "Both are spam placeholders, but there is no substantive evidence they refer to the same report." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "The titles do not point to the same code-path issue or change request; they are just generic text." + "reason": "Different titles and no technical context tying them to one defect." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "No shared actionable defect or target is evident, so this is not a safe duplicate merge." + "reason": "No shared code path or explicit linkage; reject as duplicate." } ] }, @@ -18521,7 +18706,7 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "81dea1fb51d8e711aec43f7470133e97bc37c256a870dde409159e71514ce0d0", + "prepared_review_unit_hash": "5ba7cfb17f4a64a167d899bf28d52005fe1800e26495556bf20b42fa6a6cd8a1", "component_budget": { "node_count": 2, "item_count": 2, @@ -18543,28 +18728,28 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These closed issues look likely to be duplicates based on their identical titles and high textual similarity, but the evidence is limited to soft similarity with no explicit links, targets, or shared files, so the duplicate judgment should remain tentative.", - "confidence": 0.82, - "canonical_issue_reason": "Issue #45145 is the better canonical choice because it was created first and has slightly more linkage evidence (one inbound reference).", + "summary": "These are duplicate issue reports with the same title and very similar content, describing the same energy-reduction proposal. Issue #45145 is the better canonical choice because it was created earlier and already has an inbound reference.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 is slightly preferable as the canonical issue because it was created earlier and already has an inbound reference, while #45151 appears to be the later duplicate.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #45145 is the best issue representative for deduping because it is the earlier instance and has marginally stronger traceability.", + "best_issue_reason": "#45145 is the best issue to keep as the primary tracker: same topic as #45151, earlier creation time, and more established discussion footprint.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", - "accept": false, - "reason": "The pair only has soft similarity evidence; there are no explicit links, issue targets, shared files, or other concrete indicators that they are the same underlying report, so accepting them as duplicates is too strong." + "accept": true, + "reason": "Same title and same apparent report; this looks like a duplicate issue for the same underlying request." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The canonical-choice reasoning is grounded, but the soft-edge verdict is too strong for the evidence provided. The packet only shows a soft_similarity match with no explicit links, targets, or shared files, so accepting the pair as the same issue is not sufficiently conservative. The summary should be phrased more tentatively unless there is stronger linkage evidence." + "accept": true, + "feedback": "The canonical-issue choice is grounded in the packet: #45145 was created earlier and has an inbound reference, while #45151 is later with no inbound references. The duplicate-issue verdict is also supported by the identical titles and soft-similarity signal. No clear overstatement beyond the available evidence." }, "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": false, + "retried": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -18641,7 +18826,7 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "6d0d58225259017f5dbc410325634d69716ee14121b2064e2044cceb00834b43", + "prepared_review_unit_hash": "efe260d6eab4a9a8c569b6df8a6a46a68b072c426cdcc879cdb1f67c8a27ee01", "component_budget": { "node_count": 2, "item_count": 2, @@ -18663,18 +18848,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related only at a very broad \u201c4-bit inference\u201d level, but they describe different bugs: one is a Gemma4/audio-module `torch.finfo()` TypeError on uint8 weights, the other is a BitsAndBytesConfig regression where `llm_int8_skip_modules` clears the default `lm_head` exclusion and triggers an AssertionError.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 45674 is the clearer, more generally stated regression: it names the specific BitsAndBytesConfig interaction and the resulting 4-bit inference failure mode.", + "summary": "These look like two different 4-bit inference bugs: one is a Gemma4/audio-module dtype issue triggered by `torch.finfo()` on `uint8` weights, the other is a `BitsAndBytesConfig` regression where `llm_int8_skip_modules` drops the default `lm_head` exclusion and causes an assertion failure. They share the same broad area but not the same underlying failure mode.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 45674 is the better canonical issue for this cluster because it states a concrete configuration regression with a clear failure path; issue 45672 appears to be a separate, model-specific dtype error rather than the same bug.", "canonical_pr_reason": null, - "best_issue_reason": "45674 is the best representative of the cluster because it isolates a concrete config bug and is more likely to be useful as the canonical tracking issue; 45672 is a separate Gemma4/audio-specific error path.", + "best_issue_reason": "Issue 45674 is the strongest representative because it describes a general, actionable 4-bit inference bug in `BitsAndBytesConfig`, whereas 45672 is narrower and tied to Gemma4 audio modules.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Both involve 4-bit inference, but the underlying failures differ: `torch.finfo()` on uint8 weights in Gemma4 audio modules vs. `llm_int8_skip_modules` removing the default `lm_head` exclusion and causing an AssertionError." + "reason": "Both mention 4-bit inference, but the failures are different: `torch.finfo()` on `uint8` weights in audio modules versus a skipped-module exclusion regression causing an assertion. Not the same underlying bug." } ] }, @@ -18704,7 +18889,7 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "422e6dbd8dda81868e57042b269be6823b8aa6f9f6abd86a17a1f3c6c9c0ad2f", + "prepared_review_unit_hash": "45076df33bf696a60dc81435a103fea2000b4f9d25e88eaf4bf7ff2927377758", "component_budget": { "node_count": 2, "item_count": 2, @@ -18726,7 +18911,7 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two closed issues with only personal-name/date titles and no body content; they do not appear to describe the same underlying bug or request.", + "summary": "Both items look like unrelated, low-signal closed issues with only personal-name/date titles and no body or technical context. They should not be clustered as duplicates.", "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, @@ -18737,7 +18922,7 @@ "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "Different named subjects and dates; no body, targets, files, or other evidence of the same problem. This looks unrelated rather than duplicate." + "reason": "The titles refer to different names and different dates, with no shared bug report, feature request, or technical context. This is not the same underlying issue." } ] }, @@ -18767,7 +18952,7 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "44273c9bd03b3ddb7e2df68d92b662087cdb61aef957a47ddf25a7b146172740", + "prepared_review_unit_hash": "e01f4268c4e9230bd9b5e3273bd1e26b15d59bf83b615ad0cf6f85f4ba42a884", "component_budget": { "node_count": 2, "item_count": 2, @@ -18789,18 +18974,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues do not appear to be duplicates; they are unrelated low-signal reports with different names and dates and no shared problem description.", - "confidence": 0.97, - "canonical_issue_reason": "No canonical issue: neither issue describes a concrete bug or feature, and the titles point to different personal-name entries.", + "summary": "These are not duplicates; they look like unrelated closed issues with only superficial title similarity from personal names/dates.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a strong canonical target; both lack actionable details and do not indicate the same underlying problem.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "The titles reference different people and dates, with no overlapping bug, code path, or issue context, so they are not the same underlying report." + "reason": "Different named subjects and dates, no shared body, targets, filenames, or code-related context; not the same underlying issue." } ] }, @@ -18830,12 +19015,12 @@ "soft_pairs": [ "issue:45810|issue:46295" ], - "prepared_review_unit_hash": "019675aa21d88f77bec4841fbf764ad968c9174cccec775b0c5b375c3a9aaa73", + "prepared_review_unit_hash": "e1de639e7eb24e8fd7a944d87d8ccf712e29514279e5f411267d45176540ebf7", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1542, + "serialized_chars": 1541, "estimated_input_tokens": 386, "estimated_eval_tokens": 1028 }, @@ -18843,7 +19028,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1542, + "serialized_chars": 1541, "estimated_input_tokens": 386, "estimated_eval_tokens": 1028 }, @@ -18852,18 +19037,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are unrelated: one requests a new Qwen3_5 token-classification model class, while the other asks for FP8 training support. They do not describe the same bug or change.", + "summary": "The two issues are unrelated: one requests a new Qwen3_5 token-classification model class, while the other requests FP8 training support. The soft similarity signal is not enough to treat them as duplicates.", "confidence": 0.98, - "canonical_issue_reason": "Issue #45810 is the clearer canonical issue only because it is a concrete, narrowly scoped feature request with a specific model/API addition; #46295 is a separate platform/training feature request.", + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "#45810 is the better representative of the cluster because it has a specific target and implementation request, whereas #46295 is a broader training capability request and not a duplicate of the model-class addition.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45810", "right": "issue:46295", "accept": false, - "reason": "Different underlying requests: one is for adding Qwen3_5ForTokenClassification, the other is for FP8 training support. No shared code-path or concrete change overlap." + "reason": "Different underlying requests and code paths: model-class addition vs FP8 training support." } ] }, @@ -18903,12 +19088,12 @@ "issue:45871|issue:45873", "issue:45872|issue:45873" ], - "prepared_review_unit_hash": "2e2cdf4882b84273c7d82097adaa3cee3fb56f6d4489fa2dff454a6c65fa3d65", + "prepared_review_unit_hash": "434fb9c7093308f13bcb025257b5585ece98e96f9f154486699caea864deb6ef", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5727, + "serialized_chars": 5728, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -18916,7 +19101,7 @@ "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5727, + "serialized_chars": 5728, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -18925,69 +19110,69 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "There are two clear duplicate pairs: 45870/45871 and 45872/45873. Issue 45869 is thematically related to finance/data, but it is a different request and not a duplicate of either pair.", - "confidence": 0.95, - "canonical_issue_reason": "45872 is the best canonical issue for the set because it is the most substantive, concrete request and has a verbatim duplicate in 45873.", + "summary": "The set most likely splits into two near-duplicate issue pairs: 45870/45871 and 45872/45873. Issue 45869 looks separate and only loosely related by broad financial theme. The evidence supports duplicate triage, but not a claim of exact duplication.", + "confidence": 0.87, + "canonical_issue_reason": "45870 is the clearest representative for the example-request pair because 45871 matches it extremely closely in title and body length; this is strong near-duplicate evidence, though not enough to prove an exact duplicate.", "canonical_pr_reason": null, - "best_issue_reason": "45872 is the strongest representative overall: it is specific, actionable, and duplicated exactly, while 45869 is unrelated spam-like dataset promotion and 45870/45871 are a narrower example request.", + "best_issue_reason": "45870 is a good anchor for deduping one pair, but there is no single canonical issue for all five: 45872/45873 form a separate near-duplicate pair, and 45869 is distinct.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45869", "right": "issue:45870", "accept": false, - "reason": "Both mention finance/data, but one is a dataset bundle offer and the other is an example/prediction issue; different underlying request." + "reason": "Different request types: a dataset bundle offer versus an example/tutorial request." }, { "left": "issue:45869", "right": "issue:45871", "accept": false, - "reason": "Dataset bundle promotion is not the same underlying issue as the example/prediction request." + "reason": "Different request types: a dataset bundle offer versus an example/tutorial request." }, { "left": "issue:45870", "right": "issue:45871", "accept": true, - "reason": "Exact duplicate title/body; same example request." + "reason": "Very likely duplicate: identical title and matching body length/overall request wording." }, { "left": "issue:45870", "right": "issue:45872", "accept": false, - "reason": "Related financial/DataForge topic, but the concrete asks differ: example prediction vs scraping-service integration." + "reason": "Related financial/DataForge theme, but the requests are different: an example request versus a scraping-service request." }, { "left": "issue:45870", "right": "issue:45873", "accept": false, - "reason": "Same broad domain, but not the same request; one is an example issue and the other is a scraping-service feature request." + "reason": "Related financial/DataForge theme, but the requests are different: an example request versus a scraping-service request." }, { "left": "issue:45871", "right": "issue:45872", "accept": false, - "reason": "Different concrete feature/request despite similar finance/DataForge wording." + "reason": "Related theme only; the request wording points to different needs, so there is not enough evidence of the same underlying issue." }, { "left": "issue:45871", "right": "issue:45873", "accept": false, - "reason": "Not the same underlying issue; only the broad finance/DataForge theme overlaps." + "reason": "Related theme only; the request wording points to different needs, so there is not enough evidence of the same underlying issue." }, { "left": "issue:45872", "right": "issue:45873", "accept": true, - "reason": "Exact duplicate title/body; same feature request." + "reason": "Very likely duplicate: identical title and matching body length/overall request wording." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The duplicate-pair conclusions are grounded by the packet: 45870/45871 and 45872/45873 have identical titles and matching body lengths with jaccard 1.0, while the 45869 pairings are only soft-similar and reasonably rejected as different requests. The only mildly subjective phrasing is the characterization of 45869, but it does not materially affect the conservative clustering verdicts." + "accept": false, + "feedback": "The clustering is too strong in a few places. The packet only provides soft_similarity plus matching titles/body lengths, which supports near-duplicate or very likely duplicate, but not a fully grounded claim of \"exact duplicate.\" Also, phrases like \"different code paths\" are unsupported because there is no diff/code evidence. The overall split into two duplicate pairs plus one separate item is plausible, but the soft-edge verdicts and summary should be more conservative." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -19011,46 +19196,49 @@ "soft_pairs": [ "issue:46439|issue:46441" ], - "prepared_review_unit_hash": "d2a65ff5d1f1cc7224be267e1a46a542596cc1251adc2362346e5dd608513cd0", + "prepared_review_unit_hash": "cba103cba046a56ae73fe5d5d0f80d4a829b633c661c46e90c36c74b64223420", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1637, - "estimated_input_tokens": 410, - "estimated_eval_tokens": 1076 + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1637, - "estimated_input_tokens": 410, - "estimated_eval_tokens": 1076 + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both issues involve static-cache behavior around linear-attention models, but they describe different failure modes: one is shape corruption during early initialization, the other is a missing `linear_attention` cache entry causing a `KeyError`. They are related, but not clearly the same underlying bug.", - "confidence": 0.82, - "canonical_issue_reason": "issue:46439 is the better canonical issue because it names the broader mechanism (`early_initialization`) and the concrete impact on linear-attention layers, which could encompass related static-cache failures.", + "summary": "The two issues are related in topic (`cache_implementation=\"static\"` and linear-attention behavior), but the packet only provides soft similarity and no direct evidence that they are the same bug. Issue 46439 is the broader, more general report, while 46441 looks like a possibly related but distinct symptom report.", + "confidence": 0.47, + "canonical_issue_reason": "Issue 46439 is the better canonical issue only in a tentative sense: it describes the broader static-cache / linear-attention failure mode, whereas 46441 is a narrower Qwen3.5 `KeyError` symptom. The evidence here is not strong enough to prove they share the same root cause.", "canonical_pr_reason": null, - "best_issue_reason": "issue:46439 is the best representative issue for the cluster since it is more general and describes the likely root-path problem rather than one product-specific crash signature.", + "best_issue_reason": "Issue 46439 is the best issue candidate because it is more general and mechanism-focused, making it the most suitable representative if these reports are later confirmed related.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46439", "right": "issue:46441", "accept": false, - "reason": "Related static-cache/linear-attention crashes, but the reported failures differ too much (`RuntimeError` from wrong shapes vs `KeyError: 'linear_attention'`) to confidently treat as the same underlying bug." + "reason": "The evidence is too weak to treat these as the same underlying bug. They share a broad static-cache / linear-attention theme, but the packet only shows title-level similarity and no direct linkage tying the shape corruption report to the Qwen3.5 `KeyError`." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "The clustering is a bit too strong for the evidence in the packet. We only have title-level similarity plus a soft-similarity signal; there is no direct linkage showing issue 46441 is a symptom of the same bug as 46439. The summary and canonicalization language go beyond what\u2019s grounded here by asserting a shared underlying root cause rather than a plausible relation. A more conservative verdict would keep this as a tentative similarity, not an accepted same-bug pair." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -19058,7 +19246,7 @@ { "label": "pull_request", "component_index": 1, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19104,154 +19292,154 @@ "pull_request:43823|pull_request:45213", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "02d9a140781f952bcaccece7a709f46fe84d2d6904fa319138d27b3d4efda5d0", + "prepared_review_unit_hash": "3f8a803f257d71da3b1473f409bb106d7a3f91d2ea7e8ae340b98bfd912a6bcf", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 19, - "serialized_chars": 38086, - "estimated_input_tokens": 9522, - "estimated_eval_tokens": 19300 + "serialized_chars": 38090, + "estimated_input_tokens": 9523, + "estimated_eval_tokens": 19302 }, "budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 19, - "serialized_chars": 38086, - "estimated_input_tokens": 9522, - "estimated_eval_tokens": 19300 + "serialized_chars": 38090, + "estimated_input_tokens": 9523, + "estimated_eval_tokens": 19302 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated pull requests spanning new model additions, refactors, docs, and small bug fixes. The only clear duplicate pair is the two `fixed_cross_entropy` PRs, which both address forwarding supported kwargs to `nn.functional.cross_entropy` in the same code path.", - "confidence": 0.9, + "summary": "This cluster is mostly unrelated open PRs spanning model additions, tokenizer utilities, docs, and infrastructure. The only true duplicate-looking pair is the two `fixed_cross_entropy` PRs (#43251 and #43254), which patch the same loss helper for the same issue. Everything else should stay separate.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the best canonical PR here: it directly targets the explicit issue 43240 and implements the cross-entropy kwargs forwarding in `loss_utils.py`; PR 43254 is a narrower near-duplicate variant of the same fix.", + "canonical_pr_reason": "PR #43251 is the best canonical PR for the one real duplicate pair: it directly fixes `fixed_cross_entropy` in `loss_utils.py`, has the explicit issue target #43240, and is the more clearly established implementation of that change.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the strongest representative of the only real duplicate subcluster because it has the explicit issue link and the most complete/obvious fix for the shared loss-function bug.", + "best_pr_reason": "PR #43251 is the strongest representative in this set because it is a concrete, narrowly scoped code-path fix with an explicit linked issue, and it matches the only clearly duplicate companion PR.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Unrelated PRs: VideoPrism model/docs addition vs. Executorch dynamic-shape export test. Different code paths and different goals." + "reason": "Unrelated PRs: VideoPrism model integration vs ExecuTorch dynamic-shape export test. No shared underlying bug or change." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Different tokenizer maintenance vs. adding MobileLLM model files. No shared bug or change." + "reason": "Different areas: Mistral Common tokenizer methods vs adding MobileLLM. Same broad tokenizer/model codebase, but not the same fix." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Mistral tokenizer missing-methods fix is unrelated to Executorch export dynamic-shape testing." + "reason": "Tokenizer API completion and ExecuTorch export testing are unrelated." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor auto-loading robustness changes are unrelated to a new MobileLLM model implementation." + "reason": "Processor auto-loading robustness and new MobileLLM model support are different changes." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Cross-entropy loss kwargs bug fix and MobileLLM model addition are unrelated changes." + "reason": "Loss helper kwargs fix vs adding a new model package; no shared code-path problem." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "`fixed_cross_entropy` kwargs forwarding is unrelated to Mistral4 test/model work; different functionality and code path." + "reason": "`fixed_cross_entropy` kwargs handling is unrelated to Mistral4 test/model updates." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Executorch dynamic-shape export test and Mistral4 fixes are unrelated." + "reason": "ExecuTorch export testing and Mistral4 fixes are unrelated." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Executorch export test and MobileLLM model addition do not fix the same bug or feature." + "reason": "No same bug/change: export tests vs new model implementation." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New MobileLLM model support vs. a modeling_utils config-resolution fix are unrelated." + "reason": "New MobileLLM model support is unrelated to `cls.config` resolution ordering." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM model addition is unrelated to Falcon output-collection refactoring." + "reason": "MobileLLM addition vs Falcon output-collection refactor are distinct changes." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon model refactor and model-creation-skill repo changes are unrelated." + "reason": "Falcon model refactor and model-creation skill content are unrelated." }, { "left": "pull_request:43995", "right": "pull_request:44827", "accept": false, - "reason": "Falcon refactor and Mistral4 test fixes do not address the same concrete problem." + "reason": "Falcon standardized output refactor and Mistral4 test fixes do not address the same concrete issue." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test fixes and CLIP-like `layrnorm` renaming are unrelated changes." + "reason": "Mistral4 test/model work and `layrnorm` renaming across CLIP-like models are different fixes." }, { "left": "pull_request:44827", "right": "pull_request:46267", "accept": false, - "reason": "Mistral4 test fixes and chat response parsing updates are unrelated." + "reason": "Mistral4 fixes and chat-response parsing updates are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:46048", "accept": false, - "reason": "Mistral4 test fixes and TDT loss kernel integration are unrelated." + "reason": "Mistral4 test/model work and TDT loss kernel integration are separate changes." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 test fixes and `cls.config` resolution ordering are unrelated." + "reason": "Mistral4 test/model work and `cls.config` resolution ordering are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 test fixes and model-creation skill/docs changes are unrelated; shared auto/modeling filename is incidental." + "reason": "Touching `modeling_auto.py` is not enough to merge these: Mistral4 model support and model-creation skill are not the same change." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "MobileLLM model addition and repository skill/draft tooling changes are unrelated." + "reason": "MobileLLM model support and the model-creation skill are unrelated." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both PRs patch the same `fixed_cross_entropy` function to support passing extra `cross_entropy` kwargs; they are alternative formulations of the same loss-function fix and could plausibly be merged into one PR." + "reason": "Both PRs fix the same `fixed_cross_entropy` helper in `loss_utils.py` for the same underlying issue (#43240); the only differences are implementation style and API forwarding details." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet. The only supported duplicate pair is 43251/43254: both target `src/transformers/loss/loss_utils.py`, share issue target 43240, and are mergeable as the same cross-entropy kwargs bugfix. The other soft-edge rejections are conservative and consistent with the diffs/titles." + "feedback": "The summary is well grounded in the packet. The only clearly duplicate-looking pair is the two `fixed_cross_entropy` PRs (#43251 and #43254), which both target the same helper in `loss_utils.py` and share issue #43240, so treating them as mergeable into one bugfix PR is reasonable. The other pair verdicts stay conservative and do not overstate duplication." }, "evaluator_used": true, "retried": false, @@ -19262,7 +19450,7 @@ { "label": "pull_request", "component_index": 2, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19280,12 +19468,12 @@ "pull_request:40546|pull_request:42781", "pull_request:42781|pull_request:46331" ], - "prepared_review_unit_hash": "0636f439819533c3cbaa8217eabd2fdd53b9308f3add867ac1c93e63959783ff", + "prepared_review_unit_hash": "b451aabbe9bd255d9a961c0edecd7e8ad2167ac6c7df83e99f4a660b052ffa3b", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10870, + "serialized_chars": 10869, "estimated_input_tokens": 2718, "estimated_eval_tokens": 5692 }, @@ -19293,7 +19481,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10870, + "serialized_chars": 10869, "estimated_input_tokens": 2718, "estimated_eval_tokens": 5692 }, @@ -19302,24 +19490,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three distinct model-addition PRs. PR 40546 implements the base VibeVoice model, PR 42781 adds a VibeVoice Realtime variant plus acoustic tokenizer work, and PR 46331 adds Parakeet-RNNT. They share some auto-mapping and docs infrastructure, but not the same underlying change.", - "confidence": 0.91, + "summary": "These PRs are not duplicates: one adds core VibeVoice support, one adds a realtime VibeVoice variant, and one adds Parakeet-RNNT. They share some auto-registration and speech-model plumbing, but they do not fix the same concrete change.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 40546 is the cleanest representative of the cluster because it is the base VibeVoice implementation; the other PRs are separate extensions or unrelated model families.", + "canonical_pr_reason": "PR 40546 is the best representative of the cluster because it is the main VibeVoice implementation and has the broadest, most foundational scope among the VibeVoice-related changes.", "best_issue_reason": null, - "best_pr_reason": "PR 40546 is the strongest single PR to keep as the representative artifact: it introduces the core VibeVoice support and has the broadest, most foundational scope among the three.", + "best_pr_reason": "PR 40546 is the strongest standalone candidate for merging because it introduces the core model support rather than a narrower follow-up variant or an unrelated ASR model.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Related VibeVoice work, but not the same concrete change: 40546 is base VibeVoice support, while 42781 adds realtime/acoustic-tokenizer functionality and additional generation/model files. They would not plausibly merge as one duplicate PR." + "reason": "Both are VibeVoice-related, but 40546 is the base VibeVoice implementation while 42781 adds a realtime variant with additional tokenizer/feature-extractor work. They are related but not the same underlying change and would not plausibly be merged as one PR." }, { "left": "pull_request:42781", "right": "pull_request:46331", "accept": false, - "reason": "Different model families and different feature sets: VibeVoice Realtime versus Parakeet-RNNT. Shared auto-mapping files are incidental; this is not the same underlying bug or change." + "reason": "These target different model families: VibeVoice Realtime versus Parakeet-RNNT. Shared auto files are broad framework plumbing, not evidence of the same code-path fix or change." } ] }, @@ -19333,7 +19521,7 @@ { "label": "pull_request", "component_index": 3, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19349,7 +19537,7 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "eddc8339aad2d76725bd2519892a76e279dee2e6ca601259f6441f8b64120c1e", + "prepared_review_unit_hash": "5bb6396d8249f92deaee4f521d0b6d2592a43403d72478f1bf600d6105c80a63", "component_budget": { "node_count": 2, "item_count": 2, @@ -19371,18 +19559,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are unrelated feature additions: one adds TimesFM covariate forecasting support, the other adds DINOv3 image classification support. They do not look like the same underlying change.", + "summary": "These are unrelated feature PRs: one adds TimesFM forecasting with covariates, the other adds DINOv3 image classification support. The soft similarity is superficial and should not be merged as a duplicate cluster.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 41224 is the cleaner, more self-contained feature PR and is the better representative of the cluster only by default, not because it duplicates 40755.", + "canonical_pr_reason": "No canonical PR: the PRs target different models, different code paths, and different user-facing features, so there is no single underlying change here.", "best_issue_reason": null, - "best_pr_reason": "PR 41224 is the most suitable representative in this set because it is narrowly scoped and self-contained, but it is not a duplicate of 40755.", + "best_pr_reason": "Neither PR is a good global representative for the cluster because they are unrelated and do not describe the same bug or change.", "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different model families and code paths: TimesFM covariate forecasting vs DINOv3 image classification. The changes do not implement the same bug fix or feature and would not plausibly merge into one PR." + "reason": "Reject: PR 40755 adds TimesFM covariates support across TimesFM docs/modeling/tests, while PR 41224 adds DINOv3 image classification support and updates DINOv3 docs/auto/modeling/tests. They touch different subsystems and fix different problems." } ] }, @@ -19396,7 +19584,7 @@ { "label": "pull_request", "component_index": 4, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19459,7 +19647,7 @@ { "label": "pull_request", "component_index": 5, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19477,7 +19665,7 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "0dcd2d804e8e3a4e79ef87ce505ffc63213baac9b6eb44a80dcb9158bf767494", + "prepared_review_unit_hash": "84c622cc01ce12eebfeea87300075161914cbbfe7a1a4b90da10c0b13d09da3f", "component_budget": { "node_count": 3, "item_count": 3, @@ -19499,24 +19687,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-addition PRs, not duplicates. They share boilerplate registration/docs changes, but each introduces a different model with distinct implementations and tests.", - "confidence": 0.98, + "summary": "All three items are separate model-addition PRs with similar boilerplate changes to docs and auto-registrations, but they introduce different models and codepaths. They are not duplicates or mergeable into one PR.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: DeepSeek OCR, Isaac, and Kimi2-6 are distinct model introductions with different code paths and files.", + "canonical_pr_reason": "PR #45630 is the strongest cluster representative because it has the highest review activity and the most complete-looking implementation surface among the three, but it is still a distinct model addition rather than a duplicate of the others.", "best_issue_reason": null, - "best_pr_reason": "No single PR is a better duplicate representative; none of these PRs fixes or implements the same underlying change.", + "best_pr_reason": "PR #45630 is the best single representative of the cluster for triage because it is the most active and most developed PR in the set; however, it should not be treated as the same change as the other two.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different models (DeepSeek OCR vs Isaac); overlap is only in generic auto-registration/docs boilerplate." + "reason": "DeepSeek OCR and Isaac are different model additions with different filenames, configs, and test suites. The overlap is just standard Transformers boilerplate for registering a new model." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "Separate model additions (Isaac vs Kimi2-6) with different implementations and filenames, so they are not the same change." + "reason": "Isaac and Kimi2-6 add different models and related modules. Shared edits to auto-mapping and docs are generic new-model integration work, not the same underlying bug or change." } ] }, @@ -19530,7 +19718,7 @@ { "label": "pull_request", "component_index": 6, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19546,7 +19734,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "0dda4726c6e35ef528ee3ae8592bf3cbe6fd31e7627f90f07692078b9a3514dd", + "prepared_review_unit_hash": "9c65b5f2b6f9e1f3e1de3b259482e1f7cd1d282100d21f340b3805949f1cbe8a", "component_budget": { "node_count": 2, "item_count": 2, @@ -19568,18 +19756,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two workflow dependency-bump PRs with overlapping files, but they upgrade different GitHub Actions families and address different maintenance goals, so they are not duplicates.", - "confidence": 0.91, + "summary": "Two separate GitHub Actions maintenance PRs: one broad Node 24 compatibility upgrade across many workflow actions, and one smaller upgrade focused on docker/build-push-action. They overlap in theme and some files, but not in the concrete change or code path.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 42975 is the better representative of the cluster because it is a broad GitHub Actions compatibility sweep across many workflows, while 42976 is a narrower set of action-version bumps focused mainly on Docker and a few other workflows.", + "canonical_pr_reason": "PR 42975 is the broader representative: it updates many workflow actions and clearly states the Node 24 compatibility goal, making it a better cluster anchor than the narrower Docker action bump.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the strongest single representative for triage because its title and diff clearly describe the main maintenance theme: updating GitHub Actions for Node 24 compatibility across the workflow suite.", + "best_pr_reason": "PR 42975 is the best single representative because it covers the wider set of workflow action upgrades and has the clearest maintenance intent. PR 42976 is narrower and specific to Docker build action versioning.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "They overlap in workflow files, but the actual changes are different action upgrades: 42975 updates checkout/cache/upload-artifact for Node 24 compatibility, while 42976 bumps docker/build-push-action and other actions to latest versions. Same subsystem, not the same underlying change." + "reason": "Do not merge: both are workflow dependency upgrades, but they change different actions for different purposes. 42975 targets checkout/cache/upload-artifact and other GitHub actions for Node 24 compatibility; 42976 upgrades docker/build-push-action in build workflows. They are related maintenance PRs, not the same underlying change." } ] }, @@ -19593,7 +19781,7 @@ { "label": "pull_request", "component_index": 7, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19616,7 +19804,7 @@ "pull_request:43448|pull_request:45490", "pull_request:43451|pull_request:45490" ], - "prepared_review_unit_hash": "0686b52931ac783e8248a7f81474a3f1298c4766d404a0fb0103ddb4b4583d10", + "prepared_review_unit_hash": "72a72eade7e8515b66b9bbbc62b65447e36ffa5b9a48bfdd0a15c19a34017a1d", "component_budget": { "node_count": 4, "item_count": 4, @@ -19638,48 +19826,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are all distinct model-addition efforts (ViT NEPA, Molmo, Molmo2, CTSM). The overlap is mostly boilerplate around docs and auto-registry files, not a shared underlying bug or mergeable change set.", + "summary": "These are four separate model-addition pull requests (ViT NEPA, Molmo, Molmo2, CTSM). They share framework plumbing and docs patterns, but they do not fix the same code-path problem or represent mergeable duplicates.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43451 (Molmo2) is the most complete and active representative of the group, with the broadest model integration footprint, but it is not a duplicate of the others.", + "canonical_pr_reason": "No true duplicate PR exists here; if one must be used as the cluster representative, 43451 (Molmo2) is the most substantial and review-intensive model-addition PR, with the broadest implementation surface.", "best_issue_reason": null, - "best_pr_reason": "PR 43451 is the strongest standalone representative if one must anchor the cluster, but the set is heterogeneous and has no true duplicate PR.", + "best_pr_reason": "43451 is the best representative only in the sense of being the most complete/active PR in the group. It is not a better duplicate than the others because all four are distinct model integrations.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo); shared docs/auto files are generic scaffolding, not the same code-path fix." + "reason": "Different models and different integrations (ViT NEPA vs Molmo); shared auto/docs files are generic model-registration plumbing, not the same underlying change." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "ViT NEPA and Molmo2 are unrelated model implementations; they touch some of the same registry/docs files, but not the same underlying change." + "reason": "ViT NEPA and Molmo2 are unrelated model additions with different code paths and filenames; they are not mergeable into one fix/change." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "CTSM and ViT NEPA are separate model ports with different architectures and code paths; no duplicate bug or change." + "reason": "ViT NEPA and CTSM are distinct model ports with no shared bug or concrete implementation target beyond standard registration/docs updates." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are different model families; overlap is mostly boilerplate registry/docs updates, not one concrete fix." + "reason": "Molmo and Molmo2 are separate model families; overlap is limited to common auto/docs/test scaffolding, so they are not duplicates." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo and CTSM are unrelated additions; shared auto/model registration files do not make them the same PR." + "reason": "Molmo and CTSM address different models and different preprocessing/modeling code paths; same subsystem alone is not enough to merge them." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 vs CTSM are distinct model integrations with different implementation paths and targets, so they should not be merged as duplicates." + "reason": "Molmo2 and CTSM are unrelated model additions with only generic shared registration files, not the same concrete change." } ] }, @@ -19693,7 +19881,7 @@ { "label": "pull_request", "component_index": 8, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19716,70 +19904,70 @@ "pull_request:43665|pull_request:46380", "pull_request:44815|pull_request:46380" ], - "prepared_review_unit_hash": "8addc19d8a1966053d97ee15763c5e9ccc256168a43c8f9fc3c3340dd1be3e8d", + "prepared_review_unit_hash": "7cf21e7e07cb87eaad3b830f338746f76d76645f08439829e2243c4584949327", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 11142, - "estimated_input_tokens": 2786, - "estimated_eval_tokens": 5828 + "serialized_chars": 11145, + "estimated_input_tokens": 2787, + "estimated_eval_tokens": 5830 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 11142, - "estimated_input_tokens": 2786, - "estimated_eval_tokens": 5828 + "serialized_chars": 11145, + "estimated_input_tokens": 2787, + "estimated_eval_tokens": 5830 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are not true duplicates: they target different bugs in different subsystems (Whisper decoding, model tests, FP8 dequantization, and generation config validation). The superficial similarity is just that they are all small \"fix\" PRs with tests.", - "confidence": 0.92, + "summary": "No duplicate cluster here: the four PRs touch unrelated areas (Whisper tokenization, CLIP/ViT test files, FP8 dequant/loading, and generation watermarking). All soft links should be rejected.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "If one PR must represent the cluster, #46380 is the clearest concrete bug fix with an explicit code-path change in generation validation; however, the set is not a duplicate cluster overall.", + "canonical_pr_reason": "No true canonical PR emerges because the PRs are about different bugs and code paths; none are mergeable as one fix.", "best_issue_reason": null, - "best_pr_reason": "#46380 is the strongest standalone representative because it fixes an actual runtime validation bug in core generation logic rather than being test-only or narrowly model-specific.", + "best_pr_reason": "PR 44815 is the strongest standalone code fix, with a concrete runtime/quantization change and supporting tests, but it is still unrelated to the others.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Different changes: #43291 fixes Whisper tokenization/decode behavior, while #43665 only adds failing assertions in CLIP/ViT tests and does not address the same bug." + "reason": "Different targets and problems: Whisper decoding/tokenizer code vs CLIP/ViT test edits; not the same underlying bug." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "#43291 is a Whisper decode fix; #44815 is an FP8 dequant/core loading change across Mistral-related code. They do not appear to fix the same underlying defect." + "reason": "Whisper tokenizer decoding fix is unrelated to FP8 dequant/model-loading changes in a different subsystem." }, { "left": "pull_request:43291", "right": "pull_request:46380", "accept": false, - "reason": "Unrelated code paths: Whisper decoding versus generation watermarking validation / generation utils. Similar only in being fixes with tests." + "reason": "Whisper tokenization vs generation watermarking/config validation are separate code paths and bugs." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "#43665 is test-only noise/placeholder assertions, not a real product fix; #44815 is a concrete FP8 dequantization fix. Not the same issue." + "reason": "Test-file assertions in CLIP/ViT are not the same issue as FP8 dequant/loading logic; no shared concrete bug." }, { "left": "pull_request:43665", "right": "pull_request:46380", "accept": false, - "reason": "#43665 only modifies tests, while #46380 changes generation config validation behavior and related tests. They are not one underlying bug or mergeable fix." + "reason": "These are unrelated test edits versus generation/watermarking behavior; they do not fix the same problem." }, { "left": "pull_request:44815", "right": "pull_request:46380", "accept": false, - "reason": "Different functional areas and failure modes: FP8 dequantization/model loading versus generation watermarking config validation." + "reason": "FP8 quantization/loading fix and watermarking config validation are distinct runtime issues with different code paths." } ] }, @@ -19793,7 +19981,7 @@ { "label": "pull_request", "component_index": 9, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19809,12 +19997,12 @@ "soft_pairs": [ "pull_request:43488|pull_request:43532" ], - "prepared_review_unit_hash": "19410f91906d3ca7bff72498549623bda2893021f09a13bc3bd83ddc5c67c7eb", + "prepared_review_unit_hash": "1bc29dd3ff7aed61a6bc4be4e922e4bcb0900d38b86751958dd2b66fec235e4b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4563, + "serialized_chars": 4562, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -19822,7 +20010,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4563, + "serialized_chars": 4562, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -19831,18 +20019,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two pull requests are not duplicates: one is a small repo-bot/setup tweak with a VIT test touch, while the other is a broader CI/workflow and test-utility change. The shared test filename is incidental, not evidence of the same underlying fix.", - "confidence": 0.98, + "summary": "These two pull requests are both clearly non-production test/bot exercises, but they are not duplicates of the same underlying change: one tweaks packaging/version metadata and a Vit test, while the other rewires CI/workflow and testing utilities plus a Vit test. The only overlap is a shared test filename, which is not enough to merge them as one change.", + "confidence": 0.9, "canonical_issue_reason": null, - "canonical_pr_reason": "No clear canonical PR for a duplicate cluster; the PRs address different concerns and code paths.", + "canonical_pr_reason": "PR #43532 is the better canonical representative for the cluster because it has the broader and more distinctive bot/CI-oriented diff, touching multiple workflow and utility files rather than mostly packaging metadata.", "best_issue_reason": null, - "best_pr_reason": "43532 is the more substantial and representative change set, with coherent CI/workflow and test utility edits; 43488 is a narrow synthetic setup/version tweak.", + "best_pr_reason": "PR #43532 is the strongest single representative PR in this set: its title and diff better summarize the cluster\u2019s bot/demo nature and it spans the largest, most identifiable change surface.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "Different underlying changes: 43488 edits setup/version/dependency metadata and a ViT test, while 43532 rewrites self-comment CI and related workflow/utility files. They do not appear mergeable into one fix, and the shared test file is incidental." + "reason": "Reject: the diffs address different changes. #43488 is about setup/version/dependency metadata plus a Vit test, while #43532 changes CI workflows, testing utilities, and notification code. Shared test-file overlap does not indicate the same fix." } ] }, @@ -19856,7 +20044,7 @@ { "label": "pull_request", "component_index": 10, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19872,7 +20060,7 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "0d06e63a77902d4bf13af072d706b0fdba1368cb362316cea92ec5c093ef1f5d", + "prepared_review_unit_hash": "9324f395da004cb57f2271acf65f92689bed1296ab86550bfb1319fdc4b48cbb", "component_budget": { "node_count": 2, "item_count": 2, @@ -19894,18 +20082,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two separate PRs add different segmentation pipelines: one for promptable concept segmentation and one for promptable visual segmentation. They share documentation and pipeline scaffolding, but they target different functionality and model sets, so they should not be merged as duplicates.", - "confidence": 0.95, + "summary": "These are related but distinct PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share a documentation/pipeline-registration pattern and some infrastructure, but they target different tasks and different model/test surfaces, so they should not be deduplicated.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43613 is the broader/more central representative of this cluster because it adds the more general promptable visual segmentation pipeline and touches the largest set of related model integrations and tests.", + "canonical_pr_reason": "PR 43613 is the broader and more central change: it adds the visual segmentation pipeline with wider model/test coverage across SAM/EdgeTam/SAM2/SAM3 tracker, making it the more representative PR of this cluster.", "best_issue_reason": null, - "best_pr_reason": "PR #43613 is the best single representative because it is a substantial end-to-end pipeline addition with broad integration coverage, making it a stronger canonical example than the narrower concept-segmentation PR.", + "best_pr_reason": "PR 43613 is the best representative because it has the larger, more complete implementation and test footprint. PR 43612 is narrower and focuses on a different segmentation task, so it is not the better cluster anchor.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different underlying changes: concept segmentation vs visual segmentation. Shared docs and pipeline boilerplate are not enough to treat them as the same fix, and they would not plausibly be merged into one PR as a duplicate." + "reason": "Different underlying features/tasks: concept segmentation vs visual segmentation. Shared docs/registry plumbing is not enough to treat them as the same change." } ] }, @@ -19919,7 +20107,7 @@ { "label": "pull_request", "component_index": 11, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19935,7 +20123,7 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "79583826a105480063a2a8b9d476e5d146da14140f5c06771c30ac99a9e6c4dd", + "prepared_review_unit_hash": "e080d2feb53b2923eb7ee59158fb69bcfb7bddd5571b047f0304244226ff6655", "component_budget": { "node_count": 2, "item_count": 2, @@ -19957,18 +20145,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs both edit `trainer.py`, but they address different, unrelated trainer behaviors: gradient-accumulation loss scaling vs. signature-column filtering for custom/PEFT trainers. They are not duplicates and should not be merged into one change.", + "summary": "Two open PRs in `trainer.py` that both add small extension points to `Trainer`, but they target different code paths and solve different problems: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering for PEFT/custom trainers. Not a duplicate pair.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: the changes are on separate code paths and solve different problems, so neither one represents the other.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "No single best PR for the cluster; the PRs are not interchangeable and do not form one coherent fix/change set.", + "best_pr_reason": "PR 43663 is the better representative of the cluster because it introduces a reusable Trainer hook with broader API impact; PR 43651 is a narrower fix for loss scaling behavior.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Different underlying trainer behaviors: one changes loss scaling during gradient accumulation, the other refactors signature-column extraction for filtering. Same file alone is not enough to treat them as duplicates." + "reason": "Different underlying changes: gradient-accumulation loss normalization vs. input-signature column filtering. They touch the same file, but not the same bug or mergeable fix path." } ] }, @@ -19982,7 +20170,7 @@ { "label": "pull_request", "component_index": 12, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19998,7 +20186,7 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "8cecf1a7c0ddc2e942eaea01dc3c2306d3c52d88b109a90dd1f06c959c776889", + "prepared_review_unit_hash": "b19de155753931771c328cd21432807f129908f505520351def2d9fad5a82283", "component_budget": { "node_count": 2, "item_count": 2, @@ -20020,18 +20208,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs touch the same CLI file and both mention TypeAdapter, but they appear to fix different NameError causes in separate code paths, so they should not be merged as duplicates.", - "confidence": 0.91, + "summary": "Two open PRs touch the same CLI server file but fix different NameError/typing problems: one addresses optional pydantic/TypeAdapter import handling, the other addresses annotation evaluation issues in serve.py more broadly. They should not be treated as duplicates.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43842 is the more direct and concrete fix: it targets the runtime failure when pydantic is unavailable by moving the TypeAdapter import behind TYPE_CHECKING and quoting the annotation.", + "canonical_pr_reason": "PR 43842 is the more specific and concrete fix: it directly targets the TypeAdapter NameError when pydantic is absent, with a narrow import/annotation guard in the affected code path.", "best_issue_reason": null, - "best_pr_reason": "PR 43842 is the best representative because it has the clearest bug statement, explicit issue target, and a narrowly scoped change aligned to a single failure mode. PR 43656 is broader annotation cleanup and appears to address a different NameError source.", + "best_pr_reason": "PR 43842 is the best representative of the cluster because its title, issue target, and diff all align on one well-defined bug and a minimal fix. PR 43656 looks like a separate typing cleanup/runtime-annotation issue rather than the same underlying problem.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": false, - "reason": "Reject: both are in the same file and mention TypeAdapter, but 43656 fixes annotation-related NameErrors/typing evaluation in CLI serve.py, while 43842 fixes the pydantic-optional import/TypeAdapter runtime NameError. They are different bugs and would not plausibly be one merged PR." + "reason": "Different underlying bugs: 43656 changes type annotations/future annotations in serve.py, while 43842 specifically guards TypeAdapter for optional pydantic. Same file, but not the same concrete fix or mergeable into one PR." } ] }, @@ -20045,7 +20233,7 @@ { "label": "pull_request", "component_index": 13, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20111,7 +20299,7 @@ { "label": "pull_request", "component_index": 14, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 2, "status": "reviewed", @@ -20224,286 +20412,226 @@ "pull_request:44044|pull_request:44154", "pull_request:44024|pull_request:44076" ], - "prepared_review_unit_hash": "050266a04335d8af0aeec05f2696c08e06ee43710e320ade7260de47e9254a66", + "prepared_review_unit_hash": "77ec5ad138f50c1204a3243aace597dc003b3a33a62bd2afb960c9699d166bcb", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84508, - "estimated_input_tokens": 21127, - "estimated_eval_tokens": 42510 + "serialized_chars": 84496, + "estimated_input_tokens": 21124, + "estimated_eval_tokens": 42504 }, "budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 72, - "serialized_chars": 76940, - "estimated_input_tokens": 19235, - "estimated_eval_tokens": 38726 + "serialized_chars": 76930, + "estimated_input_tokens": 19233, + "estimated_eval_tokens": 38722 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a broad wave of model-specific refactors toward standardized output tracing, not a set of exact duplicates across the whole cluster. The clearest duplicate subcluster is GPT-J/CodeGen, with a possible overlap on RegNet between a RegNet-only PR and a broader multi-model PR.", - "confidence": 0.82, - "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44722 is the strongest representative of the clearest duplicate subcluster: it covers the same GPT-J output-tracing refactor as the other GPT-J PRs and also includes CodeGen, making it the broadest overlapping implementation.", - "best_issue_reason": null, - "best_pr_reason": "PR 44722 is the best overall PR candidate because it is the most comprehensive version of the shared GPT-J/CodeGen tracing refactor and overlaps the strongest duplicate evidence in the set.", + "summary": "This cluster is mostly a shared #43979 output-tracing refactor set, but it is not limited to GPT-Neo/GPT-J. The packet contains several true mergeable subclusters, including the GPT-J/CodeGen line centered on #44722 and the FocalNet/ImageGPT pair (44024/44076), plus a few other issue-linked output-tracing pairs; many remaining edges are only theme-similar and should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "#43979 is the common umbrella tracker repeatedly linked by these refactor PRs and is the clearest issue-level anchor for the cluster.", + "canonical_pr_reason": "PR #44722 is the strongest canonical PR candidate because it is issue-linked and directly overlaps with both GPT-J/CodeGen refactor PRs on the same concrete code path.", + "best_issue_reason": "No separate issue artifact is needed here; #43979 is the best global issue anchor for the whole refactor cluster.", + "best_pr_reason": "PR #44722 is the best overall representative because it sits at the center of the GPT-J/CodeGen subcluster and has the cleanest direct duplicate links.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model/code paths (CVT/FNet vs GPT-J); they only share the generic output-tracing migration theme and tracking issue." + "reason": "Different model families; the shared issue is only a broad refactor umbrella." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Different model implementations (CVT/FNet vs DeBERTa) with only a shared standardized-tracing theme." + "reason": "Different backbones and different concrete forward paths; same tracking issue is not enough." }, { "left": "pull_request:44066", "right": "pull_request:44085", - "accept": true, - "reason": "Both touch the GPT-J output-tracing refactor in the same file; one is a broader/superset variant of the same change." + "accept": false, + "reason": "Same broader GPT-J area, but 44066 also rewrites CodeGen and the PRs are not the same concrete change." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "Both are output-tracing refactors, but they affect different model code paths (RegNet/ResNet/RT-DETR vs EfficientNet)." + "reason": "Different vision model code paths; they share only the standardized output-tracing pattern." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "Different models and implementations; shared issue target is not enough to make them the same change." + "reason": "Unrelated model families and code paths; only the refactor theme overlaps." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen work versus EfficientNet work; same theme, different concrete code paths." + "reason": "GPT-J/CodeGen vs EfficientNet are different concrete changes." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTa are unrelated implementations; only the refactor theme matches." + "reason": "Different model implementations; both are output-tracing refactors, but not the same bug/change." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen output-tracing refactor versus MGP-STR output-tracing refactor; different model code paths." + "reason": "GPT-J/CodeGen and MGP-STR are unrelated code paths." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen and MPT are different concrete implementations, so this is only a thematic match." + "reason": "Different model stacks; same decorator migration is too broad to merge as one PR." }, { "left": "pull_request:44018", "right": "pull_request:44068", - "accept": true, - "reason": "Both are GPT-Neo output-tracing refactors touching the same model/file and look like overlapping versions of the same change." + "accept": false, + "reason": "Same GPT-Neo family, but the diffs are not the same concrete rewrite and do not look mergeable as one PR." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen versus GPT-Neo; similar refactor pattern, but not the same model or code path." + "reason": "GPT-J/CodeGen and GPT-Neo are distinct model code paths." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are different model implementations, so this is not the same underlying change." + "reason": "GPT-Neo and GPT-J/CodeGen are different concrete changes." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet output tracing is a different code path from the GPT-J/CodeGen refactor." + "reason": "ResNet and GPT-J/CodeGen are unrelated implementations." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPT-J/CodeGen are unrelated model implementations; shared theme only." + "reason": "MGP-STR and GPT-J/CodeGen do not share the same code path or bug." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPT-J are different model code paths despite both being output-tracing refactors." + "reason": "ResNet and GPT-J are unrelated model implementations." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo and ResNet are different concrete implementations; this is not the same bug/change." + "reason": "GPT-Neo and ResNet are different models with different output-handling code." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet versus MPT; same refactor family, different model code paths." + "reason": "ResNet and MPT are separate implementations with no shared code path." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo and Flaubert/XLM are unrelated implementations; only the standardized-tracing theme overlaps." + "reason": "GPT-Neo and Flaubert are unrelated model implementations." }, { - "left": "pull_request:44068", + "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo output tracing is distinct from the GPT-J/CodeGen refactor." + "reason": "MPT and GPT-J/CodeGen are different model code paths." }, { - "left": "pull_request:44013", - "right": "pull_request:44044", + "left": "pull_request:44018", + "right": "pull_request:44085", "accept": false, - "reason": "MobileNetV2 and DeBERTa are unrelated code paths; not the same underlying change." + "reason": "GPT-Neo and GPT-J are different model files and forward paths." }, { - "left": "pull_request:44066", + "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen and MGP-STR are different model implementations, so they should not be merged as duplicates." + "reason": "GPT-Neo and MGP-STR are unrelated concrete changes." }, { - "left": "pull_request:44071", - "right": "pull_request:44722", + "left": "pull_request:44068", + "right": "pull_request:44086", "accept": false, - "reason": "MPT and GPT-J/CodeGen are different code paths; same tracing style does not make them duplicates." + "reason": "Different model code paths; same output-tracing style is not enough for duplication." }, { - "left": "pull_request:44018", - "right": "pull_request:44085", + "left": "pull_request:44071", + "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and GPT-J are different models, even though both are output-tracing refactors." + "reason": "MPT and MGP-STR are separate implementations with separate output handling." }, { - "left": "pull_request:44018", + "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated model implementations." + "reason": "GPT2 and MGP-STR do not share the same code path or bug." }, { - "left": "pull_request:44068", + "left": "pull_request:44085", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are different concrete changes." + "reason": "GPT-J and MGP-STR are unrelated; only the output-tracing initiative overlaps." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT and Flaubert/XLM are different model families; only the refactor style matches." + "reason": "ViLT and Flaubert are different model code paths." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 output tracing is a different code path from GPT-J/CodeGen." + "reason": "GPT2 and GPT-J/CodeGen are distinct model implementations and cannot be merged as one PR." }, { "left": "pull_request:44018", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo versus MPT; same broad theme, not the same change." + "reason": "GPT-Neo and MPT are different concrete changes despite similar decorator migration." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and LongT5 are unrelated model implementations." + "reason": "SegFormer and LongT5 are unrelated model code paths." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer output tracing is not the same concrete change as GPT-J/CodeGen." + "reason": "SegFormer and GPT-J/CodeGen are not the same underlying bug/change." }, { "left": "pull_request:44017", "right": "pull_request:44073", "accept": false, - "reason": "SegFormer and VisualBert are different code paths; shared output-tracing pattern is too broad." - }, - { - "left": "pull_request:44068", - "right": "pull_request:44071", - "accept": false, - "reason": "GPT-Neo and MPT are different model-specific refactors." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44073", - "accept": false, - "reason": "RegNet and VisualBert are different models; only the migration theme overlaps." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44086", - "accept": false, - "reason": "MPT and MGP-STR are unrelated concrete implementations." - }, - { - "left": "pull_request:44073", - "right": "pull_request:44101", - "accept": false, - "reason": "VisualBert and Flaubert/XLM are different code paths." - }, - { - "left": "pull_request:44017", - "right": "pull_request:44071", - "accept": false, - "reason": "SegFormer and MPT are unrelated model implementations." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44071", - "accept": false, - "reason": "GPT-2 and MPT are different underlying changes." - }, - { - "left": "pull_request:44056", - "right": "pull_request:44101", - "accept": false, - "reason": "MPNet and Flaubert/XLM are different model-specific refactors." - }, - { - "left": "pull_request:44018", - "right": "pull_request:44059", - "accept": false, - "reason": "GPT-Neo and GPT-2 are different code paths." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44068", - "accept": false, - "reason": "GPT-2 and GPT-Neo are distinct model implementations." - }, - { - "left": "pull_request:44101", - "right": "pull_request:44161", - "accept": false, - "reason": "Flaubert/XLM and LongT5 are unrelated changes." + "reason": "SegFormer and VisualBert are separate implementations; same refactor pattern only." }, { "left": "pull_request:43997", "right": "pull_request:44017", "accept": false, - "reason": "RegNet and SegFormer are different model code paths." + "reason": "RegNet and SegFormer are different backbone code paths." }, { "left": "pull_request:43997", @@ -20513,51 +20641,27 @@ }, { "left": "pull_request:44071", - "right": "pull_request:44076", - "accept": false, - "reason": "MPT and ImageGPT are different concrete changes." - }, - { - "left": "pull_request:44074", - "right": "pull_request:44101", - "accept": false, - "reason": "TextNet and Flaubert/XLM are different model-specific refactors." - }, - { - "left": "pull_request:44056", - "right": "pull_request:44161", - "accept": false, - "reason": "MPNet and LongT5 are unrelated code paths." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44018", - "accept": false, - "reason": "RegNet and GPT-Neo are different model implementations." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44161", + "right": "pull_request:44129", "accept": false, - "reason": "RegNet and LongT5 are unrelated changes." + "reason": "MPT and SpeechT5 are different model families with different output-handling code." }, { "left": "pull_request:44074", "right": "pull_request:44161", "accept": false, - "reason": "TextNet and LongT5 are different code paths." + "reason": "TextNet and LongT5 are unrelated model implementations." }, { "left": "pull_request:44018", "right": "pull_request:44161", "accept": false, - "reason": "GPT-Neo and LongT5 are unrelated model-specific refactors." + "reason": "GPT-Neo and LongT5 are different concrete changes." }, { "left": "pull_request:44010", "right": "pull_request:44071", "accept": false, - "reason": "SqueezeBert and MPT are different underlying changes." + "reason": "SqueezeBert and MPT are separate model implementations with no shared code path." }, { "left": "pull_request:44010", @@ -20569,25 +20673,19 @@ "left": "pull_request:44018", "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo and SpeechT5 are unrelated implementations." + "reason": "GPT-Neo and SpeechT5 are unrelated models." }, { "left": "pull_request:43997", "right": "pull_request:44074", "accept": false, - "reason": "RegNet and TextNet are different model-specific changes." + "reason": "RegNet and TextNet are different implementations and not a duplicate fix." }, { "left": "pull_request:43997", "right": "pull_request:44129", "accept": false, - "reason": "RegNet and SpeechT5 are unrelated changes." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44129", - "accept": false, - "reason": "MPT and SpeechT5 are different concrete code paths." + "reason": "RegNet and SpeechT5 are unrelated concrete changes." }, { "left": "pull_request:44010", @@ -20599,96 +20697,90 @@ "left": "pull_request:43996", "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and LongT5 are unrelated models; only the output-tracing theme is shared." + "reason": "FNet/CVT and LongT5 are unrelated code paths despite the shared output-tracing initiative." }, { "left": "pull_request:43996", "right": "pull_request:44017", "accept": false, - "reason": "CVT/FNet and SegFormer are different code paths." + "reason": "FNet/CVT and SegFormer are different model implementations." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBert and DeBERTa are unrelated model-specific refactors." + "reason": "SqueezeBert and DeBERTa V2 are distinct code paths." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa and TextNet are different concrete changes." + "reason": "DeBERTa V2 and TextNet are unrelated model implementations." }, { "left": "pull_request:44066", "right": "pull_request:44722", "accept": true, - "reason": "Both include the same GPT-J output-tracing rewrite; 44722 is the broader superset that also adds CodeGen." + "reason": "Both target the same GPT-J/CodeGen output-tracing rewrite, with overlapping files and the same concrete change." }, { "left": "pull_request:44085", "right": "pull_request:44722", "accept": true, - "reason": "Both overlap on the GPT-J output-tracing refactor; 44722 extends the same change set with CodeGen." + "reason": "Both modify the GPT-J output-tracing path in modeling_gptj.py, so they look like the same underlying change." }, { "left": "pull_request:44085", "right": "pull_request:44129", - "accept": false, - "reason": "GPT-J and SpeechT5 are different model code paths." + "accept": true, + "reason": "Both are issue-linked standardized output-tracing refactors in the same umbrella initiative, and the packet treats them as mergeable." }, { "left": "pull_request:43996", "right": "pull_request:44007", - "accept": false, - "reason": "CVT/FNet and RegNet/ResNet/RT-DETR are different model-specific refactors." + "accept": true, + "reason": "Both are standardized output-tracing refactors under the same issue, with the same concrete refactor pattern across the model code." }, { "left": "pull_request:43996", "right": "pull_request:44129", - "accept": false, - "reason": "CVT/FNet and SpeechT5 are unrelated model implementations." + "accept": true, + "reason": "Both are issue-linked output-tracing refactors that apply the same standardized capture pattern." }, { "left": "pull_request:44044", "right": "pull_request:44129", - "accept": false, - "reason": "DeBERTa and SpeechT5 are different concrete changes." + "accept": true, + "reason": "Both are issue-linked output-tracing refactors in the same initiative and are close enough to be merged as one change." }, { "left": "pull_request:44044", "right": "pull_request:44154", - "accept": false, - "reason": "DeBERTa and VITS are different model code paths." + "accept": true, + "reason": "Both implement the same standardized output-collection refactor pattern on compatible model code paths." }, { "left": "pull_request:44024", "right": "pull_request:44076", - "accept": false, - "reason": "FocalNet and ImageGPT are unrelated model implementations." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44007", "accept": true, - "reason": "Both touch RegNet's output-tracing rewrite in the same file; 44007 is a broader PR that also includes other models." + "reason": "Both are the same standardized output-tracing refactor applied to image-model encoder code, so they are plausibly mergeable into one PR." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative. The accepted soft edges are all mergeable same-bugfix overlaps (GPT-J/CodeGen, GPT-Neo, RegNet), while the rejections mostly reflect distinct model-specific refactors despite a shared tracing theme. No overstatement of evidence stands out." + "accept": false, + "feedback": "The cluster summary is not fully grounded in the packet: it says only the GPT-Neo/GPT-J pairs look like true duplicates, but the packet also marks pull_request:44024 and pull_request:44076 as a true mergeable pair. That makes the summary too strong. The accepted soft-edge verdicts that are present are mostly fine, but the narrative overstates how exclusive the GPT-Neo/GPT-J duplication is." }, "evaluator_used": true, - "retried": false, - "accepted_nontrivial_soft_edge": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, { "label": "pull_request", "component_index": 14, - "component_count": 36, + "component_count": 35, "review_unit_index": 2, "review_unit_count": 2, "status": "reviewed", @@ -20712,20 +20804,20 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "fba696a9b9516ae93e3dc15b0f78fb2d3fe402648eba11e8bb78608de5ada4ca", + "prepared_review_unit_hash": "8aa08d8eee1c022c7e562169f98083749ab13fd0401ecbff32566135143804f3", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84508, - "estimated_input_tokens": 21127, - "estimated_eval_tokens": 42510 + "serialized_chars": 84496, + "estimated_input_tokens": 21124, + "estimated_eval_tokens": 42504 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 5, - "serialized_chars": 13068, + "serialized_chars": 13066, "estimated_input_tokens": 3267, "estimated_eval_tokens": 6790 }, @@ -20734,42 +20826,42 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These PRs are all similar model-local refactors to standardized output tracing/capture_outputs, but they target different architectures and code paths, so they should not be deduplicated.", + "summary": "This cluster is a collection of similar-looking but model-specific PRs that all refactor output tracing/capture in different architectures. They are not the same underlying change, so the soft duplicate links should be rejected.", "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44129 is the most canonical of the set because it has an explicit issue target and a focused, concrete SpeechT5 output-tracing refactor.", + "canonical_pr_reason": "PR 44129 is the most concrete anchor because it has an explicit issue target and a substantial SpeechT5-specific output-tracing refactor, but it is still not a duplicate of the others.", "best_issue_reason": null, - "best_pr_reason": "PR 44129 is the best anchor for the cluster: it is issue-linked and clearly scoped, while the others are parallel model-specific migrations rather than the same change.", + "best_pr_reason": "PR 44129 is the best representative of the cluster because it is the most fully specified and ties to an explicit issue, while the others are smaller per-model refactors.", "soft_edge_verdicts": [ { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Both are standardized output-tracing refactors, but they touch different models and separate code paths (VisualBert vs TextNet), so they are not the same mergeable change." + "reason": "Both are output-tracing refactors, but they touch different model code paths (VisualBert vs TextNet) and are independent per-model migrations, not one mergeable fix." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "MPNet and VisualBert are distinct implementations with different layers and tracing hooks; this is a similar migration pattern, not a duplicate fix." + "reason": "MPNet and VisualBert are separate model implementations; these are parallel refactors of the same framework pattern, not the same underlying bug or change." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "These are analogous refactors in different model files and subsystems, with no shared concrete bug or patchable code path." + "reason": "Different models and different code paths; similarity is only in adopting the same tracing utility, which is too broad to count as a duplicate." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model architectures; the PRs share a theme but not the same underlying change." + "reason": "RegNet and SqueezeBert are unrelated model implementations, and the PRs refactor output handling in different forward paths rather than fixing one shared defect." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and SpeechT5 are different model code paths; both refactor output handling, but they do not fix the same concrete issue." + "reason": "These affect different architectures and different output-tracing code; they are not the same concrete change and would not plausibly be merged into one PR." } ] }, @@ -20783,7 +20875,7 @@ { "label": "pull_request", "component_index": 15, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20791,7 +20883,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-43998-14", + "cluster_id": "cluster-43998-13", "nodes": [ "pull_request:43998", "pull_request:43999", @@ -20805,8 +20897,7 @@ "pull_request:44027", "pull_request:44028", "pull_request:44029", - "pull_request:44030", - "pull_request:46438" + "pull_request:44030" ], "soft_pairs": [ "pull_request:43998|pull_request:44002", @@ -20821,16 +20912,13 @@ "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:44000|pull_request:44026", "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", "pull_request:44001|pull_request:44025", - "pull_request:44001|pull_request:44026", "pull_request:44002|pull_request:44029", - "pull_request:44002|pull_request:44025", - "pull_request:44001|pull_request:44003", "pull_request:44002|pull_request:44003", - "pull_request:44000|pull_request:44003", "pull_request:44003|pull_request:44028", "pull_request:44003|pull_request:44004", "pull_request:43998|pull_request:44004", @@ -20851,329 +20939,276 @@ "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:46438", - "pull_request:44000|pull_request:46438", - "pull_request:44029|pull_request:46438", - "pull_request:44028|pull_request:46438", - "pull_request:43998|pull_request:46438" + "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "c67c84246cfc16e8398b867c30de0022e82b2c685b6dc3374f4749147293306c", + "prepared_review_unit_hash": "a1e406186f2e2e2f0ba25661a1689e74e7b6b34c1f2768d08c618f5ac3de4927", "component_budget": { - "node_count": 14, - "item_count": 14, - "soft_pair_count": 48, - "serialized_chars": 45118, - "estimated_input_tokens": 11280, - "estimated_eval_tokens": 22816 + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40174, + "estimated_input_tokens": 10044, + "estimated_eval_tokens": 20344 }, "budget": { - "node_count": 14, - "item_count": 14, - "soft_pair_count": 48, - "serialized_chars": 45118, - "estimated_input_tokens": 11280, - "estimated_eval_tokens": 22816 + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40174, + "estimated_input_tokens": 10044, + "estimated_eval_tokens": 20344 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This is a loose topical cluster of independent PRs that all refactor output-tracing behavior in different model implementations, plus one unrelated utility speedup PR. They share a pattern, but not the same concrete bug or change, so they should not be merged as duplicates.", - "confidence": 0.93, + "summary": "These PRs share a generic \"refactor output tracing\" theme, but they touch different model implementations and different forward/output code paths, so they are not duplicates of one another.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44003 is the broadest and most infrastructure-heavy example here: it spans both mamba and falcon_mamba and adds capture_outputs/OutputRecorder plumbing, making it the best anchor for the shared output-tracing theme.", + "canonical_pr_reason": "No single canonical PR: each item is a separate model-specific refactor, not the same underlying code change.", "best_issue_reason": null, - "best_pr_reason": "44003 is the most representative single PR for this cluster because it goes beyond a one-file signature tweak and actually introduces the core output-capture machinery.", + "best_pr_reason": "No PR is globally best for duplicate triage here; the cluster is a set of parallel refactors across different architectures, not one mergeable change.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Same refactor pattern, but different model implementations and call paths." + "reason": "Different model families and forward paths; same refactor theme only." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Different models; only the output-tracing theme overlaps." + "reason": "Different architectures and files; not the same underlying change." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Different model code paths; not the same underlying change." + "reason": "Different model code paths; only the wording of the refactor is similar." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Separate model families and distinct refactor targets." + "reason": "Different model families (timm_backbone vs mamba/falcon_mamba); separate fixes." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "Different architectures; same broad theme is not enough." + "reason": "Different model implementations and output plumbing; not mergeable as one PR." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "Separate model files and distinct output behavior." + "reason": "Mobilenet_v1 and UnivNet are unrelated model paths; same refactor pattern only." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different concrete model path, even if the refactor style matches." + "reason": "Different model backbones and forward signatures; not the same change." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "Different models and different wrapper logic." + "reason": "MobileNetV1 and Depth Anything are separate model changes, not duplicates." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "Different code paths; both are refactors, not duplicates." + "reason": "Different architectures and files; only a shared output-tracing refactor theme." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Different model families and separate implementation changes." + "reason": "Different model code paths; not the same bug or feature change." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "Same stylistic refactor, but unrelated model implementations." + "reason": "Vision-text dual encoder and UnivNet are unrelated implementations." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Different model wrappers; not one shared bug fix." + "reason": "Different models and output paths; same naming pattern only." }, { "left": "pull_request:44000", - "right": "pull_request:44025", + "right": "pull_request:44003", "accept": false, - "reason": "Both are output-tracing refactors, but for different models." + "reason": "Different model families and code paths; not one underlying change." }, { "left": "pull_request:44000", - "right": "pull_request:44026", + "right": "pull_request:44025", "accept": false, - "reason": "Distinct encoder/decoder wrapper code paths." + "reason": "Vision-text dual encoder and Depth Anything are separate model refactors." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "Different model architectures and forward signatures." + "reason": "UnivNet and UperNet are unrelated model implementations." }, { "left": "pull_request:44001", - "right": "pull_request:44025", + "right": "pull_request:44003", "accept": false, - "reason": "Same theme, different model-specific change." + "reason": "Different architectures and forward methods; not the same fix." }, { "left": "pull_request:44001", - "right": "pull_request:44026", + "right": "pull_request:44025", "accept": false, - "reason": "Not the same underlying code path or bug." + "reason": "Different model families; no shared concrete code-path problem." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "Different model families; no shared concrete fix." - }, - { - "left": "pull_request:44002", - "right": "pull_request:44025", - "accept": false, - "reason": "Separate models with independent output-tracing updates." - }, - { - "left": "pull_request:44001", - "right": "pull_request:44003", - "accept": false, - "reason": "Different model families and separate tracing refactors." + "reason": "UperNet and RWKV are unrelated code paths; only the refactor theme matches." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "Distinct code paths; only the refactor theme is shared." - }, - { - "left": "pull_request:44000", - "right": "pull_request:44003", - "accept": false, - "reason": "Different architectures and different forward implementations." + "reason": "Different model families and distinct output handling." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "Different model families; not the same concrete change." + "reason": "Mamba/FalconMamba refactor vs SuperPoint refactor; separate changes." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "Different models and different output behavior." + "reason": "Different model code paths (mamba vs codegen); not the same underlying issue." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "Unrelated model code paths; same refactor motif only." + "reason": "Timm backbone and CodeGen are unrelated implementations." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "Different model implementations and distinct forward contracts." + "reason": "UnivNet and CodeGen fix different forward/output paths." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Separate model-specific refactors, not one merged fix." + "reason": "UperNet and CodeGen are separate model refactors." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "Different model families and separate code paths." + "reason": "Vision-text dual encoder and CodeGen are different architectures." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Parallel wrapper refactors, but different files and model families." + "reason": "Depth Anything and VisionEncoderDecoder are different model families." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Same broad pattern, but not the same concrete bug/change." + "reason": "Depth Anything and SpeechEncoderDecoder are unrelated code paths." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "Different models; no shared code path." + "reason": "Depth Anything and SuperPoint are separate model-specific refactors." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "Distinct model implementations with separate tracing updates." + "reason": "Depth Anything and RWKV are unrelated implementations." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Different models and unrelated forward paths." + "reason": "Depth Anything and DPR touch different model code paths." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "Separate models; not the same underlying fix." + "reason": "VisionEncoderDecoder and RWKV are unrelated model changes." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "Closely related wrapper style, but different concrete code paths." + "reason": "VisionEncoderDecoder and SpeechEncoderDecoder are different code paths." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "Different model implementations and output contracts." + "reason": "Different model families; not the same concrete bug/change." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "Unrelated model-specific refactors." + "reason": "VisionEncoderDecoder and DPR are separate refactors." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "Different model families and distinct files." + "reason": "SpeechEncoderDecoder and SuperPoint are unrelated model implementations." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "Separate model wrappers; same theme is insufficient." + "reason": "Different model families and output paths." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "Different concrete code paths and output behavior." + "reason": "SpeechEncoderDecoder and DPR are separate changes." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "Different model implementations; not a duplicate change." + "reason": "SuperPoint and RWKV are different code paths; same refactor style only." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "Distinct model-specific refactors." + "reason": "SuperPoint and DPR are unrelated model changes." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "RWKV and DPR are unrelated models with separate changes." - }, - { - "left": "pull_request:44027", - "right": "pull_request:46438", - "accept": false, - "reason": "Speech encoder-decoder refactor vs unrelated modular-conversion utility speedup." - }, - { - "left": "pull_request:44000", - "right": "pull_request:46438", - "accept": false, - "reason": "Completely different subsystem and purpose." - }, - { - "left": "pull_request:44029", - "right": "pull_request:46438", - "accept": false, - "reason": "Unrelated utility script change; no shared bug or code path." - }, - { - "left": "pull_request:44028", - "right": "pull_request:46438", - "accept": false, - "reason": "Different model refactor versus a build/tooling optimization." - }, - { - "left": "pull_request:43998", - "right": "pull_request:46438", - "accept": false, - "reason": "No relation beyond both being PRs in the same repo." + "reason": "RWKV and DPR are different model families; not mergeable as one PR." } ] }, @@ -21187,7 +21222,7 @@ { "label": "pull_request", "component_index": 16, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21203,12 +21238,12 @@ "soft_pairs": [ "pull_request:44662|pull_request:45401" ], - "prepared_review_unit_hash": "56f9662899fb2fe5699c1ddb952a37741da48c59d1f329af8351a3b87c81a51d", + "prepared_review_unit_hash": "fd7a8f449d550442a2145d659c12fba948a24af60377cc13b37f293835146b67", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6552, + "serialized_chars": 6550, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -21216,7 +21251,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6552, + "serialized_chars": 6550, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -21225,18 +21260,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated model-addition PRs that only overlap on shared Transformers integration files and docs. They do not look like the same underlying change, so they should not be merged as duplicates.", - "confidence": 0.96, + "summary": "These are unrelated model-addition PRs: one adds PenguinVL, the other adds Voxtral TTS. They only overlap on generic transformer integration files and docs infrastructure, not on the same underlying code path or fix.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45401 is the more self-contained candidate to represent the cluster because it has an explicit issue target and a narrowly scoped Voxtral TTS implementation; PR 44662 is a separate PenguinVL model addition.", + "canonical_pr_reason": "PR 44662 is the more complete and mature standalone addition: it is not draft, has much broader implementation coverage, and more review activity. PR 45401 is a separate draft PR scoped to a different model family and explicit issue.", "best_issue_reason": null, - "best_pr_reason": "PR 45401 is the best representative among the provided PRs for the same reasons: explicit issue linkage and a focused model-specific change. PR 44662 is unrelated rather than another version of the same fix.", + "best_pr_reason": "PR 44662 is the stronger representative PR in this cluster because it is a fuller implementation with more discussion; PR 45401 is a distinct Voxtral TTS support change and should not be merged with it.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different models and different feature goals: PenguinVL implementation vs Voxtral TTS support. Shared registry/docs files are incidental to adding any new model, not evidence of the same underlying change." + "reason": "Different models and different functionality (vision-language vs text-to-speech). Shared auto-registration/docs changes are generic and do not indicate the same underlying bug or change." } ] }, @@ -21250,7 +21285,7 @@ { "label": "pull_request", "component_index": 17, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21270,12 +21305,12 @@ "pull_request:45363|pull_request:46339", "pull_request:45666|pull_request:46339" ], - "prepared_review_unit_hash": "b669a0c1f31c4ece4f0b103fa922eed7a304dff26a959348cb232203409119a0", + "prepared_review_unit_hash": "479dd366bd16587335484f727771d6c7a04d898225b71f8c26ec3102aa9f32e9", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 3, - "serialized_chars": 9634, + "serialized_chars": 9633, "estimated_input_tokens": 2409, "estimated_eval_tokens": 5074 }, @@ -21283,7 +21318,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 3, - "serialized_chars": 9634, + "serialized_chars": 9633, "estimated_input_tokens": 2409, "estimated_eval_tokens": 5074 }, @@ -21292,36 +21327,36 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs all revolve around the same n-to-1 kernel fusion / KernelConfig feature, with later PRs extending and simplifying earlier ones. PR 46339 is the broadest and most complete follow-up, so it is the best canonical representative.", - "confidence": 0.84, + "summary": "This cluster is centered on the KernelConfig / n-to-1 kernel fusion feature line. PR 46339 looks like the strongest representative because it is the latest and most comprehensive iteration of that implementation, while PR 44979 is a separate foundational API PR and not a duplicate of the kernel-fusion PRs.", + "confidence": 0.79, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46339 covers the core KernelConfig fusion implementation plus follow-up simplifications in core loading/hub kernel code and tests, making it the most complete representative of the feature cluster.", + "canonical_pr_reason": "PR 46339 is the best canonical PR because it is the most complete and mature iteration of the KernelConfig-based n-to-1 fusion work, with broader implementation and tests than the earlier drafts.", "best_issue_reason": null, - "best_pr_reason": "PR 46339 is the strongest cluster anchor: it spans the shared fusion-loading codepaths, has the broadest file coverage, and appears to subsume the earlier KernelConfig fusion work.", + "best_pr_reason": "PR 46339 best represents the cluster\u2019s concrete feature set: it continues the same KernelConfig fusion path as the earlier PRs and adds the broadest, most refined implementation changes.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Related feature area, but 44979 is the initial Module Fusion API while 45363 is a separate KernelConfig-based fusion integration; they are not the same concrete change set." + "reason": "44979 is a new Module Fusion API; 45363 is a KernelConfig-based n-to-1 fusion PR. They are related, but not the same concrete change and would not be merged as one duplicate PR." }, { "left": "pull_request:45363", "right": "pull_request:46339", "accept": true, - "reason": "Both target the same n-to-1 kernel fusion via KernelConfig codepath, and 46339 reads as a direct extension/simplification of the earlier implementation." + "reason": "Both target the same KernelConfig-based n-to-1 kernel fusion code path, touch overlapping kernel-loading files, and 46339 reads as a later refactor/extension of the same feature rather than a separate change." }, { "left": "pull_request:45666", "right": "pull_request:46339", "accept": true, - "reason": "These are the same KernelConfig n-to-1 fusion feature family; 46339 appears to refine and simplify the same implementation rather than introduce a different change." + "reason": "These are clearly the same feature family: both implement extended n-to-1 fusion via KernelConfig with substantial overlap in the same loading/integration files, so they could plausibly be unified into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster summary is grounded in the packet: 46339 has the broadest overlapping file coverage and its title/diff preview support it as a later extension/simplification of the KernelConfig fusion work. The pair verdicts are conservative overall: 44979 and 45363 look like related but distinct change sets, while 45363\u219446339 and 45666\u219446339 are plausibly mergeable within the same n-to-1 kernel fusion feature family based on shared files and titles. No clear overstatement of evidence." + "feedback": "The summary is grounded in the packet and stays mostly conservative. The rejection of 44979 vs 45363 is well supported because they are different concrete changes (Module Fusion API vs KernelConfig-based n-to-1 fusion). The accepted soft pairs are also plausible: 45363 and 46339, and 45666 and 46339, share the same KernelConfig/n-to-1 fusion line and overlapping integration files, so treating them as mergeable iterations of one feature is reasonable. The canonical PR choice of 46339 is also supported by its later timestamp and broader shared implementation surface, though the wording could stay a bit more tentative." }, "evaluator_used": true, "retried": false, @@ -21332,7 +21367,7 @@ { "label": "pull_request", "component_index": 18, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21348,7 +21383,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "9240f482a90be394ff4a5e6d934bd24888fb3542df1f2bc1a44e39d84e6c0e58", + "prepared_review_unit_hash": "eb56340009b2f194d26ffa5049b6e5a90dafbf30c504ab64c815c1b118a39b30", "component_budget": { "node_count": 2, "item_count": 2, @@ -21370,18 +21405,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs fix the same Mistral regex patching AttributeError in the same code path; 45317 is the more complete version because it adds tests and clearer framing, while 45086 is a smaller earlier variant.", - "confidence": 0.97, + "summary": "Both pull requests address 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 adds tests and more explicit coverage.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45317 is the stronger canonical representative: same fix as the other PR, same issue target, and it includes test coverage and clearer problem description.", + "canonical_pr_reason": "`pull_request:45317` is the best canonical PR because it fixes the same concrete bug as `45086` while also adding test coverage and clearer validation of `fix_mistral_regex=True`.", "best_issue_reason": null, - "best_pr_reason": "PR 45317 is best suited as the cluster representative because it addresses the concrete bug with the most complete implementation and validation, while still matching the shared fix precisely.", + "best_pr_reason": "`pull_request:45317` is the strongest PR overall: same code-path fix, broader validation, and extra tests make it the most complete and merge-ready representation of the change.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Accept: both PRs patch the same AttributeError in _patch_mistral_regex by changing the same tokenizer pre-tokenizer access in the same file, with the same issue target; 45317 just adds tests and more context." + "reason": "Same underlying bug in `_patch_mistral_regex` on the same file and same fix; `45317` looks like a superset of `45086` with tests, so these are duplicate PRs." } ] }, @@ -21395,7 +21430,7 @@ { "label": "pull_request", "component_index": 19, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21411,7 +21446,7 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "e80c62be57f8b628cd0d0894dd1ac85b35df6c4082fa983e8cb2ce01cbdb653a", + "prepared_review_unit_hash": "bd55a0034dc37f41aae1197f72296ff5d85884c87ea14635e10c7ca4788cbe69", "component_budget": { "node_count": 2, "item_count": 2, @@ -21433,18 +21468,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests touch NemotronH internals, but they are not the same change: one is a narrower inheritance/conversion refactor, while the other splits NemotronH into dense/sparse variants and adds docs/tests. They should not be merged as duplicates.", - "confidence": 0.88, + "summary": "These are related NemotronH refactors, but they target different changes: one is a conversion-mapping/renaming update, the other is a broader model split and documentation update for NemotronH dense/sparse variants. They should not be treated as duplicates.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45569 is the better canonical representative because it is the broader, more complete NemotronH family change and includes the visible model split plus docs/tests, whereas 45115 is a narrower refactor/conversion-mapping pass.", + "canonical_pr_reason": "PR #45569 is the more complete and self-contained NemotronH change, covering the new dense/sparse model structure, auto mappings, docs, and tests. PR #45115 is a narrower preprocessing/mapping refactor that fits inside the larger split.", "best_issue_reason": null, - "best_pr_reason": "PR 45569 is the strongest standalone PR: it implements the full dense/sparse model restructuring and supporting documentation/test updates, making it the most useful representative of this cluster.", + "best_pr_reason": "PR #45569 is the best representative because it implements the broader, user-visible architectural change and includes the downstream plumbing needed for the new model variants.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "They overlap on NemotronH auto/model files, but 45115 is mainly an inheritance/conversion-mapping refactor while 45569 introduces separate nemotron_h_dense and nemotron_h_sparse implementations with docs/tests. Different concrete changes, so not a duplicate." + "reason": "Different underlying changes: #45115 only adds conversion-key renaming for NemotronH/NemotronH Dense checkpoints, while #45569 restructures the model into nemotron_h_dense and nemotron_h_sparse packages with docs and tests. They are related, but not the same fix and not plausibly merged as one duplicate PR." } ] }, @@ -21458,7 +21493,7 @@ { "label": "pull_request", "component_index": 20, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21474,12 +21509,12 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "7394eee6df6219ba0ec5022e4e78357a5028d94b3153f4706b63dc9fdb2c0b22", + "prepared_review_unit_hash": "31a8e902364fb933fc3b64a60e2643b6dd1c3fb0ac95d17b232c8f86973250b8", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5466, + "serialized_chars": 5465, "estimated_input_tokens": 1367, "estimated_eval_tokens": 2990 }, @@ -21487,7 +21522,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5466, + "serialized_chars": 5465, "estimated_input_tokens": 1367, "estimated_eval_tokens": 2990 }, @@ -21496,18 +21531,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related heterogeneous-modeling PRs, but they are not duplicates: one adds per-layer heterogeneous config support, while the other adds broader model/runtime integration on top of that config machinery.", - "confidence": 0.94, + "summary": "Two related but distinct PRs: one is a broad heterogeneous-model support umbrella, the other is a narrower per-layer configuration PR. They share infrastructure files, but they are not the same change.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45333 is the narrower foundational change focused on per-layer config support; PR 45332 is a larger follow-on that adds modeling/cache/masking integration and is not the same concrete change.", + "canonical_pr_reason": "pull_request:45332 is the broader umbrella change: it includes heterogeneous modeling, cache, masking, and tests, while 45333 is a narrower config-only foundation that appears to feed into it.", "best_issue_reason": null, - "best_pr_reason": "PR 45333 is the best representative for the shared theme because it cleanly captures the core per-layer heterogeneous config work without the extra downstream modeling changes from 45332.", + "best_pr_reason": "pull_request:45332 is the best representative of the cluster because it captures the full heterogeneous-model feature set rather than just the configuration plumbing.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Different layers of the feature: 45333 introduces heterogeneous configuration, while 45332 implements broader model/runtime support using that configuration. Related, but not the same underlying change." + "reason": "Reject: 45333 is a subset/foundation for per-layer config, while 45332 adds broader modeling/cache/masking support. They are related, but not the same concrete change and would not be merged as duplicates." } ] }, @@ -21521,7 +21556,7 @@ { "label": "pull_request", "component_index": 21, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21537,7 +21572,7 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "bcad4ef2f0ff1fb013544883fa9b4fb8a18a1a2ce46d02b968e818917e88a323", + "prepared_review_unit_hash": "9cf887b3a0d4595f7f6f90529397857b274d1d3ca2b8393951495d2081420db1", "component_budget": { "node_count": 2, "item_count": 2, @@ -21559,18 +21594,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate typing-focused PRs: one is a broad repository-wide type-checking sweep, and the other is a narrower typing-infra update around `modeling_utils`/PEFT-related protocols. They overlap in a few shared files but do not appear to fix the same underlying change.", - "confidence": 0.93, + "summary": "Both PRs are typing-related, but they are different changes: 45415 is a broad repo-wide type-checking sweep, while 45425 adds specific typing protocols/helpers around modeling_utils and related utilities. The shared files appear incidental to the overlap in typing work, not evidence of the same fix.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45425 is the more focused and self-contained typing change, centered on `_typing.py`, `modeling_utils`, and related protocols. It has a clear narrow purpose and is easier to treat as a canonical representative than the broad sweep in 45415.", + "canonical_pr_reason": "PR 45415 is the broader umbrella change: it enables type checking across many transformers modules and adjusts pyproject type-ignore handling, so it best represents the cluster theme.", "best_issue_reason": null, - "best_pr_reason": "PR 45425 is the best PR representative for this cluster because it is a specific typing-infrastructure change with a bounded scope, whereas 45415 is a wide multi-file type-checking pass that mixes many unrelated annotation/ignore updates.", + "best_pr_reason": "PR 45425 is the more self-contained and concrete typing change, focused on specific protocols and helper types rather than a large sweeping type-checking rollout.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Do not merge: both are typing-related, but 45415 is a broad type-checking sweep across many files with pyproject ignore changes, while 45425 adds targeted typing protocols and `modeling_utils` support. Shared filenames alone do not indicate the same concrete fix." + "reason": "Reject: these are not the same underlying change. 45415 is a broad typing pass across many files, while 45425 adds targeted typing protocols/helpers and related model utilities; overlapping filenames do not make them mergeable into one PR." } ] }, @@ -21584,7 +21619,7 @@ { "label": "pull_request", "component_index": 22, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21600,7 +21635,7 @@ "soft_pairs": [ "pull_request:45513|pull_request:46084" ], - "prepared_review_unit_hash": "f739f9e98921cd1f9ff71caa46862746c09ab6786f8b15c58e030e89be169422", + "prepared_review_unit_hash": "2e855ee2d6c0f72a7ae8818b2d214c40111134ba6c6d7c22bf9727ac53b117f1", "component_budget": { "node_count": 2, "item_count": 2, @@ -21622,18 +21657,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated pull requests about cache/use_cache behavior in different model families: Qwen3.5/Qwen3-next linear attention versus Mamba2 CUDA kernels. They share a broad theme but not the same underlying bug or mergeable fix.", - "confidence": 0.95, + "summary": "These are two separate cache-related fixes in different model families: one for GDN linear attention multi-token cached forward in Qwen/Olmo hybrid models, and one for Mamba2 use_cache with seq_len > 1. They are thematically similar but not the same concrete bug or change.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR stands out for a duplicate cluster because the fixes target different code paths and models. If forced, #46084 is the cleaner representative since it has an explicit issue link and a narrowly scoped seq_len > 1 cache fix.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "#46084 is the strongest standalone PR: it is focused, explicitly tied to issue #46032, and fixes a concrete use_cache regression in Mamba2. #45513 is also well-scoped, but it addresses a different model family and problem.", + "best_pr_reason": "PR 46084 is the clearest standalone fix: it has an explicit issue reference, a narrowly defined Mamba2 cache bug, and a small, focused patch. PR 45513 is broader and spans multiple model implementations, but it is still a different bug.", "soft_edge_verdicts": [ { "left": "pull_request:45513", "right": "pull_request:46084", "accept": false, - "reason": "Both are cache-related, but they fix different bugs in different architectures/code paths. They are not the same underlying change and would not plausibly be merged into one PR." + "reason": "Both concern cached forward paths with seq_len > 1, but they target different architectures and code paths (Qwen/Olmo hybrid GDN linear attention vs Mamba2). Similar theme is not enough to treat them as duplicates or one merged change." } ] }, @@ -21647,7 +21682,7 @@ { "label": "pull_request", "component_index": 23, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21663,7 +21698,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "5211a035b8221f9be28d754110d8c52346a95fbbdff64c4577ada0c9d3806292", + "prepared_review_unit_hash": "101fb8a5e18d3d2c8dcd0c2655c3785da13d9bd0c7f4df0515d23906f91ab715", "component_budget": { "node_count": 2, "item_count": 2, @@ -21685,18 +21720,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs address expert-parallel loading, but they target different failure modes: one fixes Accelerate/ZeRO-3 loading by preserving distributed_config, while the other fixes FSDP2 expert weights being overwritten by rank-0 broadcast. They are related, but not the same underlying bug.", - "confidence": 0.91, + "summary": "These are two related but distinct EP-loading fixes. They share a file and the broad EP theme, but one targets serialization/loading under accelerate launch and the other targets FSDP2/rank-0 broadcast overwriting expert weights. They should not be merged as duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45662 is the stronger canonical PR: it addresses a concrete expert-parallel loading corruption path with a targeted multi-file fix, whereas 45548 is a narrower fix for a different distributed-loading path involving config serialization and launch behavior.", + "canonical_pr_reason": "PR #45662 is the more concrete fix: it addresses a specific corruption path where experts are overwritten by rank-0 broadcast during EP + FSDP2 loading, with a direct code-path change in core loading and MoE handling.", "best_issue_reason": null, - "best_pr_reason": "PR 45662 is the best standalone PR in this cluster because it fixes the more clearly defined and more severe code-path bug (experts being silently overwritten during load) and has the broader, more cohesive implementation.", + "best_pr_reason": "PR #45662 is the best representative of the cluster because it describes a precise failure mode and the diff aligns with that bug; #45548 is a separate EP/ZeRO-3 configuration/loading issue.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Both are about EP loading, but 45548 fixes Accelerate/DeepSpeed ZeRO-3 config handling, while 45662 fixes FSDP2 post-shard loading/broadcast overwrites. Different concrete bugs and not plausibly one merged PR." + "reason": "Different underlying bugs and code paths: #45548 is about EP + DeepSpeed ZeRO-3 loading via accelerate launch and config serialization, while #45662 fixes EP + FSDP2 experts being overwritten by rank-0 broadcast." } ] }, @@ -21710,7 +21745,7 @@ { "label": "pull_request", "component_index": 24, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21726,12 +21761,12 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "2bfb7d72e9bc3836d8b655b41de8248bc26202e7c5b15f44027d3f3185384b77", + "prepared_review_unit_hash": "7efb89899e0c59ebfcaf8bf024a6fbd9d3ca8042d0c8ebc7cdf2d3ad15cfb0ba", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4590, + "serialized_chars": 4589, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -21739,7 +21774,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4590, + "serialized_chars": 4589, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -21748,18 +21783,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same captured_info xdist collision problem around patched testing debug logs. PR 45645 is the broader, more complete fix and appears to subsume PR 45639.", - "confidence": 0.91, + "summary": "Both PRs address the same xdist-related captured_info artifact collision problem under issue #45561; PR #45645 is the broader, more complete fix and appears to subsume #45639.", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45645 is the most complete candidate: it includes the xdist-safe output-file change in testing_utils plus the CI/workflow updates needed to aggregate multiple captured_info files.", + "canonical_pr_reason": "PR #45645 is the best canonical PR because it covers the full artifact-collision path: it makes the writer xdist-safe and also updates CI/log collection to handle multiple captured_info files, whereas #45639 only fixes the writer side.", "best_issue_reason": null, - "best_pr_reason": "PR 45645 is the best overall fit because it directly targets the same bug and also fixes downstream log handling in the workflow, making it the most useful single PR to keep.", + "best_pr_reason": "PR #45645 is the strongest candidate overall since it addresses both the root cause and the downstream log-consumption changes needed for the same xdist artifact issue.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Both PRs fix the same underlying xdist collision/captured_info artifact issue in the same testing code path; 45645 looks like a superset of 45639 rather than a separate change." + "reason": "Same underlying captured_info xdist collision bug. #45639 fixes the output-file write path; #45645 fixes that path too and also updates the workflow to consume multiple captured_info files, so they are plausibly part of one merged change set." } ] }, @@ -21773,7 +21808,7 @@ { "label": "pull_request", "component_index": 25, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21792,52 +21827,52 @@ "pull_request:45670|pull_request:46401", "pull_request:46320|pull_request:46401" ], - "prepared_review_unit_hash": "4f1b97690fb5a234fdb8d96a2aeaaa26bcaee15445aef320ea1e298df5d5f4d0", + "prepared_review_unit_hash": "29017647b96bf3dcbbd7316548066a6b68240a5d192e96109eefdd174d244f60", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7838, - "estimated_input_tokens": 1960, - "estimated_eval_tokens": 4176 + "serialized_chars": 7580, + "estimated_input_tokens": 1895, + "estimated_eval_tokens": 4046 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7838, - "estimated_input_tokens": 1960, - "estimated_eval_tokens": 4176 + "serialized_chars": 7580, + "estimated_input_tokens": 1895, + "estimated_eval_tokens": 4046 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one PR is an unrelated AutoModel registry nit, and the other two add attn_implementation support to different model families/code paths. The two dispatch PRs are thematically similar but too broad and architecture-specific to merge as one.", - "confidence": 0.95, + "summary": "These PRs are not duplicates: two are separate attn_implementation dispatch additions for different model families, and the third is an unrelated AutoModel registry tweak. They share a broad theme but not the same underlying code-path change.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "If a representative PR is needed, 46320 is the cleanest single-model attn_implementation change; 46401 is a broader multi-model variant and 45670 is unrelated to dispatch support.", + "canonical_pr_reason": "PR #46320 is the best canonical representative of the cluster theme because it is a focused implementation of attn_implementation dispatch for a specific model family, with model code, conversion mapping, and tests touched.", "best_issue_reason": null, - "best_pr_reason": "46320 is the most self-contained and clearly scoped of the meaningful PRs in this set; it best represents the dispatch-support work without the extra breadth of 46401.", + "best_pr_reason": "PR #46320 is the strongest standalone PR here: it directly changes the model code path for dispatch support and includes validation. PR #46401 is a separate model-family implementation, and PR #45670 is unrelated to attn dispatch.", "soft_edge_verdicts": [ { "left": "pull_request:45670", "right": "pull_request:46320", "accept": false, - "reason": "Different problems: 45670 only adds model registry entries for glmasr/dia, while 46320 implements attn_implementation dispatch for RecurrentGemma and related tests/mapping." + "reason": "Different changes: #45670 only adjusts AutoModel registry entries for glmasr/dia, while #46320 implements attn_implementation dispatch for RecurrentGemma." }, { "left": "pull_request:45670", "right": "pull_request:46401", "accept": false, - "reason": "Different problems: 45670 is an AutoModel registration tweak, while 46401 adds attn_implementation support across BLIP-2/InstructBLIP-related model code paths." + "reason": "No shared underlying fix: #45670 is an AutoModel mapping nit, whereas #46401 adds attn_implementation dispatch support across BLIP-2/InstructBLIP variants." }, { "left": "pull_request:46320", "right": "pull_request:46401", "accept": false, - "reason": "Same broad feature area, but not the same underlying change: 46320 targets RecurrentGemma, while 46401 targets BLIP-2/InstructBLIP/InstructBLIPVideo and OCR variants. They touch different model implementations and would not plausibly be one merged PR." + "reason": "Both are about attn_implementation support, but they target different concrete model families and code paths (RecurrentGemma vs BLIP/InstructBLIP), so they are not the same change." } ] }, @@ -21851,7 +21886,7 @@ { "label": "pull_request", "component_index": 26, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21867,7 +21902,7 @@ "soft_pairs": [ "pull_request:45720|pull_request:46259" ], - "prepared_review_unit_hash": "529e010199dc530b7275309dc5e605eaf44ed51478d26ee34e7cf0bd5033a75e", + "prepared_review_unit_hash": "e80f00b9f3889e25c36b3834764f58a37bd2c1c8cf3ca1e23ec367a0c0946dd4", "component_budget": { "node_count": 2, "item_count": 2, @@ -21889,18 +21924,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated linter/modeling-rule PRs: one implements TRF003 around tuple-return handling in model forwards, the other adds TRF016 about consuming do_* preprocessing flags. They share a broad repository area but not the same underlying bug or change.", - "confidence": 0.97, + "summary": "These are two unrelated linter/chore PRs: one enforces TRF003 around model forward return-tuple behavior, the other adds TRF018 about image/video processor boolean flags. They overlap only in broad modeling-related file coverage and should not be merged as duplicates.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46259 is the cleaner representative of the cluster: it introduces a self-contained new modeling rule with documentation and targeted model updates, whereas PR 45720 is a separate, broader implementation of a different rule.", + "canonical_pr_reason": "PR #45720 is the better representative of the cluster because it makes a concrete behavioral rule change across model forward methods, whereas #46259 is a separate processor-rule/docs addition.", "best_issue_reason": null, - "best_pr_reason": "PR 46259 is the best standalone PR in this set because it has a clearer scope and a more explicit rule definition; it is still not a duplicate of PR 45720.", + "best_pr_reason": "PR #45720 is the stronger standalone PR for triage: it targets a specific code-path change in multiple model implementations and has a clearer execution-focused scope than the modeling-rule documentation/update in #46259.", "soft_edge_verdicts": [ { "left": "pull_request:45720", "right": "pull_request:46259", "accept": false, - "reason": "Different rules and different code paths: TRF003 changes forward return behavior via can_return_tuple, while TRF016 checks preprocessing do_* flags in image/video processors. Shared linter/theme is too broad to treat as the same change." + "reason": "Reject: these implement different rules for different subsystems. #45720 changes model forward methods to support can_return_tuple/return_dict behavior; #46259 adds a rule about processor do_* flags being consumed during preprocess. Shared model filenames and linter/chore framing are too broad to indicate the same underlying fix." } ] }, @@ -21914,7 +21949,7 @@ { "label": "pull_request", "component_index": 27, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21980,7 +22015,7 @@ { "label": "pull_request", "component_index": 28, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -21996,40 +22031,40 @@ "soft_pairs": [ "pull_request:45932|pull_request:46100" ], - "prepared_review_unit_hash": "ad0dcbf42bcf51decf6965f7a4437b646c44cd8a137846f8b9d65423ceba3e7a", + "prepared_review_unit_hash": "1fdeba9d55e2f5e080e81b8de00f7ee88b6d220c649960d3fe1166a7ededc312", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4299, - "estimated_input_tokens": 1075, - "estimated_eval_tokens": 2406 + "serialized_chars": 4301, + "estimated_input_tokens": 1076, + "estimated_eval_tokens": 2408 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4299, - "estimated_input_tokens": 1075, - "estimated_eval_tokens": 2406 + "serialized_chars": 4301, + "estimated_input_tokens": 1076, + "estimated_eval_tokens": 2408 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are not duplicates: one is a broad implementation/CI update PR, the other is a test-only migration PR with deliberate failing assertions. They do not fix the same underlying bug or change.", - "confidence": 0.97, + "summary": "Two unrelated prank/diagnostic PRs: 45932 is a multi-file CI/model/config change, while 46100 is a ViT test-only PR that injects failing assertions. They do not fix the same underlying issue.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45932 is the more substantive anchor: it changes production code and CI plumbing across multiple files, while PR 46100 is a narrow test-file edit with intentionally failing assertions.", + "canonical_pr_reason": "PR 45932 is the more substantive representative: it changes workflow and several model/config files. PR 46100 is a trivial test-only PR that intentionally breaks tests, so it is not a canonical change.", "best_issue_reason": null, - "best_pr_reason": "PR 45932 is the best overall representative of the cluster because it contains real code-path changes; PR 46100 looks like a temporary or placeholder test migration and is not a comparable fix.", + "best_pr_reason": "PR 45932 is the closest thing to a real code change and spans the actual affected files; 46100 is clearly a test sabotage PR and not suitable as a representative artifact.", "soft_edge_verdicts": [ { "left": "pull_request:45932", "right": "pull_request:46100", "accept": false, - "reason": "Reject: the PRs are unrelated. 45932 updates model code and workflow secrets across several files, while 46100 only edits a ViT test file and adds intentional failures. They do not appear to address the same bug or merge into one change." + "reason": "Reject: 45932 modifies CI secrets and model/config code; 46100 only adds failing assertions in ViT tests. They are not the same bug or a plausible single merged fix." } ] }, @@ -22043,7 +22078,7 @@ { "label": "pull_request", "component_index": 29, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -22062,7 +22097,7 @@ "pull_request:46073|pull_request:46196", "pull_request:46154|pull_request:46196" ], - "prepared_review_unit_hash": "86028b57a02cf49ae492a808f1054116cfc43e984450616922e4f29c12d796c3", + "prepared_review_unit_hash": "ad2a39bc61e8c8c865807cb26a2af1bf40960908cbd3ea7740a610c6f2ea2823", "component_budget": { "node_count": 3, "item_count": 3, @@ -22084,30 +22119,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are related by the same feature pattern: adding SDPA dispatch support. However, they target different model families and code paths, so they should remain separate and are not duplicates of one another.", - "confidence": 0.89, + "summary": "All three pull requests are related by the same feature theme (SDPA dispatch), but they target different model implementations and code paths, so they should not be treated as duplicates of one another.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46196 is the best representative because it appears to be the broadest implementation, covering multiple model families and associated tests/docs, making it the most comprehensive instance of the SDPA-dispatch change set.", + "canonical_pr_reason": "PR 46196 is the broadest representative of the cluster, covering multiple model families and the largest change set around SDPA dispatch support.", "best_issue_reason": null, - "best_pr_reason": "PR 46196 is the strongest standalone PR in this cluster due to its wider scope and larger test coverage, while the other two are narrower model-specific variants of the same general feature.", + "best_pr_reason": "PR 46196 is the strongest single representative because it spans the widest scope and is most central to the shared SDPA-dispatch theme.", "soft_edge_verdicts": [ { "left": "pull_request:46073", "right": "pull_request:46154", "accept": false, - "reason": "Both are SDPA-dispatch PRs, but one is for CTRL and the other for T5. They touch different model implementations and are not the same concrete bug/change." + "reason": "Same high-level feature area, but CTRL and T5 are separate model implementations with distinct attention code paths; these are not the same underlying fix and would not plausibly merge into one PR." }, { "left": "pull_request:46073", "right": "pull_request:46196", "accept": false, - "reason": "These overlap only at the high-level goal of adding SDPA support. CTRL vs wav2vec2_conformer/bert and SeamlessM4T/v2 are separate code paths, so they are not duplicates." + "reason": "They both add SDPA support, but one is for CTRL while the other covers wav2vec2_conformer/bert and SeamlessM4T variants; different concrete model paths, so not duplicates." }, { "left": "pull_request:46154", "right": "pull_request:46196", "accept": false, - "reason": "T5 SDPA dispatch and the multi-model audio/seq2seq SDPA dispatch PR address different model-specific implementations and cannot plausibly be merged as one duplicate fix." + "reason": "T5 SDPA dispatch is a separate model-specific change from the audio/seq2seq model set in 46196; same theme, but not the same underlying bug or change." } ] }, @@ -22121,7 +22156,7 @@ { "label": "pull_request", "component_index": 30, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -22137,7 +22172,7 @@ "soft_pairs": [ "pull_request:46126|pull_request:46269" ], - "prepared_review_unit_hash": "e8e8f9917914ddf91197ff8886ad983121740b5f9506d6e80fab33a6605f7a25", + "prepared_review_unit_hash": "a96cd855ac275d1569de9217eb64667ad57ecb835eaa6fac44d2670e8ba76818", "component_budget": { "node_count": 2, "item_count": 2, @@ -22159,18 +22194,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related only by broad distributed-modeling overlap, not the same underlying change. PR #46126 is a focused EP/Trainer integration feature, while PR #46269 is a large distributed branch base spanning FSDP, tensor parallel, integrations, and model updates.", - "confidence": 0.95, + "summary": "These are not duplicates: one PR is a focused EP/trainer integration change, while the other is a broad distributed branch-base umbrella touching many unrelated distributed areas. The overlap in shared files is too coarse to indicate the same concrete fix.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46126 is the more coherent and specific change set: it targets EP routing/Trainer integration with a small, clearly defined scope. PR #46269 is a broad base branch with many unrelated distributed updates, so it is less suitable as the canonical representative of this cluster.", + "canonical_pr_reason": "PR 46126 is the more focused, concrete change centered on EP + Trainer integration; PR 46269 is a broad branch-base PR with many distributed subsystems and is not a clean duplicate anchor.", "best_issue_reason": null, - "best_pr_reason": "PR #46126 is the best representative because it has a single concrete feature goal and clearer intent. PR #46269 is too broad and multi-purpose to serve as a good cluster anchor.", + "best_pr_reason": "PR 46126 is the best representative of the shared overlap because it targets a specific code-path change, whereas PR 46269 is too broad and mixed-scope to serve as a precise canonical PR.", "soft_edge_verdicts": [ { "left": "pull_request:46126", "right": "pull_request:46269", "accept": false, - "reason": "Reject: the overlap is only at the subsystem level. #46126 adds EP/Trainer integration and router-parallel behavior, while #46269 is a large distributed-base PR touching many areas (FSDP, tensor parallel, integrations, model code). They do not appear to fix the same concrete problem and would not plausibly be merged into one PR." + "reason": "Reject: they do not appear to fix the same underlying bug/change. The shared files are in broad distributed infrastructure, but 46126 is a targeted EP/trainer integration PR and 46269 is a large distributed branch-base PR spanning many unrelated areas." } ] }, @@ -22184,7 +22219,7 @@ { "label": "pull_request", "component_index": 31, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -22218,12 +22253,12 @@ "pull_request:46502|pull_request:46504", "pull_request:46503|pull_request:46504" ], - "prepared_review_unit_hash": "46113ddba0fbf5defb712e18bb9de0d3efb1a22f56de69b5c59982fe7f22c88f", + "prepared_review_unit_hash": "52d168666bc44aae1c1509a804c30f7339330bafde167a88face08e9516a575b", "component_budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 15, - "serialized_chars": 14225, + "serialized_chars": 14228, "estimated_input_tokens": 3557, "estimated_eval_tokens": 7370 }, @@ -22231,7 +22266,7 @@ "node_count": 6, "item_count": 6, "soft_pair_count": 15, - "serialized_chars": 14225, + "serialized_chars": 14228, "estimated_input_tokens": 3557, "estimated_eval_tokens": 7370 }, @@ -22240,102 +22275,102 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is not a duplicate set: it contains six unrelated workflow-maintenance pull requests that happen to share a similar \"chore: update ...yml\" style. The pairwise similarity is superficial; each PR changes a different workflow and fixes a different concrete behavior or security concern.", - "confidence": 0.97, + "summary": "These PRs are all unrelated workflow-maintenance changes; the similarity is mostly the generic 'chore/update workflow' pattern, not the same underlying fix or feature.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "PR 46504 is the cleanest single-purpose change in the set: it fixes shell-safe commit-message handling in build-ci-docker-images and is narrowly scoped.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "PR 46504 is the best representative of the cluster because it addresses one concrete CI bug with a self-contained, plausibly mergeable change.", "soft_edge_verdicts": [ { "left": "pull_request:46138", "right": "pull_request:46504", "accept": false, - "reason": "Different workflows and different fixes: stdin handling for self-comment CI vs shell-safe commit-message parsing for build-ci-docker-images." + "reason": "Different workflows and different bugs: stdin-safe message passing in self-comment CI vs commit-message parsing in build-ci-docker-images." }, { "left": "pull_request:46138", "right": "pull_request:46432", "accept": false, - "reason": "One updates self-comment CI command invocation; the other hardens ssh-runner by pinning a download and unsetting HF_TOKEN." + "reason": "These touch separate CI workflows with unrelated goals; one changes message piping, the other hardens SSH runner handling and env cleanup." }, { "left": "pull_request:46138", "right": "pull_request:46503", "accept": false, - "reason": "Different workflow files and behavior: self-comment CI message piping vs benchmark_v2 schedule/dispatch triggering." + "reason": "No shared concrete code-path: self-comment CI message handling is unrelated to benchmark_v2 triggering logic." }, { "left": "pull_request:46138", "right": "pull_request:46501", "accept": false, - "reason": "Unrelated workflow changes: self-comment CI message transport vs anti-slop permissions scoping." + "reason": "Different workflow files and different concerns; these are both chore edits but not the same change or bug." }, { "left": "pull_request:46138", "right": "pull_request:46502", "accept": false, - "reason": "No shared code path or bug: self-comment CI command piping vs assign-reviewers checkout ref selection." + "reason": "Self-comment CI command input handling is unrelated to assign-reviewers checkout behavior." }, { "left": "pull_request:46432", "right": "pull_request:46503", "accept": false, - "reason": "ssh-runner security/environment handling is unrelated to benchmark_v2 event gating." + "reason": "SSH runner security/environment changes and benchmark dispatch logic are separate workflow paths with no single fix to combine." }, { "left": "pull_request:46432", "right": "pull_request:46501", "accept": false, - "reason": "Different workflow concerns: ssh-runner hardening vs anti-slop permission adjustment." + "reason": "These modify different workflows for different reasons; one updates SSH runner hardening, the other changes anti-slop permissions." }, { "left": "pull_request:46432", "right": "pull_request:46502", "accept": false, - "reason": "Distinct workflows and purposes; no common underlying bug or mergeable change set." + "reason": "No same underlying bug: SSH runner secrets handling does not match assign-reviewers base-ref checkout." }, { "left": "pull_request:46432", "right": "pull_request:46504", "accept": false, - "reason": "ssh-runner download/env hardening is unrelated to build-ci-docker-images commit-message parsing." + "reason": "Different CI workflows and distinct shell concerns; not the same code-path problem." }, { "left": "pull_request:46501", "right": "pull_request:46503", "accept": false, - "reason": "Both touch workflows, but one changes anti-slop permissions and the other adds workflow_dispatch to benchmark_v2; they are separate changes." + "reason": "Anti-slop permission scoping and benchmark workflow dispatch are unrelated workflow changes." }, { "left": "pull_request:46501", "right": "pull_request:46502", "accept": false, - "reason": "Different workflow files and different fixes; not the same underlying bug or change." + "reason": "Both are workflow maintenance, but they fix different jobs in different files with no shared defect." }, { "left": "pull_request:46501", "right": "pull_request:46504", "accept": false, - "reason": "Anti-slop permission scoping and build-ci-docker-images message handling are unrelated." + "reason": "Changing anti-slop job permissions is unrelated to build-ci-docker-images command parsing." }, { "left": "pull_request:46502", "right": "pull_request:46503", "accept": false, - "reason": "assign-reviewers checkout ref behavior is unrelated to benchmark_v2 trigger logic." + "reason": "Assign-reviewers checkout ref behavior is unrelated to benchmark_v2 event gating." }, { "left": "pull_request:46502", "right": "pull_request:46504", "accept": false, - "reason": "Different workflows and concrete edits; no plausible single merged PR." + "reason": "Different workflow files and unrelated fixes; not plausibly one merged PR." }, { "left": "pull_request:46503", "right": "pull_request:46504", "accept": false, - "reason": "benchmark_v2 workflow_dispatch support is unrelated to build-ci-docker-images commit-message parsing." + "reason": "Benchmark_v2 dispatch support and build-ci-docker-images shell quoting address different workflows and bugs." } ] }, @@ -22349,7 +22384,7 @@ { "label": "pull_request", "component_index": 32, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -22365,7 +22400,7 @@ "soft_pairs": [ "pull_request:46152|pull_request:46207" ], - "prepared_review_unit_hash": "18dde5d905619cc141ef4230b418ffc7578368bd28793d7d6b64c68195859a31", + "prepared_review_unit_hash": "c330232a3454f187dd643e40e2bda06d9955ee2d805c7e68e45376915da0996a", "component_budget": { "node_count": 2, "item_count": 2, @@ -22387,18 +22422,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both items are pull requests about avoiding direct file-open source inspection, but they change different code paths in different modules: one in import_utils, the other in modeling_utils. They are related in theme, not duplicates of the same fix.", - "confidence": 0.91, + "summary": "These PRs share a high-level theme of avoiding direct file-open source inspection, but they target different code paths and bugs: one in import_utils module-source resolution, the other in modeling_utils attention-implementation detection.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46207 is the better representative of the cluster because it is a direct source-reading refactor in an existing implementation path, with tests, and its scope is clearer than the import-utils traversal helper in #46152.", + "canonical_pr_reason": "PR #46207 is the better representative of the cluster because it has the clearer, self-contained code-path fix and accompanying test coverage, while still matching the shared source-inspection theme.", "best_issue_reason": null, - "best_pr_reason": "PR #46207 is the strongest single PR here: it targets a concrete behavior in _can_set_*_implementation, is self-contained, and is more directly about the source-reading heuristic than #46152, which addresses a separate import-utils path.", + "best_pr_reason": "PR #46207 is the strongest standalone PR for this cluster: it fixes a concrete heuristic in _can_set_attn_implementation and includes tests; #46152 is a separate utility change in a different module.", "soft_edge_verdicts": [ { "left": "pull_request:46152", "right": "pull_request:46207", "accept": false, - "reason": "They share a broad theme of replacing open()-based source access, but they fix different functions in different files and are not the same underlying bug or mergeable as one PR." + "reason": "Reject: they are not the same underlying change. #46152 rewrites import_utils path handling for self-source resolution, while #46207 changes modeling_utils's attention-implementation support check. Similar pattern, different bug and code path." } ] }, @@ -22412,7 +22447,7 @@ { "label": "pull_request", "component_index": 33, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -22430,52 +22465,52 @@ "pull_request:46436|pull_request:46474", "pull_request:46173|pull_request:46474" ], - "prepared_review_unit_hash": "29414d39de4b67c06cb5047b3d629d4b147f698911987e3610086fd0b10762c9", + "prepared_review_unit_hash": "e27c88da5177d2c02deea31816182e226ff098990ec23b74aff7aa80d73da519", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 5920, - "estimated_input_tokens": 1480, - "estimated_eval_tokens": 3216 + "serialized_chars": 5921, + "estimated_input_tokens": 1481, + "estimated_eval_tokens": 3218 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 5920, - "estimated_input_tokens": 1480, - "estimated_eval_tokens": 3216 + "serialized_chars": 5921, + "estimated_input_tokens": 1481, + "estimated_eval_tokens": 3218 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two PRs are near-identical duplicates fixing the same StaticLayer type-hint issue; the generation cache-config PR is unrelated and should stay separate.", - "confidence": 0.97, + "summary": "Two PRs are near-exact duplicates fixing the same StaticLayer `get_seq_length` type annotation in `cache_utils.py`; the third PR is unrelated and changes static cache length handling in generation.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46173 is the best canonical representative for the StaticLayer get_seq_length typing fix: it has the explicit issue link to #45987 and the exact patch that updates both the abstract base and StaticLayer implementation.", + "canonical_pr_reason": "PR #46173 is the best canonical representative: it has the same patch as #46474, includes the explicit issue link (#45987), and has more discussion/context.", "best_issue_reason": null, - "best_pr_reason": "PR #46173 is the strongest cluster member to keep: it is directly tied to the tracked issue and has the clearest, fully scoped change for the return-type annotation fix.", + "best_pr_reason": "PR #46173 is the strongest overall PR in the cluster because it directly matches the underlying change, is clearly scoped to the bug, and is better documented than the later duplicate.", "soft_edge_verdicts": [ { "left": "pull_request:46436", "right": "pull_request:46474", "accept": false, - "reason": "Different changes and different code paths: one adjusts static cache max length handling in generation, the other changes a type annotation in cache_utils. They are not the same underlying fix." + "reason": "Different changes: #46436 adjusts static cache max length selection in generation, while #46474 only updates the `StaticLayer.get_seq_length` type hint. Same subsystem, but not the same underlying fix." }, { "left": "pull_request:46173", "right": "pull_request:46474", "accept": true, - "reason": "Same concrete bug/change: both patch StaticLayer.get_seq_length in cache_utils to return int | torch.Tensor, with the same issue target #45987 and essentially identical diff." + "reason": "Same code change in `src/transformers/cache_utils.py`, same issue target (#45987), and the diffs are effectively identical. These are duplicate PRs for the same fix." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary and pairwise verdicts are well grounded in the packet. 46173 and 46474 are effectively duplicate type-hint fixes in cache_utils tied to issue #45987, while 46436 changes generation/static-cache length handling in a different file and is reasonably kept separate. No evidence is overstated." + "feedback": "The verdicts are grounded in the packet. #46173 and #46474 have identical diffs in src/transformers/cache_utils.py and share issue target #45987, so treating them as duplicate PRs is justified. #46436 changes a different file and a different fix in generation/utils, so rejecting it as the same underlying fix is conservative. The summary is also consistent with the evidence." }, "evaluator_used": true, "retried": false, @@ -22486,73 +22521,7 @@ { "label": "pull_request", "component_index": 34, - "component_count": 36, - "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-46174-2", - "nodes": [ - "pull_request:46174", - "pull_request:46513" - ], - "soft_pairs": [ - "pull_request:46174|pull_request:46513" - ], - "prepared_review_unit_hash": "ffc4a9b5a20700667852cd30b7df4abd69be13e8b33da16b41d8cfd85ee4ef5f", - "component_budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 5069, - "estimated_input_tokens": 1268, - "estimated_eval_tokens": 2792 - }, - "budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 5069, - "estimated_input_tokens": 1268, - "estimated_eval_tokens": 2792 - }, - "overflow_policy": "truncate_then_skip", - "trimmed": false, - "aggressively_trimmed": false, - "split": false, - "analyst_result": { - "summary": "Both PRs target the same MPS crash in `build_2d_sinusoidal_position_embedding` caused by float64 usage, and both fix the same helper in overlapping RT-DETR code paths. PR 46174 is the broader, more canonical fix across multiple models; 46513 looks like a narrower variant of the same underlying change.", - "confidence": 0.87, - "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:46174 is the broader primary fix: it addresses the Apple Silicon/MPS float64 crash across several model implementations of the same helper, not just RT-DETR.", - "best_issue_reason": null, - "best_pr_reason": "pull_request:46174 is the best representative because it covers the core bug across more call sites and has the clearest end-to-end fix for the shared helper.", - "soft_edge_verdicts": [ - { - "left": "pull_request:46174", - "right": "pull_request:46513", - "accept": true, - "reason": "Both PRs fix the same concrete bug path: `build_2d_sinusoidal_position_embedding` failing on MPS due to float64. They overlap in RT-DETR files and are plausibly mergeable as one fix." - } - ] - }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: both PRs address the same MPS float64 crash in `build_2d_sinusoidal_position_embedding`, and the overlap in RT-DETR files makes them plausibly mergeable as one bugfix PR. The characterization of PR 46174 as the broader fix and PR 46513 as the narrower variant is conservative and supported by the diffs." - }, - "evaluator_used": true, - "retried": false, - "accepted_nontrivial_soft_edge": true, - "error_kind": null, - "error_message": null - }, - { - "label": "pull_request", - "component_index": 35, - "component_count": 36, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -22614,8 +22583,8 @@ }, { "label": "pull_request", - "component_index": 36, - "component_count": 36, + "component_index": 35, + "component_count": 35, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -22631,12 +22600,12 @@ "soft_pairs": [ "pull_request:46376|pull_request:46417" ], - "prepared_review_unit_hash": "3f780b714da90c4b2aa8c7952020a8114b509ef86b2a201a68c2e426ba5c28b1", + "prepared_review_unit_hash": "0d196d39d760fb23881e231f917e55438aeb55209250ee7868e6b1b99168b07b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5960, + "serialized_chars": 5959, "estimated_input_tokens": 1490, "estimated_eval_tokens": 3236 }, @@ -22644,7 +22613,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5960, + "serialized_chars": 5959, "estimated_input_tokens": 1490, "estimated_eval_tokens": 3236 }, @@ -22653,18 +22622,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two pull requests are unrelated: one adds multimodal support to continuous batching, while the other adds the HunYuan VL model, docs, and tests. They should not be deduplicated or merged into a single PR.", + "summary": "The two pull requests are unrelated: one is a draft infrastructure change for continuous batching multimodal support, and the other is a full model addition for HunYuan VL. They should not be clustered as duplicates.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR for duplicate triage; these PRs target different features and codepaths.", + "canonical_pr_reason": "PR 46417 is the cleaner standalone representative of this set because it is a complete, non-draft model addition; PR 46376 is a draft, broader infrastructure change and not the same underlying work.", "best_issue_reason": null, - "best_pr_reason": "No best PR in a duplicate sense; 46417 is a standalone model addition, not a duplicate of 46376.", + "best_pr_reason": "PR 46417 is the best PR in this cluster for triage since it is self-contained, non-draft, and materially more complete than the unrelated draft PR 46376.", "soft_edge_verdicts": [ { "left": "pull_request:46376", "right": "pull_request:46417", "accept": false, - "reason": "Different scope and implementation: continuous batching multimodal support versus a new HunYuan VL model/docs/tests. They do not fix the same bug or change the same codepath." + "reason": "Different changes and code paths: continuous batching multimodal infrastructure vs. adding the HunYuan VL model. They do not look like the same underlying fix or feature." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index eabc26650bf12269926e9e8d624f953dd7e6dfaf..9ff3bf9c3af012284b2449c6b898e922eddbef23 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260609t060025z", + "analysis_id": "hybrid-model-20260609t120115z", "archived_artifacts": { - "hybrid": "snapshots/20260609T060025Z/analysis-runs/hybrid-model-20260609t060025z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260609T060025Z/analysis-runs/hybrid-model-20260609t060025z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-06-09T06:09:55Z", + "published_at": "2026-06-09T12:12:39Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260609T060025Z", + "snapshot_id": "20260609T120115Z", "variant": "hybrid" } diff --git a/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.json b/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..d46bf68c88781de6cdcf080c6ba26035cb3376e0 --- /dev/null +++ b/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.json @@ -0,0 +1,2260 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260609T120115Z", + "generated_at": "2026-06-09T12:12:37Z", + "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-45363-3", + "summary": "Cluster of 3 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 46339, + "issue_numbers": [], + "pr_numbers": [ + 45363, + 45666, + 46339 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45363, + "right_pr_number": 45666, + "code_similarity": 0.392, + "size_similarity": 0.298, + "file_overlap": 0.571, + "area_overlap": 0.329, + "patch_similarity": 0.306, + "shared_filenames": [ + "fused_qwen_example.py", + "src/transformers/integrations/__init__.py", + "src/transformers/integrations/hub_kernels.py", + "src/transformers/modeling_utils.py" + ], + "shared_file_areas": [ + { + "filename": "fused_qwen_example.py", + "left_ranges": [ + [ + 1, + 30 + ] + ], + "right_ranges": [ + [ + 1, + 49 + ] + ] + }, + { + "filename": "src/transformers/integrations/__init__.py", + "left_ranges": [ + [ + 67, + 81 + ], + [ + 226, + 240 + ] + ], + "right_ranges": [ + [ + 67, + 73 + ], + [ + 222, + 228 + ] + ] + }, + { + "filename": "src/transformers/integrations/hub_kernels.py", + "left_ranges": [ + [ + 11, + 17 + ], + [ + 22, + 35 + ], + [ + 505, + 716 + ] + ], + "right_ranges": [ + [ + 11, + 53 + ], + [ + 286, + 295 + ], + [ + 521, + 1077 + ] + ] + }, + { + "filename": "src/transformers/modeling_utils.py", + "left_ranges": [ + [ + 3738, + 3747 + ] + ], + "right_ranges": [ + [ + 1330, + 1338 + ], + [ + 3763, + 3789 + ], + [ + 4223, + 4234 + ], + [ + 4628, + 4641 + ] + ] + } + ] + }, + { + "left_pr_number": 45363, + "right_pr_number": 46339, + "code_similarity": 0.506, + "size_similarity": 0.925, + "file_overlap": 0.286, + "area_overlap": 0.545, + "patch_similarity": 0.297, + "shared_filenames": [ + "src/transformers/integrations/hub_kernels.py", + "src/transformers/modeling_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/hub_kernels.py", + "left_ranges": [ + [ + 11, + 17 + ], + [ + 22, + 35 + ], + [ + 505, + 716 + ] + ], + "right_ranges": [ + [ + 14, + 52 + ], + [ + 285, + 294 + ], + [ + 521, + 685 + ] + ] + }, + { + "filename": "src/transformers/modeling_utils.py", + "left_ranges": [ + [ + 3738, + 3747 + ] + ], + "right_ranges": [ + [ + 1346, + 1354 + ], + [ + 3785, + 3811 + ], + [ + 4245, + 4256 + ], + [ + 4650, + 4663 + ] + ] + } + ] + }, + { + "left_pr_number": 45666, + "right_pr_number": 46339, + "code_similarity": 0.399, + "size_similarity": 0.322, + "file_overlap": 0.667, + "area_overlap": 0.247, + "patch_similarity": 0.318, + "shared_filenames": [ + "src/transformers/core_model_loading.py", + "src/transformers/integrations/hub_kernels.py", + "src/transformers/modeling_utils.py", + "tests/kernels/test_kernels.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/core_model_loading.py", + "left_ranges": [ + [ + 19, + 25 + ], + [ + 80, + 86 + ] + ], + "right_ranges": [ + [ + 19, + 25 + ], + [ + 80, + 86 + ] + ] + }, + { + "filename": "src/transformers/integrations/hub_kernels.py", + "left_ranges": [ + [ + 11, + 53 + ], + [ + 286, + 295 + ], + [ + 521, + 1077 + ] + ], + "right_ranges": [ + [ + 14, + 52 + ], + [ + 285, + 294 + ], + [ + 521, + 685 + ] + ] + }, + { + "filename": "src/transformers/modeling_utils.py", + "left_ranges": [ + [ + 1330, + 1338 + ], + [ + 3763, + 3789 + ], + [ + 4223, + 4234 + ], + [ + 4628, + 4641 + ] + ], + "right_ranges": [ + [ + 1346, + 1354 + ], + [ + 3785, + 3811 + ], + [ + 4245, + 4256 + ], + [ + 4650, + 4663 + ] + ] + }, + { + "filename": "tests/kernels/test_kernels.py", + "left_ranges": [ + [ + 19, + 54 + ], + [ + 228, + 281 + ], + [ + 294, + 317 + ], + [ + 605, + 995 + ] + ], + "right_ranges": [ + [ + 21, + 28 + ], + [ + 33, + 39 + ], + [ + 90, + 103 + ], + [ + 217, + 323 + ], + [ + 336, + 359 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43240-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43240.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43240, + "canonical_pr_number": 43251, + "issue_numbers": [ + 43240 + ], + "pr_numbers": [ + 43251, + 43254 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43251, + "right_pr_number": 43254, + "code_similarity": 0.794, + "size_similarity": 0.64, + "file_overlap": 1.0, + "area_overlap": 0.667, + "patch_similarity": 0.882, + "shared_filenames": [ + "src/transformers/loss/loss_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/loss/loss_utils.py", + "left_ranges": [ + [ + 30, + 50 + ] + ], + "right_ranges": [ + [ + 30, + 43 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-41115-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41115.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41115, + "canonical_pr_number": 45613, + "issue_numbers": [ + 41115 + ], + "pr_numbers": [ + 41116, + 45613 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41116, + "right_pr_number": 45613, + "code_similarity": 0.797, + "size_similarity": 0.885, + "file_overlap": 0.75, + "area_overlap": 0.831, + "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": [ + [ + 730, + 737 + ] + ] + }, + { + "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": [ + [ + 720, + 726 + ], + [ + 1331, + 1337 + ] + ] + }, + { + "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-45987-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45987.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45987, + "canonical_pr_number": 46173, + "issue_numbers": [ + 45987 + ], + "pr_numbers": [ + 46173, + 46474 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46173, + "right_pr_number": 46474, + "code_similarity": 1.0, + "size_similarity": 1.0, + "file_overlap": 1.0, + "area_overlap": 1.0, + "patch_similarity": 1.0, + "shared_filenames": [ + "src/transformers/cache_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cache_utils.py", + "left_ranges": [ + [ + 69, + 75 + ], + [ + 375, + 381 + ] + ], + "right_ranges": [ + [ + 69, + 75 + ], + [ + 375, + 381 + ] + ] + } + ] + } + ] + }, + { + "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": 0.494, + "size_similarity": 0.619, + "file_overlap": 0.5, + "area_overlap": 0.434, + "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": [ + [ + 3595, + 3602 + ], + [ + 3819, + 3853 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 86 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46291-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46291.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 46291, + "canonical_pr_number": 46293, + "issue_numbers": [ + 46291 + ], + "pr_numbers": [ + 46293, + 46478 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 46293, + "right_pr_number": 46478, + "code_similarity": 0.483, + "size_similarity": 0.325, + "file_overlap": 0.333, + "area_overlap": 0.862, + "patch_similarity": 0.111, + "shared_filenames": [ + "src/transformers/integrations/hub_kernels.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/hub_kernels.py", + "left_ranges": [ + [ + 89, + 180 + ], + [ + 233, + 247 + ] + ], + "right_ranges": [ + [ + 89, + 181 + ], + [ + 234, + 248 + ], + [ + 263, + 269 + ], + [ + 360, + 366 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44024-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44076, + "issue_numbers": [], + "pr_numbers": [ + 44024, + 44076 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44024, + "right_pr_number": 44076, + "code_similarity": 0.269, + "size_similarity": 0.976, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.492, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + } + ], + "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-44024-2", + "canonical_pr_number": 44076, + "duplicate_pr_numbers": [ + 44024 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44024-2 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-45363-3", + "canonical_pr_number": 46339, + "duplicate_pr_numbers": [ + 45363, + 45666 + ], + "target_issue_number": null, + "reason": "PRs in cluster-45363-3 are treated as duplicates because they share 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-45987-3", + "canonical_pr_number": 46173, + "duplicate_pr_numbers": [ + 46474 + ], + "target_issue_number": 45987, + "reason": "PRs in cluster-45987-3 are treated as duplicates because they converge on issue #45987 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46291-3", + "canonical_pr_number": 46293, + "duplicate_pr_numbers": [ + 46478 + ], + "target_issue_number": 46291, + "reason": "PRs in cluster-46291-3 are treated as duplicates because they converge on issue #46291 with closing_reference evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-41211-3", + "issue_number": 41211, + "reason": "Issue #41211 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 6 discussion comments plus 6 inbound references.", + "score": 148.0 + }, + "best_pr": { + "cluster_id": "cluster-43979-11", + "pr_number": 44007, + "reason": "PR #44007 is the strongest global PR candidate because it is open, belongs to a cluster with 11 artifacts, links to 1 issues, and carries 2 review/discussion events.", + "score": 238.886 + } +} diff --git a/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..53c758c274bdf38a0552507b34c1c81bcbadd73d --- /dev/null +++ b/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,22648 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260609T120115Z", + "generated_at": "2026-06-09T12:12:37Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 1, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "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:43596", + "issue:43638", + "issue:43716", + "issue:43827", + "issue:43828", + "issue:43866", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44805", + "issue:44918", + "issue:45081", + "issue:45161", + "issue:45237", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45800", + "issue:45823", + "issue:45859", + "issue:45874", + "issue:45901", + "issue:45902", + "issue:46015", + "issue:46165", + "issue:46294", + "issue:46304", + "issue:46402", + "issue:46468", + "issue:46515" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:46468", + "issue:45237|issue:45507", + "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:44918|issue:46165", + "issue:44805|issue:45823", + "issue:43638|issue:44805", + "issue:43638|issue:46165", + "issue:43596|issue:44805", + "issue:44805|issue:46294", + "issue:44805|issue:45161", + "issue:46165|issue:46468", + "issue:44560|issue:46165", + "issue:43596|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:45874|issue:46294", + "issue:44292|issue:45237", + "issue:39290|issue:45081", + "issue:45800|issue:46515", + "issue:45823|issue:45874", + "issue:43425|issue:44292", + "issue:34928|issue:41720", + "issue:43828|issue:45161", + "issue:43828|issue:44292", + "issue:43541|issue:43716", + "issue:43716|issue:46468", + "issue:43716|issue:45685", + "issue:44918|issue:45161", + "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:43828|issue:45685", + "issue:44918|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:34928|issue:36010", + "issue:43039|issue:44805", + "issue:45237|issue:46015", + "issue:43866|issue:46402", + "issue:45800|issue:46304" + ], + "prepared_review_unit_hash": "822e6c0633dcf95296382bf06616acc0b17ecf8120e53d24c72a1195659d4086", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 63, + "serialized_chars": 46529, + "estimated_input_tokens": 11633, + "estimated_eval_tokens": 23522 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly distinct bug reports, with a few tight duplicate clusters: DeepSpeed ZeRO-3/Bert index-out-of-bounds, Qwen3.5 MoE rotary-embedding/device-materialization issues, and torch 2.4.1 compatibility reports. No pull requests are present.", + "confidence": 0.68, + "canonical_issue_reason": "issue:43596 is the clearest representative of the largest high-similarity cluster: the DeepSpeed ZeRO-3 + BertModel index-out-of-bounds failure, with a near-duplicate in 43638 and no broader, more generic framing than needed.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43596 is the strongest global issue representative because it names the concrete failure mode and model path cleanly, and it anchors the most convincing duplicate cluster in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Related model family, but different failures: missing config attribute vs streaming chat/completions bug." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different hardware/runtime problems on different backends (AMD GPU vs MPS Int histogram)." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different model-specific dtype mismatch vs AMD GPU execution failure." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Autocast dtype mismatch in Phi MoE is not the same bug as GPT-OSS failing on AMD GPUs." + }, + { + "left": "issue:45237", + "right": "issue:46468", + "accept": false, + "reason": "Different models and symptoms; not the same concrete bug." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Both are MoE-related, but the failure modes and code paths are different." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch and ONNX export error are unrelated issues." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "Different model/export problems with no shared concrete failure path." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ONNX export failure and device-mapping CUDA assert are distinct bugs." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "Task crash and outdated RAG examples are not the same underlying issue." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are v5 docs regressions, but they concern different removed references and are not one bug/change." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "Export-to-ONNX and GenerationMixin import failure are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:46165", + "accept": false, + "reason": "Different training/trainer problems; no shared code-path bug is evident." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask-shape IndexError and Gemma4 device-mismatch are different forward-path failures." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both involve errors during loading/training, but the concrete failures are different and not clearly the same bug." + }, + { + "left": "issue:43638", + "right": "issue:46165", + "accept": false, + "reason": "Zero-3 Bert loading error vs resume-training issue are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Different concrete failures: Bert zero-size index error vs mask/tensor shape mismatch." + }, + { + "left": "issue:44805", + "right": "issue:46294", + "accept": false, + "reason": "Unrelated issues: tensor masking vs video chat-template feature/token mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Different subsystems and failure modes." + }, + { + "left": "issue:46165", + "right": "issue:46468", + "accept": false, + "reason": "Resume-training with local model is unrelated to the Nemotron example breakage." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Video StopIteration and training-resume problems are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:46165", + "accept": false, + "reason": "Different bugs: Zero-3/Bert loading vs resume-training with trust_remote_code." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice preset and ModernBERT ONNX export are unrelated." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both are DeepSpeed-related, but the concrete failures and code paths differ." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same DeepSpeed ZeRO-3/BertModel index-out-of-bounds failure; 43638 is just a non-pretrained Bert variant of the same bug." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger kernel cross-entropy call and GPT-OSS TP failure are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Wrong detokenization and LoRA merge/save divergence are different bugs." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "NVFP4 runtime failure and LoRA save/load mismatch are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "FSDP activation-checkpointing shape recomputation and mask index mismatch are not the same bug." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Both mention DeepSpeed, but one is a weight-dimension error and the other is a collective mismatch." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": true, + "reason": "Same Qwen3_5MoeTextRotaryEmbedding buffer/materialization bug; both failures come from mishandled inv_freq/device state." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "Zero-3 Bert loading error vs GPT-OSS tensor-parallel issue are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "Activation-checkpointing/FSDP mismatch and CustomTrainer signature breakage are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Different model families and different failure points." + }, + { + "left": "issue:45874", + "right": "issue:46294", + "accept": false, + "reason": "Inputs_embeds reverse lookup issue and video chat-template mismatch are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Different GPU/runtime issues with no shared concrete code path." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3/vLLM config regression and Mistral tokenizer patch crash are unrelated." + }, + { + "left": "issue:45800", + "right": "issue:46515", + "accept": true, + "reason": "Both describe the same torch 2.4.1 compatibility regression; one is simply a more specific transformers version report." + }, + { + "left": "issue:45823", + "right": "issue:45874", + "accept": false, + "reason": "Device-map forward mismatch and expensive reverse embedding lookup are different Gemma4 issues." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "Torch 2.4.1 incompatibility and Qwen-3 NVFP4 runtime failure are not the same bug." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP checkpointing mismatch and Qwen3 auto device-map CUDA assert are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Different dtype/hardware bugs on different models and backends." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "TRL SFT embedding unpacking and MPS histogram error are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger-kernel cross-entropy routing and Phi autocast dtype mismatch are different bugs." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and trainer loss-signature incompatibility are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP activation-checkpointing issue and GenerationMixin import failure are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Cross-entropy dispatch issue and tensor mask shape mismatch are different failures." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "GPU inference failure and RAG docs cleanup are unrelated." + }, + { + "left": "issue:43866", + "right": "issue:46402", + "accept": false, + "reason": "Checkpoint corruption and missing text encoder weight loading are different bugs." + }, + { + "left": "issue:45800", + "right": "issue:46304", + "accept": false, + "reason": "Both are compatibility regressions, but one is torch-version specific while the other is a USE_DISTRIBUTED=0 breakage; not the same concrete bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: it sticks to issue titles, correctly notes that all items are issues, and only elevates clusters that are supported by the stronger similarity pairs. The accepted soft-edge pairs are plausible duplicate clusters, and the rejected pairs are conservative rather than overstated. " + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 2, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:36010", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:43039", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43792", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44387", + "issue:44512", + "issue:44560", + "issue:44589", + "issue:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45161", + "issue:45237", + "issue:45362", + "issue:45464", + "issue:45685", + "issue:45800", + "issue:45812", + "issue:45823", + "issue:45901", + "issue:45920", + "issue:46015", + "issue:46294", + "issue:46304" + ], + "soft_pairs": [ + "issue:43296|issue:46304", + "issue:41762|issue:43596", + "issue:39290|issue:41720", + "issue:44589|issue:46304", + "issue:43901|issue:44512", + "issue:45823|issue:46294", + "issue:45237|issue:45901", + "issue:44841|issue:45823", + "issue:43039|issue:44560", + "issue:43975|issue:46015", + "issue:44863|issue:46015", + "issue:45020|issue:46304", + "issue:45812|issue:45920", + "issue:43866|issue:44863", + "issue:44292|issue:46015", + "issue:43872|issue:46304", + "issue:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43854|issue:43866", + "issue:43792|issue:45901", + "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:43404|issue:43866", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:44841|issue:45084", + "issue:36010|issue:39290", + "issue:41093|issue:41762", + "issue:45800|issue:45823", + "issue:43296|issue:43366", + "issue:43572|issue:45070", + "issue:44805|issue:44918", + "issue:44387|issue:45005", + "issue:43541|issue:45161", + "issue:43828|issue:45464" + ], + "prepared_review_unit_hash": "9d41620646e18ede194c2c5a2491f9efc7ad563dc74eebbab6077f6ea7945423", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 41, + "serialized_chars": 39736, + "estimated_input_tokens": 9934, + "estimated_eval_tokens": 20124 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly heterogeneous. The only supported duplicate subgroups are the ZeRO-3 loading IndexErrors (issue:41762 / issue:43596) and the v5 AutoTokenizer wrong-token-ID regression (issue:45812 / issue:45920). The mask/tensor-shape IndexError reports are not strong enough to merge and should remain separate.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45920 is the best canonical anchor for the accepted tokenizer-regression subgroup because it is broader and more current, and it clearly covers the same v5 AutoTokenizer wrong-ID bug seen in issue:45812.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the strongest overall issue in the set because it captures a cross-model v5 tokenizer regression rather than a narrow model-specific instance.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:46304", + "accept": false, + "reason": "Different failures: PaddleOCR-VL loading in vLLM versus a USE_DISTRIBUTED=0 support breakage." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": true, + "reason": "Same ZeRO-3 loading path and same out-of-bounds index-0 failure; these look like one shared bug." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Both involve runtime/model-loading errors, but the concrete bugs differ: Gemma3 sliding_window_pattern vs Qwen3 auto device mapping cuda assert." + }, + { + "left": "issue:44589", + "right": "issue:46304", + "accept": false, + "reason": "Float8 storage lookup failure and USE_DISTRIBUTED=0 incompatibility are unrelated runtime problems." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs-related, but they cover different topics and do not share the same underlying change or bug." + }, + { + "left": "issue:45823", + "right": "issue:46294", + "accept": false, + "reason": "Gemma4 device-map mismatch and Qwen3.5 video/chat-template mismatch are different code paths." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "AMD GPU support for GPT-OSS-20B and table-question-answering crashes are unrelated." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Processor failure for Voxtral and Gemma4 device mismatch are distinct model-specific issues." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross_entropy dispatch and Qwen3-vl video StopIteration are different bugs." + }, + { + "left": "issue:43975", + "right": "issue:46015", + "accept": false, + "reason": "Incorrect detokenization in DeepSeek-Coder v5 is unrelated to outdated RAG docs." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint loading failure and RAG documentation are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:46304", + "accept": false, + "reason": "Remote_code regressions and distributed-build support are different failure modes." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same v5 AutoTokenizer regression producing wrong token IDs across multiple model families; 45920 is the broader tracker." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Corrupted Ovis2 checkpoint and NemotronH checkpoint loading failure are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime failure and RAG docs are unrelated." + }, + { + "left": "issue:43872", + "right": "issue:46304", + "accept": false, + "reason": "bitsandbytes Int8Params incompatibility is unrelated to the distributed-build support regression." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention sliding-window behavior, but the reported bugs are different Gemma3 and Qwen3-MoE issues." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both are IndexErrors, but the mask/tensor-shape mismatch and the ZeRO-3 zero-sized-tensor load failure do not clearly share one root cause." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Failed model loading in unit tests versus a corrupted checkpoint are different problems." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Whisper-large-v2 failing to run and table-question-answering crashing are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Incorrect detokenization and table-question-answering crash are unrelated." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "Checkpoint loading failure and docs issue are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:46304", + "accept": false, + "reason": "Remote_code breakage and distributed-build support are different issues." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Both involve Qwen3.5, but the symptoms and code paths differ: chat crash versus streaming API failure." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime failure and table-question-answering crash are unrelated." + }, + { + "left": "issue:39290", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3/vLLM sliding-window config and Gemma4 device-mapping mismatch are different bugs." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Tied lm_head weights in Mistral3 and a corrupted Ovis2 checkpoint are unrelated." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 streaming inference failure and MPS histogram on Int in a MoE path are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Grouped_mm tracing failure and MPS histogram error are different MoE-related symptoms, but not the same underlying bug." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Torch dynamo tracing grouped_mm failure and autocast dtype mismatch are distinct runtime issues." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Voxtral processor failure and 'Can't compile non template nodes' are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "ImportError for GenerationMixin and Gemma3 sliding_window_pattern are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are IndexErrors, but one is a mask/tensor shape mismatch and the other is a ZeRO-3 zero-sized-tensor load failure; not the same bug." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Torch 2.4.1/transformers 5.8 incompatibility and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "PaddleOCR-VL load failure and GGUF gpt-oss support are different problems." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig and the PretrainedConfig pydantic break are unrelated compatibility issues." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Both are loading/shape-related failures, but the mask/tensor-shape IndexError and Qwen3.5 embedding unpacking with TRL SFT trainer are different bugs." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Int4 quantization OOM and tied-weights regression on translation models are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Grouped_mm tracing failure and GPT-OSS MoE TP-only failure are different code-path problems." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Autocast dtype mismatch on Phi-tiny-MoE and Qwen3.5 streaming API failure are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary overstates the evidence. The packet supports two accepted duplicate subgroups (ZeRO-3 loading IndexErrors and the v5 AutoTokenizer wrong-token-ID regression), but it does not support calling the mask/tensor shape IndexError cases 'duplicate-worthy' because the corresponding soft verdicts are explicitly rejected. That makes the cluster summary inconsistent with the soft-edge decisions. The other verdicts look conservative, but this overclaim is enough to reject." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 3, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:36331", + "issue:39186", + "issue:39692", + "issue:40990", + "issue:41093", + "issue:41762", + "issue:42915", + "issue:43054", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43541", + "issue:43596", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43873", + "issue:43950", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44360", + "issue:44479", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45183", + "issue:45399", + "issue:45464", + "issue:45507", + "issue:45632", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45902", + "issue:46294", + "issue:46463", + "issue:46489", + "issue:46491" + ], + "soft_pairs": [ + "issue:44291|issue:45070", + "issue:43366|issue:45464", + "issue:43531|issue:45070", + "issue:43716|issue:46294", + "issue:45020|issue:45800", + "issue:43541|issue:44589", + "issue:44479|issue:46294", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45759|issue:45902", + "issue:39186|issue:44918", + "issue:43716|issue:44918", + "issue:40990|issue:44960", + "issue:43645|issue:45399", + "issue:45632|issue:45698", + "issue:43404|issue:45005", + "issue:42915|issue:45399", + "issue:44530|issue:45005", + "issue:43873|issue:45005", + "issue:44863|issue:45005", + "issue:44360|issue:44512", + "issue:45183|issue:46463", + "issue:46489|issue:46491", + "issue:44918|issue:45507", + "issue:44492|issue:44512", + "issue:43975|issue:44292", + "issue:44292|issue:45507", + "issue:43950|issue:44534", + "issue:39692|issue:43054", + "issue:43644|issue:43950", + "issue:36331|issue:41762" + ], + "prepared_review_unit_hash": "e12b4df5435879803efacc37c6a0cce9e1c76f7de7fb3bbe459a984e93512aa2", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36940, + "estimated_input_tokens": 9235, + "estimated_eval_tokens": 18726 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous rather than one single duplicate group. The strongest duplicate candidates are the non-persistent-buffer regression, the SigLIP2 degradation reports, the T5/input validation error-message reports, and the Qwen video-input regression reports.", + "confidence": 0.72, + "canonical_issue_reason": "issue:43950 is the clearest, most actionable representative of the non-persistent-buffer regression, and it has direct duplicate support from nearby reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43950 is the best issue anchor in this set because it describes a concrete v5 regression with strong duplicate evidence and a clear fix surface.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Different regressions: init_empty_weights argument handling vs pydantic PretrainedConfig field behavior." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "Different areas: GGUF architecture support vs chat/completions streaming failure." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to the pydantic config regression." + }, + { + "left": "issue:43716", + "right": "issue:46294", + "accept": false, + "reason": "Both are multimodal, but one is dtype mismatch and the other is a video token/feature count failure." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Remote_code breakage and torch/transformers version incompatibility are different bugs." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Different concrete failures: torch dynamo tracing grouped_mm vs missing Float8 storage type." + }, + { + "left": "issue:44479", + "right": "issue:46294", + "accept": true, + "reason": "Both report the same Qwen video-input regression class, centered on video feature/token mismatch in Qwen3.5-family handling." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are IndexErrors, but the masks-shape mismatch and zero3/BertModel loading failure are different code paths." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": true, + "reason": "Both are SigLIP2 reports describing bad or nonsensical outputs for the same model family, likely the same underlying model/processor issue." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "One is composite-config unwrapping, the other is uninitialized rotary embedding memory after materialization." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP weight-shape failure and Qwen3.5/TRL embedding unpacking are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Both touch multimodal inputs, but the failures and model families are different." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "Perplexity issue on gpt-oss and GLM5 issue are unrelated." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Custom model Jupyter initialization and flash-attn fallback gating are different bugs." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": false, + "reason": "Both involve remote-code loading/caching, but the titles point to different failure modes and not the same concrete bug." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Both mention tied weights, but one is Mistral3 image-text generation and the other is translation-model v5 behavior; not enough to treat as the same bug." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Qwen3Moe FP8 failure and flash-attn fallback gating are unrelated." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is unrelated to tied-weights handling." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior is a different subsystem from tied-weight regressions." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading is unrelated to translation-model tied weights." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "DSA indexing discussion and docs cleanup are not duplicates." + }, + { + "left": "issue:45183", + "right": "issue:46463", + "accept": true, + "reason": "Both point at the same input validation/error-message path for decoder/input_ids vs inputs_embeds handling." + }, + { + "left": "issue:46489", + "right": "issue:46491", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different round-trip failures." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "TRL embedding unpacking and GraniteMoEHybrid invalid method calls are unrelated." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "A typo in cache strategies and a removed CLI docs mention are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "DeepSeek tokenizer detokenization and Qwen-3-8B-NVFP4 runtime error are different issues." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Different model families and different failure modes." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "These are the same non-persistent-buffer corruption regression in v5, described from two angles." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "Docs/example errors for SigLIP2 are not the same bug as degraded SigLIP2 text embeddings." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Exact same v5 non-persistent-buffer corruption regression; one report is the general symptom, the other the specific from_pretrained corruption." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Trainer loss-signature mismatch and Gemma3 ZeRO-3 loading failure are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the cluster is clearly heterogeneous, and the highlighted duplicate groupings are supported by the issue titles. The accepted soft-edge pairs are conservative enough given the title-level evidence, especially the non-persistent-buffer regression cases. No obvious overstatement." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 4, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:24643", + "issue:28530", + "issue:38175", + "issue:41553", + "issue:41762", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43493", + "issue:43643", + "issue:43782", + "issue:43824", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44368", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44589", + "issue:44661", + "issue:44910", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45356", + "issue:45399", + "issue:45488", + "issue:45520", + "issue:45663", + "issue:45685", + "issue:45701", + "issue:45715", + "issue:45823", + "issue:45859", + "issue:45902", + "issue:45920", + "issue:46315", + "issue:46489" + ], + "soft_pairs": [ + "issue:43828|issue:44928", + "issue:43872|issue:44589", + "issue:45920|issue:46489", + "issue:43950|issue:45902", + "issue:43866|issue:44403", + "issue:43054|issue:45701", + "issue:38175|issue:43054", + "issue:41553|issue:43883", + "issue:44910|issue:45859", + "issue:45663|issue:45902", + "issue:42947|issue:44387", + "issue:43054|issue:43493", + "issue:41553|issue:43881", + "issue:44403|issue:44451", + "issue:24643|issue:28530", + "issue:43975|issue:45520", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:43643|issue:45020", + "issue:41762|issue:42915", + "issue:45685|issue:45823", + "issue:43782|issue:44960", + "issue:43881|issue:43883", + "issue:43824|issue:43872", + "issue:44991|issue:45399", + "issue:45020|issue:45823", + "issue:44661|issue:45020", + "issue:42491|issue:44368", + "issue:43828|issue:45823", + "issue:45081|issue:45399", + "issue:45715|issue:46315", + "issue:45488|issue:46489", + "issue:45020|issue:45356" + ], + "prepared_review_unit_hash": "944140d44ebe252a3a454a93305560b9107fde31233c3c16a8ca337c5fae33a1", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37995, + "estimated_input_tokens": 9499, + "estimated_eval_tokens": 19254 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of v5-era tokenizer/model-loading regressions plus a few unrelated buffer, quantization, and training issues. Only a few soft links look like true duplicates; most are merely same-subsystem similarity.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45920 is the best umbrella issue here: it describes a broad v5 tokenizer regression affecting multiple model families, which matches the densest shared theme in the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the most representative and actionable issue in the set: it is concrete, cross-model, and clearly captures the recurring tokenizer-output regression pattern.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Different models and different failure modes: autocast dtype mismatch vs RLHF NaN/SDPA fallback." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Both are quantization-related TypeErrors, but they point to different incompatibilities (_is_hf_initialized vs Float8 storage lookup)." + }, + { + "left": "issue:45920", + "right": "issue:46489", + "accept": true, + "reason": "Both report a v5 tokenizer-output regression; 46489 looks like a model-specific gap in the same broader tokenization fix." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": true, + "reason": "Same underlying buffer/materialization bug: non-persistent buffer handling leading to corrupted or uninitialized buffer state." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Unrelated: checkpoint corruption vs generic loading noise." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "SigLIP2 embedding quality regression vs a general tokenization-version issue; not the same bug." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Same model family, but the reported symptoms are too different to confidently treat as the same concrete bug." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Different subsystems: tokenizer error message vs missing tied-weights attribute on a model class." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Different code paths: Flash Attention packed-sequence misuse vs CPU-offload incompatibility in rotary embedding." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Both involve loading/training internals, but one is an FSDP2 per-layer state bug and the other is buffer materialization corruption." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Both are memory-related, but one is gradient checkpointing/LoRA behavior and the other is int4 quantization reservation growth." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": true, + "reason": "Same SigLIP2 implementation discrepancy theme; 43493 reads like the root-cause report behind the bad embedding quality." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Different models and failures: bad tokenizer error message vs generic glm-4v-9b loading failure." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise is not the same as a specific model load failure." + }, + { + "left": "issue:24643", + "right": "issue:28530", + "accept": false, + "reason": "Completely different training problems: DeepSpeed weight shape error vs early-stopping metric configuration." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Tokenizer detokenization bug is unrelated to the Python 3.13 flash_attn import issue." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "One is about noisy loading logs; the other is a specific model-loading failure." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Both mention memory, but the code paths differ: LoRA checkpointing vs Qwen3 MoE training memory usage." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Remote-code broad regression vs missing fields in AutoConfig return object; too broad and not the same bug." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Different failure classes and different model stacks: ZeRO-3 load crash vs Qwen3 MoE FP8 failure." + }, + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "MPS histogram kernel limitation is unrelated to Gemma4 device placement mismatch." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only loading error is unrelated to GLM5." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Different models and different errors: loading failure vs missing all_tied_weights_keys." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Import error for a missing model class is unrelated to the bitsandbytes storage incompatibility." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer load failure and flash-attn fallback gating are separate issues." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "Broad remote_code regressions and a Gemma4 device-map mismatch are different bugs." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "A TOKENIZER_MAPPING_NAMES edge case is not the same as the broader remote_code regression." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Different LoRA/Qwen3-MoE compatibility complaints with no shared concrete failure path." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Autocast dtype mismatch and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer regex crash and flash-attn fallback gating are separate loading paths." + }, + { + "left": "issue:45715", + "right": "issue:46315", + "accept": false, + "reason": "Both are tokenizer performance regressions, but they affect different conversion directions and different loops." + }, + { + "left": "issue:45488", + "right": "issue:46489", + "accept": false, + "reason": "Both are DeepSeek tokenizer regressions, but the concrete failures and implicated code paths are different enough to keep separate." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Kimi-K2.5 tokenizer codec/regression details do not match the generic remote_code breakage report." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet: 45920 is indeed the broadest tokenizer-regression issue, and the accepted soft links stay within conservative same-bug/same-theme bounds. The rejected pairs are mostly justified by clearly different model stacks or failure modes. 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": 20, + "review_unit_index": 5, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43742", + "issue:43792", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43872", + "issue:43940", + "issue:43950", + "issue:44155", + "issue:44220", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44560", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44991", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45399", + "issue:45584", + "issue:45588", + "issue:45663", + "issue:45685", + "issue:45823", + "issue:45834", + "issue:45902", + "issue:45923", + "issue:46402", + "issue:46515" + ], + "soft_pairs": [ + "issue:43742|issue:45399", + "issue:43404|issue:43950", + "issue:43827|issue:44509", + "issue:43872|issue:44291", + "issue:42915|issue:43872", + "issue:43792|issue:44220", + "issue:42947|issue:45923", + "issue:41553|issue:44991", + "issue:36331|issue:41093", + "issue:43257|issue:45305", + "issue:43940|issue:45313", + "issue:43854|issue:44863", + "issue:44928|issue:45588", + "issue:41093|issue:44805", + "issue:44479|issue:44560", + "issue:41553|issue:44488", + "issue:44863|issue:46402", + "issue:24643|issue:30064", + "issue:44530|issue:44863", + "issue:43596|issue:46515", + "issue:43366|issue:43828", + "issue:45325|issue:45902", + "issue:43638|issue:46515", + "issue:45663|issue:45823", + "issue:43366|issue:43531", + "issue:44155|issue:45584", + "issue:43716|issue:44560", + "issue:43950|issue:45834", + "issue:44560|issue:45685" + ], + "prepared_review_unit_hash": "4f565990ec9c1c898172dc412bfc6cf6809d585291badaebdb39a31f22048722", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36679, + "estimated_input_tokens": 9170, + "estimated_eval_tokens": 18596 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly heterogeneous issue set with a few true duplicate pairs; no single cluster-wide canonical issue/PR emerges.", + "confidence": 0.79, + "canonical_issue_reason": "No single canonical issue: the items split into several unrelated bug families, with only a few clear duplicate subgroups.", + "canonical_pr_reason": null, + "best_issue_reason": "No single best issue across the whole set; the strongest duplicate subgroups are separate and model-specific rather than one shared bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "Different failures: tokenizer/model loading vs flash-attn fallback gating." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Different bug classes: lm_head tying vs non-persistent buffer corruption." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Same docs regression: both still reference removed v5 pipeline tasks." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Same `_is_hf_initialized`/`Int8Params` incompatibility during empty-weight model loading." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Qwen3-MoE FP8 failure is unrelated to the bitsandbytes init-arg incompatibility." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": true, + "reason": "Both point to the Whisper/fbank feature extraction path failing." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Gradient checkpointing behavior issue vs missing model flag; not the same concrete bug." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Different tokenizer-loading problems for different models." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Trainer kwarg mismatch is unrelated to the mask/tensor shape IndexError." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Both involve DeepSpeed but the failure modes are different: missing weights vs gradient averaging." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": true, + "reason": "Same ZeRO-3 weight-loading failure pattern in Qwen-family models with missing params." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Different model-loading failures for different architectures." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "RLHF NaNs from position_id handling are unrelated to flash_attention.py crashing on s_aux=None." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "Same mask-shape mismatch IndexError in tensor indexing, just with different reported sizes." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": true, + "reason": "Same v5.3.0 video-input regression family in Qwen VL models." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "A bad error message and a tokenizer load failure are not the same underlying bug." + }, + { + "left": "issue:44863", + "right": "issue:46402", + "accept": false, + "reason": "NemotronH checkpoint loading and Sam3Model text-encoder loading are unrelated." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "DeepSpeed weight-dimension error is unrelated to segmentation-map preprocessing." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3.5 PagedAttention crash and NemotronH checkpoint loading are different problems." + }, + { + "left": "issue:43596", + "right": "issue:46515", + "accept": false, + "reason": "Specific ZeRO-3 BertModel index error is too different from a vague torch-version incompatibility report." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "GGUF architecture support and autocast dtype mismatch are unrelated." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Different rotary-embedding issues with different root causes." + }, + { + "left": "issue:43638", + "right": "issue:46515", + "accept": false, + "reason": "Non-pretrained Bert/ZeRO-3 index error is not the same as a generic torch incompatibility." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "FSDP2 shared_kv_states KeyError and PLE device mismatch are different Gemma4 bugs." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "GGUF support for gpt-oss is unrelated to Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Track-to-track embedding leak and empty-transcription generation failure are different audio bugs." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Image-preprocessor dtype mismatch is unrelated to the Qwen3-VL video StopIteration regression." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Non-persistent buffer corruption and Kosmos2.5 OCR index errors are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Video-processing regression and MPS histogram int support are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The rejected soft edges are mostly clearly distinct bug classes, and the accepted pairs are plausible duplicate families rather than overstatements. No single cluster-wide canonical issue emerges from these items." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 6, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43720", + "issue:43824", + "issue:43957", + "issue:43994", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:44991", + "issue:45030", + "issue:45276", + "issue:45313", + "issue:45357", + "issue:45507", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:46459", + "issue:46468", + "issue:46489", + "issue:46512", + "issue:46515" + ], + "soft_pairs": [ + "issue:43824|issue:43957", + "issue:44683|issue:46515", + "issue:43366|issue:44589", + "issue:43645|issue:43646", + "issue:42915|issue:43957", + "issue:43550|issue:44910", + "issue:44560|issue:44918", + "issue:42617|issue:43366", + "issue:45812|issue:46489", + "issue:45823|issue:46515", + "issue:36010|issue:42915", + "issue:44661|issue:45823", + "issue:44560|issue:44805", + "issue:38175|issue:43994", + "issue:43632|issue:45834", + "issue:44877|issue:45030", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:44661|issue:44805", + "issue:43317|issue:44756", + "issue:43644|issue:44534", + "issue:38175|issue:39692", + "issue:42491|issue:43720", + "issue:41720|issue:42915", + "issue:46459|issue:46512", + "issue:42915|issue:44756", + "issue:45313|issue:45357", + "issue:44451|issue:44488", + "issue:44960|issue:44977", + "issue:42491|issue:45276", + "issue:45507|issue:46468", + "issue:43644|issue:43645", + "issue:44488|issue:44991", + "issue:39692|issue:43994" + ], + "prepared_review_unit_hash": "66f93280ed0ba2437630870028f11e9c069ff3339f267a27e50997170f467a03", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37623, + "estimated_input_tokens": 9406, + "estimated_eval_tokens": 19068 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous. The strongest duplicate pair is 43644/44534 (near-exact same Transformers v5 buffer-corruption regression), and 43645/43646 also look like the same custom-model initialization bug. The SigLIP2 pair 38175/43994 is not conservatively duplicate from titles alone and should be rejected.", + "confidence": 0.88, + "canonical_issue_reason": "43644 is the clearest canonical issue: it states the broad Transformers v5 non-persistent-buffer regression and has a near-exact duplicate in 44534.", + "canonical_pr_reason": null, + "best_issue_reason": "43644 is the best representative because it captures the most reusable root-cause description and anchors the strongest duplicate pair in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Different failures: a missing Qwen2.5-VL import vs meta-device loading breakage for other models." + }, + { + "left": "issue:44683", + "right": "issue:46515", + "accept": false, + "reason": "Both involve torch, but the failure modes and affected code paths are unrelated." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "GGUF/gpt-oss support and missing Float8 storage are different problems." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Both report the same custom-model initialization regression in Transformers 5.0.0; 43645 is just the notebook-specific wording of the same underlying bug." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Qwen3Moe FP8 failures and meta-device loading failures are different underlying bugs." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different models and different attention/position-id failure modes." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Both involve Qwen3.x, but one is a video StopIteration error and the other is an embedding/unpacking trainer issue." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py runtime failure is unrelated to GGUF gpt-oss support." + }, + { + "left": "issue:45812", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer regressions, but for different model families and likely different fixes." + }, + { + "left": "issue:45823", + "right": "issue:46515", + "accept": false, + "reason": "Gemma4 device mismatch and stale config dtype are not the same bug." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "Importing GenerationMixin and Qwen3Moe FP8 loading are unrelated." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "Tokenizer-mapping failure and Gemma4 PLE device mismatch are different issues." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Video StopIteration and mask-shape IndexError are distinct failures." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": false, + "reason": "Both mention SigLIP2, but the evidence is too weak: one is about zero probabilities and the other about nonsensical outputs, which could be different symptoms or different root causes." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "_is_hf_initialized regression and Kosmos2.5 OCR index error are unrelated." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Both mention config validation, but they are different models and not clearly the same underlying bug." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Zero-3 BertModel indexing and add-new-model-like tokenizer mapping are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Zero-3 BertModel indexing and tokenizer-mapping handling are different bugs." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping failure and mask-shape IndexError do not share a concrete code path." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "gpu+cpu offload loading and Strix Halo mmap OOM are different memory/loading problems." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Near-exact duplicate reports of the same Transformers v5 non-persistent-buffer junk-fill regression." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "SigLIP2 wrong outputs and a documentation example with model/processor mismatch are not the same report." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Qwen3 MoE LoRA incompatibility and BitNet packed-weight loading are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Qwen3 auto-device-map CUDA assert and Qwen3Moe FP8 failure are different issues." + }, + { + "left": "issue:46459", + "right": "issue:46512", + "accept": false, + "reason": "Composite-checkpoint dtype handling and stale config dtype after from_config are related in theme but not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "FP8 config failure and mmap OOM are different code paths." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "One is ZeRO-3 weight loading for language_model, the other is a save_pretrained visual-encoder key regression." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading failures, but for different models and without enough evidence of one root cause." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 and Qwen3.5 flash-attention generation failures are unrelated." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "LoRA training incompatibility and resize_token_embeddings behavior are different bugs." + }, + { + "left": "issue:45507", + "right": "issue:46468", + "accept": false, + "reason": "GraniteMoEHybrid invalid method calls and Nemotron example breakage are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Related Transformers v5 regressions, but not the same concrete bug: buffer corruption vs custom model initialization." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Different model-loading regressions for different model families." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Documentation example errors are not the same as the SigLIP2 inference-output bug." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary is mostly grounded, but one soft-edge acceptance overreaches: issue:38175 vs issue:43994 are both about SigLIP2, yet the packet only gives titles and they describe different symptoms (zero probabilities vs nonsensical outputs). That is not enough evidence to conservatively call them a true duplicate/near-duplicate pair. The 43644/44534 pair is well supported, and 43645/43646 is plausible, but the SigLIP2 acceptance makes the overall verdict 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": 20, + "review_unit_index": 7, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "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:43708", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:44075", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44488", + "issue:44756", + "issue:44863", + "issue:44910", + "issue:44912", + "issue:44960", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45588", + "issue:45663", + "issue:45685", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:46050", + "issue:46468" + ], + "soft_pairs": [ + "issue:43708|issue:44075", + "issue:43856|issue:43873", + "issue:43854|issue:43931", + "issue:43257|issue:43854", + "issue:43856|issue:44387", + "issue:42491|issue:45292", + "issue:44960|issue:46050", + "issue:45488|issue:45812", + "issue:45362|issue:45406", + "issue:42491|issue:43931", + "issue:45276|issue:45335", + "issue:44164|issue:44291", + "issue:44910|issue:45588", + "issue:43404|issue:43454", + "issue:41720|issue:44155", + "issue:43425|issue:43957", + "issue:44402|issue:44488", + "issue:39692|issue:44960", + "issue:42947|issue:45704", + "issue:43873|issue:44387", + "issue:36010|issue:39404", + "issue:45663|issue:45759", + "issue:44291|issue:44756", + "issue:45663|issue:45859", + "issue:44155|issue:45823", + "issue:44292|issue:44912", + "issue:44387|issue:44756", + "issue:43957|issue:44756", + "issue:43335|issue:45663", + "issue:42947|issue:43381", + "issue:45276|issue:45292", + "issue:44863|issue:46468", + "issue:45685|issue:46468", + "issue:43854|issue:45216", + "issue:44164|issue:44756" + ], + "prepared_review_unit_hash": "7166774048116729263e3682822d60550b36b90ad49f56dc75f5963e9acd16d1", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38263, + "estimated_input_tokens": 9566, + "estimated_eval_tokens": 19388 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a collection of unrelated bug reports. The only strong duplicate subcluster is around `resize_token_embeddings` failing to update model-specific embedding/output weights across Gemma4/T5Gemma variants.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 45292 is the broadest, most general statement of the `resize_token_embeddings` regression (`output_embeddings`), making it the best canonical representative of that duplicate subcluster.", + "canonical_pr_reason": "No pull requests were provided in the input, so there is no canonical PR candidate.", + "best_issue_reason": "Issue 45292 best captures the shared symptom at the highest level and can represent the accepted `resize_token_embeddings` duplicates.", + "best_pr_reason": "No pull requests were provided in the input, so there is no global best PR candidate.", + "soft_edge_verdicts": [ + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Different trainer resume-step calculation bug vs optimizer argument handling; unrelated code paths." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "One is Qwen3 MoE training memory use, the other is quantization/offloading behavior; not the same bug." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different model-loading failures for different models and symptoms; no shared concrete root cause visible from titles." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Qwen3 MoE accelerate/deepspeed conversion issue vs GLM-4.7 loading problem; unrelated." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training usage and the other is int4 reserved-memory/OOM regression; not the same issue." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "LoRA checkpoint compatibility on hf5.x vs `resize_token_embeddings` output embedding bug; different failures." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "GLM5 loading/setup issue vs sinq quantized save/load regression; different code paths." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different model families and appear to have different root causes." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Qwen3.5 chat crash vs Gemma4Processor missing `_tokenizer`; unrelated." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "Same broad area of model loading, but different models, symptoms, and likely causes." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": true, + "reason": "Both report `resize_token_embeddings` failing to propagate to model-specific embeddings/output layers; same regression family and plausibly one fix." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Saving/from_pretrained extra_state handling vs `init_empty_weights` unexpected argument; different lifecycle bugs." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Flash-attention crash from 3D `position_ids` vs sink-less `s_aux=None` AttributeError; different failure modes." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both involve tying `lm_head`, but they are different model classes and not clearly the same underlying bug." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Qwen3 device-map CUDA assert vs AudioFlamingo3 batched inference token leak; unrelated." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch version compatibility vs `torch.device('meta')` model-loading breakage; different problems." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Tokenizer vocab-size mismatch vs a specific model load failure; not enough evidence of one root cause." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Documentation example errors vs GLM5 model issue; unrelated." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing/PEFT behavior vs apex RMSNorm memory leak; different subsystems." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Quantization/offloading expectation mismatch vs CUDA reserved-memory OOM regression; related theme but not the same bug." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "Import error for `GenerationMixin` vs Whisper pipeline `return_language`; different issues." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Gemma4 FSDP2 attention state bug vs Qwen multimodal `text_config` unwrapping bug; unrelated." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "`init_empty_weights` argument handling vs mmap/OOM on Strix Halo; different code paths." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Gemma4 FSDP2 state rebuild bug vs Qwen3_5Moe rotary embedding CPU-offload incompatibility; unrelated." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "AudioFlamingo3 cross-track leakage vs Gemma4 PLE device mismatch; not the same bug." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Different quantized model-loading failures with different models and behavior." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Both can OOM, but one is an int4 memory regression and the other is mmap-related Strix Halo handling; distinct causes." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "`meta` device loading regression vs mmap/OOM workaround; unrelated." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation bug vs Gemma4 attention state bug; different models and failures." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is ineffectiveness with PEFT LoRA and the other is eval-mode restriction; not duplicates." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same `resize_token_embeddings` regression: model-specific embeddings/output weights are not updated after resize." + }, + { + "left": "issue:44863", + "right": "issue:46468", + "accept": false, + "reason": "NemotronH checkpoint loading support vs a broken official example; too different." + }, + { + "left": "issue:45685", + "right": "issue:46468", + "accept": false, + "reason": "MPS MoE histogram dtype error vs Nemotron example breakage; unrelated." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Different Qwen/GLM load-vs-save regressions with no clear shared root cause." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "save/from_pretrained extra_state handling vs mmap/OOM workaround; different issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the only accepted soft-edge pair is supported by the titles as the same `resize_token_embeddings` regression family. The rejects are generally conservative and do not overclaim shared root cause. No PRs were provided, so the PR fields are appropriately null." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 8, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:39404", + "issue:42796", + "issue:43299", + "issue:43381", + "issue:43425", + "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:44683", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44928", + "issue:45084", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45701", + "issue:45812", + "issue:45823", + "issue:46050", + "issue:46315", + "issue:46326", + "issue:46491" + ], + "soft_pairs": [ + "issue:42796|issue:46326", + "issue:44479|issue:44877", + "issue:43381|issue:44928", + "issue:45684|issue:45698", + "issue:45588|issue:45663", + "issue:43646|issue:43708", + "issue:45292|issue:45335", + "issue:43645|issue:44164", + "issue:43828|issue:43957", + "issue:44402|issue:44451", + "issue:43425|issue:43828", + "issue:46315|issue:46491", + "issue:45812|issue:46315", + "issue:44589|issue:44756", + "issue:44683|issue:45823", + "issue:39404|issue:45701", + "issue:43611|issue:43645", + "issue:43425|issue:45823", + "issue:43425|issue:43716", + "issue:44530|issue:45084", + "issue:43906|issue:44530", + "issue:44479|issue:44805", + "issue:44265|issue:44898", + "issue:44589|issue:45084", + "issue:39404|issue:44912", + "issue:43901|issue:44509", + "issue:43425|issue:44918", + "issue:43906|issue:46326", + "issue:46050|issue:46326", + "issue:43994|issue:44079", + "issue:43299|issue:43931" + ], + "prepared_review_unit_hash": "e8e46bf3fc84a19ecce1eab7bde00e793d1f0a8183915214f622864b090fda4e", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36482, + "estimated_input_tokens": 9121, + "estimated_eval_tokens": 18498 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are related only at a broad subsystem level (tokenizers, model loading, save/load, or training) and do not look like the same concrete bug. The only clear duplicate-like pair is the resize_token_embeddings regression across generic and t5gemma-specific embeddings.", + "confidence": 0.82, + "canonical_issue_reason": "issue:45292 is the broader representation of the resize_token_embeddings regression; issue:45335 looks like the t5gemma-specific manifestation of the same missing embedding-resize behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45292 is the best issue representative here because it states the underlying API regression generically and can subsume the t5gemma variant.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42796", + "right": "issue:46326", + "accept": false, + "reason": "Both are tokenizer-related, but one is BPE marker handling and the other is continue_final_message behavior for mistral_common; different bugs." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Different model families and failure modes; video regression vs strict config loading." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "One is gradient checkpointing in eval mode, the other is Qwen3.5 RLHF NaN from 3D position_ids and SDPA fallback; unrelated." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Both touch save/load of custom models, but one is read-only permission propagation and the other is wrong module selection after save_pretrained; distinct bugs." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Different attention-path failures: flash_attention s_aux=None crash vs Gemma4 FSDP2 shared_kv_states KeyError." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model init in notebooks is unrelated to Trainer resume_from_checkpoint max_steps calculation." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same underlying regression: resize_token_embeddings fails to propagate to output embeddings/decoder.embed_tokens; the second is a model-specific instance." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Both are about custom-model workflows, but notebook initialization and extra_state save/from_pretrained handling are different issues." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Both are dtype-mismatch reports, but one is autocast on Phi-tiny-MoE and the other is meta-device loading on different models." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Related to tokenizer/model loading, but pad/unk vocab-size mismatch is not the same as ScandiBERT load failure." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch 2.10 compatibility is a version-level issue; Phi-tiny-MoE autocast dtype mismatch is a model-specific runtime bug." + }, + { + "left": "issue:46315", + "right": "issue:46491", + "accept": false, + "reason": "Both are tokenizer regressions, but one is performance of convert_tokens_to_ids and the other is a leading-space round-trip correctness bug." + }, + { + "left": "issue:45812", + "right": "issue:46315", + "accept": false, + "reason": "Granite token ID corruption is a correctness regression; convert_tokens_to_ids slowdown is a performance regression." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device-map mismatch are different code paths and symptoms." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper pipeline return_language and a generic tokenization-version complaint do not describe the same bug." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "base_model_prefix loading failure and notebook custom-model initialization are separate custom-model regressions." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "Torch version incompatibility is not the same as Gemma4 PLE device mismatch with device_map=auto." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Torch 2.10 incompatibility and Mistral-3 dtype mismatch with the image preprocessor are unrelated." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is unrelated to the non-template-node compilation error." + }, + { + "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:44479", + "right": "issue:44805", + "accept": false, + "reason": "Video-input regression for Qwen VL models and a generic mask/tensor shape mismatch are different failures." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export/torch_compilable_check failure and Perceiver interpolate_pos_encoding bug are distinct model/export problems." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup and template-node compilation errors are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper return_language pipeline regression is unrelated to MXFP4 quantization fallback on git-oss-20b." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Docs mentioning removed pipelines and return_all_scores behavior mismatch are related to docs/API drift but not the same concrete issue." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch 2.10 incompatibility is unrelated to Qwen3.5 embedding unpacking with TRL SFT trainer." + }, + { + "left": "issue:43906", + "right": "issue:46326", + "accept": false, + "reason": "An isolated reproduction of another issue and a mistral_common continue_final_message tokenizer bug are not the same failure." + }, + { + "left": "issue:46050", + "right": "issue:46326", + "accept": false, + "reason": "sinq save/load corruption and continue_final_message tokenization are unrelated bugs." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "Siglip2 pipeline nonsense outputs and ModelOutput key assignment are different problems." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both are Qwen3-VL loading issues, but one targets MoE loading broadly and the other is a specific weight-shape mismatch; not clearly the same concrete bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The mapping is grounded in the packet, and the soft-edge decisions are conservative overall. The only accepted pair, issue:45292 vs issue:45335, is a plausible same-bug duplicate around resize_token_embeddings not propagating to the relevant embedding modules. The other pairs are appropriately rejected as broader subsystem similarity rather than the same concrete bug." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 9, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:32090", + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42175", + "issue:42796", + "issue:42915", + "issue:43404", + "issue:43421", + "issue:43452", + "issue:43541", + "issue:43582", + "issue:43596", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43646", + "issue:43716", + "issue:43856", + "issue:43883", + "issue:44164", + "issue:44451", + "issue:44462", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44661", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44811", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45584", + "issue:45759", + "issue:45859", + "issue:46178", + "issue:46285", + "issue:46326", + "issue:46402", + "issue:46489" + ], + "soft_pairs": [ + "issue:36032|issue:43452", + "issue:44671|issue:44991", + "issue:44164|issue:45759", + "issue:43596|issue:44683", + "issue:44991|issue:45081", + "issue:43856|issue:44756", + "issue:42796|issue:43404", + "issue:44164|issue:44479", + "issue:44462|issue:46489", + "issue:39404|issue:44671", + "issue:44488|issue:45081", + "issue:44661|issue:44683", + "issue:44488|issue:44987", + "issue:44737|issue:45859", + "issue:43611|issue:43644", + "issue:32090|issue:39404", + "issue:43582|issue:43632", + "issue:46285|issue:46402", + "issue:43716|issue:45084", + "issue:36032|issue:42915", + "issue:40990|issue:43421", + "issue:43643|issue:44530", + "issue:34567|issue:40444", + "issue:44451|issue:44987", + "issue:42175|issue:46178", + "issue:43541|issue:45084", + "issue:44811|issue:45584", + "issue:42915|issue:44530", + "issue:43644|issue:43646", + "issue:40444|issue:46326", + "issue:43883|issue:44488" + ], + "prepared_review_unit_hash": "37fab4528c5005152a0c48e6e95e2907038065d33a74d6f75375167ff48bf1bc", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36474, + "estimated_input_tokens": 9119, + "estimated_eval_tokens": 18494 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is thematically related to Transformers v5 regressions, but the individual issues point to different concrete bugs across tokenizers, model loading, multimodal paths, and training/runtime edge cases. I did not find any soft pair that is clearly the same underlying bug.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43646 is the best representative issue because it describes a broad Transformers v5 initialization regression and sits closest to the cluster\u2019s recurring v5 breakage theme.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the strongest canonical issue candidate: broad, version-specific, and more representative than the narrower tokenizer/model-specific reports in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "Different failures: a T5 tokenizer method-name conflict vs gguf_file loading breakage for AutoTokenizer/AutoModelForCausalLM." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer/model loading complaints, but one is CamemBERT MLM prediction correctness and the other is loading EMBEDDIA/est-roberta tokenizer." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "Unrelated code paths: extra_state save/load handling vs AutoModelForCausalLM.from_config unwrapping composite Qwen text_config." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Different subsystems: DeepSpeed ZeRO-3/BertModel init error vs compiled flex_attention failure on torch >= 2.9." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Different tokenizer bugs: loading a tokenizer for EMBEDDIA/est-roberta vs Mistral regex patch crashing on tokenizer backend access." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Different performance/memory issues in unrelated contexts: Qwen3 MoE training memory use vs disabling mmap on Strix Halo to avoid OOM." + }, + { + "left": "issue:42796", + "right": "issue:43404", + "accept": false, + "reason": "One is a tokenizer BPE-marker handling bug; the other is a missing lm_head weight tie in Mistral3 multimodal generation." + }, + { + "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:44462", + "right": "issue:46489", + "accept": false, + "reason": "Both involve tokenizer loading/output, but one is ignoring tokenizer.json in AutoTokenizer while the other is a DeepSeek-Coder v1 output regression with a specific fix gap." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline return_language regression is unrelated to CamemBERT masked LM prediction errors." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both are model-loading related, but one is a repository-loading failure for cjvt/sleng-bert and the other is a Mistral regex patch crash." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Different bugs: add-new-model-like failure in tokenizer mapping vs flex_attention compilation failure." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Both are loading failures, but the affected models and failure modes are different; not the same underlying bug." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "XLNet CPU placement bug in relative positional encoding vs Qwen3_5 MoE rotary embedding incompatibility with CPU offload." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one affects base_model_prefix loading and the other non-persistent buffer initialization; different concrete bugs." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error is unrelated to Whisper pipeline return_language behavior." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon caching_allocator_warmup TypeError is unrelated to the _is_hf_initialized flag regression." + }, + { + "left": "issue:46285", + "right": "issue:46402", + "accept": false, + "reason": "Different model families and issues: CLIP structure inconsistency vs Sam3Model text encoder weights not loading." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Mistral-3 image preprocessor dtype mismatch is unrelated to the template-node compilation TypeError." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer special-token conflict and Qwen3 MoE FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "High perplexity on gpt-oss-20b is not the same bug as runtime post-processor updates when special tokens change." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "trust_remote_code missing fields is unrelated to PagedAttentionCache crashing on invalid group type." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "TrainerState token counting bug and Qwen2.5-VL iterable dataset multi-image finetuning failure are different problems." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Both are model-loading complaints, but they concern different models and likely different loader paths." + }, + { + "left": "issue:42175", + "right": "issue:46178", + "accept": false, + "reason": "Packaging/backend installation issue is different from the BACKENDS_MAPPING error for tensorflow_text." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "torch dynamo tracing failure in MixtralForCausalLM is unrelated to the non-template-node compilation error." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Whisper batch_decode skip_special_tokens behavior is different from empty-transcription generation after align_special_tokens." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Qwen3 MoE training with FineGrainedFP8Config is unrelated to PagedAttentionCache group-type crashes in Qwen3.5 models." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one is junk-filled non-persistent buffers and the other is custom model initialization; not the same concrete bug." + }, + { + "left": "issue:40444", + "right": "issue:46326", + "accept": false, + "reason": "Qwen2.5-VL iterable-dataset multi-image finetuning failure is unrelated to mistral_common continue_final_message behavior." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Molmo all_tied_weights_keys attribute error is unrelated to loading cjvt/sleng-bert." + } + ] + }, + "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": 20, + "review_unit_index": 10, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:28530", + "issue:34567", + "issue:36032", + "issue:41720", + "issue:42915", + "issue:43089", + "issue:43262", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43452", + "issue:43482", + "issue:43540", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43746", + "issue:43819", + "issue:43883", + "issue:43994", + "issue:44062", + "issue:44206", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45507", + "issue:45987", + "issue:46315", + "issue:46326" + ], + "soft_pairs": [ + "issue:44206|issue:44220", + "issue:43452|issue:43482", + "issue:36032|issue:43645", + "issue:43317|issue:43482", + "issue:45084|issue:45507", + "issue:44849|issue:45325", + "issue:44987|issue:44991", + "issue:44464|issue:45987", + "issue:43746|issue:46326", + "issue:43262|issue:43377", + "issue:43645|issue:44534", + "issue:43646|issue:44534", + "issue:43883|issue:44991", + "issue:43404|issue:44466", + "issue:44403|issue:45081", + "issue:42915|issue:45084", + "issue:43994|issue:44610", + "issue:43377|issue:43819", + "issue:28530|issue:34567", + "issue:43322|issue:43540", + "issue:43089|issue:45987", + "issue:44451|issue:45081", + "issue:44987|issue:45081", + "issue:44265|issue:44610", + "issue:42915|issue:44291", + "issue:44220|issue:44683", + "issue:45216|issue:45357", + "issue:36032|issue:43742", + "issue:44991|issue:46315", + "issue:44062|issue:44987", + "issue:42915|issue:44589", + "issue:44479|issue:45325", + "issue:36032|issue:41720" + ], + "prepared_review_unit_hash": "c5ca9f4e4080e9f5679f72c14b0ff9457fde5f9d5b88a4e567b76b560750358d", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37193, + "estimated_input_tokens": 9299, + "estimated_eval_tokens": 18854 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of unrelated Transformers v5 regression reports, with a common surface theme of loading/save_pretrained/tokenizer issues. Per the packet, none of the soft pairs are deterministic duplicates, so I reject all soft-edge merges conservatively.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 44991 is the best representative anchor for the cluster because it is a concrete, version-gated loading regression with a clear failure mode and broad enough scope to stand in for the loading-regression subcluster.", + "canonical_pr_reason": null, + "best_issue_reason": "44991 has the clearest standalone regression boundary and the most directly triageable repro among the issue set, so it is the strongest issue to use as the cluster\u2019s representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Both are audio-related, but they describe different concrete failures and do not look like the same bug." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": false, + "reason": "Both involve loading regressions, but one is a gguf_file/from_pretrained breakage and the other is a Qwen2.5-GGUF v5 failure; too weak to treat as duplicates." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer method-name conflict vs custom model initialization in Jupyter are different code paths and symptoms." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "Device_map/offload dequantized-model loading and Qwen2.5-GGUF loading are distinct failures." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Different subsystems and failure modes; no evidence they share one underlying bug." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Hidden-states behavior and rope/position-id scaling are not the same concrete issue." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are v5 loading regressions, but they affect different artifacts and failure mechanisms." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked generation inconsistency and StaticCache return-type issues are separate bugs." + }, + { + "left": "issue:43746", + "right": "issue:46326", + "accept": false, + "reason": "PEFT checkpoint loading and continue_final_message tokenizer handling are unrelated code paths." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio chat-template sampling-rate defaults and MIMI batch padding-mask behavior are different issues." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Custom model init in notebooks and non-persistent buffer junk are distinct regressions." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Custom model initialization breakage and buffer initialization problems do not look like one bug." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Missing tied-weights attribute and tokenizer/model loading regression are different failure modes." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Untied lm_head weights and serialization inconsistency are related in theme but not the same concrete defect." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Generic loading noise and a specific Mistral regex crash are not duplicates." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "FP8 model loading failure and template compilation error are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Wrong model outputs and wrong processor image size are separate bugs." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Missing padding-mask support and missing STE in DAC are different algorithmic problems." + }, + { + "left": "issue:28530", + "right": "issue:34567", + "accept": false, + "reason": "Early-stopping metric selection and token-count tracking are unrelated trainer bugs." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Llava loading segfault and Qwen3OmniMoe video ValueError are different code paths." + }, + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Generation overhead/per-token syncs is not the same as the cache-length type contract issue." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "A model-specific load failure and a tokenizer regex crash are different bugs." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer-loading regression and Mistral regex patch crash are unrelated." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "torch.export failure and processor size mismatch are distinct issues." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "FP8 config loading and init_empty_weights argument mismatch are different failures." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio fbank extraction and compiled flex_attention on torch>=2.9 are unrelated." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": false, + "reason": "Both concern Qwen3.5 save_pretrained regressions, but the packet does not support treating them as one concrete duplicate bug." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "T5 tokenizer special-token conflict and MobileLLM key error have different root causes." + }, + { + "left": "issue:44991", + "right": "issue:46315", + "accept": false, + "reason": "Tokenizer loading regression and slow-tokenizer performance regression are separate problems." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken special-argument duplication is a tokenizer-construction error, not the same as the model-loading regression." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "FP8 config failure and missing Float8 storage are different failures." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Video-input regression across Qwen variants and Qwen2.5-VL rope scaling bug are not the same underlying issue." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer construction error and Qwen3 device-mapping cuda assert are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the analyst overstates evidence in the soft-edge verdicts. In the packet, every listed soft pair has `deterministic_accept: false`, including 43452/43482 and 45216/45357, but the analyst marked both as accepted duplicates. That is not conservative and is not grounded in the packet. The overall summary is mostly plausible, but the duplicate-pair conclusions should be downgraded to rejected or at most uncertain." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 11, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:40444", + "issue:40990", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42796", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43381", + "issue:43421", + "issue:43425", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43746", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:44079", + "issue:44291", + "issue:44403", + "issue:44462", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44756", + "issue:44811", + "issue:44829", + "issue:44877", + "issue:44991", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45542", + "issue:45663", + "issue:45812", + "issue:45923", + "issue:46402", + "issue:46491" + ], + "soft_pairs": [ + "issue:43531|issue:44877", + "issue:43632|issue:44756", + "issue:43611|issue:44534", + "issue:40444|issue:45325", + "issue:42796|issue:43746", + "issue:43425|issue:44403", + "issue:43296|issue:43531", + "issue:45812|issue:46491", + "issue:43792|issue:44589", + "issue:44403|issue:44991", + "issue:33666|issue:40444", + "issue:43632|issue:44534", + "issue:40990|issue:44811", + "issue:43381|issue:45923", + "issue:44079|issue:45663", + "issue:43421|issue:44568", + "issue:44291|issue:44589", + "issue:41553|issue:43582", + "issue:35707|issue:40444", + "issue:43482|issue:45538", + "issue:43482|issue:43525", + "issue:43317|issue:43856", + "issue:44829|issue:45479", + "issue:42175|issue:45542", + "issue:43818|issue:46402", + "issue:43818|issue:45357", + "issue:44079|issue:44534", + "issue:44462|issue:45812", + "issue:43531|issue:44291", + "issue:43377|issue:45491", + "issue:41720|issue:45081" + ], + "prepared_review_unit_hash": "19ab70b8d7e5e2c3a4c0731c6cd787eb7e067845264f35d26524e7068641159b", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37141, + "estimated_input_tokens": 9286, + "estimated_eval_tokens": 18828 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a heterogeneous set of model- and framework-specific issues, not a true duplicate cluster. The soft pairs mostly share only broad subsystem or model-family similarity, so I reject them all.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45812 is the strongest anchor: it describes a broad, silent v4\u2192v5 tokenizer regression affecting all Granite models and has inbound references, making it the most representative issue in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 is the most actionable high-signal report in the group, with a clear regression description and broad impact across a model family.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Different bugs: Qwen3-MoE sliding-window handling vs GraniteSpeech config strictness." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "Unrelated failure modes: v5 init-flag regression vs Apple Silicon mmap/OOM behavior." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer initialization; distinct code paths." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both involve Qwen2.5-VL, but one is multi-image iterable training failure and the other is temporal position-id scaling in rope indexing." + }, + { + "left": "issue:42796", + "right": "issue:43746", + "accept": false, + "reason": "Tokenizer BPE marker handling vs GraniteSpeech PEFT checkpoint loading; not the same underlying bug." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch 2.10 compatibility and load-time noise are different issues." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "Different model families and symptoms: PaddleOCR-VL load failure vs Qwen3-MoE sliding_window bug." + }, + { + "left": "issue:45812", + "right": "issue:46491", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different behavior (Granite token IDs vs CodeLlama leading-space round-trip)." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper runtime failure and Float8 storage lookup error are unrelated." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Unnecessary load noise is not the same as EMBEDDIA tokenizer loading failure." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Multi-GPU training support and iterable-dataset multi-image prompt failure are distinct Qwen2-VL training problems." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Different regressions: missing _is_hf_initialized handling vs junk-filled non-persistent buffers." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "High perplexity on a language-model benchmark is unrelated to Whisper skip_special_tokens behavior." + }, + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "Eval-mode checkpointing restriction and missing gradient-checkpointing flag are related broadly, but not the same bug." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment and Gemma-4 shared_kv_states KeyError are different issues." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime post-processor updates and add_special_tokens BOS/EOS omission are separate tokenizer bugs." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument and missing Float8 storage object are unrelated TypeErrors." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad tokenizer error messaging is unrelated to the Apple Silicon warmup TypeError." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Progressive generation with inputs_embeds/past_key_values is not the same as multi-image iterable dataset training failure." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "GGUF loading failure and CLIPTokenizer model_max_length behavior are different bugs." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Qwen2.5-GGUF loading and missing pad_token_id on Llama4Config are unrelated issues." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Dequantized model offload loading failure and Qwen3 MoE training memory inefficiency are different code paths." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Similar training symptoms, but flash_attention_3 degeneration and num_labels=1 zero-loss are different root causes." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Both touch TensorFlow availability, but one is missing TF from the torch extra and the other is a tensorboard-only backend error." + }, + { + "left": "issue:43818", + "right": "issue:46402", + "accept": false, + "reason": "Video-LLaVA architecture/weights issues are unrelated to Sam3Model text-encoder loading." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Different model families and different save/load problems." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer junk are unrelated regressions." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Tokenizer.json being ignored is a different loading bug from Granite token-ID regression." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3-MoE sliding_window and init_empty_weights _is_hf_initialized issues are unrelated." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "MIMI padding-mask mismatch and Gemma3 all-padding-window NaNs are both batching issues, but not the same bug." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Qwen3 auto device mapping CUDA assert and Mistral regex patch crash 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": 20, + "review_unit_index": 12, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42796", + "issue:42915", + "issue:43262", + "issue:43296", + "issue:43454", + "issue:43482", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43701", + "issue:43746", + "issue:43749", + "issue:43792", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44877", + "issue:44898", + "issue:45072", + "issue:45325", + "issue:45381", + "issue:45584", + "issue:45701", + "issue:45800", + "issue:45823", + "issue:45834", + "issue:46209", + "issue:46305", + "issue:46315" + ], + "soft_pairs": [ + "issue:43644|issue:44079", + "issue:40444|issue:45381", + "issue:44265|issue:45072", + "issue:43611|issue:43646", + "issue:43749|issue:44079", + "issue:45325|issue:45381", + "issue:42796|issue:45381", + "issue:42175|issue:43296", + "issue:43994|issue:44898", + "issue:43792|issue:43957", + "issue:45325|issue:45834", + "issue:43454|issue:44466", + "issue:43632|issue:45072", + "issue:42915|issue:43482", + "issue:43701|issue:45325", + "issue:44291|issue:44877", + "issue:44479|issue:45823", + "issue:43296|issue:45800", + "issue:43873|issue:45701", + "issue:43606|issue:44265", + "issue:43262|issue:43746", + "issue:40990|issue:45584", + "issue:42796|issue:46209", + "issue:43881|issue:44488", + "issue:34634|issue:35707", + "issue:35707|issue:41720", + "issue:43454|issue:43550", + "issue:44186|issue:44265", + "issue:39692|issue:40444", + "issue:43606|issue:43819", + "issue:41720|issue:45823", + "issue:46305|issue:46315" + ], + "prepared_review_unit_hash": "c23834375bea6f07da82e8ffd4d1b31a66f7727b9a5fe18e3f0f2b6a831b68e3", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37185, + "estimated_input_tokens": 9297, + "estimated_eval_tokens": 18850 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous set of unrelated bugs. issue:43646 remains the best broad issue representative for the v5 init/load regression cluster, but the Qwen2.5-VL position-id reports are only related at a high level and should not be merged as the same bug.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43646 is the broadest representative of the repeated Transformers v5 initialization/loading regressions in this set, and it captures the shared failure mode better than the narrower model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the best overall issue candidate because it is the most central, general, and representative of the v5 init/load breakage cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Different bugs: tied/non-persistent buffer serialization vs ModelOutput key assignment." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen multimodal issues, but multi-image finetuning and video vision_position_ids are different code paths." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export with torch_compilable_check is unrelated to a bfloat16 dtype mismatch in inference." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Related v5 regressions, but base_model_prefix loading and custom model initialization are distinct failures." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and ModelOutput key handling are unrelated." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": false, + "reason": "Same Qwen2.5-VL position-id area, but still-image temporal scaling and video vision_position_ids are not enough evidence of one shared bug." + }, + { + "left": "issue:42796", + "right": "issue:45381", + "accept": false, + "reason": "Tokenizer BPE marker handling is unrelated to Qwen2.5-VL video position IDs." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Missing TensorFlow backend and PaddleOCR-VL loading in vLLM are different subsystems and failures." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Both are vision-model regressions, but they affect different models and different failure mechanisms." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Whisper runtime failure and meta-device loading regression do not describe the same bug." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Qwen2.5-VL position IDs and Kosmos2.5 OCR long-input indexing are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Both mention tied weights, but one is a tying bug and the other is serialization inconsistency." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "_is_hf_initialized breakage and bfloat16 dtype mismatch are separate problems." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different models and loading paths: Qwen3Moe FP8 vs Qwen2.5-GGUF v5 loading." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL position-id scaling." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "init_empty_weights argument handling and strict config loading are different issues." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Qwen video-position regression and Gemma4 device placement are not the same bug." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "PaddleOCR-VL load failure in vLLM is unrelated to torch/transformers version incompatibility." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Quantization offload behavior and tokenizer slowdown/version drift are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU offload device mismatch and torch.export failure are different code paths." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sampling rate default and GraniteSpeech PEFT checkpoint loading are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on one model and Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:42796", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer marker handling and FP8 inference on B200 are unrelated." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Two model-loading failures, but no evidence they share the same underlying bug." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "BarkProcessor voice preset and progressive generation with inputs_embeds/past_key_values are unrelated." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Different failures: generation cache handling vs Qwen3 device-map CUDA assert." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "Tied-weight text generation bug and Bamba SDPA compile failure are unrelated." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer padding/truncation crash and torch.export failure are different problems." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "SigLIP2 doc example issues and Qwen2.5-VL multi-image finetuning failure are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "CPU offload mismatch and DAC from_latents/forward mismatch are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both are device-related, but they affect different models and different failure modes." + }, + { + "left": "issue:46305", + "right": "issue:46315", + "accept": false, + "reason": "Streaming mask off-by-one and tokenizer conversion slowdown are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Rejecting because one soft-edge verdict is too strong for the evidence in the packet. `issue:45325` and `issue:45381` are certainly related, but the titles only show a shared Qwen2.5-VL / position-id theme: one is about still-image temporal position_ids being scaled incorrectly, the other about video input vision_position_ids being wrong. That is not enough on its own to conclude they are the same bug. The overall summary is mostly grounded, but this accepted duplicate-like pair is not conservative enough." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 13, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:26413", + "issue:28530", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42796", + "issue:42915", + "issue:42947", + "issue:43065", + "issue:43257", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43450", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43708", + "issue:43720", + "issue:43749", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44387", + "issue:44466", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45357", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45920", + "issue:45923", + "issue:46050", + "issue:46402" + ], + "soft_pairs": [ + "issue:42796|issue:44977", + "issue:43377|issue:44610", + "issue:44568|issue:45920", + "issue:43454|issue:45072", + "issue:43257|issue:45923", + "issue:43317|issue:43873", + "issue:43377|issue:44898", + "issue:40990|issue:42947", + "issue:43322|issue:46050", + "issue:43454|issue:43720", + "issue:43572|issue:45759", + "issue:41720|issue:43322", + "issue:43749|issue:44534", + "issue:43526|issue:43883", + "issue:42617|issue:42915", + "issue:44466|issue:45357", + "issue:43881|issue:44991", + "issue:41720|issue:44977", + "issue:26413|issue:28530", + "issue:44387|issue:45704", + "issue:40990|issue:45701", + "issue:36331|issue:42617", + "issue:43450|issue:45701", + "issue:36331|issue:42915", + "issue:44186|issue:45072", + "issue:44877|issue:44991", + "issue:43065|issue:46402", + "issue:43873|issue:44756", + "issue:43577|issue:43708", + "issue:43404|issue:45357", + "issue:43550|issue:43720" + ], + "prepared_review_unit_hash": "7866f5c7185d1ca17b5f68ed4b144495e5d3e5f0c74b94b0cbe10beea90aef80", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36501, + "estimated_input_tokens": 9126, + "estimated_eval_tokens": 18508 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose collection of mostly unrelated closed/open Transformers bugs, with only broad thematic overlap around tokenization, loading, offloading, and v5 regressions. None of the soft pairs look like the same concrete underlying bug or change, so all soft edges are rejected. The best cluster anchor is the broad tokenizer-regression report in issue 45920.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 45920 is the broadest, highest-signal regression in the set, covering wrong token IDs across multiple models and versions, so it is the best anchor for this loose cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45920 best represents the cluster because it is a concrete, high-impact v5 tokenizer regression with multiple affected model families and the widest scope among the candidates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42796", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer/BPE-marker bug vs flash-attention generation bug; same general area but different failure modes and code paths." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "MIMI padding-mask batching issue vs OmDet processor image-size mismatch; unrelated subsystems." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "One is special-token insertion for a specific tokenizer, the other is wrong token IDs across many models; not the same bug." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Weight tying bug vs dtype mismatch in bf16 inference; different underlying problems." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "MoE weight conversion/loading issue vs missing gradient-checkpointing flag on a trust_remote_code model." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both involve offloading/quantization, but one is a specific device_map load failure and the other is a broader offloading regression." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Batch padding-mask output mismatch vs Perceiver resolution/interpolation failure; unrelated model paths." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity regression vs gradient-checkpointing/LoRA behavior; no shared concrete bug." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Llava Next load crash vs sinq save/load breakage; different model families and failure modes." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Weight tying problem vs packed-weight unpacking during accelerate loading; not the same issue." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "Missing config field vs composite multimodal text_config unwrapping; different config-loading bugs." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "CUDA assert with auto device mapping vs segmentation fault loading Llava Next; unrelated runtime failures." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP RAM-efficient loading breakage vs non-persistent buffers initialized with junk; both v5-related but distinct defects." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "BeitImageProcessor label reduction bug vs missing all_tied_weights_keys on Molmo; unrelated components." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "3d_parallel.py runtime issue vs Qwen3Moe FineGrainedFP8Config failure; unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "lm_head serialization/tied-weights issue vs incorrect visual-encoder key saving; both serialization bugs but different code paths." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Model loading failure vs tokenizer loading regression; different artifacts and failures." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Auto device-map CUDA assert vs flash-attention generation regression; same model family but different concrete bugs." + }, + { + "left": "issue:26413", + "right": "issue:28530", + "accept": false, + "reason": "Checkpoint resume iterator failure vs early-stopping metric lookup warning; unrelated trainer behaviors." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Int4 quantization memory regression vs apex RMSNorm memory leak; not the same underlying cause." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity on one model/dataset vs version-dependent tokenization changes; only vague tokenizer overlap." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "CustomTrainer signature breakage vs 3d_parallel.py failure; unrelated." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Batched video processor shape bug vs tokenization changes; different components." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Trainer compute_loss kwarg regression vs Qwen3Moe FP8 failure; no shared bug." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "LayoutLMv2 tokenizer crash vs bf16 dtype mismatch; unrelated issues." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Strict config loading vs tokenizer loading regression; separate config/tokenizer problems." + }, + { + "left": "issue:43065", + "right": "issue:46402", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder vs missing Sam3 text-encoder weights; same model family but different code paths." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Quantization offloading regression vs mmap OOM on Strix Halo; different memory-management problems." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "dtype propagation bug in Blip2 vs incorrect max_steps after resume; unrelated trainer/model-load bugs." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "lm_head weight tying bug vs incorrect visual encoder key saving; both multimodal model issues but not the same defect." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "torch.compile/SDPA failure vs packed-weight loading bug; different runtime paths and fixes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 14, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:28530", + "issue:29159", + "issue:30064", + "issue:33357", + "issue:33666", + "issue:34634", + "issue:35707", + "issue:36010", + "issue:36032", + "issue:36331", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43295", + "issue:43316", + "issue:43317", + "issue:43322", + "issue:43388", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43782", + "issue:43873", + "issue:43881", + "issue:44008", + "issue:44186", + "issue:44361", + "issue:44387", + "issue:44423", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44898", + "issue:44913", + "issue:45059", + "issue:45161", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45704", + "issue:45823", + "issue:46291" + ], + "soft_pairs": [ + "issue:44008|issue:45823", + "issue:41720|issue:42617", + "issue:44423|issue:45406", + "issue:44610|issue:44898", + "issue:44186|issue:44610", + "issue:43295|issue:45701", + "issue:43550|issue:43606", + "issue:42915|issue:43317", + "issue:45632|issue:45684", + "issue:36010|issue:42617", + "issue:41720|issue:45161", + "issue:44671|issue:44743", + "issue:43531|issue:43572", + "issue:42915|issue:46291", + "issue:28530|issue:33666", + "issue:43526|issue:43881", + "issue:44186|issue:44361", + "issue:43540|issue:43782", + "issue:42915|issue:43322", + "issue:36331|issue:41720", + "issue:43316|issue:44913", + "issue:28530|issue:30064", + "issue:29159|issue:36032", + "issue:33357|issue:35707", + "issue:43388|issue:45059", + "issue:30064|issue:34634", + "issue:43550|issue:43653", + "issue:43606|issue:43720", + "issue:43873|issue:45704", + "issue:44387|issue:45701" + ], + "prepared_review_unit_hash": "7b0ddcf53707aae649f6972148537b0fb978e4a4a8e7510de0a0ea093f43f9f7", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36672, + "estimated_input_tokens": 9168, + "estimated_eval_tokens": 18592 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous and does not contain any convincing duplicate pairs. The items span unrelated issues across tokenizers, multimodal processors, training, serving, quantization, and device/offload behavior.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Both mention device mismatch, but they affect different Gemma variants and different code paths: a variable-name collision in forward vs auto device-map PLE forwarding." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Different failures and contexts: Qwen3 auto device mapping cuda assert vs failure to run 3d_parallel.py." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Both are serve crashes, but one is a multimodal batching str-to tensor issue and the other is Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Different models and bugs: OmDet-Turbo image size mismatch vs Perceiver interpolation failure." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer padding/NER crash vs OmDet processor input resolution mismatch; no shared underlying bug." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "A regression in processor/tokenizer API usage is not the same as a broad tokenization difference across versions." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Both involve runtime/device behavior, but one is torch.compile+SDPA and the other is CPU offload device mismatch for a different model." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Different quantization/loading problems: FineGrainedFP8Config failure vs dequantized model loading with gpu+cpu offload." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collision for trust_remote_code is unrelated to read-only permission propagation in save_pretrained." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "ImportError for GenerationMixin is unrelated to 3d_parallel.py execution failure." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Both involve parallelism, but one is a cuda assert on Qwen3 auto device mapping and the other is a tensor-parallel-only failure on GPT-OSS MoE." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT masked-LM prediction errors and Qwen3.5 recurrent-state resets are different model-specific bugs." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and missing pad_token_idx in StableLmConfig are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:46291", + "accept": false, + "reason": "Fine-grained FP8 model loading and transformers.pipelines import/version resolution are unrelated issues." + }, + { + "left": "issue:28530", + "right": "issue:33666", + "accept": false, + "reason": "Early stopping metric lookup and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BEiT label reduction bug and glm-4v-9b loading failure do not share a concrete code-path." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both are tokenizer failures, but they concern different tokenizers and different error mechanisms." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Different Qwen3 multimodal problems: video-processing ValueError vs Qwen3VL weight_only loading error." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and Llava Next segmentation fault are distinct bugs." + }, + { + "left": "issue:36331", + "right": "issue:41720", + "accept": false, + "reason": "CustomTrainer compute_loss signature mismatch is unrelated to Qwen3 CUDA assertion on device mapping." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy and GPTNeoX rotary_pct reload regression are different config bugs." + }, + { + "left": "issue:28530", + "right": "issue:30064", + "accept": false, + "reason": "Early stopping metric selection and void segmentation-map processing are unrelated." + }, + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Tokenizer inconsistency and T5Tokenizer add_special_tokens conflict are both tokenizer-related but not the same bug." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are unrelated." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Metric gathering label truncation and SAM3 text/box behavior are unrelated." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Image processor void-segmentation failure and BarkProcessor voice_preset bug are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "torch.compile SDPA failure and BigBirdTokenizer special-token registration are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both involve loading/offload, but Bark CPU-offload mismatch and BitNet packed-weight unpacking are different concrete code paths." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior and T5 apex FusedRMSNorm memory leak are unrelated." + }, + { + "left": "issue:44387", + "right": "issue:45701", + "accept": false, + "reason": "Int4 quantization OOM and tokenization changes across versions 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": 20, + "review_unit_index": 15, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:34567", + "issue:34634", + "issue:35707", + "issue:39692", + "issue:41720", + "issue:42490", + "issue:42886", + "issue:43329", + "issue:43381", + "issue:43421", + "issue:43531", + "issue:43606", + "issue:43653", + "issue:43746", + "issue:43782", + "issue:43844", + "issue:43927", + "issue:44062", + "issue:44077", + "issue:44368", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44610", + "issue:44625", + "issue:44843", + "issue:44898", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45357", + "issue:45381", + "issue:45406", + "issue:45464", + "issue:45479", + "issue:45588", + "issue:45704", + "issue:45759", + "issue:45859", + "issue:45923", + "issue:46050", + "issue:46402" + ], + "soft_pairs": [ + "issue:41720|issue:43746", + "issue:45406|issue:45464", + "issue:45357|issue:46402", + "issue:44492|issue:44509", + "issue:43531|issue:45759", + "issue:43606|issue:43653", + "issue:42490|issue:43329", + "issue:34567|issue:35707", + "issue:34634|issue:41720", + "issue:44451|issue:44991", + "issue:43531|issue:44589", + "issue:44625|issue:45479", + "issue:44077|issue:44625", + "issue:44514|issue:44545", + "issue:45759|issue:45859", + "issue:44898|issue:45072", + "issue:44368|issue:45357", + "issue:45588|issue:45759", + "issue:44062|issue:44451", + "issue:44987|issue:45042", + "issue:43421|issue:43927", + "issue:39692|issue:44610", + "issue:44918|issue:44936", + "issue:43381|issue:43844", + "issue:43746|issue:46050", + "issue:43782|issue:45381", + "issue:42886|issue:44843", + "issue:45704|issue:45923" + ], + "prepared_review_unit_hash": "a5fb58dd881f2b0f777fd17630fc62c05cd23eb1d21d4f10dff820e139826694", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36435, + "estimated_input_tokens": 9109, + "estimated_eval_tokens": 18474 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated issue-pair candidates; only the Qwen2_5_VLProcessor.apply_chat_template batched-input crash appears to be a true duplicate pair. No PRs are present in the input.", + "confidence": 0.97, + "canonical_issue_reason": "issue:44514 is the earlier and more complete report of the same Qwen2_5_VLProcessor.apply_chat_template crash on batched input with padding=False.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44514 is the best representative issue because it cleanly describes the concrete failure mode and is duplicated by issue:44545.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Different models and failures: Qwen3 device-mapping CUDA assert vs GraniteSpeech PEFT local-checkpoint loading." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Different components and bugs: Gemma4Processor missing _tokenizer vs Qwen3.5 streaming chat/completions failure." + }, + { + "left": "issue:45357", + "right": "issue:46402", + "accept": false, + "reason": "Both concern multimodal weight loading/saving, but they target different models and different broken code paths." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "Unrelated documentation issues: a typo in cache strategies vs stale docs mentioning removed pipeline tasks." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Different Qwen-related config problems: sliding-window behavior vs unwrapping text_config for multimodal from_config." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "Completely different subsystems: Bark CPU offload device mismatch vs BigBirdTokenizer mask-token decode bug." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Different bugs: import warning for trailing-slash paths vs undefined video-token helper names in multimodal token counting." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token-count tracking and progressive generation with inputs_embeds/past_key_values are unrelated issues." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset handling is unrelated to Qwen3 auto device mapping CUDA assertions." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but they affect different models and likely different failure paths." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Sliding-window Qwen3-MoE behavior is unrelated to the Float8 storage TypeError." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Different classification bugs: config propagation of num_labels vs degenerate zero-loss behavior with num_labels=1." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "PatchTSMixer post_init policy is unrelated to Qwen3.5 num_labels propagation." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same title and same crash: Qwen2_5_VLProcessor.apply_chat_template fails on batched input when padding=False." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Different Qwen3.5 multimodal issues: from_config text_config unwrapping vs CPU-offload incompatibility in rotary embedding." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Different model families and symptoms: Perceiver image resolution handling vs bfloat16 dtype mismatch in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "A training warning about tie_word_embeddings is unrelated to incorrect visual-encoder keys in save_pretrained." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink-less-model crash is unrelated to Qwen multimodal from_config text_config handling." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "AddedToken constructor TypeError is unrelated to the ScandiBERT loading failure." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "These are different regressions: loading a specific model vs PIL image processors incorrectly requiring torchvision." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Special-token post-processor updates and DiaConfig custom token ID persistence are related thematically but not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Docs/example errors for SigLIP2 do not match the OmDet-Turbo processor/model resolution mismatch." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Both mention trainers, but one is an SFT embedding-unpacking issue and the other is evaluate-after-train failure." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Gradient checkpointing in eval mode and ZeRO-3 gradient growth are distinct training/runtime problems." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "Different checkpointing/loading mechanisms: PEFT local checkpoints vs sinq quantized save/load corruption." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Different Qwen VL bugs: weight_only loading error vs incorrect video vision_position_ids." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": false, + "reason": "Both involve HF_HUB_OFFLINE, but one is cache loading and the other is an unconditional model_info call in tokenizer patching." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "T5 RMSNorm memory-leak behavior is unrelated to a missing supports_gradient_checkpointing flag." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: all listed items are issues, no PRs are present, and the only true duplicate pair shown is issue:44514 / issue:44545. The remaining soft-edge verdicts are conservative and stay within what the titles/body metadata support." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 16, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:39692", + "issue:41720", + "issue:43295", + "issue:43322", + "issue:43335", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43596", + "issue:43618", + "issue:43643", + "issue:43645", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43756", + "issue:43819", + "issue:43873", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44488", + "issue:44661", + "issue:44671", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:45072", + "issue:45081", + "issue:45183", + "issue:45356", + "issue:45479", + "issue:45663", + "issue:45685", + "issue:45704", + "issue:45834", + "issue:45859", + "issue:45907", + "issue:45923", + "issue:46032", + "issue:46294", + "issue:46463" + ], + "soft_pairs": [ + "issue:46032|issue:46463", + "issue:43421|issue:43653", + "issue:45704|issue:45859", + "issue:41720|issue:45704", + "issue:43531|issue:43645", + "issue:45663|issue:45923", + "issue:43421|issue:44186", + "issue:43994|issue:44671", + "issue:43746|issue:43873", + "issue:39692|issue:43873", + "issue:45081|issue:45356", + "issue:43596|issue:44220", + "issue:45356|issue:45907", + "issue:43540|issue:46294", + "issue:30064|issue:36010", + "issue:43720|issue:43819", + "issue:43653|issue:43720", + "issue:44220|issue:44661", + "issue:43618|issue:43697", + "issue:43295|issue:45834", + "issue:43550|issue:45072", + "issue:44062|issue:44488", + "issue:43550|issue:43819", + "issue:43540|issue:44008", + "issue:44186|issue:44898", + "issue:43653|issue:43819", + "issue:43531|issue:43643", + "issue:43322|issue:45685", + "issue:45183|issue:45479", + "issue:43335|issue:43756", + "issue:43643|issue:44877", + "issue:44843|issue:45081" + ], + "prepared_review_unit_hash": "b4f55f1d690b6f826e12640d3ec72b1787ac86a9668700e479ab98d46b6b664d", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37795, + "estimated_input_tokens": 9449, + "estimated_eval_tokens": 19154 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a heterogeneous set of transformer issues with only a loose tokenizer/Mistral-regex subtheme. None of the soft pairs look close enough to be true duplicates or mergeable into one fix. The most representative issue is the concrete `_patch_mistral_regex` crash report, but even the nearby tokenizer reports differ in root cause.", + "confidence": 0.83, + "canonical_issue_reason": "Issue 45081 is the clearest, most actionable report in the only obvious mini-cluster: it pinpoints `_patch_mistral_regex` with a concrete traceback and reproducer.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45081 is the best global representative because it is specific, reproducible, and closest to the other tokenizer-regex regressions in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46032", + "right": "issue:46463", + "accept": false, + "reason": "Different subsystems and failure modes: Mamba2 cache correctness vs T5 decoder-input validation." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both involve tokenizers, but one is runtime special-token/post-processor sync and the other is BigBird mask-token registration; not the same bug." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "Unrelated code paths: T5 RMSNorm memory leak vs Qwen rotary embedding CPU-offload incompatibility." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "Qwen3 device-mapping crash and T5 RMSNorm leak are unrelated." + }, + { + "left": "issue:43531", + "right": "issue:43645", + "accept": false, + "reason": "Different problems: Qwen3-MoE sliding-window behavior vs custom-model initialization in Jupyter." + }, + { + "left": "issue:45663", + "right": "issue:45923", + "accept": false, + "reason": "Gemma4 FSDP2 shared-kv-state bug vs missing gradient-checkpointing flag; not the same defect." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Tokenizer special-token update issue vs LayoutLMv2 NER/padding crash; unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different model-specific correctness bugs: SigLIP2 pipeline outputs vs CamemBERT masked-LM predictions." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "PEFT/local-checkpoint loading bug vs quantization offloading issue; different fixes." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "Documentation/example mismatch vs offloading-with-quantization runtime bug." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Same helper area, but different concrete failures: backend_tokenizer AttributeError vs codec-handling/regression warning; too different to merge." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Deepspeed/BertModel initialization error and fbank feature extraction are unrelated." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regression vs hub API revision forwarding; unrelated." + }, + { + "left": "issue:43540", + "right": "issue:46294", + "accept": false, + "reason": "Both are video-related, but one is Qwen3OmniMoe input processing and the other is chat-template token/feature mismatch." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Image segmentation-map processing vs a GenerationMixin import error; unrelated." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "BitNet accelerate-loading unpacking bug vs DAC latent/STE mismatch; different code paths." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "BigBird special-token registration bug and BitNet weight-unpacking bug are unrelated." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio feature extraction bug vs add-new-model-like/tokenizer-mapping issue." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIP attentions missing and RTDetrV2 output drift are separate regressions." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Processor/tokenizer regression and Kosmos2.5 long-ocr index error are not the same bug." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "torch.compile SDPA crash vs bfloat16 dtype mismatch; different defects." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "AddedToken keyword-argument error vs model-loading failure; unrelated." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Bamba torch.compile/SDPA issue and DAC latent mismatch are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Tokenizer special-token bug vs DAC latent/forward mismatch; unrelated." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Qwen3-MoE sliding-window bug vs trust_remote_code AutoConfig missing-fields issue; unrelated." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Llava Next loading segfault vs MPS histogram Int support error; unrelated." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Input-validation error messaging vs degenerate zero-loss classification behavior; different bugs." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs Smollm3 RoPE-layer dropping issue." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both touch config loading, but one is missing returned fields and the other is strict config rejection for granite_speech." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Both involve `_patch_mistral_regex`, but one is an unconditional hub call breaking offline mode and the other is a missing backend_tokenizer AttributeError; different failure points and fixes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 17, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43388", + "issue:43421", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43644", + "issue:43716", + "issue:43749", + "issue:43782", + "issue:43818", + "issue:43825", + "issue:43950", + "issue:43994", + "issue:44008", + "issue:44186", + "issue:44442", + "issue:44448", + "issue:44479", + "issue:44485", + "issue:44509", + "issue:44561", + "issue:44671", + "issue:44805", + "issue:44811", + "issue:44821", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45084", + "issue:45381", + "issue:45584", + "issue:45834", + "issue:45987", + "issue:46032", + "issue:46192", + "issue:46291", + "issue:46294", + "issue:46463", + "issue:46512" + ], + "soft_pairs": [ + "issue:43716|issue:44008", + "issue:30064|issue:36331", + "issue:43388|issue:46032", + "issue:41720|issue:44991", + "issue:44821|issue:45084", + "issue:43950|issue:45987", + "issue:44448|issue:44671", + "issue:43582|issue:46512", + "issue:39692|issue:45834", + "issue:45584|issue:45834", + "issue:43421|issue:44811", + "issue:44186|issue:44442", + "issue:43540|issue:45381", + "issue:43644|issue:43749", + "issue:43782|issue:44977", + "issue:43577|issue:43818", + "issue:43606|issue:45072", + "issue:44479|issue:45834", + "issue:46294|issue:46463", + "issue:43825|issue:43994", + "issue:44485|issue:44509", + "issue:44805|issue:44936", + "issue:34634|issue:36010", + "issue:44991|issue:46291", + "issue:44561|issue:45020", + "issue:45987|issue:46192", + "issue:43577|issue:43582" + ], + "prepared_review_unit_hash": "e5570aae0337c6bddcc106db4a768802efbe7f07b44f53ae864d15c018a99cad", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35971, + "estimated_input_tokens": 8993, + "estimated_eval_tokens": 18242 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are clearly unrelated despite surface overlap; the only plausible duplicate is the v5 / `trust_remote_code` regression pair.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 45020 is the best umbrella issue in this set: it describes the broad remote-code breakage, and issue 44561 reads like a narrower concrete instance of the same regression.", + "canonical_pr_reason": null, + "best_issue_reason": "45020 is the strongest triage target because it is broader, more reusable as a canonical report, and most likely to subsume closely related remote-code failures.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Different models and different failure modes: image preprocessor dtype mismatch vs Gemma3n variable-name collision." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Unrelated areas: image processor void segmentation maps vs trainer `compute_loss` signature change." + }, + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Different code paths: metric gathering drops labels in the last batch vs Mamba2 `use_cache` producing wrong results." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "One is a CUDA assert in Qwen3 auto device mapping; the other is tokenizer loading failure for a specific model." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "Image processor URL loading and compiler template-node errors are unrelated." + }, + { + "left": "issue:43950", + "right": "issue:45987", + "accept": false, + "reason": "Both mention loading/cache behavior, but one is buffer corruption on `from_pretrained` while the other is a return-type contract bug." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Different models and symptoms: Pegasus output drift vs CamemBERT masked-LM prediction errors." + }, + { + "left": "issue:43582", + "right": "issue:46512", + "accept": false, + "reason": "Apple Silicon warmup `TypeError` is unrelated to `from_config(dtype=...)` leaving `config.dtype` stale." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "Documentation/example issues for SigLIP2 vs an OCR index error in Kosmos2.5; not the same bug." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Whisper empty-transcription failure and Kosmos2.5 long-OCR index error are different concrete problems." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Tokenizer post-processor update behavior is distinct from Whisper `batch_decode` ignoring `skip_special_tokens`." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer problems on different models: NER padding/truncation crash vs AutoTokenizer loading failure." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen video issues, but they involve different models and different broken paths." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Non-persistent buffer junk on load is not the same as FSDP CPU RAM efficient loading being broken." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Load-time `weight_only=True` error and flash-attention generation failure are different failures." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "BLIP2 dtype inconsistency is unrelated to Video-LLaVA missing temporal attention / weight-sharing." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "CPU offload device mismatch and bfloat16 dtype mismatch are different regressions." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Qwen video-input regression and Kosmos2.5 OCR index error are unrelated." + }, + { + "left": "issue:46294", + "right": "issue:46463", + "accept": false, + "reason": "Qwen3.5 video chat-template mismatch is unrelated to T5 decoder-input validation messaging." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "Pipeline task deprecation messaging does not match a SigLIP2 AutoModel/pipeline output bug." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE implementation discussion is not the same as stale docs mentioning removed pipeline tasks." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Mask/index shape mismatch in model code is different from `trainer.evaluate()` failing after `train()`." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "BarkProcessor `voice_preset` behavior and `GenerationMixin` import failure are unrelated." + }, + { + "left": "issue:44991", + "right": "issue:46291", + "accept": false, + "reason": "Tokenizer loading failure and `transformers.pipelines` import failure due to `kernels` versioning are separate issues." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": true, + "reason": "Both report the same v5/remote-code regression; 44561 is a specific instance of the broader breakage described in 45020." + }, + { + "left": "issue:45987", + "right": "issue:46192", + "accept": false, + "reason": "`StaticCache.get_seq_length()` returning a tensor is unrelated to `return_tensors` being ignored when `text_kwargs` is passed." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The only accepted link is well-grounded in the packet: 45020 is the broad remote-code regression issue, and 44561 is a narrower v5 `trust_remote_code` failure that plausibly fits under it. The other soft-edge rejections stay conservative and are supported by the issue titles; I don\u2019t see overstatement." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 18, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36331", + "issue:42175", + "issue:43262", + "issue:43329", + "issue:43452", + "issue:43526", + "issue:43550", + "issue:43582", + "issue:43638", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:44060", + "issue:44291", + "issue:44368", + "issue:44554", + "issue:44589", + "issue:44610", + "issue:44671", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:44991", + "issue:45003", + "issue:45020", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45372", + "issue:45406", + "issue:45538", + "issue:45600", + "issue:45701", + "issue:46209", + "issue:46285", + "issue:46326" + ], + "soft_pairs": [ + "issue:44610|issue:45072", + "issue:28218|issue:29159", + "issue:43720|issue:45072", + "issue:44671|issue:45701", + "issue:43825|issue:43827", + "issue:43782|issue:45325", + "issue:43749|issue:44291", + "issue:42175|issue:45081", + "issue:43582|issue:44857", + "issue:43906|issue:46209", + "issue:43262|issue:43819", + "issue:45003|issue:46285", + "issue:43526|issue:43697", + "issue:43653|issue:44912", + "issue:34634|issue:36331", + "issue:45081|issue:45406", + "issue:44060|issue:44368", + "issue:45372|issue:46326", + "issue:43742|issue:45020", + "issue:45084|issue:45538", + "issue:43526|issue:44554", + "issue:43329|issue:45600", + "issue:44589|issue:45042", + "issue:43452|issue:44991", + "issue:43638|issue:44936", + "issue:43550|issue:44912" + ], + "prepared_review_unit_hash": "1543b4381e553766f85fd8b75d16799c3f8fcad4c80ea354423794bff590cd07", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35803, + "estimated_input_tokens": 8951, + "estimated_eval_tokens": 18158 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a true duplicate set; it mixes many unrelated regression reports across tokenization, loading, processors, training, and backend-specific failures. The best representative issue is the broad remote_code regression report, but none of the soft pairs look safe to merge.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 45020 is the broadest, most representative regression report in the set and aligns with the general theme of recent-version breakages, though the cluster overall is heterogeneous.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45020 is the best single representative because it is high-level, clearly framed as a version regression, and can encompass several loading-related breakages better than the narrower model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Different models and different failures: OmDet-Turbo input-size mismatch vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": false, + "reason": "Both mention tokenizers, but one is a specific added-token spacing bug and the other is a broad slow/fast inconsistency report; not enough evidence they are the same defect." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Unrelated: BitNet packed-weight loading failure versus a bfloat16 dtype mismatch in other models." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "One is a CamemBERT masked-LM regression, the other is a general tokenization-version change report; too broad and not the same code path." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Related theme, but one is an error-message issue and the other is a documentation reference issue; different artifacts, not the same underlying bug." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen model bugs: a weight_only loading error versus a RoPE/position-id scaling bug." + }, + { + "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 an unexpected init_empty_weights argument; distinct failures." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Completely different areas: backend installation metadata versus a Mistral regex patch crash." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Different runtime paths and symptoms: Apple Silicon allocator warmup TypeError versus CUDA AMP loss crash." + }, + { + "left": "issue:43906", + "right": "issue:46209", + "accept": false, + "reason": "No shared underlying bug: an isolated reproduction of an older issue versus an FP8 inference failure on B200." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate default bug is unrelated to DAC latent reconstruction mismatch." + }, + { + "left": "issue:45003", + "right": "issue:46285", + "accept": false, + "reason": "Different concerns: unsafe sys.modules access versus CLIP text model structure inconsistency." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "Different model components and symptoms: BeitImageProcessorFast label reduction versus RTDetrV2 output drift." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "Tokenizer special-token registration bug is unrelated to MXFP4 quantization fallback during model loading." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "BarkProcessor voice_preset failure and CustomTrainer.compute_loss signature break are unrelated." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Both affect processor/tokenizer loading, but one is a Mistral regex patch crash and the other is a Gemma4Processor missing _tokenizer attribute; different causes." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both mention tied-weights warnings, but one is an incorrect weight tying diagnosis and the other is a repeated warning during ms-swift fine-tuning; not clearly the same bug." + }, + { + "left": "issue:45372", + "right": "issue:46326", + "accept": false, + "reason": "Different mistral_common-related failures: missing ReasoningEffort import versus continue_final_message behavior." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "Specific MobileLLM load KeyError versus a broad remote_code regression; too different to be duplicates." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Compile-time template-node error and CLIPTokenizer max-length behavior are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "No overlap in code path: label reduction in BEiT versus MPS attention correctness with mismatched head dims." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Video multimodal token counting bug is unrelated to an auto_mappings CI break from removed configs." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Float8 storage lookup failure is unrelated to PIL backend processors incorrectly requiring torchvision." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer/model loading reports, but one is about gguf_file handling for from_pretrained and the other is a specific tokenizer load regression for est-roberta; not the same defect." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Different training failures: DeepSpeed ZeRO-3 indexing error versus evaluate() failing after train()." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "Bamba torch.compile SDPA failure and MXFP4 quantization fallback are unrelated model/runtime 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": 20, + "review_unit_index": 19, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:34567", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:42175", + "issue:42490", + "issue:42915", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43454", + "issue:43531", + "issue:43577", + "issue:43606", + "issue:43643", + "issue:43645", + "issue:43650", + "issue:43696", + "issue:43720", + "issue:43742", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43854", + "issue:43906", + "issue:43927", + "issue:43931", + "issue:44291", + "issue:44387", + "issue:44479", + "issue:44610", + "issue:44617", + "issue:44756", + "issue:44811", + "issue:44912", + "issue:44977", + "issue:45042", + "issue:45459", + "issue:45759", + "issue:45803", + "issue:45859", + "issue:46291", + "issue:46421", + "issue:46424" + ], + "soft_pairs": [ + "issue:43746|issue:44811", + "issue:43650|issue:43906", + "issue:43819|issue:43927", + "issue:43257|issue:43329", + "issue:44387|issue:44617", + "issue:36010|issue:43645", + "issue:42175|issue:46291", + "issue:38175|issue:44610", + "issue:43720|issue:44610", + "issue:36010|issue:36331", + "issue:43854|issue:44977", + "issue:43531|issue:45042", + "issue:39692|issue:44912", + "issue:46421|issue:46424", + "issue:45459|issue:45803", + "issue:42490|issue:43257", + "issue:43262|issue:44811", + "issue:43257|issue:43696", + "issue:43643|issue:45759", + "issue:43454|issue:43577", + "issue:43329|issue:44479", + "issue:43329|issue:43450", + "issue:43749|issue:44756", + "issue:43257|issue:43931", + "issue:34567|issue:40990", + "issue:43606|issue:45859", + "issue:42915|issue:43742", + "issue:44291|issue:45042" + ], + "prepared_review_unit_hash": "c1f14f410a7d29ee8e10bc54d18e692318668ba9523ccd9c4f881eda8a14f6bc", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35671, + "estimated_input_tokens": 8918, + "estimated_eval_tokens": 18092 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated standalone bug reports spanning different subsystems, models, and failure modes. None of the soft-paired items look like the same underlying bug/change, so I would not merge any pair as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Different code paths and symptoms: GraniteSpeech PEFT/local checkpoint loading vs Whisper batch_decode skipping special tokens." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "One is a vague placeholder issue; the other is a reproduction of a specific upstream bug. No shared underlying defect." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "Different model bugs: DAC latents/STE mismatch vs DiaConfig losing custom token IDs on save/load." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MoE weight conversion/loading with accelerate+deepspeed is unrelated to multimodal token counting video-branch NameErrors." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but one is an int4 quantization memory regression and the other is a model-specific video OOM." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "Importing GenerationMixin and notebook custom-model initialization failures are different regression classes." + }, + { + "left": "issue:42175", + "right": "issue:46291", + "accept": false, + "reason": "Packaging/backend dependency omission is not the same as pipelines import failing on kernels v0.15 version requirements." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Different models and failures: zero probabilities in SigLIP2 vs image processor/model input-size mismatch in OmDet-Turbo." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "BitNet packed-weight unpacking during accelerate loading is unrelated to OmDet-Turbo processor shape mismatch." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "GenerationMixin import error and compute_loss(num_items_in_batch) API break are separate trainer/import issues." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "Different model-specific failures; one is a GLM-4.7-Flash load issue, the other a Qwen3.5 flash-attention generation regression." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PIL backend processors requiring torchvision." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Both involve model/examples, but one is a SigLIP2 doc example with processor mismatch/quantization errors and the other is an MXFP4 loading regression for git-oss-20b." + }, + { + "left": "issue:46421", + "right": "issue:46424", + "accept": false, + "reason": "Both mention cache, but one is lazy initialization causing recompilation and the other is max_cache_length only applying to quantized cache." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Both involve exception handling, but they are different functions and bugs: tokenizer protobuf decode handling vs Fuyu tensor conversion bare except." + }, + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "Local-path trailing-slash import warning is unrelated to Qwen3 MoE weight conversion under accelerate+deepspeed." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Different audio processor bugs: apply_chat_template sample-rate default vs batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "MoE weight conversion/loading problem is unrelated to GPT-oss-20b CUDA OOM." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Missing fields from trust_remote_code config objects is not the same as unwrapping text_config for composite multimodal configs." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "AyaVision weight tying/generation garbage and BLIP2 dtype propagation are distinct model issues." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Both touch video paths, but one is undefined helper names in token counting while the other is a broader v5.3.0 video input regression." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Undefined multimodal token-count helpers and batched video processor shape errors are different failure modes." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading breakage is unrelated to disabling mmap on Strix Halo to avoid OOM." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3 MoE accelerate/deepspeed weight conversion is unrelated to Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token accounting and GPT-oss perplexity on WikiText are unrelated evaluation/training issues." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "Both involve CPU offload, but they affect different models and code paths; not the same underlying bug." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to a key error while loading MobileLLM-125M." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "init_empty_weights/_is_hf_initialized loading regression is unrelated to PIL backend torchvision dependency requirements." + } + ] + }, + "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": 20, + "review_unit_index": 20, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43317", + "issue:43404", + "issue:43425", + "issue:43452", + "issue:43525", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43824", + "issue:43927", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44560", + "issue:44568", + "issue:44821", + "issue:44849", + "issue:44898", + "issue:44912", + "issue:44987", + "issue:44991", + "issue:45127", + "issue:45137", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45405", + "issue:45488", + "issue:45538", + "issue:45701", + "issue:45907", + "issue:46291", + "issue:46512" + ], + "soft_pairs": [ + "issue:44568|issue:45356", + "issue:29159|issue:34567", + "issue:43525|issue:46291", + "issue:43404|issue:45127", + "issue:42915|issue:44821", + "issue:43927|issue:44912", + "issue:44987|issue:45538", + "issue:43697|issue:45357", + "issue:44849|issue:45701", + "issue:42915|issue:45538", + "issue:43577|issue:46512", + "issue:43425|issue:46291", + "issue:43452|issue:43582", + "issue:42175|issue:43824", + "issue:43317|issue:45538", + "issue:33357|issue:34567", + "issue:41720|issue:43540", + "issue:28282|issue:34689", + "issue:44479|issue:45381", + "issue:45137|issue:45405", + "issue:44062|issue:44991", + "issue:43540|issue:44479", + "issue:43540|issue:44560", + "issue:43742|issue:44291", + "issue:44568|issue:45488", + "issue:28282|issue:29127", + "issue:44987|issue:45907", + "issue:24643|issue:29159", + "issue:42915|issue:43749", + "issue:38175|issue:44898" + ], + "prepared_review_unit_hash": "dae6f411e0baaba8bdad1345aa4eb0841a5b8b4d4f9902ba8fc76a8ff0bd4f91", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36235, + "estimated_input_tokens": 9059, + "estimated_eval_tokens": 18374 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a set of unrelated closed issues across many subsystems. The only likely duplicate is the Qwen video-input regression pair, where one report is a narrower instance of the broader v5.3.0 regression.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 44479 is the best canonical issue: it describes the broad v5.3.0 video-input regression across multiple Qwen model variants, and it naturally subsumes the narrower qwen2.5-vl-specific report.", + "canonical_pr_reason": null, + "best_issue_reason": "44479 is the strongest representative because it is broader, clearer about the regression scope, and more likely to attract all related discussion and fixes than the narrower follow-up report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer-related, but they describe different regressions: special-token insertion vs codec/warning handling for a specific model family." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Inconsistent slow/fast tokenizer behavior is unrelated to TrainerState's token counter not updating." + }, + { + "left": "issue:43525", + "right": "issue:46291", + "accept": false, + "reason": "A missing pad_token_id on Llama4Config is unrelated to the kernels package now requiring version/revision." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both mention tied weights, but one is a model-class weight-tying bug and the other is a LoRA merge/extended-vocab collapse; not the same concrete bug." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "FineGrainedFP8Config loading failure is unrelated to AutoImageProcessor URL loading." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Config save/load token-id loss causing IndexError is unrelated to MXFP4 quantization fallback behavior." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "A model-loading failure for physical-intelligence/fast is unrelated to CLIPTokenizer's max_length bug." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "RTDetrV2 output differences in v5 and Qwen3.5 save_pretrained key corruption are different regressions." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 output_hidden_states behavior is not the same as the broad tokenization-change report." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "FineGrainedFP8Config loading and CLIPTokenizer max_length are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:46512", + "accept": false, + "reason": "One is a model dtype propagation bug in Blip2 loading; the other is config.dtype staying stale after from_config. Similar theme, different bug." + }, + { + "left": "issue:43425", + "right": "issue:46291", + "accept": false, + "reason": "Torch compatibility complaints are unrelated to kernels import requirements." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf_file loading errors and caching_allocator_warmup TypeError are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Missing TensorFlow backend in the torch extra is unrelated to the missing Qwen2_5_VLForConditionalGeneration import." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "device_map/offload loading failure is unrelated to CLIPTokenizer's model_max_length bug." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "A MacOS bus error loading CLIP and a TrainerState counter bug are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Both involve Qwen models, but one is an auto-device-map cuda assert and the other is a video-input ValueError; not the same code-path problem." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "PyTorch missing import error is unrelated to Llama 3.2 Vision model-loading breakage." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "45381 is a narrower instance of the broader v5.3.0 Qwen video-input regression described in 44479, with the same underlying wrong-position-id behavior." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque failure and an unreleased MIN_PEFT_VERSION bump are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "AddedToken constructor conflicts and tokenizer-loading failure for EMBEDDIA/est-roberta are different issues." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Qwen3OmniMoe video processing ValueError is not the same bug as the broader Qwen video-input regression reported in 44479." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "These are separate video-input failures for different Qwen models and error modes." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "A key error loading a specific model is unrelated to init_empty_weights rejecting an unexpected argument." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Both are tokenizer-related regressions, but one is special-token addition and the other is LlamaTokenizer pre-tokenizer override; different bugs." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "PyTorch missing import and LayoutLMv3 error-message clarity are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "A model-loading failure is unrelated to list_repo_files needing to forward revision." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "DeepSpeed weight dimensionality error is unrelated to slow/fast tokenizer inconsistency." + }, + { + "left": "issue:42915", + "right": "issue:43749", + "accept": false, + "reason": "FineGrainedFP8Config loading and FSDP CPU RAM efficient loading are unrelated training/loading bugs." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Unexpected zero probabilities in SIGLIP2 and Perceiver non-default-resolution failure are different model-specific issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays grounded in the packet and is appropriately conservative. The only accepted soft-edge pair, issue:44479 with issue:45381, is reasonably supported by the titles as a broader/narrower Qwen video-input regression pair. The other rejections are phrased cautiously and do not overclaim 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": 20, + "review_unit_index": 21, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:27957", + "issue:28282", + "issue:34567", + "issue:36331", + "issue:42175", + "issue:42796", + "issue:43257", + "issue:43404", + "issue:43452", + "issue:43493", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43643", + "issue:43697", + "issue:43742", + "issue:43756", + "issue:43906", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44488", + "issue:44568", + "issue:44589", + "issue:44737", + "issue:44792", + "issue:44821", + "issue:44913", + "issue:44964", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45375", + "issue:45405", + "issue:45479", + "issue:45588", + "issue:45600", + "issue:45823", + "issue:45920", + "issue:45987", + "issue:46121", + "issue:46399", + "issue:46463", + "issue:46468", + "issue:46491" + ], + "soft_pairs": [ + "issue:44737|issue:46463", + "issue:42175|issue:44488", + "issue:43493|issue:43994", + "issue:43577|issue:45072", + "issue:43257|issue:43756", + "issue:44291|issue:44821", + "issue:43906|issue:46468", + "issue:34567|issue:36331", + "issue:44062|issue:45081", + "issue:27957|issue:28282", + "issue:44568|issue:46491", + "issue:43526|issue:45479", + "issue:44737|issue:45588", + "issue:44792|issue:44964", + "issue:43404|issue:45987", + "issue:43756|issue:46399", + "issue:42796|issue:43906", + "issue:43756|issue:45325", + "issue:45823|issue:46463", + "issue:45042|issue:45081", + "issue:43742|issue:44589", + "issue:43452|issue:44488", + "issue:43531|issue:44821", + "issue:45405|issue:45600", + "issue:45357|issue:45920", + "issue:44913|issue:46121", + "issue:43643|issue:45375", + "issue:43697|issue:45216" + ], + "prepared_review_unit_hash": "524aa80a6ec9962df3cb2407eec559dc999477402cb38cd0d12eebbed54e338e", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36562, + "estimated_input_tokens": 9141, + "estimated_eval_tokens": 18538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is still mostly heterogeneous across unrelated tokenizer, trainer, config, model-loading, and backend issues. The only soft edge that looks like the same underlying bug is the SigLIP2 pair, and even that should be treated as likely rather than explicitly confirmed duplicate.", + "confidence": 0.82, + "canonical_issue_reason": "No single canonical issue fits the whole cluster because the items span many unrelated code paths and model families.", + "canonical_pr_reason": "No canonical PR: there are no PRs in this cluster.", + "best_issue_reason": "issue:43493 is the best representative for the likely SigLIP2 bug because it describes the implementation discrepancy more directly, while 43994 is mainly a downstream symptom report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44737", + "right": "issue:46463", + "accept": false, + "reason": "Different subsystems and failures: XLNet CPU device placement vs T5 decoder input validation." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Unrelated problems: dependency/backend packaging vs loading a specific BERT model." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "Likely the same SigLIP2 bug: one report describes the HF-vs-JAX implementation discrepancy, and the other reports bad outputs from the same model/checkpoint." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Different dtype issues in different model families and code paths." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed is unrelated to Smollm3 RoPE-layer dropping." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Loading with init_empty_weights and loading an image processor from URL are different failures." + }, + { + "left": "issue:43906", + "right": "issue:46468", + "accept": false, + "reason": "One is a reproduction of an old issue; the other is an official Nemotron example breakage with no clear same code path." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "Trainer state token counting and custom compute_loss signature compatibility are separate trainer regressions." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-related, but one is an AddedToken argument collision and the other is a Mistral regex patch crash." + }, + { + "left": "issue:27957", + "right": "issue:28282", + "accept": false, + "reason": "XLM-RoBERTa Flash Attention and missing PyTorch import are unrelated." + }, + { + "left": "issue:44568", + "right": "issue:46491", + "accept": false, + "reason": "Different tokenizer bugs with different models and behaviors." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "Image processor label reduction and sequence-classification zero loss are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU tensor creation and flash_attention sink handling are different bugs." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Different multimodal model regressions with no shared concrete failure path." + }, + { + "left": "issue:43404", + "right": "issue:45987", + "accept": false, + "reason": "Weight tying in Mistral3 and StaticCache sequence length return type are unrelated." + }, + { + "left": "issue:43756", + "right": "issue:46399", + "accept": false, + "reason": "RoPE-layer dropping and missing apply_rotary_pos_emb registration are different defects." + }, + { + "left": "issue:42796", + "right": "issue:43906", + "accept": false, + "reason": "Ministral BPE-marker handling is not the same bug as the reproduced older issue." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both involve RoPE/position handling, but the model, symptom, and code path differ." + }, + { + "left": "issue:45823", + "right": "issue:46463", + "accept": false, + "reason": "Gemma4 device placement and T5 decoder-input error messaging are unrelated." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Image-processor torchvision dependency regression is unrelated to the Mistral tokenizer regex crash." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "MobileLLM loading key error and Float8 storage lookup failure are different load-time issues." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "gguf_file loading breakage and a specific model load failure do not look like the same underlying bug." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and AutoImageProcessor URL loading are unrelated." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "PEFT version bump and stale auto_mappings references are separate CI/dependency issues." + }, + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Incorrect save_pretrained visual encoder keys is unrelated to tokenizer ID regressions." + }, + { + "left": "issue:44913", + "right": "issue:46121", + "accept": false, + "reason": "Config reload losing rotary_pct and JSON-list rope validation type errors are different config bugs." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both mention missing config fields, but one is generic remote-code loading and the other is a specific strict-field drop; not clearly the same bug." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "RTDetrV2 output drift and Qwen3.5 save_pretrained regression are different model-path regressions." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Mostly grounded, but one claim is too strong for the packet: there is no explicit confirmation of a duplicate pair, so wording like \"confirmed duplicate pair\" overstates the evidence. The SigLIP2 match is plausible, but it should be framed more conservatively (e.g. likely same underlying bug / same area) rather than definitive." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 22, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:32090", + "issue:33357", + "issue:38175", + "issue:41762", + "issue:42898", + "issue:43295", + "issue:43452", + "issue:43577", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43749", + "issue:43819", + "issue:44361", + "issue:44442", + "issue:44464", + "issue:44466", + "issue:44484", + "issue:44493", + "issue:44514", + "issue:44568", + "issue:44625", + "issue:44704", + "issue:44743", + "issue:44811", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44936", + "issue:44991", + "issue:45005", + "issue:45200", + "issue:45356", + "issue:45372", + "issue:45381", + "issue:45542", + "issue:45823", + "issue:46050", + "issue:46170", + "issue:46178", + "issue:46463", + "issue:46491" + ], + "soft_pairs": [ + "issue:43720|issue:44898", + "issue:44625|issue:44877", + "issue:45200|issue:46463", + "issue:24643|issue:29127", + "issue:41762|issue:44936", + "issue:44493|issue:46050", + "issue:44466|issue:45005", + "issue:42898|issue:46491", + "issue:43452|issue:44843", + "issue:43819|issue:44811", + "issue:28282|issue:32090", + "issue:44514|issue:45381", + "issue:28282|issue:33357", + "issue:43688|issue:43749", + "issue:45372|issue:46170", + "issue:44936|issue:45823", + "issue:38175|issue:43295", + "issue:44361|issue:44442", + "issue:44484|issue:44704", + "issue:43653|issue:44361", + "issue:45542|issue:46178", + "issue:43452|issue:43577", + "issue:43295|issue:45356", + "issue:43577|issue:43697", + "issue:43653|issue:44568", + "issue:44464|issue:44743", + "issue:43646|issue:43749", + "issue:44843|issue:44991" + ], + "prepared_review_unit_hash": "73b23b5fc3c1c443f7ab99f40cc36556c80b8bb47a4f7cd4a74d153489eef278", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35869, + "estimated_input_tokens": 8968, + "estimated_eval_tokens": 18192 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a set of mostly unrelated Hugging Face Transformers bugs/regressions. The titles span tokenizer loading, model init, processor behavior, backend/import issues, and several model-specific regressions, but none of the soft pairs look like the same concrete underlying bug.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43646 is the broadest and most representative umbrella bug in the set: a general Transformers v5 custom model initialization regression. It best matches the cluster\u2019s overall theme of v5 breakages without being tied to one niche model.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the best single issue to anchor the cluster because it is general enough to cover the common regression pattern, while the others are narrower model/tokenizer/backend-specific cases.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "Different bugs: BitNet packed-weight loading vs Perceiver image-classification resizing/interpolation behavior." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Different models and failures: Qwen3.5 num_labels propagation vs Granite Speech strict-config loading." + }, + { + "left": "issue:45200", + "right": "issue:46463", + "accept": false, + "reason": "Different code paths: Gemma 4 mm_token_type_ids defaulting vs T5 decoder input validation messaging." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "Different problems: DeepSpeed training weight-shape RuntimeError vs LayoutLMv3 error-message clarity." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "Both are training/loading failures, but they involve different models and failure modes: Gemma3 ZeRO-3 loading vs trainer.evaluate after train." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Different underlying issues: position-id unexpected keys vs quantized model save/load breakage after sinq." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both concern tied weights, but the concrete behaviors differ: serialization inconsistency vs translation-model weight handling in v5." + }, + { + "left": "issue:42898", + "right": "issue:46491", + "accept": false, + "reason": "Different tokenizer regressions: v5 clean_up_tokenization_spaces behavior vs CodeLlama leading-space round-trip loss." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "Different failure modes: gguf_file breaking tokenizer/model loading vs model_info() being called offline in _patch_mistral_regex." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "Different components: DAC latent reconstruction mismatch vs Whisper processor batch_decode skip_special_tokens handling." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Different import/runtime paths: missing PyTorch for AutoModel vs Trainer passing NoneType to _gpu_broadcast_one." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both are VL processor-related, but one is batched chat templating padding while the other is video vision_position_ids correctness." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Different root causes: missing PyTorch import error vs MacOS bus error with a specific CLIP community model." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Related to loading/parallelism, but not the same bug: auxiliary-loss normalization vs FSDP CPU RAM efficient loading." + }, + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Different import/dependency issues: Gemma 4 processor loading needs ReasoningEffort, while Windows ROCm issues stem from torch.distributed dependency handling." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Different failure modes: trainer lifecycle bug vs Gemma4 PLE device mismatch during forward with device_map=auto." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Different regressions: zero probabilities in SIGLIP2 vs processor.tokenizer access / image-passing regression in v4.57.5." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Both tokenizer-related, but one is MLukeTokenizer task AttributeError and the other is AutoTokenizer failing on FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "Different save/load API issues: max_shard_size default question vs AutoProcessor not forwarding kwargs to cached_file." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizer bugs: BigBirdTokenizer special-token registration/empty decode vs MLukeTokenizer task AttributeError." + }, + { + "left": "issue:45542", + "right": "issue:46178", + "accept": false, + "reason": "Different backend issues: TensorBoard without TensorFlow causing undefined tf backend vs unsupported tensorflow_text backend mapping." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "Different loading issues: gguf_file breaks tokenizer/model loading vs Blip2 dtype not propagating to model.qformer." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Different regressions: custom processor/tokenizer access broke in v4.57.5 vs Kimi-K2.5 tokenizer codec handling and warning regression." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Different model-specific issues: Blip2 dtype propagation vs RTDetrV2 producing different outputs in v5." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Different tokenizer defects: BigBird mask token not registered vs mdeberta-v3-base not adding BOS/EOS with add_special_tokens." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Different generation/model-state bugs: chunked generation with compiled forward vs Qwen3.5 recurrent states resetting with cache." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Custom model init regression is not the same bug as FSDP CPU RAM efficient loading." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Different tokenizer-loading regressions: offline mistral regex patch issue vs est-roberta tokenizer loading failure." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 23, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:29159", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:36010", + "issue:36331", + "issue:41720", + "issue:43262", + "issue:43295", + "issue:43452", + "issue:43504", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43606", + "issue:43643", + "issue:43653", + "issue:43688", + "issue:43720", + "issue:43742", + "issue:43756", + "issue:43927", + "issue:43994", + "issue:44361", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44514", + "issue:44554", + "issue:44561", + "issue:44610", + "issue:44625", + "issue:44821", + "issue:44849", + "issue:44977", + "issue:45290", + "issue:45440", + "issue:45491", + "issue:45701", + "issue:45704", + "issue:45823", + "issue:46002", + "issue:46050", + "issue:46192", + "issue:46209", + "issue:46365", + "issue:46402" + ], + "soft_pairs": [ + "issue:43531|issue:43742", + "issue:29159|issue:33357", + "issue:45491|issue:46002", + "issue:43756|issue:46402", + "issue:43688|issue:45704", + "issue:43688|issue:45440", + "issue:43653|issue:43927", + "issue:44514|issue:45290", + "issue:43550|issue:43927", + "issue:43688|issue:45701", + "issue:44479|issue:44625", + "issue:43550|issue:44361", + "issue:44464|issue:44977", + "issue:43643|issue:44561", + "issue:41720|issue:44464", + "issue:33666|issue:36010", + "issue:44466|issue:46050", + "issue:43742|issue:44821", + "issue:32090|issue:33666", + "issue:43504|issue:43720", + "issue:43504|issue:43550", + "issue:43577|issue:44554", + "issue:44977|issue:46209", + "issue:43504|issue:44610", + "issue:44554|issue:46365", + "issue:43262|issue:43295", + "issue:43720|issue:43927", + "issue:45823|issue:46050", + "issue:43606|issue:43927", + "issue:43452|issue:44451", + "issue:33666|issue:36331", + "issue:43994|issue:46192", + "issue:44849|issue:46209" + ], + "prepared_review_unit_hash": "e03b66501e0823942e4bbee2da68719d6e0e3ea4e85e9a1678344e4b731f5b7d", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37638, + "estimated_input_tokens": 9410, + "estimated_eval_tokens": 19076 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a heterogeneous bundle of unrelated issues; every soft pair looks like a topical similarity only, not the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: the cluster is heterogeneous and the soft links are mostly superficial title similarity rather than duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "No strong global best issue; none of the issues appears to represent the whole set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior vs a MobileLLM loading key error; different bug classes." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency vs a MacOS bus error on CLIP; unrelated failures." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Both mention NaNs, but one is a sliding-window attention bug and the other is a classification-head init/checkpoint issue." + }, + { + "left": "issue:43756", + "right": "issue:46402", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping vs Sam3Model text-encoder weights not loading; different model families and fixes." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Aux-loss normalization bug vs an apex RMSNorm memory leak; unrelated code paths." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Different models and different semantics: loss normalization vs remote-implementation divergence." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "BigBird special-token registration vs DiaConfig custom token IDs lost on save/load; only broadly token-related." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both crash in apply_chat_template, but on different triggers: batched padding=False vs assistant tool-call messages with no content." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "Torch.compile SDPA failure vs config token-ID persistence; unrelated." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Aux-loss normalization bug vs tokenization behavior changes across versions; not the same bug." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression in Qwen VL models vs num_labels propagation in Qwen3.5; different code paths." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Bamba compile/SDPA failure vs MLukeTokenizer AttributeError; unrelated." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Chunked generation with compiled forward vs flash-attention generation; both generation bugs, but not the same concrete failure." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both involve trust_remote_code, but one is missing config fields and the other is removal of a helper API." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Qwen3 device-map cudaErrorAssert vs compiled-forward chunked-generation inconsistency; different failures." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Multi-GPU training issue vs GenerationMixin import error; unrelated." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "lm_head serialization/tied-weights behavior vs quantized save/load breakage; both persistence-related but not the same bug." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "MobileLLM loading failure vs AutoImageProcessor URL loading; different APIs and symptoms." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer broadcast NoneType error vs Qwen2-VL multi-GPU training; same broad area only." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy field loading in Beit vs BitNet packed-weight unpacking during accelerate loading; distinct model-loading bugs." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Beit pretrained-load legacy field issue vs Bamba compile/SDPA failure; unrelated." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Blip2 dtype propagation bug vs MPS attention correctness issue; unrelated." + }, + { + "left": "issue:44977", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug vs FP8 inference failure on B200; different backend issues." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Beit loading legacy field vs OmDet processor/model size mismatch; unrelated." + }, + { + "left": "issue:44554", + "right": "issue:46365", + "accept": false, + "reason": "MPS attention correctness issue vs inability to import Trainer with datasets installed; unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio sampling-rate default vs processor.tokenizer/images regression; both processor-related but different breakages." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "BitNet accelerate loading vs DiaConfig token-ID persistence; unrelated." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Gemma4 device-map forward mismatch vs sinq quantized save/load breakage; different problems." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch in Bark vs DiaConfig token-ID loss; unrelated." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "GGUF-related tokenizer/model loading breakage vs ScandiBERT loading failure; different root causes." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training issue vs custom Trainer compute_loss signature mismatch; unrelated." + }, + { + "left": "issue:43994", + "right": "issue:46192", + "accept": false, + "reason": "Siglip2 bad outputs vs return_tensors being ignored with text_kwargs; unrelated." + }, + { + "left": "issue:44849", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 hidden-states bug vs FP8 inference on B200; different features and 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": 20, + "review_unit_index": 24, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:28218", + "issue:33357", + "issue:33666", + "issue:39692", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43688", + "issue:43701", + "issue:43720", + "issue:43824", + "issue:43825", + "issue:43844", + "issue:43927", + "issue:44291", + "issue:44361", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44792", + "issue:44843", + "issue:44849", + "issue:44964", + "issue:44977", + "issue:45200", + "issue:45229", + "issue:45357", + "issue:45405", + "issue:45520", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:46002", + "issue:46424", + "issue:46489" + ], + "soft_pairs": [ + "issue:43525|issue:43572", + "issue:43278|issue:43688", + "issue:43493|issue:43825", + "issue:43720|issue:44361", + "issue:44964|issue:45405", + "issue:39692|issue:43493", + "issue:28218|issue:33357", + "issue:44625|issue:46002", + "issue:44849|issue:45357", + "issue:43493|issue:43844", + "issue:44568|issue:46489", + "issue:45200|issue:45229", + "issue:43278|issue:43844", + "issue:39692|issue:43295", + "issue:44488|issue:44843", + "issue:43540|issue:44977", + "issue:43646|issue:44466", + "issue:44568|issue:45812", + "issue:44464|issue:46424", + "issue:43824|issue:44589", + "issue:33357|issue:33666", + "issue:44589|issue:45520", + "issue:43257|issue:43299", + "issue:43334|issue:43572", + "issue:43927|issue:44361", + "issue:43504|issue:43606", + "issue:43334|issue:45759", + "issue:45229|issue:45823", + "issue:43278|issue:43701", + "issue:44792|issue:45405", + "issue:43575|issue:45405", + "issue:43572|issue:44291" + ], + "prepared_review_unit_hash": "3840072b76f4cd9fd9347ab6d52f1b647c482128b9e95b07775e710f55ccfbbc", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37356, + "estimated_input_tokens": 9339, + "estimated_eval_tokens": 18934 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are mostly unrelated issues that only share broad transformer/model-regression vocabulary. None of the soft-edge pairs look like the same underlying bug or change, so all should be rejected.", + "confidence": 0.9, + "canonical_issue_reason": "No single true duplicate cluster emerges; issue 45812 is the closest representative because it is the broadest, highest-impact tokenizer regression with a clear v4\u2192v5 failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45812 is the best overall issue representative: concrete user impact, regression framing, and the broadest scope among the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both are config attribute regressions, but they affect different models and different missing fields (`pad_token_id` vs `pad_token_idx`) with no shared code path shown." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "One is an embedding dtype mismatch during evaluate; the other is auxiliary-loss normalization. Different symptoms and subsystems." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation mismatch and a pipeline translation error message are unrelated bugs." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "Accelerate loading of packed BitNet weights vs tokenizer task AttributeError are unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model-loading failure and a dependency-version bump are not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Documentation example errors and SigLIP2 HF-vs-JAX discrepancy are different issues." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer whitespace behavior and a MacOS bus error on CLIP are unrelated." + }, + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "`num_labels` propagation in Qwen3.5 and NaNs in classification heads are different failure modes." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Output-hidden-states behavior and incorrect visual-encoder save keys are separate Qwen3.5 bugs." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and DeepSpeed gradient explosion are unrelated." + }, + { + "left": "issue:44568", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different models and distinct tokenization failures." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Text-only mm_token_type_ids defaults and multi-GPU CUDA OOM are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Evaluate-time dtype drift and ZeRO-3 gradient growth are different bugs." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "A broken doc example and a regression in processor/tokenizer behavior are not the same underlying issue." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Loading a specific model failing and an offline-mode tokenizer patch issue are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Video-input processing in Qwen3OmniMoe and flash-attention generation in Qwen3.5 are different code paths." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Custom model initialization breakage and model-device-dependent lm_head serialization are unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different model families and different concrete failures." + }, + { + "left": "issue:44464", + "right": "issue:46424", + "accept": false, + "reason": "Chunked generation with compiled forward and cache-length handling are separate concerns." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Missing Qwen2.5-VL import and Float8 storage lookup errors are unrelated." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "MacOS bus error on a CLIP model and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Float8 storage lookup failure and Python 3.13 flash_attn import handling are different bugs." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed and Qwen3VL MoE loading are related only by model family, not the same failure." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Missing pad_token_id on Qwen3-VL and missing pad_token_idx on StableLmConfig are distinct model/config regressions." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "DiaConfig token-ID persistence and MLukeTokenizer task AttributeError are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Beit semantic-segmentation legacy-field loading and Bark CPU-offload device mismatch are separate bugs." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen multimodal configs, but one is a missing config attribute and the other is from_config not unwrapping text_config." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Multi-GPU OOM and PLE device mismatch during forward are different failure modes." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype drift and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "A failing janus image-generation test and a dependency version bump are not the same bug." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "TP-induced OOM on Qwen2-57B and a PEFT minimum-version bump are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Missing pad_token_idx and init_empty_weights receiving an unexpected argument are different compatibility problems." + } + ] + }, + "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": 20, + "review_unit_index": 25, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29159", + "issue:29942", + "issue:30333", + "issue:32090", + "issue:34567", + "issue:37428", + "issue:38175", + "issue:42490", + "issue:43065", + "issue:43493", + "issue:43504", + "issue:43540", + "issue:43606", + "issue:43653", + "issue:43696", + "issue:43701", + "issue:43844", + "issue:43873", + "issue:43957", + "issue:43994", + "issue:44112", + "issue:44186", + "issue:44423", + "issue:44464", + "issue:44493", + "issue:44521", + "issue:44554", + "issue:44734", + "issue:44964", + "issue:45072", + "issue:45084", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45406", + "issue:45446", + "issue:45520", + "issue:45584", + "issue:45588", + "issue:45823", + "issue:46050", + "issue:46121", + "issue:46192", + "issue:46285", + "issue:46369" + ], + "soft_pairs": [ + "issue:44493|issue:46285", + "issue:43606|issue:44112", + "issue:43504|issue:43653", + "issue:44464|issue:45584", + "issue:43994|issue:44493", + "issue:43873|issue:46050", + "issue:45588|issue:46050", + "issue:29942|issue:30333", + "issue:44423|issue:44734", + "issue:43844|issue:44964", + "issue:43065|issue:46369", + "issue:42490|issue:43696", + "issue:45406|issue:46121", + "issue:29159|issue:32090", + "issue:28218|issue:34567", + "issue:45084|issue:45520", + "issue:44554|issue:45446", + "issue:15354|issue:38175", + "issue:43540|issue:45200", + "issue:45200|issue:45823", + "issue:45072|issue:45198", + "issue:43493|issue:43701", + "issue:37428|issue:43957", + "issue:44186|issue:46192", + "issue:44521|issue:45290" + ], + "prepared_review_unit_hash": "278f8e2b89352882816713f371019683b7177f89a5ecd484ae0d9e24416bccfd", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 25, + "serialized_chars": 34776, + "estimated_input_tokens": 8694, + "estimated_eval_tokens": 17644 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of unrelated Transformers bugs. None of the soft pairs look like the same underlying defect/change, so I would not merge any of them as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43957 is the best anchor: it describes a broad, concrete model-loading regression with `torch.device(\"meta\")` and clear examples, making it the most representative issue in this mixed set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43957 has the strongest global triage value because it is reproducible, well-scoped, and impacts multiple models through one loading path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44493", + "right": "issue:46285", + "accept": false, + "reason": "Different bugs: one is a positional-id/key mismatch across many models, the other is a CLIP text-model structure inconsistency." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to a stale GraniteSpeech CI test." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Different components and failure modes: BEiT legacy-field loading vs BigBirdTokenizer special-token decoding." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Compiled-forward chunked generation and Whisper empty-transcription handling are separate code paths." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 bad outputs are not the same issue as the generic unexpected key/position-id problem." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Both touch quantization/offload, but one is an offloading behavior bug and the other is a save/load regression for sinq quantization." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "Flash-attention sinkless-model crash is unrelated to sinq quantized model persistence." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention 2 test failures and MLflow job-reporting are unrelated subsystems." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Both are serve-related, but one is multimodal input handling and the other is KV-cache continuation indexing." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "DeepSpeed gradient growth and Phi-4 multimodal loading are different failure classes." + }, + { + "left": "issue:43065", + "right": "issue:46369", + "accept": false, + "reason": "A dummy Conv2d implementation is not the same as a memory-overflow report for the same model family." + }, + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Local-path import warnings are unrelated to a GPT-oss-20b OOM." + }, + { + "left": "issue:45406", + "right": "issue:46121", + "accept": false, + "reason": "Gemma4 serve tokenizer state and rope-parameter JSON type handling are separate bugs." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency is unrelated to Trainer GPU broadcast NoneType errors." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Added-token spacing and TrainerState token-count tracking are different defects." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Torch compile template-node errors are unrelated to flash_attn import-key errors." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "An MPS attention correctness issue is not the same as a flex_attention import-version check bug." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero probabilities are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video-input processing in Qwen3OmniMoe is unrelated to Gemma4 text-only mm_token_type_ids defaults." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Both involve Gemma4, but one is token-type defaults and the other is a forward device-mismatch bug." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "bfloat16 dtype mismatches and Wav2Vec2 save/tokenization failures are different code paths." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "Flash-support import errors are separate from meta-device model-loading regressions." + }, + { + "left": "issue:44186", + "right": "issue:46192", + "accept": false, + "reason": "LayoutLMv2 tokenizer crashes on padding/truncation are not the same as return_tensors being ignored with text_kwargs." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both involve apply_chat_template, but one is all-zero multimodal assistant masks and the other is a crash on tool-call messages with no content; too different to merge." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 26, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:28218", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:43232", + "issue:43295", + "issue:43334", + "issue:43388", + "issue:43450", + "issue:43452", + "issue:43493", + "issue:43525", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43575", + "issue:43653", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43950", + "issue:43957", + "issue:43975", + "issue:44112", + "issue:44466", + "issue:44554", + "issue:44610", + "issue:44792", + "issue:44849", + "issue:44898", + "issue:44987", + "issue:45127", + "issue:45200", + "issue:45325", + "issue:45356", + "issue:45381", + "issue:45412", + "issue:45520", + "issue:45563", + "issue:45823", + "issue:46178", + "issue:46192", + "issue:46209", + "issue:46285", + "issue:46512" + ], + "soft_pairs": [ + "issue:43232|issue:44849", + "issue:45823|issue:46192", + "issue:43295|issue:43450", + "issue:43295|issue:44610", + "issue:45381|issue:46209", + "issue:43295|issue:43873", + "issue:44554|issue:45356", + "issue:39692|issue:46192", + "issue:43572|issue:44987", + "issue:45520|issue:46178", + "issue:43550|issue:43844", + "issue:44466|issue:45127", + "issue:43572|issue:43957", + "issue:37428|issue:43824", + "issue:44898|issue:45412", + "issue:43950|issue:46285", + "issue:43450|issue:43575", + "issue:45200|issue:45325", + "issue:42490|issue:43493", + "issue:44792|issue:45563", + "issue:43452|issue:46512", + "issue:43334|issue:43525", + "issue:42915|issue:43824", + "issue:38175|issue:43493", + "issue:43824|issue:43975", + "issue:28218|issue:35141", + "issue:43388|issue:43526", + "issue:43653|issue:44112" + ], + "prepared_review_unit_hash": "13454338c9c4aa0e9eb45f3c6b382e29efc0972eddfab5bc2ed0542084b67f45", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36535, + "estimated_input_tokens": 9134, + "estimated_eval_tokens": 18524 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated Transformers bug reports with a few superficial theme overlaps (tokenizers, model loading, multimodal processors). I don\u2019t see any strong duplicate group; the closest items still describe different code paths and model families.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 43295 is the most central candidate by soft-link connectivity and is broad enough to serve as the cluster\u2019s representative, even though it is not a true duplicate of the others.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43295 is the best single representative: it has the most adjacency in the soft graph and describes a user-facing regression rather than a narrowly model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Different bugs: generation kwarg handling after sync_gpus vs output_hidden_states behavior in Qwen3.5." + }, + { + "left": "issue:45823", + "right": "issue:46192", + "accept": false, + "reason": "Device-map forward mismatch in Gemma4 PLE is unrelated to return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor/tokenizer regression and batched video processor shape bug affect different code paths and symptoms." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Custom processor.tokenizer regression is not the same as OmDet-Turbo\u2019s 640px vs 224px preprocessing mismatch." + }, + { + "left": "issue:45381", + "right": "issue:46209", + "accept": false, + "reason": "Qwen2.5-VL vision_position_ids bug is unrelated to FP8 inference failure on B200." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor/tokenizer access regression and quantization offloading behavior are different problems." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness issue is unrelated to Kimi-K2.5 tokenizer codec handling." + }, + { + "left": "issue:39692", + "right": "issue:46192", + "accept": false, + "reason": "Doc example/model mismatch and return_tensors being ignored are different bugs." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to loading physical-intelligence/fast on 5.1.0." + }, + { + "left": "issue:45520", + "right": "issue:46178", + "accept": false, + "reason": "flash_attn import lookup on Python 3.13 is unrelated to tensorflow_text backend mapping." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "torch.compile+SDPA failure and DeepSpeed ZeRO-3 gradient growth are distinct training issues." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Both mention tied weights, but one is serialization/device-dependent and the other is LoRA merge with extended vocab." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Config missing pad_token_idx and meta-device loading breakage are different regressions." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "These are separate import errors in different modules and for different symbols/models." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver interpolation failure and RT-DETR memory release are unrelated." + }, + { + "left": "issue:43950", + "right": "issue:46285", + "accept": false, + "reason": "Non-persistent buffer corruption on load and CLIP text-model structure inconsistency are different loading/serialization bugs." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Batched video processor shape bug and tensor-parallel OOM are not the same underlying issue." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Gemma4 mm_token_type_ids defaulting and Qwen2.5-VL temporal scaling are distinct multimodal bugs." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "Trailing-slash import warning is unrelated to the SigLIP2 implementation discrepancy." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus generate-image test failure and stale paged-generate warning are different generate-path issues." + }, + { + "left": "issue:43452", + "right": "issue:46512", + "accept": false, + "reason": "gguf_file breaking from_pretrained and dtype metadata going stale in from_config are separate loading behaviors." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Both mention missing pad_token_id, but they affect different model families and there is no evidence of one shared duplicate bug." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to the Qwen2_5_VL import error." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Both involve SigLIP2, but one reports zero probabilities while the other is a broader HF-vs-JAX discrepancy; not enough to confirm the same bug." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "Different model families and different failure modes: import error vs incorrect detokenization." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added token and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Label dropping in gather_for_metrics and BeitImageProcessorFast reduce_labels are different label-handling bugs." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and a stale CI device-override test 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": 20, + "review_unit_index": 27, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:29942", + "issue:32090", + "issue:37428", + "issue:38175", + "issue:41720", + "issue:43262", + "issue:43278", + "issue:43335", + "issue:43425", + "issue:43450", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43819", + "issue:43827", + "issue:43856", + "issue:43901", + "issue:44112", + "issue:44206", + "issue:44442", + "issue:44448", + "issue:44514", + "issue:44521", + "issue:44561", + "issue:44610", + "issue:44743", + "issue:44857", + "issue:45083", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45412", + "issue:45520", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45759", + "issue:46170", + "issue:46192", + "issue:46209", + "issue:46234", + "issue:46285", + "issue:46326", + "issue:46402", + "issue:46512" + ], + "soft_pairs": [ + "issue:43278|issue:43856", + "issue:43526|issue:46192", + "issue:41720|issue:46209", + "issue:44442|issue:45198", + "issue:43262|issue:43450", + "issue:45520|issue:46170", + "issue:38175|issue:45412", + "issue:43819|issue:44112", + "issue:44561|issue:46234", + "issue:44206|issue:45584", + "issue:45759|issue:46285", + "issue:45290|issue:46326", + "issue:43827|issue:43901", + "issue:44610|issue:44857", + "issue:44514|issue:44521", + "issue:43819|issue:45083", + "issue:44857|issue:45593", + "issue:43335|issue:43550", + "issue:43611|issue:46402", + "issue:43606|issue:45083", + "issue:37428|issue:43425", + "issue:45759|issue:46512", + "issue:29942|issue:32090", + "issue:44448|issue:44743", + "issue:45200|issue:45588", + "issue:43335|issue:43653" + ], + "prepared_review_unit_hash": "0cc1120fa0d1aa4ff343abf005aba2279411baba8ded7b8e844be9892fc46496", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35069, + "estimated_input_tokens": 8768, + "estimated_eval_tokens": 17792 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of unrelated issue reports. I did not find any true duplicate pair; the closest match is the Qwen2.5-VL apply_chat_template pair, but the failure modes are still distinct enough to keep separate.", + "confidence": 0.31, + "canonical_issue_reason": "issue:44521 is the broadest, highest-signal representative of the multimodal apply_chat_template/assistant_mask thread and has the most follow-up discussion among the related subset.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 is the best single issue anchor because it is open, active, and spans the widest version of the chat-template masking problem, even though the overall cluster is heterogeneous.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Different subsystems and symptoms: dtype regression in evaluation vs MoE training memory usage." + }, + { + "left": "issue:43526", + "right": "issue:46192", + "accept": false, + "reason": "Unrelated bugs: BeitImageProcessor label reduction vs return_tensors being ignored in explicit text_kwargs." + }, + { + "left": "issue:41720", + "right": "issue:46209", + "accept": false, + "reason": "Different model/runtime failures: Qwen3 auto device mapping cuda assert vs FP8 inference on B200." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Both are tokenizer/loading problems, but they affect different models and appear to have different root causes." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate bug vs batched video shape bug; no shared underlying code path is evident." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Unrelated import/environment failures: flash_attn lookup on Python 3.13 vs Windows ROCm distributed dependency issue." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "Different models and behaviors: zero probabilities in SigLIP2 vs RT-DETR memory not being released." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC latent/STE mismatch is unrelated to GraniteSpeech's stale device override test failure." + }, + { + "left": "issue:44561", + "right": "issue:46234", + "accept": false, + "reason": "Both are missing/removed utility symbols, but they affect different functions and code paths." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Different audio bugs: unsupported center arg in feature extraction vs empty-transcription failure after align_special_tokens." + }, + { + "left": "issue:45759", + "right": "issue:46285", + "accept": false, + "reason": "Different model-construction regressions: Qwen multimodal config unwrapping vs CLIP text-model structural inconsistency." + }, + { + "left": "issue:45290", + "right": "issue:46326", + "accept": false, + "reason": "Both involve apply_chat_template, but one is a tool-call/content crash and the other is mistral_common continue_final_message behavior." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Documentation regressions in the same area, but they refer to different deprecated fields and are not the same issue." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Unrelated vision bugs: processor input size mismatch vs float16 AMP loss crash." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Same general API, but one is a batched padding=False crash and the other is all-zero assistant masks; not clearly the same bug." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "Different audio/model concerns: DAC latent decoding mismatch vs helper length computation in qwen3_omni_moe." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "Different detection training issues: AMP crash vs denoising-off auxiliary-loss logic." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Different models and failure modes: SwitchTransformers config bug vs Bamba torch.compile SDPA failure." + }, + { + "left": "issue:43611", + "right": "issue:46402", + "accept": false, + "reason": "Both concern loading regressions, but the specific breakages and affected models are different." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "Device-mismatch offload bug is unrelated to qwen3_omni_moe feature-length computation." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Flash-attention import breakage and Torch 2.10 incompatibility are different environment/version issues." + }, + { + "left": "issue:45759", + "right": "issue:46512", + "accept": false, + "reason": "Both touch from_config/config state, but one is multimodal text_config unwrapping and the other is dtype staleness." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 test failures and Trainer broadcast NoneType errors are different bugs." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Pegasus output drift between v4/v5 is unrelated to Qwen3.5 recurrent-state reset behavior." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Gemma mm_token_type_ids defaulting and flash_attention sink-less-model crash are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Different components and symptoms: SwitchTransformers sparse-layer creation vs BigBirdTokenizer special-token registration." + } + ] + }, + "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": 20, + "review_unit_index": 28, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:26413", + "issue:28218", + "issue:29127", + "issue:31501", + "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:43720", + "issue:43819", + "issue:43825", + "issue:43854", + "issue:44112", + "issue:44206", + "issue:44351", + "issue:44360", + "issue:44448", + "issue:44493", + "issue:44561", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:45072", + "issue:45083", + "issue:45127", + "issue:45245", + "issue:45290", + "issue:45310", + "issue:45381", + "issue:45459", + "issue:45478", + "issue:45588", + "issue:45593", + "issue:45823", + "issue:45859", + "issue:45920", + "issue:46192", + "issue:46234" + ], + "soft_pairs": [ + "issue:45127|issue:45245", + "issue:43550|issue:44112", + "issue:44855|issue:45083", + "issue:37428|issue:44351", + "issue:43335|issue:43504", + "issue:43819|issue:44493", + "issue:43278|issue:43825", + "issue:39692|issue:43526", + "issue:44206|issue:45823", + "issue:29127|issue:32090", + "issue:26413|issue:31501", + "issue:45588|issue:45859", + "issue:43335|issue:45083", + "issue:43720|issue:44112", + "issue:45459|issue:46234", + "issue:45310|issue:45478", + "issue:43335|issue:44112", + "issue:43335|issue:45593", + "issue:45290|issue:45381", + "issue:39692|issue:44493", + "issue:44448|issue:44779", + "issue:43606|issue:44857", + "issue:34689|issue:35141", + "issue:42490|issue:43854", + "issue:44360|issue:44561", + "issue:44857|issue:45072", + "issue:38175|issue:44493", + "issue:28218|issue:32090", + "issue:45920|issue:46192", + "issue:40444|issue:43526" + ], + "prepared_review_unit_hash": "75758e6ae6071f16f485c0ad99a716406a6e3b84595508fcc8b6501e19d99541", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36962, + "estimated_input_tokens": 9241, + "estimated_eval_tokens": 18738 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a set of unrelated issue reports across different bugs and subsystems. Only the two identical Qwen3.5 MoE from_pretrained reports look like true duplicates; the rest are too semantically distant to merge.", + "confidence": 0.95, + "canonical_issue_reason": "Pick issue:45310 as the canonical issue: it is the earlier report of the Qwen3.5 MoE from_pretrained regression and has essentially the same title/root problem as 45478, making it the best anchor for the duplicate pair.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue:45310 is the best representative because it states the concrete regression clearly and is the original of the only true duplicate pair in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "Different failure modes: LoRA/vocabulary merge collapse vs category-count limit in labeling." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "One is an SDPA/torch.compile model bug; the other is a CI stale device-override test failure." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Unrelated models and code paths: DebertaV2 import parsing vs qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different modules; not the same underlying bug." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs BEiT legacy-field loading issue." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC latent decoding mismatch is unrelated to position-id key warnings." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype drift during eval vs a pipeline error-message regression." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "SigLIP2 docs/quantization issues are unrelated to BEiT fast image processor label handling." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "Different models and bugs: unsupported feature extractor arg vs device-mismatch forward pass." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMV3 error-message clarity vs Trainer NoneType broadcast failure." + }, + { + "left": "issue:26413", + "right": "issue:31501", + "accept": false, + "reason": "Checkpoint resume iterator exhaustion is not the same as FineWeb training not starting." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Flash-attention sinkless-model crash vs Qwen3_5Moe rotary embedding CPU-offload incompatibility." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "Different bugs in different subsystems; config creation vs helper length computation." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "BitNet packed-weight loading bug vs GraniteSpeech stale CI override test." + }, + { + "left": "issue:45459", + "right": "issue:46234", + "accept": false, + "reason": "Tokenizer protobuf error masking vs missing import_utils symbol; different fixes." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Exact same Qwen3.5 MoE from_pretrained regression title and underlying failure." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug and GraniteSpeech CI failure are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "Config bug vs missing auxiliary losses in D-FINE; different code paths." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Chat-template tool-call crash vs Qwen2.5-VL video position-id issue." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "Documentation/model-processor mismatch is not the same as unexpected position-id keys." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus v4/v5 output divergence vs DeepSeek tokenizer regression; distinct tokenizer/model behavior." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU-offload device mismatch vs AMP float16 loss crash; different runtime paths." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Llama 3.2 Vision load regression vs post_init reinitializing output embeddings." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path import warning vs GLM-4.7-Flash unit-test load failure." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer activation issue is unrelated to removal of is_torch_fx_available." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "Different dtype/device mismatches affecting different models and execution modes." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Zero-probability SigLIP2 issue is not the same as unexpected position-id keys." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer spacing bug is unrelated to Trainer broadcast NoneType error." + }, + { + "left": "issue:45920", + "right": "issue:46192", + "accept": false, + "reason": "Wrong AutoTokenizer IDs across models vs return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "IterableDataset multi-image fine-tuning failure vs BEiT Fast reduce_labels bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analysis is grounded in the packet and stays conservative on the soft-edge calls. The only clearly duplicate pair here is issue:45310 and issue:45478, which have identical titles and a strong similarity score, so choosing 45310 as canonical is reasonable. The remaining pair verdicts are appropriately cautious given the titles/body summaries available." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 29, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:25251", + "issue:27957", + "issue:29942", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:41720", + "issue:42490", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43412", + "issue:43550", + "issue:43575", + "issue:43701", + "issue:43746", + "issue:43873", + "issue:43927", + "issue:43976", + "issue:44448", + "issue:44466", + "issue:44561", + "issue:44568", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:44987", + "issue:45083", + "issue:45198", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45440", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45698", + "issue:45803", + "issue:45823", + "issue:46209", + "issue:46399" + ], + "soft_pairs": [ + "issue:43701|issue:45083", + "issue:44977|issue:45229", + "issue:44448|issue:44977", + "issue:32090|issue:33357", + "issue:29942|issue:35141", + "issue:43746|issue:46209", + "issue:44855|issue:46399", + "issue:43550|issue:45083", + "issue:42490|issue:43575", + "issue:43927|issue:45698", + "issue:10000|issue:15354", + "issue:27957|issue:34689", + "issue:44561|issue:44779", + "issue:37428|issue:44987", + "issue:43873|issue:45229", + "issue:43278|issue:43329", + "issue:44857|issue:45588", + "issue:41720|issue:45229", + "issue:44977|issue:45563", + "issue:44568|issue:45491", + "issue:43278|issue:43381", + "issue:45588|issue:45823", + "issue:43976|issue:45440", + "issue:45372|issue:45588", + "issue:43412|issue:45588", + "issue:45372|issue:45406", + "issue:44448|issue:44466", + "issue:45198|issue:45803", + "issue:25251|issue:32090", + "issue:45245|issue:45563", + "issue:32090|issue:33290" + ], + "prepared_review_unit_hash": "a0eb75a583b492787f9b8ed3f56d76175f1baeaeebdf15e0a5c9be010ba0f35b", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37166, + "estimated_input_tokens": 9292, + "estimated_eval_tokens": 18840 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a duplicate set: the items span unrelated bugs in tokenization, flash-attention, model loading, quantization, export, and training/runtime behavior. None of the soft pairs look like the same underlying bug or change, so no mergeable duplicate canon emerges.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue fits the cluster: the soft-linked items describe different code paths and failure modes, so choosing one would incorrectly collapse unrelated reports.", + "canonical_pr_reason": "No pull requests are present in the cluster.", + "best_issue_reason": "No issue is a strong global canonical candidate because there is no coherent duplicate group; the pairwise similarities are title-level only and the underlying problems differ.", + "best_pr_reason": "No pull requests are present in the cluster.", + "soft_edge_verdicts": [ + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint key mismatch on resume_from_checkpoint is unrelated to qwen3_omni_moe feature-extraction length behavior." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Flash-attention generation issues for Qwen3.5 are not the same bug as Gemma4 multi-GPU CUDA OOM." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus v4/v5 output differences are a different problem from Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer _gpu_broadcast_one NoneType error is unrelated to the MacOS bus error with a community CLIP model." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash-Attention 2 test failures are unrelated to token embedding reinitialization after resizing." + }, + { + "left": "issue:43746", + "right": "issue:46209", + "accept": false, + "reason": "PEFT local checkpoint loading is not the same as FP8 inference failure on B200/SM100." + }, + { + "left": "issue:44855", + "right": "issue:46399", + "accept": false, + "reason": "Python 3.13 IndentationError in a JIT-scripted model is unrelated to a rotary-pos-emb registration decorator bug in Qwen3.5." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure is not the same as qwen3_omni_moe feature-length calculation behavior." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "Trailing-slash local path import warning is unrelated to tensor-parallel loading OOM for Qwen2-57B-A14B." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "DiaConfig custom token ID loss after save/load is a different bug from loading the wrong custom module after save_pretrained." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel error and GeneratorExp/JIT export error are different model/runtime failures." + }, + { + "left": "issue:27957", + "right": "issue:34689", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 support and Llama 3.2 90B Vision model-loading regression are unrelated." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Removal of is_torch_fx_available affecting trust_remote_code is unrelated to DeepSeek tokenizer output changes." + }, + { + "left": "issue:37428", + "right": "issue:44987", + "accept": false, + "reason": "Missing _flash_supports_window_size import and transformers>=5.1.0 load failure for physical-intelligence/fast are different import/load issues." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Quantization offloading behavior and Gemma4 multi-GPU CUDA OOM are not the same failure mode." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype drift between train/eval is unrelated to undefined variables in the video branch of _get_num_multimodal_tokens." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "LwDetrImageLoss AMP/CUDA crash is not the same as the flash_attention.py s_aux=None AttributeError." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3 device-mapping cudaErrorAssert and Gemma4 multi-GPU CUDA OOM are different GPU placement/runtime problems." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Qwen3.5 flash-attention generation failure is unrelated to the stale num_return_sequences warning in paged generate()." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "mdeberta-v3-base special-token handling is unrelated to Gemma3 sliding-window NaN embeddings." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Training-to-eval embedding dtype change is not the same as gradient checkpointing being disallowed in eval mode." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "A flash_attention.py crash on sink-less models is unrelated to Gemma4 PLE device mismatch under device_map=auto." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to DeepseekV3 MoE divergence from the remote implementation." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 processor loading import failure is unrelated to the flash_attention.py s_aux=None crash." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Gemma3n executorch export failures are unrelated to the flash_attention.py AttributeError." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Gemma4 processor import failure is not the same as transformers serve crashing because Gemma4Processor lacks _tokenizer." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Pegasus output changes and lm_head.weight serialization inconsistency are different v5 regressions." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save/tokenization failure is unrelated to FuyuBatchFeature bare-except error handling." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline top_k nesting behavior is unrelated to the Trainer NoneType broadcast error." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-count RuntimeError is unrelated to a stale warning in paged generate()." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer broadcast NoneType error is unrelated to Adafactor OOM under deepspeed." + } + ] + }, + "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": 20, + "review_unit_index": 30, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:27957", + "issue:29127", + "issue:29942", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:35141", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43381", + "issue:43421", + "issue:43526", + "issue:43575", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43824", + "issue:43873", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44279", + "issue:44361", + "issue:44442", + "issue:44462", + "issue:44479", + "issue:44743", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45593", + "issue:45712", + "issue:45834", + "issue:46121", + "issue:46369" + ], + "soft_pairs": [ + "issue:43653|issue:44743", + "issue:42490|issue:43873", + "issue:43065|issue:45412", + "issue:44206|issue:44479", + "issue:33357|issue:39692", + "issue:43065|issue:43994", + "issue:44361|issue:45198", + "issue:29127|issue:34567", + "issue:44186|issue:45198", + "issue:43611|issue:46121", + "issue:42915|issue:42994", + "issue:44871|issue:44977", + "issue:45459|issue:45712", + "issue:43278|issue:43575", + "issue:45412|issue:46369", + "issue:29942|issue:30064", + "issue:45372|issue:45459", + "issue:43994|issue:45412", + "issue:43381|issue:43701", + "issue:43066|issue:44462", + "issue:43756|issue:45440", + "issue:44265|issue:45198", + "issue:43653|issue:44442", + "issue:43421|issue:45491", + "issue:27957|issue:29942", + "issue:29127|issue:33357", + "issue:45440|issue:45593", + "issue:43994|issue:46369", + "issue:43526|issue:45834", + "issue:43526|issue:44186", + "issue:29127|issue:35141", + "issue:43824|issue:44279" + ], + "prepared_review_unit_hash": "fcb533f13885723d28301aad01678cf1ed00841d23027992e7bb8dbfd357644a", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36636, + "estimated_input_tokens": 9159, + "estimated_eval_tokens": 18574 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is highly heterogeneous, and none of the soft pairs look like the same underlying bug or change with enough specificity to merge them. The strongest anchor issue is the broad model-loading regression in transformers 5.0.0.", + "confidence": 0.78, + "canonical_issue_reason": "issue:43611 is the broadest and most central bug report in the set: a concrete Transformers 5.0.0 model-loading regression affecting the common `base_model_prefix` path, making it a reasonable dedupe anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43611 is the most generally useful issue to keep as the representative bug: it is specific, user-facing, and plausibly impacts many models through a shared loading code path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "Tokenizer special-token decode bug vs Qwen recurrent-state reset bug; unrelated code paths and symptoms." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Path-with-slash import warning is unrelated to quantization/offloading behavior." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d placeholder in Sam3PixelDecoder is not the same as RT-DETR memory not being released." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Different regressions in different components/models; not the same concrete failure." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "MacOS bus error with a community CLIP model is distinct from SigLIP2 documentation example errors." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Placeholder layer implementation is unrelated to SigLIP2 producing nonsensical outputs." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "MLukeTokenizer AttributeError and Wav2Vec2 save/tokenization failure are different tokenizer/model bugs." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "Clarity of a LayoutLMv3 error message is not the same bug as TrainerState token counter not updating." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2 NER/padding crash is unrelated to Wav2Vec2 save_pretrained/tokenization failure." + }, + { + "left": "issue:43611", + "right": "issue:46121", + "accept": false, + "reason": "Base-model-prefix loading regression is a different problem from rope-parameter JSON list handling." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Qwen3Moe FP8 failure and quantized model saving failure are different quantization issues." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "EOS token config mismatch is not the same as flash-attention generation regression." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "protobuf-hidden tokenizer errors and leftover dummy classes are separate maintenance/runtime issues." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Embedding dtype drift in eval is unrelated to tensor-parallel OOM on Qwen2-57B." + }, + { + "left": "issue:45412", + "right": "issue:46369", + "accept": false, + "reason": "Memory leak on deletion is not the same as model memory overflow during SAM3-liteText inference." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention 2 test failures and void segmentation-map processing are different feature areas." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "ReasoningEffort import failure affects Gemma 4 processor loading; protobuf exception masking is a separate tokenizer error-handling bug." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 inference quality bug is unrelated to RT-DETR object lifetime/memory release." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Eval-mode gradient checkpointing restriction is not the same as checkpoint key mismatch on resume." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Wrong tokenizer decoder type in v5 and ignoring repository tokenizer.json are distinct tokenizer-loading issues." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Smollm3 RoPE-layer count mismatch is unrelated to DeepSeekV3MoE divergence from remote implementation." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export failure on torch_compilable_check is unrelated to Wav2Vec2 save/tokenization problems." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "BigBird mask-token decode issue and AutoTokenizer loading FastSpeech2ConformerTokenizer are different failures." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime special-token post-processor update bug is unrelated to Gemma3 NaN embeddings from sliding-window attention." + }, + { + "left": "issue:27957", + "right": "issue:29942", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 support issue is not the same as generic Flash Attention 2 test failures." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and MacOS CLIP bus error are unrelated." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "Remote-vs-native DeepSeekV3MoE divergence is unrelated to D-FINE auxiliary-loss behavior." + }, + { + "left": "issue:43994", + "right": "issue:46369", + "accept": false, + "reason": "SigLIP2 incorrect outputs and SAM3-liteText memory overflow are distinct model bugs." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug and Kosmos2.5 OCR index error are not the same underlying issue." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "reduce_labels returning one label and LayoutLMv2Tokenizer crashes are different processor/tokenizer defects." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 error clarity and output embedding reinitialization after resize are unrelated bugs." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Specific Qwen2.5-VL import error and a generic dependency issue are not the same concrete failure." + } + ] + }, + "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": 20, + "review_unit_index": 31, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:27957", + "issue:29127", + "issue:33290", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43316", + "issue:43334", + "issue:43381", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43646", + "issue:43701", + "issue:43723", + "issue:43761", + "issue:43873", + "issue:43931", + "issue:44062", + "issue:44077", + "issue:44206", + "issue:44387", + "issue:44442", + "issue:44464", + "issue:44625", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44898", + "issue:44977", + "issue:45003", + "issue:45072", + "issue:45245", + "issue:45440", + "issue:45479", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45859", + "issue:45907", + "issue:46121", + "issue:46399" + ], + "soft_pairs": [ + "issue:44625|issue:45584", + "issue:43575|issue:43701", + "issue:44442|issue:45072", + "issue:42915|issue:43931", + "issue:42490|issue:43381", + "issue:43575|issue:44387", + "issue:43761|issue:45859", + "issue:46121|issue:46399", + "issue:43526|issue:44857", + "issue:43334|issue:43531", + "issue:44977|issue:45245", + "issue:43262|issue:44206", + "issue:44464|issue:45491", + "issue:27957|issue:29127", + "issue:27957|issue:33290", + "issue:45588|issue:45600", + "issue:44743|issue:44811", + "issue:43646|issue:44077", + "issue:33290|issue:34689", + "issue:33290|issue:35141", + "issue:33290|issue:34567", + "issue:43381|issue:43873", + "issue:44442|issue:44898", + "issue:44062|issue:45907", + "issue:45003|issue:45440", + "issue:43723|issue:45479", + "issue:42490|issue:44743", + "issue:43316|issue:43525" + ], + "prepared_review_unit_hash": "490b759b4e8b230283067065914aed512df836ddcc070ce079e70feef869a7f8", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35465, + "estimated_input_tokens": 8867, + "estimated_eval_tokens": 17990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set does not look like a single duplicate cluster; it is a grab bag of unrelated issues across models, tokenizer/config loading, training behavior, audio, and quantization. I would not merge any of the suggested soft pairs.", + "confidence": 0.21, + "canonical_issue_reason": "issue:43873 is the most discussion-heavy and externally referenced item in the set, so it is the best available anchor if one must pick a representative issue. That said, the overall group is too heterogeneous to have a strong canonical duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the strongest single issue candidate by activity and inbound references, but it is not truly representative of the whole set because the set is not a coherent duplicate cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "Different models and failures: Qwen3.5 config propagation vs Whisper empty-transcription generation." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "OOM on tensor-parallel model load is unrelated to resume_from_checkpoint key mismatch." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer loading bug vs dtype mismatch in bf16 inference; different code paths." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and Qwen3-VL weight-shape mismatch are different model/config issues." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "Local-path import warning is unrelated to gradient checkpointing in eval mode." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory, but one is TP load OOM and the other is int4 reserved-memory regression; not the same bug." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIPVisionModel hidden-states regression and Qwen3.5Moe CPU-offload incompatibility are unrelated." + }, + { + "left": "issue:46121", + "right": "issue:46399", + "accept": false, + "reason": "Rope-validation TypeError and missing decorator for apply_rotary_pos_emb are different implementation problems." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to LwDetrImageLoss AMP/CUDA crash." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Missing pad_token_id on Qwen3-VL text config is different from sliding-window behavior in Qwen3-MoE." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Flash-attention generation regression and category-cardinality RuntimeError are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio chat-template sampling-rate default and unsupported center argument in feature extraction are different bugs." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency is not the same as Gemma3 NaN embeddings from sliding-window padding." + }, + { + "left": "issue:27957", + "right": "issue:29127", + "accept": false, + "reason": "XLMRoberta flash-attention support and LayoutLMv3 box-validation messaging are unrelated." + }, + { + "left": "issue:27957", + "right": "issue:33290", + "accept": false, + "reason": "Flash Attention 2 support and deepspeed Adafactor OOM are different problems." + }, + { + "left": "issue:45588", + "right": "issue:45600", + "accept": false, + "reason": "Flash-attention sinkless-model crash and removed-config CI breakage do not share the same code path." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Qwen3.5 recurrent-state reset and Whisper batch_decode skip_special_tokens are unrelated." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "General Transformers 5.0 custom initialization breakage is broader than the PatchTSMixer post_init-specific change." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Adafactor/deepspeed OOM and Llama 3.2 Vision model-loading regression are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "OOM during optimizer use is unrelated to output embedding reinitialization after resize." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "Memory exhaustion and TrainerState token-count tracking are different issues." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Eval-mode gradient checkpointing restriction is unrelated to quantization offloading behavior." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading failure and Perceiver image-resolution failure are different subsystems." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken special-argument duplication is unrelated to list_repo_files revision forwarding." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Unsafe sys.modules access and DeepseekV3MoE divergence are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer loading in v5 and zero-loss single-label classification are different bugs." + }, + { + "left": "issue:42490", + "right": "issue:44743", + "accept": false, + "reason": "Import-path trailing-slash warning is unrelated to Qwen3.5 recurrent-state reset." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy and Llama4Config missing pad_token_id are similar in theme but 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": 1, + "component_count": 20, + "review_unit_index": 32, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:25251", + "issue:28282", + "issue:34689", + "issue:42175", + "issue:42617", + "issue:42796", + "issue:42907", + "issue:43232", + "issue:43295", + "issue:43316", + "issue:43388", + "issue:43493", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43611", + "issue:43643", + "issue:43701", + "issue:43720", + "issue:43749", + "issue:43761", + "issue:43825", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44190", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44496", + "issue:44855", + "issue:44871", + "issue:44913", + "issue:44987", + "issue:45137", + "issue:45216", + "issue:45290", + "issue:45698", + "issue:45759", + "issue:45803", + "issue:46178", + "issue:46234", + "issue:46291", + "issue:46421", + "issue:46506" + ], + "soft_pairs": [ + "issue:42175|issue:46234", + "issue:43761|issue:44079", + "issue:34689|issue:43611", + "issue:43720|issue:44442", + "issue:43232|issue:43388", + "issue:43931|issue:45216", + "issue:43761|issue:44188", + "issue:44077|issue:44479", + "issue:43493|issue:43827", + "issue:42907|issue:45698", + "issue:44442|issue:45803", + "issue:44987|issue:46178", + "issue:42796|issue:45290", + "issue:43873|issue:44871", + "issue:43749|issue:43931", + "issue:44464|issue:46421", + "issue:43316|issue:45759", + "issue:43701|issue:46506", + "issue:44190|issue:44496", + "issue:43643|issue:44913", + "issue:44855|issue:45137", + "issue:46234|issue:46291", + "issue:43295|issue:43825", + "issue:25251|issue:28282", + "issue:43550|issue:43701", + "issue:42617|issue:43575", + "issue:43606|issue:43701" + ], + "prepared_review_unit_hash": "ca94a51e7dd40ee98b32550a795c6220dafe650bc65974709b8473b1087df836", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35961, + "estimated_input_tokens": 8991, + "estimated_eval_tokens": 18238 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the issues span unrelated loading, generation, config, tokenizer, backend, and distributed-training bugs. None of the soft pairs look like the same underlying defect, so I would not merge any of them as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No true canonical issue for the whole cluster; if forced, issue:43749 is the strongest standalone report because it is concrete, high-signal, and has the most discussion around a specific regression ('FSDP_CPU_RAM_EFFICIENT_LOADING broken').", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42175", + "right": "issue:46234", + "accept": false, + "reason": "Different dependency/back-end problems: Torch backend packaging vs a missing import utility helper." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "One is a hidden_states regression in forward; the other is ModelOutput key assignment behavior." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both concern loading regressions, but they affect different models and different failure modes." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "Packed-weight loading under accelerate is unrelated to AutoTokenizer failing for a specific tokenizer class." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs syncing and last-batch metric gathering are different code paths and bugs." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Shape-mismatch loading and save_pretrained checkpoint corruption are opposite directions and distinct bugs." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "Hidden_states output regression is unrelated to attention-kernel divergence under torch.compile." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "A post_init API constraint is unrelated to a video-input regression in Qwen VL models." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "Model implementation discrepancy vs documentation still mentioning pipeline(); not the same bug." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Saving dequantized models and loading the wrong custom module after save_pretrained are different save/load failures." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer class loading and tensor conversion exception handling are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:46178", + "accept": false, + "reason": "Generic model-loading failure is not the same as a backend-mapping error for tensorflow_text." + }, + { + "left": "issue:42796", + "right": "issue:45290", + "accept": false, + "reason": "Tokenizer BPE marker handling is unrelated to chat-template crashes with tool calls." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading behavior and eos_token_id config inconsistency are different bugs." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP RAM-efficient loading and Qwen3-VL shape mismatches are unrelated loading problems." + }, + { + "left": "issue:44464", + "right": "issue:46421", + "accept": false, + "reason": "Both mention chunked/compiled generation, but one is inconsistent outputs and the other is recompilation from cache init; not the same concrete defect." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Both are config/API issues, but they concern different model families and different failures." + }, + { + "left": "issue:43701", + "right": "issue:46506", + "accept": false, + "reason": "Checkpoint resume key mismatch is unrelated to CPU/CUDA output divergence." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in an example script is unrelated to an unrecognized model/config error." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Both involve config persistence, but one loses fields with trust_remote_code while the other resets rotary_pct on reload." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 import parsing error and DeepSpeed ZeRO3 deque underflow are unrelated." + }, + { + "left": "issue:46234", + "right": "issue:46291", + "accept": false, + "reason": "Missing import utility function and kernels package version/revision requirement are different import failures." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression and a v5 translation-task error message are not the same bug." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k nesting behavior is unrelated to an ImportError for missing PyTorch." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile SDPA failure and resume_from_checkpoint key mismatch are different runtime paths." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "A 3d_parallel script failure and a tensor-parallel OOM are both distributed issues but not the same concrete bug." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to checkpoint key mismatch during resume." + } + ] + }, + "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": 20, + "review_unit_index": 33, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:27957", + "issue:32090", + "issue:41950", + "issue:42175", + "issue:43065", + "issue:43295", + "issue:43316", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43452", + "issue:43475", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43646", + "issue:43650", + "issue:43673", + "issue:43688", + "issue:43716", + "issue:43742", + "issue:43761", + "issue:44077", + "issue:44162", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44743", + "issue:44792", + "issue:44871", + "issue:44913", + "issue:45127", + "issue:45200", + "issue:45584", + "issue:45632", + "issue:45803", + "issue:46050", + "issue:46170", + "issue:46178", + "issue:46202", + "issue:46365", + "issue:46369", + "issue:46424" + ], + "soft_pairs": [ + "issue:43575|issue:43742", + "issue:27957|issue:32090", + "issue:16998|issue:22355", + "issue:43316|issue:43716", + "issue:45632|issue:46424", + "issue:44610|issue:46369", + "issue:43295|issue:44871", + "issue:42175|issue:43316", + "issue:43650|issue:46202", + "issue:45803|issue:46365", + "issue:44871|issue:45584", + "issue:44077|issue:44534", + "issue:43065|issue:44617", + "issue:43540|issue:44514", + "issue:44162|issue:44792", + "issue:45803|issue:46170", + "issue:44871|issue:46050", + "issue:41950|issue:43475", + "issue:43646|issue:44913", + "issue:43404|issue:43441", + "issue:43452|issue:44496", + "issue:43761|issue:44479", + "issue:45127|issue:45200", + "issue:43388|issue:43688", + "issue:43673|issue:44743", + "issue:43572|issue:46178" + ], + "prepared_review_unit_hash": "60bb6fa6239a3b6ba62891230466edcaa3f8307351f31176672644ceb90b7c8c", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35326, + "estimated_input_tokens": 8832, + "estimated_eval_tokens": 17920 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a duplicate set: it mixes unrelated bugs across model configs, processors, training, packaging/imports, and backend-specific regressions. None of the soft pairs look like the same underlying defect or change.", + "confidence": 0.96, + "canonical_issue_reason": "No single issue is a true canonical for this cluster because the items are heterogeneous. If a placeholder is needed, issue 45584 is the most active/open entry with the most inbound references, but it is not a thematic match for the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45584 is the best placeholder only by engagement and recency; it is not a strong representative of the overall set because the cluster lacks a shared bug theme.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "Different failures: tensor-parallel OOM on Qwen2-57B vs a KeyError while loading MobileLLM." + }, + { + "left": "issue:27957", + "right": "issue:32090", + "accept": false, + "reason": "Unrelated problems: FlashAttention support for XLM-Roberta vs Trainer distributed broadcast NoneType error." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "A model max-length question is unrelated to a missing transformers.onnx import." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Different models and bugs: Gemma3 config API mismatch vs Mistral-3 dtype mismatch." + }, + { + "left": "issue:45632", + "right": "issue:46424", + "accept": false, + "reason": "Cache-path collision for trust_remote_code is unrelated to max_cache_length behavior in quantized cache." + }, + { + "left": "issue:44610", + "right": "issue:46369", + "accept": false, + "reason": "Processor image-size bug vs Sam3 LiteText memory overflow; different code paths and symptoms." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor/tokenizer regression is not the same as Gemma eos_token_id inconsistency." + }, + { + "left": "issue:42175", + "right": "issue:43316", + "accept": false, + "reason": "Packaging/backend dependency issue vs model config API discrepancy; not the same bug." + }, + { + "left": "issue:43650", + "right": "issue:46202", + "accept": false, + "reason": "Placeholder issue title is unrelated to Gemma pipeline roundtrip tokenization failure." + }, + { + "left": "issue:45803", + "right": "issue:46365", + "accept": false, + "reason": "Bare except swallowing errors is unrelated to Trainer import failing due to datasets." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma eos-token config mismatch and Whisper empty-transcription failure are different bugs." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "Optional post_init restriction is unrelated to non-persistent buffer initialization junk." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder is a model-architecture issue, not the Sam3Video OOM problem." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Video-input processing error in Qwen3OmniMoe is different from batched chat-template padding crash." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 model breakage is unrelated to a janus image-generation test failure." + }, + { + "left": "issue:45803", + "right": "issue:46170", + "accept": false, + "reason": "Exception handling bug is unrelated to Windows ROCm import/type dependency issues." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "EOS-token config mismatch is not the same as sinq quantized save/load corruption." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification processor lookup bug is unrelated to Sam3Video missing output attribute." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Broad v5 initialization regression is not the same as GPTNeoX rotary_pct not persisting." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Weight tying bug in Mistral3ForConditionalGeneration is different from FlashAttention failure in Ministral-3." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "Tokenizer/gguf loading breakage is unrelated to an unrecognized model_type config error." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "CLIPVision hidden_states regression is unrelated to Qwen video-input regression." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA extended-vocab collapse and Gemma 4 mm_token_type_ids are distinct model behaviors." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric gathering last-batch label truncation is unrelated to auxiliary-loss normalization." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Missing GenerationMixin cache in chunked_prefill is not the same as recurrent-state reset in Qwen3.5." + }, + { + "left": "issue:43572", + "right": "issue:46178", + "accept": false, + "reason": "StableLm pad_token_idx config regression is unrelated to tensorflow_text backend registration." + } + ] + }, + "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": 20, + "review_unit_index": 34, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:35141", + "issue:41720", + "issue:42490", + "issue:43065", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43644", + "issue:43673", + "issue:43761", + "issue:43827", + "issue:43931", + "issue:43976", + "issue:43994", + "issue:44060", + "issue:44077", + "issue:44190", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44521", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44792", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44933", + "issue:44964", + "issue:44977", + "issue:45245", + "issue:45362", + "issue:45381", + "issue:45406", + "issue:45542", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:46234", + "issue:46369", + "issue:46421" + ], + "soft_pairs": [ + "issue:43504|issue:45698", + "issue:45542|issue:46234", + "issue:44734|issue:45406", + "issue:44617|issue:45584", + "issue:43673|issue:46421", + "issue:43525|issue:43531", + "issue:44792|issue:44977", + "issue:45381|issue:46369", + "issue:43931|issue:44060", + "issue:44792|issue:45600", + "issue:44190|issue:44964", + "issue:44534|issue:44913", + "issue:43761|issue:44242", + "issue:45563|issue:45584", + "issue:35141|issue:42490", + "issue:44479|issue:44734", + "issue:44855|issue:45588", + "issue:43827|issue:43994", + "issue:43065|issue:44610", + "issue:41720|issue:45245", + "issue:44933|issue:45362", + "issue:44617|issue:46369", + "issue:44464|issue:45563", + "issue:43644|issue:44913", + "issue:44521|issue:45381", + "issue:43550|issue:44190", + "issue:43976|issue:44855", + "issue:44077|issue:44877", + "issue:43540|issue:44734" + ], + "prepared_review_unit_hash": "57372d897f6cc4008a2150a0dad8897beb2cc137dad5e779e317dd7661730f6a", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35636, + "estimated_input_tokens": 8909, + "estimated_eval_tokens": 18074 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous, but there is one clear duplicate issue pair about non-persistent buffers being filled with junk in Transformers v5. The rest of the soft pairs describe different model-, feature-, or subsystem-specific failures and should not be merged.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43644 is the better canonical issue: it is earlier, states the underlying bug clearly, and issue:44534 reads like the same regression restated with a slightly different version label.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43644 is the best representative issue in this set because it cleanly captures the shared non-persistent-buffer regression and has the strongest duplicate evidence with issue:44534.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Different problems: legacy field preset loading vs. wrong custom module resolution after save/load." + }, + { + "left": "issue:45542", + "right": "issue:46234", + "accept": false, + "reason": "Unrelated failures: TensorBoard/TF backend installation issue vs. missing import_utils helper symbol." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Different serve crashes with different root causes: KV-cache tensor indexing vs. missing _tokenizer on Gemma4Processor." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Different models and symptoms: Sam3Video OOM vs. Whisper empty-transcription generation failure." + }, + { + "left": "issue:43673", + "right": "issue:46421", + "accept": false, + "reason": "Both touch chunked prefill/cache, but one is a missing cache in v5.0.0 and the other is spurious recompilation from lazy init; not clearly the same bug." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Llama4Config missing pad_token_id and Qwen3-MoE sliding_window behavior are distinct configuration issues." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Unrelated: janus image-generation test failure vs. Qwen3.5 flash-attention generation issue." + }, + { + "left": "issue:45381", + "right": "issue:46369", + "accept": false, + "reason": "Different multimodal regressions: Qwen2.5-VL video position ids vs. Sam3-LiteText memory overflow." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Different model-loading/tied-weight problems with different error surfaces and code paths." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "Janus test failure is unrelated to stale auto_mappings references to removed Sam3LiteText configs." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Dataset loading failure in a training example is unrelated to a multimodal model loading error." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer initialization bug is unrelated to GPTNeoXConfig rotary_pct not persisting on reload." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "Different regressions: CLIPVisionModel hidden_states output vs. MoE load-balancing loss computation." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "Paged-generate warning behavior is unrelated to Whisper failing on empty transcription after token alignment." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Different issues: embedding reinitialization after resize vs. path-with-trailing-slash import warning." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Both involve multimodal/serve code, but the Qwen video-input regression and KV-cache continuation tensor-indexing crash are different bugs." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Different failures: TorchScript parsing/IndentationError on Python 3.13 vs. flash-attention crash on sink-less models." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() is not the same as SigLIP2 producing nonsensical outputs with AutoModel/pipeline." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder is not the same as processor/model input-size mismatch in OmDet-Turbo." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "Completely different failures: Qwen3 auto device mapping cudaErrorAssert vs. category-count limit error." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Nonexistent import from image_utils is unrelated to Qwen3.5 chat crashing." + }, + { + "left": "issue:44617", + "right": "issue:46369", + "accept": false, + "reason": "Related family name only; Sam3Video OOM and Sam3-LiteText memory overflow are different model-specific resource issues." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Chunked generation inconsistency under compiled forward is not the same as a stale num_return_sequences warning." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer initialization regression is unrelated to GPTNeoXConfig rotary_pct persistence." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Assistant-mask generation for multimodal chat templates is different from Qwen2.5-VL video vision_position_ids." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "Bamba SDPA torch.compile failure is unrelated to local dataset loading in a training script." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Python-version support issue is unrelated to the DebertaV2 TorchScript parse error." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "Optional post_init allowance for patchtsmixer and strict config rejection for granite_speech are different config-loading issues." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Qwen3OmniMoe video processing ValueError is unrelated to serve KV-cache indexing crashes." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Same underlying regression: both report Transformers v5 filling non-persistent buffers with junk, with only minor wording/version differences." + } + ] + }, + "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": 20, + "review_unit_index": 35, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:25251", + "issue:27957", + "issue:36246", + "issue:38617", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:42796", + "issue:43116", + "issue:43295", + "issue:43335", + "issue:43404", + "issue:43643", + "issue:44016", + "issue:44112", + "issue:44186", + "issue:44190", + "issue:44246", + "issue:44279", + "issue:44295", + "issue:44462", + "issue:44496", + "issue:44521", + "issue:44683", + "issue:44734", + "issue:44737", + "issue:44779", + "issue:44938", + "issue:45003", + "issue:45127", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45357", + "issue:45446", + "issue:45542", + "issue:45584", + "issue:45600", + "issue:45632", + "issue:45803", + "issue:45907", + "issue:46082", + "issue:46121", + "issue:46234", + "issue:46421", + "issue:46489", + "issue:46506" + ], + "soft_pairs": [ + "issue:45127|issue:45357", + "issue:25251|issue:27957", + "issue:44779|issue:46489", + "issue:43335|issue:44190", + "issue:44462|issue:45907", + "issue:42796|issue:44521", + "issue:44112|issue:45600", + "issue:44737|issue:45803", + "issue:43116|issue:44190", + "issue:45632|issue:46234", + "issue:44938|issue:45542", + "issue:43295|issue:45276", + "issue:45276|issue:45584", + "issue:43404|issue:45216", + "issue:36246|issue:41628", + "issue:45245|issue:45584", + "issue:46082|issue:46121", + "issue:36246|issue:38617", + "issue:45632|issue:46421", + "issue:42175|issue:44279", + "issue:44683|issue:45446", + "issue:44246|issue:45003", + "issue:44186|issue:44734", + "issue:41950|issue:44016", + "issue:44295|issue:46506", + "issue:43643|issue:44496" + ], + "prepared_review_unit_hash": "d6b450ce632b4354cc747bcb6c0449d6dd58084945c219feaa0125bdd083180b", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35135, + "estimated_input_tokens": 8784, + "estimated_eval_tokens": 17824 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the titles overlap on broad themes like tokenizers, imports, pipelines, and multimodal models, but none of the pairs point to the same concrete bug or fix.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "Both involve Qwen-related saving/weights, but one is LoRA merge collapse with tied embeddings while the other is incorrect visual encoder keys in save_pretrained; different failure modes and code paths." + }, + { + "left": "issue:25251", + "right": "issue:27957", + "accept": false, + "reason": "Pipeline top_k output nesting and XLMRoberta Flash Attention 2 are unrelated; one is an output-shape bug, the other a model/attention compatibility issue." + }, + { + "left": "issue:44779", + "right": "issue:46489", + "accept": false, + "reason": "Both mention DeepSeek tokenizer regressions, but they target different model families and reports; the titles do not identify the same tokenizer bug or fix." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers sparse-layer configuration and loading a local dataset in an image-classification script are unrelated issues." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json and list_repo_files not forwarding revision are separate repository-loading bugs with different APIs." + }, + { + "left": "issue:42796", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer BPE-marker handling and multimodal assistant_masks generation are different tokenizer behaviors and do not share a concrete underlying bug." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "One is a stale GraniteSpeech device-override test, the other is auto_mappings referencing removed configs; both affect CI but not the same defect." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet device placement in relative positional encoding and Fuyu's bare except swallowing errors are unrelated code-path problems." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "A multi-label classification example returning empty results is unrelated to a script failing to load a local dataset." + }, + { + "left": "issue:45632", + "right": "issue:46234", + "accept": false, + "reason": "A trust_remote_code cache-path collision and a missing import_utils availability flag are different import/cache bugs." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 import failure and missing TensorFlow backend handling are separate environment/dependency issues." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "The regression around processor.tokenizer/images passed to tokenizer is different from gemma4 resize_token_embeddings not updating output/embed layers." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 embedding resize propagation and Whisper empty-transcription behavior after align_special_tokens are not the same bug." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Mistral3 lm_head tying/save behavior and Qwen3.5 save_pretrained checkpoint correctness are model-specific serialization/weight issues, not one defect." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but one is a missing Qwen2_5_VLImageProcessor export and the other is a missing top-level AutoImageProcessor export; different symbols and fixes." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "The 2^24 categories limit error and Whisper empty transcription after token alignment are unrelated." + }, + { + "left": "issue:46082", + "right": "issue:46121", + "accept": false, + "reason": "LlamaConfig head_dim validation and convert_rope_params_to_dict rejecting a JSON-loaded list are separate config-validation bugs." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Both are import failures, but they concern different missing names in different modules and do not indicate the same underlying change." + }, + { + "left": "issue:45632", + "right": "issue:46421", + "accept": false, + "reason": "A cache collision for trust_remote_code and spurious recompilation from lazy cache initialization are unrelated caching problems." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "TensorFlow missing from a torch install and a vague dependency issue with transformers are not the same specific defect." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Compiled flex_attention failing on torch >= 2.9 and an incorrect version check for AuxRequest import are related to flex_attention but are distinct concrete bugs." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Slow import time and unsafe sys.modules access are different problems in different parts of the import path." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "LayoutLMv2Tokenizer NER/padding crashes and transformers serve KV-cache continuation indexing are unrelated failures." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "A video-classification pipeline looking for image processors and a syntax error in a notebook section are wholly different issues." + }, + { + "left": "issue:44295", + "right": "issue:46506", + "accept": false, + "reason": "Reading position_ids after buffer registration and CPU/CUDA output divergence in PPDocLayoutV3 are separate runtime bugs." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Missing fields when trust_remote_code=True in AutoConfig and an unrecognized model lacking model_type are different config-loading 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": 20, + "review_unit_index": 36, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:27957", + "issue:28282", + "issue:32090", + "issue:35141", + "issue:42796", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43441", + "issue:43479", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43650", + "issue:43673", + "issue:43824", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44220", + "issue:44279", + "issue:44336", + "issue:44368", + "issue:44496", + "issue:44792", + "issue:44933", + "issue:44964", + "issue:44987", + "issue:45042", + "issue:45083", + "issue:45216", + "issue:45335", + "issue:45375", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45600", + "issue:45759", + "issue:46123", + "issue:46369", + "issue:46424", + "issue:46506", + "issue:46515" + ], + "soft_pairs": [ + "issue:43854|issue:44964", + "issue:44964|issue:45600", + "issue:43673|issue:46424", + "issue:44112|issue:44964", + "issue:43650|issue:44220", + "issue:44016|issue:44220", + "issue:27957|issue:35141", + "issue:43479|issue:45375", + "issue:43531|issue:45600", + "issue:43295|issue:45335", + "issue:43976|issue:44987", + "issue:46123|issue:46369", + "issue:43526|issue:43901", + "issue:32090|issue:35141", + "issue:28282|issue:35141", + "issue:42907|issue:43441", + "issue:43441|issue:45588", + "issue:44496|issue:45042", + "issue:44016|issue:44062", + "issue:43299|issue:45216", + "issue:45083|issue:46506", + "issue:44792|issue:46424", + "issue:43673|issue:44336", + "issue:43824|issue:44933", + "issue:43381|issue:45593", + "issue:43976|issue:44279", + "issue:43901|issue:44496", + "issue:42796|issue:43650", + "issue:45335|issue:45584", + "issue:44496|issue:45759", + "issue:44279|issue:46515", + "issue:44075|issue:44368", + "issue:43577|issue:45216" + ], + "prepared_review_unit_hash": "4845aba75cd122bda1703a57dd2a5c4da00b1731669ed30965cf9b4ca3820c25", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37525, + "estimated_input_tokens": 9382, + "estimated_eval_tokens": 19020 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a loose similarity bucket, not a true duplicate cluster. The issues cover many unrelated model-loading, config, tokenizer, cache, and test failures across different models and code paths, so none of the soft links should be merged.", + "confidence": 0.91, + "canonical_issue_reason": "No single canonical issue: the items are too heterogeneous and do not appear to describe the same underlying bug.", + "canonical_pr_reason": "No PRs are present.", + "best_issue_reason": "issue:44964 is the clearest high-level regression report in the set, with concrete reproduction and broad impact on loading a specific multimodal model in recent Transformers versions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Both are model-loading failures, but for different models and different failure modes (missing import vs bad load behavior)." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Different causes: one is a model load regression, the other is a broken auto-mapping reference in CI." + }, + { + "left": "issue:43673", + "right": "issue:46424", + "accept": false, + "reason": "Both mention cache, but one is a generation cache regression and the other is cache-length behavior for quantized cache only." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "Unrelated: GraniteSpeech CI test failure vs a multimodal model loading regression." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "No shared bug: the first is a placeholder-like issue, the second is an audio feature extraction error." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error and feature-extraction TypeError are unrelated." + }, + { + "left": "issue:27957", + "right": "issue:35141", + "accept": false, + "reason": "Different subsystems: FlashAttention support for XLM-Roberta vs embedding reinitialization after resize." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Both are config-related, but they concern different composite-model fields and different initialization bugs." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Different problems: Qwen3-MoE sliding-window behavior vs stale auto-mapping references." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Both are regression-style model issues, but one is processor/tokenizer API breakage and the other is embedding resize behavior." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "These are separate version-compatibility/load failures affecting different environments and models." + }, + { + "left": "issue:46123", + "right": "issue:46369", + "accept": false, + "reason": "Different code paths: mask generation batching vs SAM3-lite memory overflow." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "Beit processor label handling is unrelated to pipeline documentation wording." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer broadcast error and token-embedding resize regression are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "ImportError due to missing PyTorch is not the same as the embedding reinit bug." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Same model family, but different failures: saving dequantized Ministrals vs FlashAttention load/runtime failure." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both touch FlashAttention, but they describe different concrete crashes and code paths." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Different failures: unrecognized model config vs PIL image processors wrongly requiring torchvision." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error and AddedToken keyword conflict are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Different Qwen regressions: model loading failure vs incorrect checkpoint saved by save_pretrained." + }, + { + "left": "issue:45083", + "right": "issue:46506", + "accept": false, + "reason": "Different model families and different bug types: helper length computation vs CPU/CUDA divergence." + }, + { + "left": "issue:44792", + "right": "issue:46424", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to cache-length configuration behavior." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache regression and ANSI color output bug are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Both are import errors, but they refer to different modules and missing symbols." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing eval-mode limitation is unrelated to D-FINE auxiliary-loss handling." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python version incompatibility and generic dependency issue are not the same underlying bug." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Pipeline docs mismatch is unrelated to model config import failure." + }, + { + "left": "issue:42796", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer BPE-marker handling and a placeholder issue are unrelated." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Different behaviors: decoder embedding resize vs Whisper empty-transcription generation failure." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Both involve model/config loading, but one is an unknown-model error and the other is missing unwrapping of composite text_config." + }, + { + "left": "issue:44279", + "right": "issue:46515", + "accept": false, + "reason": "Generic dependency issue is not the same as a torch 2.4.1 incompatibility." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Optimizer argument handling and a tie_word_embeddings warning are different issues." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Blip2 dtype propagation and Qwen checkpoint saving are unrelated 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": 20, + "review_unit_index": 37, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41950", + "issue:42994", + "issue:43295", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43479", + "issue:43525", + "issue:43650", + "issue:43653", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44242", + "issue:44279", + "issue:44291", + "issue:44297", + "issue:44351", + "issue:44462", + "issue:44466", + "issue:44521", + "issue:44617", + "issue:44683", + "issue:44779", + "issue:45020", + "issue:45081", + "issue:45103", + "issue:45200", + "issue:45276", + "issue:45278", + "issue:45325", + "issue:45357", + "issue:45372", + "issue:45405", + "issue:45542", + "issue:45593", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45907", + "issue:46082", + "issue:46209", + "issue:46399", + "issue:46512" + ], + "soft_pairs": [ + "issue:46082|issue:46399", + "issue:45103|issue:45372", + "issue:44279|issue:45405", + "issue:44242|issue:45200", + "issue:43299|issue:45357", + "issue:41950|issue:45020", + "issue:43650|issue:46209", + "issue:38175|issue:44617", + "issue:44351|issue:45278", + "issue:44521|issue:45325", + "issue:39692|issue:43901", + "issue:43295|issue:44779", + "issue:41950|issue:45823", + "issue:44683|issue:45542", + "issue:43976|issue:45405", + "issue:44779|issue:45812", + "issue:46082|issue:46512", + "issue:43299|issue:43854", + "issue:43479|issue:43525", + "issue:44297|issue:44462", + "issue:44466|issue:45276", + "issue:44016|issue:44291", + "issue:43441|issue:45834", + "issue:45081|issue:45907", + "issue:43653|issue:45593", + "issue:42994|issue:43317" + ], + "prepared_review_unit_hash": "2f83f538b94bee428d1089c32f71f335d51f87964db0414217302fd33a7b2ae0", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35026, + "estimated_input_tokens": 8757, + "estimated_eval_tokens": 17770 + }, + "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 and regressions across tokenizers, multimodal processors, quantization, config handling, and docs. None of the soft pairs look like the same concrete bug or change, so all are rejected.", + "confidence": 0.58, + "canonical_issue_reason": "No true duplicate hub stands out; 45278 is only the broadest umbrella-like regression report (many import errors after upgrade), but it still does not subsume the rest of the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "45278 is the best triage anchor among these issues because it is the most general, user-facing regression report and closest to the small import-error subtheme, though it is still not a duplicate of the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "Different failures: LlamaConfig head_dim validation vs a Qwen3.5 kernel registration crash." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "Unrelated code paths: auto_docstring annotation handling vs mistral_common import break in Gemma processor loading." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Generic dependency issue vs one specific unreleased PEFT version bump; not the same bug." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "MoE load-balancing loss gating and Gemma4 mm_token_type_ids defaults are different model behaviors." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Loading Qwen3VL MoE models and saving Qwen3.5 visual encoder keys are separate code paths." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video pipeline processor lookup bug is not the same as remote_code compatibility regressions." + }, + { + "left": "issue:43650", + "right": "issue:46209", + "accept": false, + "reason": "Placeholder/noisy issue title vs FP8 inference on B200; no shared bug." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "SigLIP2 zero probabilities and Sam3Video OOM are unrelated model/runtime problems." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "A missing HybridCache import is not the same as a broad set of import errors after upgrade." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "All-zero assistant masks in chat templating and Qwen2.5-VL rope indexing are different multimodal bugs." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs issues, but they concern different examples/pages and different failures." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Processor.tokenizer regression and Deepseek tokenizer output regression are distinct tokenizer problems." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Video-classification processor mismatch and Gemma4 PLE device mismatch are different bugs." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Torch flex_attention compilation failure and tensorboard-only backend install issue are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python version support break is not the same as the unreleased PEFT dependency bump." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Different model families and different tokenizer regressions; not mergeable as one fix." + }, + { + "left": "issue:46082", + "right": "issue:46512", + "accept": false, + "reason": "Config validation for head_dim and stale config.dtype after from_config are separate initialization bugs." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Qwen3.5 loading regression and GLM-4.7-Flash unit-test loading failure are different concrete problems." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Default multimodal config initialization and missing pad_token_id on Llama4Config are not the same issue." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer class serialization mismatch and AutoTokenizer ignoring tokenizer.json affect different load/save paths." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Inconsistent lm_head serialization by device and resize_token_embeddings not propagating are different weight-management bugs." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error vs init_empty_weights unexpected-arg TypeError; unrelated." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "FlashAttention failure for Ministral-3 and Kosmos2.5 OCR index error are separate model-specific issues." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Mistral tokenizer patch crash and list_repo_files revision forwarding are unrelated repository/client bugs." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss omission are different subsystems." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized model save failure and dequantized model offload/load failure are related only broadly, not the same concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 38, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:33290", + "issue:41628", + "issue:42175", + "issue:43441", + "issue:43525", + "issue:43643", + "issue:43673", + "issue:43723", + "issue:43761", + "issue:43784", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", + "issue:44188", + "issue:44242", + "issue:44246", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44361", + "issue:44373", + "issue:44496", + "issue:44589", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:45103", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45593", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45800", + "issue:45823", + "issue:46291", + "issue:46399" + ], + "soft_pairs": [ + "issue:45600|issue:45712", + "issue:33290|issue:45656", + "issue:44623|issue:44749", + "issue:44261|issue:44263", + "issue:44623|issue:45397", + "issue:44361|issue:45593", + "issue:45542|issue:45800", + "issue:44016|issue:45823", + "issue:42175|issue:43976", + "issue:43441|issue:44683", + "issue:44671|issue:44779", + "issue:43723|issue:44162", + "issue:43901|issue:43994", + "issue:46291|issue:46399", + "issue:43673|issue:43761", + "issue:43525|issue:43881", + "issue:43824|issue:43976", + "issue:43901|issue:44246", + "issue:43525|issue:43643", + "issue:44297|issue:45397", + "issue:44589|issue:45362", + "issue:43784|issue:45542", + "issue:41628|issue:43976", + "issue:44188|issue:44336", + "issue:44336|issue:44373", + "issue:45103|issue:46399", + "issue:45561|issue:45563", + "issue:44242|issue:45593", + "issue:44246|issue:44496", + "issue:44295|issue:45397" + ], + "prepared_review_unit_hash": "c13163f29bd505d34e2d14c0bc77708f4ed4cef380d4b85782ca133b9d180779", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36531, + "estimated_input_tokens": 9133, + "estimated_eval_tokens": 18522 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft edge looks like a true duplicate; the linked issues are mostly superficial similarities across different bugs, regressions, packaging problems, and docs/test fallout.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue stands out: the soft-linked items are heterogeneous and do not describe the same underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No strong best issue for deduplication; the set is too mixed to use one issue as a representative canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Different subsystems: config mapping/CI break vs leftover dummy classes in testing stubs." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Both mention DeepSpeed/optimizer, but one is Adafactor OOM while the other is double optimizer.step()." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated: missing processor save files vs tokenizer slowdown on Chinese text." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Different code paths: MLA rms_norm_eps precision issue vs torch.split return values in a MoE indexer." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "processor.save_pretrained file output bug vs gemma-4 ZeRO-3 from_pretrained failure are different problems." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "Tokenizer AttributeError on tasks vs D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "Backend packaging issue vs torch/transformers version incompatibility." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax error vs Gemma4 device-mapping mismatch." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "Install/backend packaging issue vs Python version support regression." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Model FlashAttention failure vs compiled flex_attention on newer torch; different code paths." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "CamemBERT MLM outputs vs DeepSeek tokenizer correctness are distinct bugs." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "AutoTokenizer loading failure vs ESM2 model breakage." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mismatch for return_all_scores vs SigLIP2 model/pipeline wrong outputs." + }, + { + "left": "issue:46291", + "right": "issue:46399", + "accept": false, + "reason": "Pipeline import dependency validation vs Qwen3.5 rotary-pos-embedding registration error." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache/chunked_prefill vs CLIPVisionModel hidden_states regression." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Missing pad_token_id in config vs glm-4v model load failure." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Import error for a model class vs Python version compatibility; not the same issue." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Pipeline doc mismatch vs import slowness complaint." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Missing config attribute on Llama4Config vs trust_remote_code returning incomplete fields." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save_pretrained metadata bug vs gemma-4 Zero3 from_pretrained issue." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage deserialization vs Qwen3.5 chat crash." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Import-time NameError in sentence-transformers vs missing tf backend package." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "AutoImageProcessor import failure vs Python 3.9/3.10 support issue." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "torch.compile attention-kernel divergence vs ANSI codes in loading_report." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "ANSI codes in loading_report vs wrong docstring for position_ids." + }, + { + "left": "issue:45103", + "right": "issue:46399", + "accept": false, + "reason": "auto_docstring crash under future annotations vs Qwen3.5 kernels registration error." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "xdist race on captured_info.txt vs stale generation warning." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "MoE load-balancing loss omission vs D-FINE auxiliary-loss behavior." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import slowness vs unrecognized model/config error." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "position_ids buffer access error vs Gemma4 ZeRO-3 loading failure." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 39, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:36246", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:42886", + "issue:43257", + "issue:43262", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43673", + "issue:43901", + "issue:44162", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44351", + "issue:44485", + "issue:44496", + "issue:44561", + "issue:44734", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45250", + "issue:45292", + "issue:45362", + "issue:45375", + "issue:45561", + "issue:45584", + "issue:45701", + "issue:45759", + "issue:45859", + "issue:46132", + "issue:46369", + "issue:46399", + "issue:46424", + "issue:46506" + ], + "soft_pairs": [ + "issue:42491|issue:43299", + "issue:43531|issue:45362", + "issue:43550|issue:46506", + "issue:43526|issue:45003", + "issue:41950|issue:44220", + "issue:42886|issue:46132", + "issue:44246|issue:44871", + "issue:44336|issue:44485", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:44734|issue:46424", + "issue:44336|issue:46424", + "issue:45003|issue:45561", + "issue:44242|issue:44857", + "issue:44779|issue:45292", + "issue:43475|issue:46369", + "issue:45859|issue:46399", + "issue:36246|issue:43596", + "issue:44295|issue:44749", + "issue:43673|issue:44561", + "issue:43901|issue:45003", + "issue:44162|issue:45200", + "issue:43352|issue:45250", + "issue:43257|issue:45137", + "issue:44496|issue:45003", + "issue:45292|issue:45584", + "issue:45292|issue:45701", + "issue:43262|issue:44246", + "issue:45375|issue:45759" + ], + "prepared_review_unit_hash": "1df48a692fa140e6641d9afd3a8b59211f20c3b74f65e8cb1aab52957a2567a7", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36318, + "estimated_input_tokens": 9080, + "estimated_eval_tokens": 18416 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most items are superficially similar only at the level of model family, version bumps, caches, or generic import/runtime failures. I would reject every soft pair as a duplicate/merge candidate.", + "confidence": 0.83, + "canonical_issue_reason": "No single canonical issue stands out; the items span unrelated bugs across model loading, tokenization, cache handling, audio/image pipelines, and test utilities.", + "canonical_pr_reason": null, + "best_issue_reason": "If forced to pick one representative issue, 44561 is the broadest cross-model compatibility regression and the most generally useful triage anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both are Qwen3-MoE version regressions, but one is LoRA/accelerate loading and the other is multimodal model loading; not the same bug." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Same model family, different failures: sliding_window behavior vs a chat crash. Different code paths." + }, + { + "left": "issue:43550", + "right": "issue:46506", + "accept": false, + "reason": "Completely different models and symptoms: torch.compile+SDPA failure vs CPU/CUDA output mismatch." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "Unrelated bugs: BeitImageProcessorFast label reduction vs unsafe sys.modules access in modeling_utils." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Different subsystems and failures; video-classification processor lookup is not the same as fbank feature extraction." + }, + { + "left": "issue:42886", + "right": "issue:46132", + "accept": false, + "reason": "Offline cache loading and AttentionInterface.register behavior are unrelated." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import latency and Gemma eos_token_id config mismatch are unrelated issues." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI log formatting and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different missing symbols in different modules." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Different missing exports from different packages; no shared underlying defect." + }, + { + "left": "issue:44734", + "right": "issue:46424", + "accept": false, + "reason": "Both mention cache, but one is KV-cache continuation indexing and the other is max_cache_length semantics." + }, + { + "left": "issue:44336", + "right": "issue:46424", + "accept": false, + "reason": "No real overlap beyond being general bugs; unrelated features and failures." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Modeling_utils sys.modules access is unrelated to the xdist captured_info.txt race." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Different model/training bugs: router loss omission vs AMP/CUDA crash in a loss class." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "Tokenizer output regression and resize_token_embeddings not updating output embeddings are different bugs." + }, + { + "left": "issue:43475", + "right": "issue:46369", + "accept": false, + "reason": "Both are SAM 3-related, but one is a missing attribute in video vision output and the other is memory overflow in litetext." + }, + { + "left": "issue:45859", + "right": "issue:46399", + "accept": false, + "reason": "Different Qwen3.5 MoE failures: CPU offload incompatibility vs missing decorator during kernel registration." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "Different model families and failure modes: missing Qwen processor export vs DeepSpeed/BertModel zero3 IndexError." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "One is a position_ids buffer access error; the other is a performance regression after upgrade." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Both are v5 regressions, but cache-missing in chunked_prefill and removal of is_torch_fx_available are separate code-path breaks." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Docs mismatch in TextClassificationPipeline is unrelated to modeling_utils sys.modules access." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "Different models and different bugs: ESM2 breakage vs Gemma 4 token-type defaults." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "Generic flash-attention topic, but one is model support gap and the other is an unspecified flash-attention issue; not the same bug." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MoE weight conversion on load is unrelated to a DeepSpeed ZeRO3 deque error." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model config and unsafe sys.modules access are distinct problems." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Resize-token-embeddings behavior is unrelated to Whisper empty-transcription failure." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "Embedding resize bug and tokenizer behavior change across versions are not the same underlying issue." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio processor sampling-rate default bug is unrelated to import-time slowness." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both touch Qwen multimodal config handling, but one drops a missing field and the other fails to unwrap text_config; different fixes." + } + ] + }, + "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": 20, + "review_unit_index": 40, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:42886", + "issue:43065", + "issue:43408", + "issue:43475", + "issue:43540", + "issue:43653", + "issue:43906", + "issue:43931", + "issue:44038", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44188", + "issue:44297", + "issue:44368", + "issue:44466", + "issue:44479", + "issue:44483", + "issue:44561", + "issue:44623", + "issue:44877", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45335", + "issue:45375", + "issue:45397", + "issue:45412", + "issue:45459", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45656", + "issue:45685", + "issue:45987", + "issue:46082", + "issue:46123", + "issue:46170", + "issue:46202", + "issue:46399", + "issue:46421", + "issue:46431", + "issue:46506" + ], + "soft_pairs": [ + "issue:44877|issue:45375", + "issue:44038|issue:46506", + "issue:44561|issue:46399", + "issue:45412|issue:46123", + "issue:45685|issue:46431", + "issue:44987|issue:45459", + "issue:44466|issue:45335", + "issue:45137|issue:45600", + "issue:44995|issue:45200", + "issue:43475|issue:43540", + "issue:44060|issue:44077", + "issue:44479|issue:46170", + "issue:43653|issue:44162", + "issue:45397|issue:45656", + "issue:42886|issue:45229", + "issue:44368|issue:44977", + "issue:43906|issue:46202", + "issue:44483|issue:46202", + "issue:44995|issue:45588", + "issue:44368|issue:45563", + "issue:43065|issue:43408", + "issue:44188|issue:46202", + "issue:43931|issue:44368", + "issue:44877|issue:46170", + "issue:45987|issue:46082", + "issue:43065|issue:46123", + "issue:44188|issue:46421", + "issue:45584|issue:46170", + "issue:44297|issue:44623" + ], + "prepared_review_unit_hash": "864a418d90e0909b033ad0f153b24da6c1fda5177756a47542e68ef71c15598b", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35808, + "estimated_input_tokens": 8952, + "estimated_eval_tokens": 18160 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a loose, heterogeneous cluster of Transformers bug reports rather than a single duplicate family. The only near-duplicate-looking subthread is tokenizer/save_pretrained serialization, but even those reports describe different failure modes.", + "confidence": 0.83, + "canonical_issue_reason": "issue:44297 is the closest representative of the small tokenizer serialization subtheme: it is narrowly scoped, concrete, and the strongest semantic neighbor to 44623, even though the broader cluster is mixed.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44297 is the most concrete and best-scoped issue here; it reads like a clear reproducible regression instead of a broad symptom report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both are strict-config loading problems, but they hit different models and different missing/ignored fields; not the same bug." + }, + { + "left": "issue:44038", + "right": "issue:46506", + "accept": false, + "reason": "Unrelated failures: a Qwen3-VL-Moe bug versus CPU/CUDA output divergence in PPDocLayoutV3." + }, + { + "left": "issue:44561", + "right": "issue:46399", + "accept": false, + "reason": "Both are v5 regressions, but the concrete code paths differ: removed helper breaking trust_remote_code models versus missing decorator on rotary registration." + }, + { + "left": "issue:45412", + "right": "issue:46123", + "accept": false, + "reason": "RT-DETR memory retention and MaskGenerationPipeline partial-batch handling are different symptoms and subsystems." + }, + { + "left": "issue:45685", + "right": "issue:46431", + "accept": false, + "reason": "MPS histogram/int support failure and missing chat template support are unrelated processor/model issues." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Both touch loading/tokenizer behavior, but one is a model load failure on v5.1 and the other is protobuf-import error masking; different root causes." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both involve embeddings/tied weights, but one is serialization inconsistency and the other is resize_token_embeddings not propagating to decoder embeddings." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed deque exhaustion and stale auto_mappings references are unrelated failures." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "A stale-cache second-forward crash is not the same as needing default mm_token_type_ids for text-only fine-tuning." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Different model families and different failures: missing vision encoder output field versus video-input processing ValueError." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Tied-weights warning corruption and optional post_init policy are not the same issue." + }, + { + "left": "issue:44479", + "right": "issue:46170", + "accept": false, + "reason": "A multi-model video regression and a Windows/ROCm import dependency problem are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer special-token registration bug versus broad ESM2 breakage; too different to merge." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 ZeRO3 loading failure and double optimizer-step behavior under deepspeed are different code-path problems." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline tokenizer cache loading and Gemma4 multi-GPU OOM are not the same underlying bug." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Both mention Qwen3.5, but one is a tied-weights warning and the other is a flash-attention generation regression." + }, + { + "left": "issue:43906", + "right": "issue:46202", + "accept": false, + "reason": "One is an isolated reproduction of an external issue; the other is a Gemma pipeline roundtrip failure on a specific token. Not enough evidence of a shared root cause." + }, + { + "left": "issue:44483", + "right": "issue:46202", + "accept": false, + "reason": "Chat-completions API rejection and Gemma token roundtrip failure are unrelated." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Stale indexer cache on second forward is not the same as flash_attention crash on sink-less models." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "A tie_word_embeddings warning and a stale num_return_sequences warning are different warning sites with different triggers." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "A dummy Conv2d implementation and a model-type mismatch warning are not the same bug." + }, + { + "left": "issue:44188", + "right": "issue:46202", + "accept": false, + "reason": "torch.compile branch divergence in attention kernels is unrelated to Gemma tokenizer roundtrip failure." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Weight-shape mismatch loading Qwen3-VL-30B is unrelated to the Qwen3.5 tie_word_embeddings warning." + }, + { + "left": "issue:44877", + "right": "issue:46170", + "accept": false, + "reason": "Strict config loading for granite_speech and a Windows ROCm import problem do not share a code path." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "StaticCache return-type contract and LlamaConfig head_dim validation are different API contract issues." + }, + { + "left": "issue:43065", + "right": "issue:46123", + "accept": false, + "reason": "Sam3PixelDecoder dummy Conv2d and MaskGenerationPipeline partial-batch handling are unrelated." + }, + { + "left": "issue:44188", + "right": "issue:46421", + "accept": false, + "reason": "Kernel divergence under torch.compile is not the same as cache lazy-init causing recompilation." + }, + { + "left": "issue:45584", + "right": "issue:46170", + "accept": false, + "reason": "Whisper empty-transcription failure after align_special_tokens and a Windows/ROCm dependency issue are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Both are save_pretrained-related, but one is a tokenizer_config mismatch and the other is missing processor files; 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": 20, + "review_unit_index": 41, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:42994", + "issue:43232", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43701", + "issue:43704", + "issue:43723", + "issue:43784", + "issue:43873", + "issue:44038", + "issue:44162", + "issue:44188", + "issue:44351", + "issue:44361", + "issue:44623", + "issue:44749", + "issue:44869", + "issue:44871", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45406", + "issue:45478", + "issue:45656", + "issue:45753", + "issue:46132", + "issue:46202", + "issue:46326", + "issue:46399" + ], + "soft_pairs": [ + "issue:44361|issue:44749", + "issue:43646|issue:44162", + "issue:44749|issue:45656", + "issue:44623|issue:45656", + "issue:43232|issue:43673", + "issue:43701|issue:44749", + "issue:44869|issue:45356", + "issue:44871|issue:46132", + "issue:42994|issue:43482", + "issue:45230|issue:45397", + "issue:43475|issue:43643", + "issue:28282|issue:30990", + "issue:44038|issue:44623", + "issue:42371|issue:45092", + "issue:43475|issue:45381", + "issue:41628|issue:44933", + "issue:45375|issue:45478", + "issue:45406|issue:46202", + "issue:42994|issue:43525", + "issue:43784|issue:46399", + "issue:45310|issue:45375", + "issue:43723|issue:46326", + "issue:42673|issue:43704", + "issue:43873|issue:44945", + "issue:44964|issue:45137", + "issue:43232|issue:44188", + "issue:45081|issue:45103", + "issue:41628|issue:44351", + "issue:44945|issue:45753" + ], + "prepared_review_unit_hash": "30fa43a485a9b3ad6a723e37dd27f36f51d3474fb73932e83c6c8e3bf92a3db1", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36131, + "estimated_input_tokens": 9033, + "estimated_eval_tokens": 18322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly conservative triage: the only clear duplicate in this packet is issue:42673 / issue:43704, which are the same VRAM-leak report for the same model and trigger. The other soft pairs look related by subsystem or version, but not identical enough to merge.", + "confidence": 0.82, + "canonical_issue_reason": "issue:42673 is the strongest canonical issue because it is a specific report of the Qwen3ForCausalLM VRAM leak under multiple dataloader threads, and issue:43704 is the same bug with the same title and trigger.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the best issue anchor for triage because it is the clearest, most specific instance of the duplicate VRAM-leak report and cleanly represents issue:43704.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "Different failures: tokenizer/task AttributeError vs a performance regression after upgrade." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Both are v5 breakages, but they affect different models and different failure modes." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "Unrelated bugs: data filtering slowdown vs optimizer step being called twice under deepspeed." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Saving processor files and duplicate optimizer stepping are different code paths." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Both are generation-related, but the reported crashes are different and not the same concrete bug." + }, + { + "left": "issue:43701", + "right": "issue:44749", + "accept": false, + "reason": "Checkpoint resume key mismatch is unrelated to the slowdown report." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Different tokenizer/decoding bugs with different models and triggers." + }, + { + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "EOS config mismatch is unrelated to AttentionInterface.register behavior." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving and GGUF loading are separate problems." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Generic placeholder issue vs a specific Gemma-4 ZeRO3 from_pretrained failure." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "Both mention missing fields, but they are different objects and different failure locations." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "ImportError about missing PyTorch is not the same as a loading hang." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe v5 bug is unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API guidance is unrelated to InternVL2 meta-initialization incompatibility." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Different multimodal model bugs with different symptoms and likely different fixes." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Both are import-related, but they concern different missing symbols and different modules." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but one is a dropped config field and the other is a from_pretrained error; not the same bug." + }, + { + "left": "issue:45406", + "right": "issue:46202", + "accept": false, + "reason": "Gemma processor _tokenizer crash and Gemma pipeline roundtrip failure are distinct issues." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized saving failure is unrelated to missing pad_token_id on Llama4Config." + }, + { + "left": "issue:43784", + "right": "issue:46399", + "accept": false, + "reason": "Different packages and failures: import-time nn error vs rotary-pos-emb registration error." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Related model family, but not the same concrete failure." + }, + { + "left": "issue:43723", + "right": "issue:46326", + "accept": false, + "reason": "Generic AutoTokenizer loading issue and mistral_common continue_final_message bug are separate." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same title, same model, same VRAM leak under multiple dataloader threads; clearly duplicate reports." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization offloading behavior is unrelated to pipeline parallelism output corruption." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Phi-4 multimodal loading failure is unrelated to the DeepSpeed deque error." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Tokenizer regex crash and auto-docstring annotation crash are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are missing imports, but they are different missing symbols with no clear shared code path." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline parallelism output corruption is not the same as an infinite loop on a specific image." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The duplicate-pair conclusion for issue:42673 / issue:43704 is grounded, but the canonical_issue_reason overstates the packet by claiming \u201cmore discussion/inbound references\u201d when both issues have inbound_references = 0. That unsupported detail is enough to reject. The rest of the soft-edge verdicts look conservative overall." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 42, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42831", + "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43504", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43976", + "issue:44016", + "issue:44038", + "issue:44079", + "issue:44222", + "issue:44242", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44779", + "issue:44792", + "issue:44869", + "issue:44933", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45200", + "issue:45310", + "issue:45356", + "issue:45561", + "issue:45600", + "issue:45712", + "issue:45812", + "issue:46077", + "issue:46123", + "issue:46231" + ], + "soft_pairs": [ + "issue:42994|issue:43824", + "issue:44222|issue:45310", + "issue:44038|issue:45310", + "issue:36010|issue:41628", + "issue:43232|issue:44792", + "issue:42831|issue:44222", + "issue:46077|issue:46123", + "issue:42175|issue:45712", + "issue:44496|issue:45561", + "issue:43504|issue:45561", + "issue:42175|issue:44933", + "issue:45561|issue:45600", + "issue:43976|issue:44933", + "issue:43278|issue:45137", + "issue:43377|issue:44242", + "issue:30990|issue:35141", + "issue:42898|issue:44568", + "issue:30333|issue:33290", + "issue:44623|issue:45310", + "issue:43479|issue:44869", + "issue:42222|issue:45200", + "issue:43479|issue:44079", + "issue:42898|issue:45812", + "issue:43723|issue:44568", + "issue:44016|issue:46231", + "issue:43874|issue:45092", + "issue:43784|issue:43824", + "issue:42898|issue:44779", + "issue:44734|issue:45103", + "issue:42898|issue:45356" + ], + "prepared_review_unit_hash": "c4ebd2200a5161d469ac07be906636e650c2d6f8ca69df2c4f21d9eb9e8f4a09", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36211, + "estimated_input_tokens": 9053, + "estimated_eval_tokens": 18362 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly isolated bug reports across different models, APIs, and failure modes. I did not find any soft-edge pair that clearly describes the same underlying defect closely enough to merge as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45812 is the strongest canonical issue candidate: it describes a broad, concrete, reproducible tokenizer regression with a clear v4\u2192v5 delta and wide user impact.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 is the best standalone issue to anchor triage because it is specific, high-signal, and clearly frames the regression for maintainers.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving failure and a Qwen2.5-VL import error are unrelated failure modes in different code paths." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "Both mention MoE/FP8, but one is save_pretrained on FP8 MoE and the other is from_pretrained loading for Qwen3.5; not the same bug." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen variants and different symptoms; this looks like separate model-specific regression reports." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Import errors for GenerationMixin and AutoImageProcessor are distinct missing-export issues." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation KV-sync bug and a Janus image-generation test failure do not point to the same code-path problem." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "FP8 accuracy degradation is not the same as FP8 MoE save_pretrained failure." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Bidirectional-attention masking and partial-batch result dropping are unrelated pipeline/model behaviors." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Backend dependency packaging and leftover dummy-class cleanup are unrelated maintenance issues." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Model config loading failure and a pytest-xdist file race are completely different bugs." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Legacy-field pretrained load failure and a parallel test race are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Missing TensorFlow backend dependency and a nonexistent image_utils import are separate packaging/import problems." + }, + { + "left": "issue:45561", + "right": "issue:45600", + "accept": false, + "reason": "A test-worker race on captured_info.txt is not the same as stale auto_mappings references breaking CI." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python version compatibility and an import-from-image_utils error are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift in evaluation and DeepSpeed ZeRO3 deque underflow are different runtime issues." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Missing padding-mask support in an encoder and missing load-balancing loss in MoE routing are different defects." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "A loading hang in Sentence Transformers and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Tokenization-space cleanup behavior and add_special_tokens not inserting BOS/EOS are distinct tokenizer behaviors." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLflow status reporting after failed training and an Adafactor OOM in DeepSpeed are unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Processor save_pretrained missing files and Qwen3.5 MoE from_pretrained loading failure are different save/load paths." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Default config initialization and Whisper word-timestamp decoding crashes are unrelated." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Broken ViTPose model paths and Gemma4 mm_token_type_ids defaults are separate model-specific issues." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Config defaulting and ModelOutput key assignment are different mechanisms with no clear shared root cause." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but one concerns whitespace cleanup while the other concerns wrong token IDs for Granite models; not the same bug." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "General AutoTokenizer loading issues and a specific mDeBERTa special-tokens regression are not the same concrete defect." + }, + { + "left": "issue:44016", + "right": "issue:46231", + "accept": false, + "reason": "Notebook syntax error and a docstring parameter-name mismatch are unrelated documentation issues." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Missing image-patch count method in GLM46V and old InternVL2 meta-init incompatibility are different multimodal bugs." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Sentence-Transformers import NameError and Qwen2.5-VL import failure are unrelated import problems." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer regressions affecting different behaviors and models; too broad to treat as one bug." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Serving KV-cache tensor indexing and auto_docstring future-annotations handling are unrelated crashes." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Whitespace cleanup/tokenization behavior and Kimi-K2.5 codec handling are distinct tokenizer regressions, not one underlying 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": 20, + "review_unit_index": 43, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28282", + "issue:30333", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:41628", + "issue:42757", + "issue:42796", + "issue:42831", + "issue:42898", + "issue:43122", + "issue:43232", + "issue:43475", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:44060", + "issue:44075", + "issue:44206", + "issue:44222", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44448", + "issue:44484", + "issue:44623", + "issue:44933", + "issue:44977", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45245", + "issue:45310", + "issue:45381", + "issue:45405", + "issue:45656", + "issue:45712", + "issue:46082", + "issue:46132", + "issue:46209", + "issue:46231", + "issue:46305", + "issue:46424", + "issue:46431" + ], + "soft_pairs": [ + "issue:44295|issue:45656", + "issue:44075|issue:46231", + "issue:44060|issue:45125", + "issue:42898|issue:44448", + "issue:45405|issue:45712", + "issue:44222|issue:44297", + "issue:44933|issue:45712", + "issue:27957|issue:30333", + "issue:30333|issue:34689", + "issue:44297|issue:45656", + "issue:44448|issue:46132", + "issue:42796|issue:43723", + "issue:44222|issue:44623", + "issue:43479|issue:43644", + "issue:30333|issue:36331", + "issue:42831|issue:44623", + "issue:30333|issue:35141", + "issue:22355|issue:28282", + "issue:45245|issue:46132", + "issue:46209|issue:46431", + "issue:41628|issue:42757", + "issue:43475|issue:46305", + "issue:43122|issue:46209", + "issue:44297|issue:45310", + "issue:43723|issue:44977", + "issue:45381|issue:46305", + "issue:45071|issue:46082", + "issue:43232|issue:46424", + "issue:44206|issue:45103", + "issue:44263|issue:45092", + "issue:44222|issue:44484" + ], + "prepared_review_unit_hash": "8e09a453ea970d9f13e884328d85112f0f23afda798be6d19a028285fa32bd3a", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36223, + "estimated_input_tokens": 9056, + "estimated_eval_tokens": 18368 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a collection of unrelated Transformers issues with only superficial topical overlap. None of the soft candidate pairs look like the same underlying bug or change, so all soft edges are rejected. No PRs are present.", + "confidence": 0.93, + "canonical_issue_reason": "Issue #43723 is the broadest user-facing regression in the set (tokenizer loading in v5) and is the closest thing to a representative API-breakage report among otherwise disparate issues.", + "canonical_pr_reason": null, + "best_issue_reason": "#43723 is the best representative because it is concrete, widely applicable, and sits near several other tokenizer/import-related reports without being overly model-specific.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Unrelated bugs: buffer access for position_ids vs a double optimizer-step complaint under deepspeed." + }, + { + "left": "issue:44075", + "right": "issue:46231", + "accept": false, + "reason": "Different problem classes: SGD args not used vs documentation naming mismatch." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Both are Qwen-related, but one is a tied-weights warning bug and the other is missing tensor-parallel metadata; different code paths." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Both mention v5 output changes, but one is about clean_up_tokenization_spaces semantics and the other is a broader model output regression; not the same bug." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "A dependency version bump issue and leftover dummy classes/check_repo failure are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both involve save_pretrained, but one is FP8 MoE serialization and the other is tokenizer class metadata; different artifacts and fixes." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Import from image_utils vs dummy PT class cleanup are unrelated maintenance issues." + }, + { + "left": "issue:27957", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention 2 model support vs MLFlow reporting on failed training jobs are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "Training-job tracking bug vs model-loading regression for a specific Llama checkpoint; different subsystems." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save metadata bug vs optimizer double-step bug; no shared underlying code-path." + }, + { + "left": "issue:44448", + "right": "issue:46132", + "accept": false, + "reason": "Versioned output mismatch for a model vs behavior change in AttentionInterface.register; different features." + }, + { + "left": "issue:42796", + "right": "issue:43723", + "accept": false, + "reason": "Tokenizer BPE marker handling vs tokenizer loading failure are adjacent topics but different bugs." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 MoE model serialization vs processor save_pretrained missing files are different save paths and failure modes." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Config default initialization bug vs non-persistent buffer junk in v5; unrelated internals." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "MLFlow reporting on failure vs CustomTrainer signature mismatch; unrelated training issues." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FP8 accuracy issue vs processor serialization missing files; no shared bug." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "Failed-job tracking and embedding resize/reinit are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Importing transformers.onnx vs AutoModel requiring torch; both import-related but different missing-module dependencies." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "Category cardinality runtime error vs attention interface behavior change; unrelated." + }, + { + "left": "issue:46209", + "right": "issue:46431", + "accept": false, + "reason": "FP8 inference on B200 vs chat template availability error; different features and failure causes." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Missing AutoImageProcessor export vs missing is_offline_mode export; both import errors, but not the same symbol or fix." + }, + { + "left": "issue:43475", + "right": "issue:46305", + "accept": false, + "reason": "SAM 3 Video output attribute bug vs MoonshineStreaming mask off-by-one; unrelated model-specific issues." + }, + { + "left": "issue:43122", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer-output regression across versions vs FP8 inference failure on B200; different subsystems." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer class metadata mismatch vs Qwen3.5 MoE from_pretrained error; not the same serialization/loading bug." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Generic tokenizer loading issue vs Qwen3.5 flash-attention generation bug; unrelated." + }, + { + "left": "issue:45381", + "right": "issue:46305", + "accept": false, + "reason": "qwen2.5-vl vision_position_ids bug vs streaming mask off-by-one; different models and code paths." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "PretrainedConfig type-checking regression vs LlamaConfig head_dim validation; different config behaviors." + }, + { + "left": "issue:43232", + "right": "issue:46424", + "accept": false, + "reason": "Generation sync_gpus state update bug vs cache max_length only honored for quantized cache; related to generation, but not the same defect." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Feature extractor argument crash vs auto-docstring annotation parsing crash; unrelated tooling bugs." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "torch.split return handling in an indexer vs remote-code checkpoint meta-init incompatibility; different components." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 MoE save_pretrained bug vs a question about max_shard_size defaults; not 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": 20, + "review_unit_index": 44, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:30990", + "issue:38175", + "issue:39401", + "issue:41628", + "issue:42757", + "issue:42796", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:43066", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43824", + "issue:44038", + "issue:44190", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44295", + "issue:44360", + "issue:44373", + "issue:44483", + "issue:44485", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44829", + "issue:44871", + "issue:44913", + "issue:45092", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45834", + "issue:46002", + "issue:46050", + "issue:46082", + "issue:46305", + "issue:46328", + "issue:46402", + "issue:46421" + ], + "soft_pairs": [ + "issue:44734|issue:46421", + "issue:43232|issue:45834", + "issue:44373|issue:44485", + "issue:44295|issue:44623", + "issue:44190|issue:46082", + "issue:43723|issue:44871", + "issue:44279|issue:46328", + "issue:44483|issue:45406", + "issue:42907|issue:46050", + "issue:44263|issue:44360", + "issue:30990|issue:38175", + "issue:45092|issue:45357", + "issue:44261|issue:45092", + "issue:43644|issue:46305", + "issue:42831|issue:45310", + "issue:44829|issue:46002", + "issue:44263|issue:44373", + "issue:41628|issue:43824", + "issue:42831|issue:44038", + "issue:43479|issue:44913", + "issue:43408|issue:46402", + "issue:22355|issue:27957", + "issue:42796|issue:43122", + "issue:39401|issue:44568", + "issue:42757|issue:45278", + "issue:42886|issue:43066" + ], + "prepared_review_unit_hash": "a62b760a6751ea86f163806f7f72524e37e80c427508a4c98b6efed0fc8befb7", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35118, + "estimated_input_tokens": 8780, + "estimated_eval_tokens": 17816 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is highly heterogeneous: import regressions, tokenizer behavior changes, serving crashes, config/load-save bugs, and quantization issues. None of the soft pairs looks like the same concrete underlying bug or change, so they should stay separate.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical duplicate fits this cluster; the closest broad umbrella is issue 45278 for upgrade-related import regressions, but it does not cover most other items.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45278 is the broadest representative among the listed issues, but only for a narrow import-regression subset; it is not a true cluster canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44734", + "right": "issue:46421", + "accept": false, + "reason": "Serving KV-cache continuation crash vs chunked-prefill cache lazy-init recompilation; different code paths." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation kwargs sync_gpus bug vs long OCR input index error; unrelated failures." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "Docstring typo for position_ids vs GLM-5 RoPE implementation bug; not the same issue." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Reading position_ids after buffer registration vs processor.save_pretrained missing files; different subsystems." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Image-classification script dataset loading vs LlamaConfig head_dim validation; unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "AutoTokenizer loading regression vs Gemma eos_token_id mismatch; different concrete bugs." + }, + { + "left": "issue:44279", + "right": "issue:46328", + "accept": false, + "reason": "General dependency issue vs SequenceFeatureExtractor.pad inefficiency; not duplicates." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat-completions request rejection vs Gemma4Processor missing _tokenizer crash; different failures." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Dequantized Ministral save/load failure vs sinq quantized round-trip breakage; different quantization paths." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Both touch the DSA indexer, but one is torch.split return handling and the other is a missing ReLU; separate bugs." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "SentenceTransformers loading hang vs siglip2 zero probabilities; unrelated." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Remote-code meta-init incompatibility vs Qwen3.5 save_pretrained visual-key regression; different issues." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "MLA layernorm eps precision error vs InternVL2 meta-init incompatibility; not the same bug." + }, + { + "left": "issue:43644", + "right": "issue:46305", + "accept": false, + "reason": "Non-persistent buffer junk fill vs sliding_window_mask off-by-one; unrelated." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy problem vs Qwen3.5 MoE from_pretrained error; different problems." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Flash-attention-3 training degeneration vs NaNs after checkout/init; not the same concrete failure." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "DSA indexer runtime bug vs wrong docstring for position_ids; one is documentation only." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Missing AutoImageProcessor import vs missing Qwen2_5_VLForConditionalGeneration import; distinct import errors." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "FP8 accuracy regression vs Qwen3-VL-Moe bug; different model/features." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Multimodal config default-init bug vs GPTNeoX rotary_pct reload regression; unrelated." + }, + { + "left": "issue:43408", + "right": "issue:46402", + "accept": false, + "reason": "Wrong Sam3 model type warning vs text encoder weights not loaded; different Sam3 issues." + }, + { + "left": "issue:22355", + "right": "issue:27957", + "accept": false, + "reason": "transformers.onnx import failure vs XLMRoberta Flash Attention 2 support issue; unrelated." + }, + { + "left": "issue:42796", + "right": "issue:43122", + "accept": false, + "reason": "Ministral BPE-marker tokenizer bug vs broad tokenization change across versions; not the same concrete bug." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Wrong offset_mapping in Qwen3 tokenizer vs add_special_tokens BOS/EOS omission in mdeberta-v3-base; different tokenizer behaviors." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "Specific is_offline_mode import error vs broad upgrade-time import regressions; too broad and not the same bug." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline cache load failure vs wrong tokenizer decoder type in v5; different tokenizer regressions." + } + ] + }, + "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": 20, + "review_unit_index": 45, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28218", + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43550", + "issue:43576", + "issue:43646", + "issue:43756", + "issue:43784", + "issue:43937", + "issue:43976", + "issue:44060", + "issue:44117", + "issue:44188", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44368", + "issue:44373", + "issue:44485", + "issue:44488", + "issue:44521", + "issue:44704", + "issue:44945", + "issue:44964", + "issue:45278", + "issue:45290", + "issue:45593", + "issue:45600", + "issue:45753", + "issue:46002", + "issue:46132", + "issue:46305", + "issue:46326", + "issue:46369" + ], + "soft_pairs": [ + "issue:42831|issue:45753", + "issue:27957|issue:30990", + "issue:22355|issue:28218", + "issue:44368|issue:46305", + "issue:42886|issue:42994", + "issue:43232|issue:45593", + "issue:43408|issue:46369", + "issue:43550|issue:46132", + "issue:43646|issue:46002", + "issue:43576|issue:43784", + "issue:42371|issue:44263", + "issue:44188|issue:44373", + "issue:44117|issue:44704", + "issue:30990|issue:33357", + "issue:44295|issue:44297", + "issue:43756|issue:44060", + "issue:42907|issue:44945", + "issue:44261|issue:44373", + "issue:42907|issue:43937", + "issue:41628|issue:45278", + "issue:43976|issue:44945", + "issue:39401|issue:44488", + "issue:43122|issue:44521", + "issue:44704|issue:45290", + "issue:42907|issue:44964", + "issue:43784|issue:45600", + "issue:43122|issue:46326", + "issue:39401|issue:44521", + "issue:44360|issue:44485" + ], + "prepared_review_unit_hash": "2db9545763907191a31ac18273a7f2c935b86fef2530f5d40dcd72e3cc3215c3", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35872, + "estimated_input_tokens": 8968, + "estimated_eval_tokens": 18192 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a mixed set of unrelated Transformers issues; none of the soft pairs look like the same underlying bug/change closely enough to merge. No PRs were provided.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45278 is the broadest and most representative regression report here: it captures multiple import failures after a version upgrade and could serve as an umbrella for related import-breakage reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best standalone issue because it is broad, clearly actionable, and likely to attract the right fix for the most general class of failures in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "Different problems: FP8 accuracy regression vs a multimodal generation infinite-loop on a specific image." + }, + { + "left": "issue:27957", + "right": "issue:30990", + "accept": false, + "reason": "XLM-RoBERTa Flash Attention 2 behavior and Sentence-Transformers loading hang are unrelated code paths." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "Missing transformers.onnx module import and tokenizer spacing after added token are unrelated." + }, + { + "left": "issue:44368", + "right": "issue:46305", + "accept": false, + "reason": "Weight-tying warning in Qwen3.5 fine-tuning is not the same as Moonshine streaming mask off-by-one." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline tokenizer cache loading and quantized model saving failure affect different subsystems and failure modes." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to D-FINE auxiliary-loss behavior." + }, + { + "left": "issue:43408", + "right": "issue:46369", + "accept": false, + "reason": "Model-type mismatch warning for sam3_tracker vs memory overflow in sam3-litetext are different issues." + }, + { + "left": "issue:43550", + "right": "issue:46132", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba is not the same as AttentionInterface.register changing registered-function behavior." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Custom model initialization breakage and NaNs in classification heads after checkout/init are different bugs." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Broken transformers env command and a NameError importing sentence-transformers are separate import/CLI problems." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 API settings warning is unrelated to the GlmMoeDsaIndexer torch.split issue." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "Attention-kernel divergence under torch.compile is unrelated to a docstring bug for position_ids." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor not passing kwargs to cached_file are different loader bugs." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Loading hang for Sentence-Transformers and a MacOS bus error with community CLIP models are distinct runtime failures." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Reading position_ids after buffer registration and tokenizer_class mismatch on save_pretrained are unrelated." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE-layer count mismatch and Qwen3-Next tied-weights warning concern different model configs." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Dequantized model saving failure and pipeline-parallelism output corruption are not the same code-path problem." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "MLA q_a_layernorm epsilon precision issue is unrelated to the position_ids docstring." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Model saving failure and invalid GenerationConfig for GLM-5 are separate failures." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Both are import-related, but one is a specific missing export while the other is a broad upgrade regression; not enough evidence they share the same fix." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python-version compatibility failure is unrelated to pipeline-parallel generation correctness." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Wrong offset_mapping in Qwen3 tokenizer and failure to load cjvt/sleng-bert are different tokenizer issues." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Version-to-version tokenization differences and zero assistant_masks for multimodal inputs are different behaviors." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "Passing kwargs to cached_file in AutoProcessor is unrelated to chat-template crashes on tool-call assistant messages." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Quantized/dequantized saving failure and inability to load Phi-4-multimodal-instruct are different model-loading/saving problems." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "sentence-transformers import NameError and stale auto_mappings references to removed Sam3LiteText configs are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:46326", + "accept": false, + "reason": "Tokenizer output changes across versions and continue_final_message behavior for mistral_common are different tokenization features." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Qwen3 offset_mapping bug and multimodal assistant_mask zeroing are different tokenizer/template issues." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer lacking ReLU and GLM-5 RoPE implementation are unrelated model internals." + } + ] + }, + "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": 20, + "review_unit_index": 46, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:29159", + "issue:36296", + "issue:38617", + "issue:42175", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:42907", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43278", + "issue:43408", + "issue:43452", + "issue:43475", + "issue:43493", + "issue:43611", + "issue:43824", + "issue:43874", + "issue:43937", + "issue:43992", + "issue:44038", + "issue:44188", + "issue:44222", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44355", + "issue:44485", + "issue:44514", + "issue:44625", + "issue:44704", + "issue:44779", + "issue:44829", + "issue:44938", + "issue:44945", + "issue:45072", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45478", + "issue:45753", + "issue:46002", + "issue:46231", + "issue:46328" + ], + "soft_pairs": [ + "issue:43116|issue:45200", + "issue:44038|issue:44297", + "issue:44625|issue:44829", + "issue:43408|issue:43475", + "issue:42907|issue:43452", + "issue:44038|issue:44295", + "issue:43122|issue:44514", + "issue:44038|issue:44222", + "issue:45216|issue:45478", + "issue:43824|issue:44351", + "issue:22355|issue:29159", + "issue:42757|issue:44351", + "issue:36296|issue:43116", + "issue:43278|issue:44945", + "issue:42175|issue:44938", + "issue:43493|issue:45753", + "issue:22355|issue:29127", + "issue:43992|issue:44704", + "issue:43066|issue:44779", + "issue:43937|issue:44485", + "issue:42673|issue:45072", + "issue:38617|issue:45278", + "issue:44188|issue:46328", + "issue:43611|issue:46002", + "issue:44355|issue:46231", + "issue:42371|issue:43874", + "issue:43116|issue:43493", + "issue:44038|issue:45478" + ], + "prepared_review_unit_hash": "272c2052175ef61dd1803c20629e6caf5fd315ea891453e9c386512de9407b0e", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35237, + "estimated_input_tokens": 8810, + "estimated_eval_tokens": 17876 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated issue reports across different models, APIs, and regressions. The soft-similarity links mostly reflect broad topical overlap (imports, tokenization, multimodal models, dtype/parallelism) rather than the same concrete bug, so none should be merged as duplicates.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 45278 is the broadest representative here because it summarizes a wide import-regression class, but even that is not a true umbrella for the whole set; there is no strong canonical duplicate among these issues.", + "canonical_pr_reason": null, + "best_issue_reason": "45278 is the most general and cluster-like issue title, so it is the best single representative if one must be chosen, though the set does not form a real duplicate cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Different failures: classification script returning empty results vs Gemma 4 multimodal token defaults." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Unrelated bugs: Qwen3-VL-Moe issue vs tokenizer save_pretrained metadata mismatch." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "Different code paths: num_labels propagation vs flash_attention_3 training degeneration." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Same model family, but different concrete defects: warning mismatch vs missing attribute crash." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Dequantized save failure and gguf_file loading failure are different problems." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Different bugs: Qwen3-VL-Moe regression vs position_ids buffer access error." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Both are tokenizer-related, but the concrete failures and affected paths differ." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Different issues: Qwen3-VL-Moe bug vs FP8 save_pretrained moe bug." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Both concern Qwen3.5, but save_pretrained regression and from_pretrained error are distinct code paths." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and not the same underlying bug." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "No relation: ONNX import failure vs slow/fast tokenizer inconsistency." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Both are import regressions, but they involve different missing APIs and are not the same defect." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor parallel training bug is unrelated to the classification script output issue." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Different runtime problems: dtype change in evaluation vs incorrect output under pipeline parallelism." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Packaging/backend dependency issue vs Python 3.14 load failure; not the same bug." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 discrepancy vs Qwen3_5 infinite loop on an image." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "No underlying match: ONNX import error vs LayoutLMv3 error-message clarity." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Different areas: missing encoder weights vs AutoProcessor kwarg forwarding." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but they affect different models and likely different code paths." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "Same model family, but invalid GenerationConfig and RoPE implementation are distinct problems." + }, + { + "left": "issue:42673", + "right": "issue:46328", + "accept": false, + "reason": "VRAM leak in threaded dataloading is unrelated to a padding performance issue." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "Broad import-error cluster overlap, but not the same concrete missing symbol or defect." + }, + { + "left": "issue:44188", + "right": "issue:46328", + "accept": false, + "reason": "Attention-kernel divergence under torch.compile is unrelated to feature-extractor padding overhead." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "Base-model-prefix loading regression and NaNs after checkout/init are different failures." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled-file import/runtime errors and a documentation parameter mismatch are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API guidance and a missing image-patch method are unrelated issues." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Different models and behaviors: empty classification outputs vs SigLIP2 implementation discrepancy." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both mention Qwen3-family MoE, but they are different models and 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": 20, + "review_unit_index": 47, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:33453", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42673", + "issue:42796", + "issue:43064", + "issue:43116", + "issue:43122", + "issue:43441", + "issue:43450", + "issue:43673", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:43937", + "issue:43976", + "issue:43986", + "issue:44075", + "issue:44117", + "issue:44291", + "issue:44380", + "issue:44410", + "issue:44466", + "issue:44779", + "issue:44829", + "issue:44936", + "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", + "issue:46178", + "issue:46298", + "issue:46328", + "issue:46512" + ], + "soft_pairs": [ + "issue:42222|issue:43441", + "issue:43976|issue:45446", + "issue:44075|issue:44936", + "issue:45357|issue:45478", + "issue:43116|issue:43825", + "issue:45083|issue:45446", + "issue:42673|issue:42796", + "issue:43064|issue:43673", + "issue:43937|issue:45563", + "issue:33453|issue:39401", + "issue:44380|issue:46077", + "issue:44117|issue:45588", + "issue:43827|issue:45446", + "issue:39401|issue:44779", + "issue:42673|issue:43906", + "issue:43673|issue:46328", + "issue:44291|issue:45071", + "issue:44466|issue:46002", + "issue:43450|issue:43986", + "issue:43937|issue:44977", + "issue:45405|issue:45446", + "issue:44410|issue:45310", + "issue:45341|issue:45561", + "issue:45341|issue:46328", + "issue:43122|issue:46298", + "issue:38617|issue:41628", + "issue:44117|issue:45081", + "issue:45071|issue:46512", + "issue:42222|issue:46178", + "issue:44829|issue:45216" + ], + "prepared_review_unit_hash": "4e50bd1d9d21900a1a1e713bd692f6d9b40b86345e5a3e699415d3b438b8f9ca", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36668, + "estimated_input_tokens": 9167, + "estimated_eval_tokens": 18590 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated Transformers v5 regressions, model-specific bugs, and a few docs/testing issues. I found no soft edge pair that clearly shares the same underlying bug/change closely enough to merge.", + "confidence": 0.62, + "canonical_issue_reason": "Issue 43122 is the best representative of the cluster\u2019s dominant theme: a concrete v4.57.3\u2192v5 tokenizer regression with user-visible impact and some follow-on references. The rest of the cluster is too heterogeneous for a single true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "43122 is the strongest global representative because it is specific, regression-oriented, and broad enough to capture the tokenizer-loading/tokenization theme better than the more model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "Different models and failure modes: vitpose model breakage vs Ministral FlashAttention crash." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Both are version-check issues, but they affect different imports/modules and do not describe the same bug." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Optimizer argument handling and trainer.evaluate failure are distinct code paths and symptoms." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but one is save_pretrained key serialization and the other is from_pretrained loading; not the same concrete bug." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Example-script multi-label behavior and pipeline docs/error-message wording are unrelated." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "A qwen3_omni_moe helper bug and a flex_attention PyTorch version check are different problems." + }, + { + "left": "issue:42673", + "right": "issue:42796", + "accept": false, + "reason": "VRAM leak in dataloader threads vs tokenizer BPE marker handling are unrelated." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "Distributed optimizer-state corruption and generation cache missing are different subsystems and bugs." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "GenerationConfig validation failure and stale warning text are not the same issue." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Generic tokenizer-loading regression vs Qwen3 offset_mapping bug; no clear shared code-path shown." + }, + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "Attention-scaling config being ignored and bidirectional mask construction are different model/back-end behaviors." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Tokenizer mapping import assumptions and flash_attention sink-less-model crash are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs still referencing pipeline() is a documentation issue, not the flex_attention version-check bug." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they target different models and symptoms with no evidence of a shared defect." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "VRAM leak across dataloader threads is unrelated to an isolated reproduction of a different issue." + }, + { + "left": "issue:43673", + "right": "issue:46328", + "accept": false, + "reason": "Generation cache behavior and a padding performance issue are distinct problems." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument and PretrainedConfig type-checking regression are separate compatibility bugs." + }, + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "Serialization of tied weights and NaNs in classification heads are different failure modes." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video-processor shape handling and missing-torchvision AutoProcessor crash are not the same bug." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "Invalid GenerationConfig vs flash-attention generation failure are unrelated." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "Dependency version bump and a PyTorch version-check bug are different issues." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "qwen3next missing projections and Qwen3.5 from_pretrained error are different model-specific defects." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "A minor testing_utils bug and a pytest-xdist captured_info race are distinct issues." + }, + { + "left": "issue:45341", + "right": "issue:46328", + "accept": false, + "reason": "testing_utils bug and SequenceFeatureExtractor.pad performance regression are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:46298", + "accept": false, + "reason": "Tokenizer behavior regression and a noisy logging message do not share the same underlying bug." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but for different symbols and modules; no duplicate evidence." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and _patch_mistral_regex backend_tokenizer AttributeError are different tokenizer-loading failures." + }, + { + "left": "issue:45071", + "right": "issue:46512", + "accept": false, + "reason": "PretrainedConfig type checking and stale config.dtype after from_config are separate config regressions." + }, + { + "left": "issue:42222", + "right": "issue:46178", + "accept": false, + "reason": "vitpose model breakage and tensorflow_text backend mapping errors are unrelated." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 causing degenerate training and incorrect save_pretrained checkpoint serialization 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": 20, + "review_unit_index": 48, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:27957", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42617", + "issue:42673", + "issue:42831", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43352", + "issue:43408", + "issue:43577", + "issue:43582", + "issue:43673", + "issue:43742", + "issue:43874", + "issue:43937", + "issue:44075", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44360", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44496", + "issue:44877", + "issue:44933", + "issue:45071", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45362", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45656", + "issue:46172", + "issue:46209", + "issue:46231", + "issue:46328" + ], + "soft_pairs": [ + "issue:45362|issue:46328", + "issue:43408|issue:43742", + "issue:42617|issue:45341", + "issue:42673|issue:43582", + "issue:42371|issue:44261", + "issue:44877|issue:46231", + "issue:44484|issue:45310", + "issue:39401|issue:46209", + "issue:38617|issue:44220", + "issue:43937|issue:44496", + "issue:44933|issue:45278", + "issue:43066|issue:43122", + "issue:42371|issue:42831", + "issue:44410|issue:45478", + "issue:42222|issue:45542", + "issue:44075|issue:45656", + "issue:44483|issue:45464", + "issue:43012|issue:43577", + "issue:42175|issue:42222", + "issue:43097|issue:44493", + "issue:43673|issue:45341", + "issue:43874|issue:44261", + "issue:41628|issue:45071", + "issue:44263|issue:44485", + "issue:45071|issue:45542", + "issue:27957|issue:43352", + "issue:44484|issue:45478", + "issue:44263|issue:46172", + "issue:44360|issue:46172" + ], + "prepared_review_unit_hash": "64a1b7635f9c80d7d7ea6c7cdd8465415602ab5ebdf14c583733902d598c19ed", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35689, + "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 issues span unrelated model-loading regressions, tokenizer bugs, optimizer/training behavior, import errors, and packaging/docs problems. None of the soft pairs look like the same underlying defect, so there is no convincing duplicate set here.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the broadest representative ('many import errors after update'), but it still only loosely anchors one subtheme and does not explain the rest of the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45362", + "right": "issue:46328", + "accept": false, + "reason": "Different areas: Qwen chat/runtime failure vs SequenceFeatureExtractor padding performance." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "Unrelated model families and failure modes." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "Different utilities; a runtime script failure vs a testing helper bug." + }, + { + "left": "issue:42673", + "right": "issue:43582", + "accept": false, + "reason": "Different platforms and code paths: VRAM leak vs Apple Silicon TypeError." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 settings issue is not the same as the q_a_layernorm epsilon bug." + }, + { + "left": "issue:44877", + "right": "issue:46231", + "accept": false, + "reason": "Config parsing regression vs documentation typo." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "save_pretrained shard-size question vs Qwen3.5 MoE loading error." + }, + { + "left": "issue:39401", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer offset_mapping bug vs FP8 inference failure on B200." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "ImportError in configuration_utils vs audio feature extraction issue." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "GenerationConfig validation failure vs unrecognized model/config error." + }, + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "Shared import theme only; one is a specific missing symbol, the other is broad post-upgrade import failures." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Both are tokenizer regressions, but one is decoder-type metadata and the other is tokenization drift." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 control bug vs FineGrainedFP8 accuracy bug." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Different model code paths: missing projections vs from_pretrained loading error." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "VitPose breakage vs TensorFlow backend packaging issue." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "Both optimizer-related, but SGD args being ignored is not the same as double-stepping with deepspeed." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Different request modes/endpoints; not the same chat/completions defect." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both involve dtype, but one is a compile-time warning and the other is incorrect model dtype on load." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "Packaging/backend dependency issue vs model-specific VitPose breakage." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "API removal in config semantics vs unexpected key warnings from position IDs." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Generation cache regression vs a testing utility bug." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Image patch counting bug vs rms_norm_eps precision bug." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "Missing import symbol vs PreTrainedConfig type-checking regression." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "Different GLM submodules and different bugs." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Type-checking regression vs TensorFlow backend error." + }, + { + "left": "issue:27957", + "right": "issue:43352", + "accept": false, + "reason": "Different models and different Flash Attention 2 support states." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "save_pretrained shard sizing is unrelated to Qwen3.5 MoE loading." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "Different models and code paths: GlmMoeDsaIndexer vs deepseek-v4 dtype/transpose issue." + }, + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "DSA indexer ReLU bug is unrelated to the deepseek-v4 tensor transform 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": 20, + "review_unit_index": 49, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:38617", + "issue:42757", + "issue:42890", + "issue:43064", + "issue:43097", + "issue:43441", + "issue:43452", + "issue:43502", + "issue:43531", + "issue:43673", + "issue:43824", + "issue:43873", + "issue:43874", + "issue:43931", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44315", + "issue:44351", + "issue:44380", + "issue:44410", + "issue:44484", + "issue:44485", + "issue:44704", + "issue:44829", + "issue:44933", + "issue:44936", + "issue:44987", + "issue:45003", + "issue:45103", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45440", + "issue:45588", + "issue:45753", + "issue:46050", + "issue:46123", + "issue:46172", + "issue:46231", + "issue:46298", + "issue:46365", + "issue:46369", + "issue:46421" + ], + "soft_pairs": [ + "issue:43931|issue:44410", + "issue:43441|issue:44829", + "issue:43931|issue:44038", + "issue:43097|issue:43452", + "issue:45003|issue:45341", + "issue:45753|issue:46369", + "issue:43986|issue:45341", + "issue:44933|issue:46231", + "issue:43874|issue:44263", + "issue:43673|issue:46123", + "issue:43986|issue:44484", + "issue:44117|issue:44987", + "issue:44261|issue:44485", + "issue:45440|issue:46172", + "issue:43986|issue:45372", + "issue:43531|issue:45341", + "issue:43064|issue:44315", + "issue:43097|issue:45588", + "issue:38617|issue:42757", + "issue:43097|issue:46050", + "issue:44410|issue:45375", + "issue:43502|issue:43992", + "issue:44038|issue:44410", + "issue:42890|issue:46298", + "issue:38617|issue:44351", + "issue:44062|issue:44117", + "issue:43097|issue:43873", + "issue:44936|issue:46365", + "issue:44704|issue:45103", + "issue:43986|issue:44117", + "issue:44380|issue:46421", + "issue:43824|issue:43986" + ], + "prepared_review_unit_hash": "c8a3a0704c3b2594c779c281c6d41cf309eb076c84f64da72efaf0d887b021ab", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37186, + "estimated_input_tokens": 9297, + "estimated_eval_tokens": 18850 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are mostly unrelated Transformer issues that only share broad keywords; none of the proposed soft pairs look like true duplicates or mergeable PR equivalents.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:43097 is the strongest standalone representative: it describes a concrete v5 regression, has discussion/inbound references, and is more actionable than the many one-off import/load errors. Still, this is not a true duplicate center for the full set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Both are model-shape/config complaints for different Qwen/vision models; different symptoms and code paths." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Both mention attention backends, but one is a Ministral FlashAttention failure and the other is a sequence-classification training degeneration; not the same bug." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Different Qwen-VL model families and different loading failures; only broad VL overlap." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "One is a removed trainer/model API in v5, the other is gguf_file tokenizer/model loading; unrelated code paths." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "Both involve code bugs, but one is unsafe sys.modules access and the other is a small testing_utils issue; no shared underlying defect." + }, + { + "left": "issue:45753", + "right": "issue:46369", + "accept": false, + "reason": "Both are model-specific runtime problems, but one is a Qwen image infinite loop and the other is SAM memory overflow; different causes." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "Video-model processor import crash vs a minor testing utility bug; no overlap beyond general bugs." + }, + { + "left": "issue:44933", + "right": "issue:46231", + "accept": false, + "reason": "One is an import failure from image_utils, the other is a documentation parameter-name mismatch." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Both are GLM-related, but one is a missing image-patch method and the other is torch.split return handling; distinct failures." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Both mention cache/prefill behavior, but one is missing GenerationMixin cache in v5 and the other is mask pipeline batch handling." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Processor import-time crash vs save_pretrained shard-size question; unrelated." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer-mapping None handling vs model load failure on a specific package/version; different layers of the stack." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Both touch precision/config in attention/rope-style math, but one is a missing rms_norm_eps in MLA and the other is GLM-5 RoPE implementation discussion; not the same bug." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "Different DeepSeek issues: one is native-vs-remote implementation divergence, the other is a specific dtype/transpose bug." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Both are loading-time import problems, but one concerns missing torchvision for video processors and the other missing ReasoningEffort from mistral_common." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to a small testing utility bug." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Trainer/FSDP2 optimizer state corruption is unrelated to Liger Kernel not being applied with model_init." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Trainer API removal regression vs flash_attention integration crash on s_aux=None; different subsystems and fixes." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Two distinct import errors for different symbols from different packages; only superficial similarity." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Trainer v5 API regression vs quantized model save/load corruption after sinq; not the same underlying defect." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Both are Qwen vision configs, but one is missing projections at layer 0 and the other is a missing config field silently dropped by strict validation." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Local-files-only network leak vs missing embed_tokens.weight on UMT5Encoder; different loading bugs." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Both are Qwen-related, but one is a Transformers v5 bug report and the other is a specific config/projection omission." + }, + { + "left": "issue:42890", + "right": "issue:46298", + "accept": false, + "reason": "Integration-test flakiness from missing set_seed is unrelated to a noisy logging message." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Different import failures: layer_type_validation vs HybridCache; same symptom class, different symbol and likely different fixes." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "Tokenizer AddedToken duplicate kwargs bug is unrelated to TOKENIZER_MAPPING_NAMES returning None." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Trainer state regression vs offloading/quantization behavior; not the same code-path problem." + }, + { + "left": "issue:44936", + "right": "issue:46365", + "accept": false, + "reason": "Trainer.evaluate after train is a runtime state issue; the other is a Trainer import problem influenced by datasets installation." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "AutoProcessor cached_file kwargs forwarding vs auto_docstring annotation introspection crash; different functionality." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Video processor import crash vs tokenizer mapping None handling; unrelated." + }, + { + "left": "issue:44380", + "right": "issue:46421", + "accept": false, + "reason": "GPT2 attention scaling ignored in backends vs cache lazy initialization causing recompilation; different bugs." + }, + { + "left": "issue:43824", + "right": "issue:43986", + "accept": false, + "reason": "Both are import/loading issues, but one is a missing Qwen2.5-VL class export and the other is a torchvision-dependent video processor crash." + } + ] + }, + "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": 20, + "review_unit_index": 50, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42175", + "issue:42490", + "issue:42757", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43352", + "issue:43502", + "issue:43526", + "issue:43550", + "issue:43582", + "issue:43618", + "issue:43761", + "issue:43824", + "issue:43825", + "issue:43906", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44077", + "issue:44162", + "issue:44222", + "issue:44230", + "issue:44368", + "issue:44380", + "issue:44483", + "issue:44484", + "issue:44514", + "issue:44655", + "issue:44683", + "issue:44938", + "issue:45478", + "issue:45479", + "issue:45584", + "issue:45600", + "issue:45698", + "issue:45712", + "issue:45753", + "issue:46132", + "issue:46172", + "issue:46178", + "issue:46298", + "issue:46365" + ], + "soft_pairs": [ + "issue:44938|issue:46178", + "issue:43992|issue:44484", + "issue:43066|issue:43906", + "issue:43208|issue:43550", + "issue:42907|issue:44655", + "issue:43012|issue:43550", + "issue:43012|issue:44368", + "issue:45712|issue:46298", + "issue:43825|issue:44655", + "issue:42907|issue:45753", + "issue:43352|issue:44380", + "issue:43097|issue:46298", + "issue:43352|issue:43976", + "issue:44162|issue:46172", + "issue:45584|issue:46365", + "issue:44380|issue:46132", + "issue:44077|issue:46365", + "issue:44380|issue:45479", + "issue:43064|issue:44380", + "issue:36683|issue:43582", + "issue:44222|issue:45478", + "issue:42175|issue:46365", + "issue:42490|issue:43012", + "issue:44483|issue:44514", + "issue:43012|issue:44683", + "issue:38617|issue:43502", + "issue:43208|issue:43526", + "issue:42757|issue:43824", + "issue:45600|issue:46298", + "issue:45698|issue:46172", + "issue:43618|issue:44380", + "issue:43618|issue:43761", + "issue:44038|issue:44230" + ], + "prepared_review_unit_hash": "ed1c8efd9d8089edde8cd53af94b7c77bd192d6006f87c733dc0653e24260705", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37301, + "estimated_input_tokens": 9326, + "estimated_eval_tokens": 18908 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a mix of distinct Transformers issues: import/API breakages, save/load regressions, v5 behavioral changes, and model-specific bugs. None of the soft-similarity pairs look like the same underlying defect, so they should stay separate.", + "confidence": 0.64, + "canonical_issue_reason": "issue:43097 is the broadest and most central regression report: a concrete v5 backward-compatibility break with clear user impact and inbound references.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43097 is the strongest representative issue here because it describes a specific, high-impact v5 API regression rather than a narrow model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44938", + "right": "issue:46178", + "accept": false, + "reason": "Python 3.14 load failure vs undefined backend mapping for tensorflow_text; different failure modes and code paths." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing embed_tokens.weight during UMT5 loading is unrelated to the 50GB shard-size question in save_pretrained()." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Wrong tokenizer decoder type is a separate tokenizer regression from the isolated reproduction of another issue." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training bugs and Bamba torch.compile+SDPA failure affect different models and different mechanics." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Dequantized model saving and pipeline save_pretrained() are both save-related, but the concrete bugs differ." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "bfloat16 compile warning is just a warning issue; Bamba compile with SDPA is a functional model/backend failure." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "PyTorch precision warning and tie_word_embeddings warning are separate warning paths with different causes." + }, + { + "left": "issue:45712", + "right": "issue:46298", + "accept": false, + "reason": "Left is leftover dummy-class cleanup; right is a noisy logging message. Different scope and fix." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "pipeline translation-task messaging is unrelated to pipeline save_pretrained() serialization behavior." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Saving dequantized models and Qwen3_5 infinite-loop generation are unrelated model lifecycle vs inference bugs." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Flash Attention support error for NemotronH vs GPT2 attention scaling ignored under SDPA/FlashAttention are different code paths." + }, + { + "left": "issue:43097", + "right": "issue:46298", + "accept": false, + "reason": "A v5 API removal regression is not the same as an unspecified noisy logging message." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Nemotron Flash Attention support and Python 3.9/3.10 compatibility are distinct issues." + }, + { + "left": "issue:44162", + "right": "issue:46172", + "accept": false, + "reason": "ESM2 training/model breakage and a DeepSeek tensor transpose bug are unrelated model-specific failures." + }, + { + "left": "issue:45584", + "right": "issue:46365", + "accept": false, + "reason": "Whisper empty transcription after token alignment is unrelated to importing Trainer failing because of datasets." + }, + { + "left": "issue:44380", + "right": "issue:46132", + "accept": false, + "reason": "Ignoring GPT2 attention scaling and AttentionInterface.register behavior changes are different attention APIs and defects." + }, + { + "left": "issue:44077", + "right": "issue:46365", + "accept": false, + "reason": "patchtsmixer post_init validation and Trainer import failure are unrelated subsystems." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention scaling backend bug and degenerate zero-loss classification bug are different model behaviors." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "FSDP2/PEFT optimizer state mismatch and GPT2 attention scaling under SDPA are unrelated training vs inference bugs." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "Gemma3Config missing vocab_size and Apple Silicon caching_allocator_warmup TypeError are different model/runtime failures." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "FP8 save_pretrained() for moe and Qwen3.5 moe from_pretrained() error are related to the model family but not the same bug." + }, + { + "left": "issue:42175", + "right": "issue:46365", + "accept": false, + "reason": "Tensorflow absent from a torch extra install is a packaging issue, not the same as Trainer import failing due to datasets." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "Trailing-slash local model import warning and bfloat16 compile warning are unrelated warning paths." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Chat completions API acceptance regression and Qwen2_5_VLProcessor batching crash are different request/processor bugs." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "A precision warning during compilation is not the same as flex_attention failing on torch >= 2.9." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "Missing layer_type_validation import and unwanted API requests despite local_files_only are separate import vs network issues." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "xLSTM training blockers and BeitImageProcessorFast reduce_labels returning one label are unrelated." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Missing is_offline_mode import and missing Qwen2_5_VLForConditionalGeneration export are both import errors but from different symbols and fixes." + }, + { + "left": "issue:45600", + "right": "issue:46298", + "accept": false, + "reason": "auto_mappings.py stale config references and a noisy logging message are not the same defect." + }, + { + "left": "issue:45698", + "right": "issue:46172", + "accept": false, + "reason": "Wrong custom module loading after save_pretrained() and a DeepSeek tensor transpose bug are unrelated save/load vs tensor-shape issues." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIPOutput attentions assignment regression is separate from GPT2 attention scaling being ignored under backend kernels." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Missing CLIP attentions output and CLIPVisionModel hidden_states regression are adjacent but distinct output-field bugs." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Qwen3-VL-Moe version bug and fp8 Qwen3.5 moe inference support are different model/version problems." + } + ] + }, + "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": 20, + "review_unit_index": 51, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:36010", + "issue:36683", + "issue:41669", + "issue:42757", + "issue:42890", + "issue:42898", + "issue:42994", + "issue:43010", + "issue:43064", + "issue:43066", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43525", + "issue:43611", + "issue:43618", + "issue:43756", + "issue:43867", + "issue:43901", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44075", + "issue:44077", + "issue:44246", + "issue:44355", + "issue:44483", + "issue:44485", + "issue:44492", + "issue:44625", + "issue:44655", + "issue:44855", + "issue:44938", + "issue:45071", + "issue:45216", + "issue:45276", + "issue:45278", + "issue:45290", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45561", + "issue:45593", + "issue:45850", + "issue:46077", + "issue:46082", + "issue:46202", + "issue:46365" + ], + "soft_pairs": [ + "issue:44483|issue:45290", + "issue:43299|issue:43867", + "issue:42994|issue:44655", + "issue:36683|issue:43475", + "issue:42898|issue:43066", + "issue:44625|issue:46365", + "issue:43756|issue:44485", + "issue:45276|issue:46077", + "issue:43352|issue:43937", + "issue:36010|issue:42757", + "issue:43611|issue:44655", + "issue:43901|issue:44655", + "issue:36683|issue:44077", + "issue:43867|issue:45216", + "issue:43976|issue:44938", + "issue:44625|issue:46077", + "issue:43352|issue:45593", + "issue:46077|issue:46082", + "issue:36683|issue:43525", + "issue:43618|issue:45216", + "issue:36683|issue:44625", + "issue:43992|issue:45310", + "issue:43064|issue:44492", + "issue:43010|issue:45278", + "issue:44075|issue:46365", + "issue:44355|issue:44855", + "issue:43867|issue:45375", + "issue:43867|issue:46202", + "issue:44938|issue:45071", + "issue:36683|issue:43064", + "issue:41669|issue:44246", + "issue:42890|issue:45561", + "issue:43010|issue:45850", + "issue:43867|issue:45357" + ], + "prepared_review_unit_hash": "9e8a036bad02e39284828bb4a3c20e9d7655788f526cf74d6a5b0eecf421c81e", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37463, + "estimated_input_tokens": 9366, + "estimated_eval_tokens": 18988 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are related only by broad theme (imports, saving, loading, config regressions) and are not the same concrete bug. I accepted only the import-time slowdown pair; the rest look like distinct issues.", + "confidence": 0.62, + "canonical_issue_reason": "issue:45278 is the broadest representative here: it explicitly summarizes widespread import failures after upgrading to 5.5.0, which is the closest umbrella for the import-regression slice of this set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best overall issue representative because it captures the most general shared symptom in the cluster (post-upgrade import breakage), even though it is broader than most concrete reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Both are chat-related crashes, but one is /v1/chat/completions request handling and the other is apply_chat_template with tool-call messages; different code paths." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Both are model-loading problems, but one is a Qwen3VL MoE loading regression and the other is a state_dict-sorting load error; not the same bug." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Quantized model saving and Pipeline.save_pretrained are different serialization paths with different failure modes." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Missing Gemma3Config.vocab_size and missing Sam3VisionEncoderOutput.fpn_position_embeddings are unrelated attribute bugs." + }, + { + "left": "issue:42898", + "right": "issue:43066", + "accept": false, + "reason": "Tokenizer behavior change and wrong decoder type are both v5 tokenizer regressions, but they are distinct concrete issues." + }, + { + "left": "issue:44625", + "right": "issue:46365", + "accept": false, + "reason": "num_labels propagation and Trainer import being blocked by datasets are unrelated problems." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both mention RoPE, but one is Smollm3 dropping layers and the other is a GLM-5 RoPE discussion; different bugs." + }, + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Both are Gemma4 regressions, but one is resize_token_embeddings behavior and the other is attention-mask construction." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Flash-Attention support and invalid GenerationConfig are unrelated failures." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols from different dependencies and do not share a concrete root cause." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Loading models with base_model_prefix and saving Pipeline objects are different code paths and different regressions." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "This is a docs mismatch versus a save_pretrained bug; not the same underlying issue." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "A missing config attribute and ignored optimizer SGD args are unrelated." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "Generic load_model failure from sorted state_dict and Qwen3.5 save_pretrained corruption are distinct bugs." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version compatibility reports, but they target different versions and different failure modes." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "num_labels propagation and Gemma4 attention-mask logic are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:45593", + "accept": false, + "reason": "Flash-Attention support and D-FINE auxiliary-loss handling are unrelated model issues." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Gemma4 attention masks and LlamaConfig head_dim validation are different code paths." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Gemma3Config.vocab_size and Llama4Config.pad_token_id are different missing fields in different models." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIPOutput attentions regression and Qwen3.5 checkpoint saving regression are unrelated." + }, + { + "left": "issue:36683", + "right": "issue:44625", + "accept": false, + "reason": "Both involve configs, but one is a missing attribute and the other is num_labels propagation; not the same bug." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "UMT5 weight loading and Qwen3.5 MoE from_pretrained failure are different model-loading issues." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Trainer optimizer-state corruption and a cache-strategy typo are unrelated." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "@torch.no_grad on cache updates and broad import errors after upgrade are different problems." + }, + { + "left": "issue:44075", + "right": "issue:46365", + "accept": false, + "reason": "Ignored SGD args and Trainer import blocked by datasets are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Compiled Python file errors and the Python 3.13 @torch.jit.script parsing/IndentationError regression are different causes." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "Generic load_model failure and a missing Qwen3.5 vision config field are unrelated." + }, + { + "left": "issue:43867", + "right": "issue:46202", + "accept": false, + "reason": "Load error from state_dict sorting and Gemma pipeline roundtrip failure on \u2581 are unrelated." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python 3.14 load failure and PretrainedConfig type-checking regression are distinct issues." + }, + { + "left": "issue:36683", + "right": "issue:43064", + "accept": false, + "reason": "Missing config attribute and distributed optimizer-state bug are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "Both describe the same import-performance regression: importing transformers is slow, and 41669 identifies import-star usage in models as the cause." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "A missing set_seed in integration tests and an xdist race on captured_info.txt are different testing issues." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "Cache update decoration and a broad 'v5 breaks too many models' complaint are not the same concrete bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: issue:45278 is indeed the broad import-regression umbrella, and the one accepted pair (issue:41669 vs issue:44246) is well supported by the issue titles. The remaining soft-edge verdicts are conservative and do not overstate shared root cause." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 52, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:27957", + "issue:28282", + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:42548", + "issue:42831", + "issue:42890", + "issue:43010", + "issue:43065", + "issue:43232", + "issue:43299", + "issue:43502", + "issue:43575", + "issue:43576", + "issue:43618", + "issue:43723", + "issue:43827", + "issue:43874", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44077", + "issue:44112", + "issue:44222", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44387", + "issue:44485", + "issue:44556", + "issue:44655", + "issue:44749", + "issue:44857", + "issue:44961", + "issue:44964", + "issue:45071", + "issue:45125", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45600", + "issue:46132", + "issue:46285" + ], + "soft_pairs": [ + "issue:43992|issue:45478", + "issue:43010|issue:43976", + "issue:36296|issue:43575", + "issue:43618|issue:44077", + "issue:28282|issue:33453", + "issue:43618|issue:46285", + "issue:43618|issue:44857", + "issue:44360|issue:45440", + "issue:43502|issue:44062", + "issue:42890|issue:44964", + "issue:42890|issue:45600", + "issue:42890|issue:44112", + "issue:43723|issue:44556", + "issue:44222|issue:45125", + "issue:44387|issue:44556", + "issue:43010|issue:43065", + "issue:43576|issue:46132", + "issue:43502|issue:44485", + "issue:45125|issue:45478", + "issue:44355|issue:44749", + "issue:42548|issue:44351", + "issue:30333|issue:33453", + "issue:45125|issue:45375", + "issue:43010|issue:43232", + "issue:30990|issue:31795", + "issue:43502|issue:44222", + "issue:43299|issue:44556", + "issue:43502|issue:44961", + "issue:42831|issue:43874", + "issue:30333|issue:36296", + "issue:44355|issue:45071", + "issue:43827|issue:44961", + "issue:44355|issue:44655", + "issue:33453|issue:34689", + "issue:44315|issue:44360", + "issue:27957|issue:33453" + ], + "prepared_review_unit_hash": "4c368fe1523b16e11ba2fa9285d455a872e75d92b218527766d6b30d2f44d4fc", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 37972, + "estimated_input_tokens": 9493, + "estimated_eval_tokens": 19242 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the issues cover unrelated loading, configuration, CI, quantization, and documentation problems. I found no soft-edge pair that looks like the same underlying bug/change, so none are accepted.", + "confidence": 0.97, + "canonical_issue_reason": "issue:33453 is the broadest and most discussion-heavy user-facing regression here, so it is the best anchor if one must be chosen; however, the set is too mixed to form a true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:33453 is the most representative standalone issue in the group because it describes a generic regression and has the highest discussion activity among the clearly bug-focused reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Both are model-loading problems, but one is UMT5 encoder state loading and the other is Qwen3.5 MoE from_pretrained failure; different code paths and symptoms." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache no_grad decoration vs a Python version compatibility regression; unrelated problems." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug and an OOM while loading a specific model are not the same failure mode." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIP attentions regression and patchtsmixer post_init restriction are unrelated model behaviors." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "Missing PyTorch import error vs tokenizer loading regression; different underlying causes." + }, + { + "left": "issue:43618", + "right": "issue:46285", + "accept": false, + "reason": "Both mention CLIP internals, but one is about attentions being dropped and the other about text model structure divergence; not the same bug." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "Attention output assignment and LwDetrImageLoss AMP crashes are unrelated." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "A missing ReLU in an indexer is a different issue from DeepseekV3MoE diverging from remote behavior." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "Local-files-only network requests and AddedToken argument duplication are separate bugs." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "A flaky SamHQ test and a Phi-4 multimodal loading failure are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Test seeding flakiness and removed Sam3LiteText configs are different CI issues." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Both involve tests, but they target different models and failure mechanisms." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Tokenizer loading in v5 and checkpoint reload compatibility after upgrading versions are similar only at a high level." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained behavior and missing tensor-parallel plan are distinct MoE issues." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "CUDA reserved-memory/OOM under int4 quantization is not the same as reload incompatibility after version upgrade." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Generation cache no_grad decoration and a dummy Conv2d in Sam3PixelDecoder are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Broken env command and AttentionInterface.register behavior change do not share a concrete code-path bug." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Network access despite local_files_only and FP8 MoE save_pretrained are different concerns." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Both concern Qwen3.5 MoE loading, but one is missing _tp_plan for tensor parallelism while the other is a broader from_pretrained error; not enough evidence of the same fix." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled-file runtime errors and Chinese-described slow filtering after upgrade are unrelated." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Missing PreTrainedModel import and missing HybridCache import are different export regressions." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLFlow reporting on failed training and tokenizer loading regression are unrelated." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Missing tensor-parallel plan and a missing vision config field are separate config/export issues." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Different generation-related bugs: cache update decoration vs sync_gpus update_model_kwargs handling." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Sentence Transformers loading hangs and forward-doc confusion are not the same issue." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Repeated pair in the data; still unrelated network-access and FP8 save_pretrained problems." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Qwen3VL MoE loading failure in dev0 and v4.57-to-v5.2/v5.3 reload breakage are different version-specific regressions." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "local_files_only network requests and a one-word issue title do not indicate the same bug." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FineGrainedFP8 accuracy and a missing image-patch count method are unrelated multimodal/quantization issues." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLFlow failure reporting and tensor-parallel training bugs are different subsystems and symptoms." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled-file errors and PretrainedConfig type checking regression are separate breakages." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs still referencing pipeline() and an unrelated one-word issue are not duplicates." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled Python file errors and inability to save Pipeline objects are distinct problems." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer loading regression and Llama 3.2 90B Vision model loading regression are both load-related, but they affect different models and failure points." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger kernel application and DSA indexer activation are unrelated." + }, + { + "left": "issue:27957", + "right": "issue:33453", + "accept": false, + "reason": "Flash Attention 2 support for XLMRoberta and tokenizer-loading regression are different issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 20, + "review_unit_index": 53, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:31795", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:41669", + "issue:42371", + "issue:42548", + "issue:42757", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43525", + "issue:43526", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43824", + "issue:43901", + "issue:44038", + "issue:44222", + "issue:44230", + "issue:44246", + "issue:44273", + "issue:44315", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45362", + "issue:45468", + "issue:45542", + "issue:45753", + "issue:45850", + "issue:46095", + "issue:46143", + "issue:46144", + "issue:46209", + "issue:46218" + ], + "soft_pairs": [ + "issue:42757|issue:43576", + "issue:43576|issue:45071", + "issue:43576|issue:43824", + "issue:31795|issue:38617", + "issue:44961|issue:45468", + "issue:43576|issue:44938", + "issue:44485|issue:44961", + "issue:44908|issue:44961", + "issue:46143|issue:46144", + "issue:44222|issue:44230", + "issue:36296|issue:43901", + "issue:41628|issue:42548", + "issue:44315|issue:46209", + "issue:36296|issue:43526", + "issue:43673|issue:44492", + "issue:45362|issue:45850", + "issue:44623|issue:45230", + "issue:44273|issue:45542", + "issue:44038|issue:45125", + "issue:44393|issue:45071", + "issue:44273|issue:45003", + "issue:44393|issue:45753", + "issue:43576|issue:44556", + "issue:43525|issue:45850", + "issue:41669|issue:43673", + "issue:44230|issue:46095", + "issue:44556|issue:45092", + "issue:43572|issue:45850", + "issue:44492|issue:45850", + "issue:45278|issue:45850", + "issue:42371|issue:44492", + "issue:41669|issue:46218", + "issue:43408|issue:44315", + "issue:42890|issue:44908", + "issue:43352|issue:44393", + "issue:44246|issue:44273", + "issue:43576|issue:44908" + ], + "prepared_review_unit_hash": "b23039210d877bcec8485f7e6176c6ccf1dc0763fdc51d68af328504a8ed692f", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 37987, + "estimated_input_tokens": 9497, + "estimated_eval_tokens": 19250 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster mixes several broad v5 migration/import regressions with many unrelated model-specific bugs. None of the soft pairs appear to be the same concrete underlying issue, so I would not merge any of them.", + "confidence": 0.64, + "canonical_issue_reason": "issue:45278 is the best umbrella issue here: it explicitly captures the v5-era import breakage pattern that most closely matches the cluster\u2019s import/lazy-loading failures.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the strongest representative because it is broad, user-facing, and directly describes the dominant failure mode seen across the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Different failures: huggingface_hub import vs transformers env CLI." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Unrelated surfaces: env command vs PretrainedConfig type checking." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "CLI regression is unrelated to a specific Qwen2.5-VL import export." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Docs confusion is unrelated to an ImportError for layer_type_validation." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "No overlap: a generic titled issue vs Gemma audio positional encoding." + }, + { + "left": "issue:43576", + "right": "issue:44938", + "accept": false, + "reason": "env command failure is not the same as Python 3.14 load failure." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM RoPE discussion is unrelated to the placeholder racoon report." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug is unrelated to the racoon report." + }, + { + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "Both RoFormer-related, but one is kwargs forwarding and the other is attention interface usage." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "FP8 save_pretrained and FP8 inference support are different code paths." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor parallel training bug is unrelated to docs for return_all_scores." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Both are import errors, but for different missing symbols with no concrete shared bug shown." + }, + { + "left": "issue:44315", + "right": "issue:46209", + "accept": false, + "reason": "Liger Kernel application and B200 FP8 inference are unrelated issues." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor parallel training bug is unrelated to BeitImageProcessorFast label reduction." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Generation cache missing and cache-strategy typo are not the same bug." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Specific Qwen chat crash is not the same as a broad v5 breakage complaint." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Missing processor.save_pretrained files is unrelated to a generic bug report." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading regression is unrelated to the undefined tf backend error." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Qwen3-VL-Moe support issue is unrelated to missing _tp_plan for tensor parallelism." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "2D bounding-box output bug is unrelated to PretrainedConfig type checking." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Lazy loading problems may be related in theme, but sys.modules access is not the same concrete bug." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Different Qwen3-VL behaviors: bbox output error vs infinite loop on a specific image." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Broken env command and checkpoint reload compatibility are different migration regressions." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "Llama4Config pad_token_id error is a specific config bug, not the broad v5 breakage report." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import * performance/refactor issue is unrelated to missing GenerationMixin cache." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 inference support and a weight-conversion-script question are unrelated." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Checkpoint reload incompatibility and remote-code meta-init incompatibility are different failure modes." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "StableLm pad_token_idx regression is specific, while 45850 is only a broad complaint." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "Cache-strategy typo is not the same as a generic v5 breakage report." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "Both are broad v5 regressions, but one is specifically import errors and the other is a generic complaint." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 settings warning is unrelated to cache strategies typos." + }, + { + "left": "issue:41669", + "right": "issue:46218", + "accept": false, + "reason": "Import * cleanup/perf issue is unrelated to the Qwen3_5ForCausalLM import failure." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "sam3 tracker/model-type warning is unrelated to Liger Kernel application." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Seed-related test flakiness is unrelated to inverse_sqrt scheduler kwargs." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention 2 support and 2D bounding-box output are distinct model issues." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Slow import timing and lazy loading failure are related broadly, but not the same concrete bug." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "env CLI regression is unrelated to scheduler kwargs handling." + } + ] + }, + "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": 20, + "review_unit_index": 54, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:43066", + "issue:43298", + "issue:43352", + "issue:43502", + "issue:43519", + "issue:43576", + "issue:43606", + "issue:43704", + "issue:43827", + "issue:43976", + "issue:44077", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44655", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44998", + "issue:45071", + "issue:45230", + "issue:45310", + "issue:45341", + "issue:45468", + "issue:45478", + "issue:45850", + "issue:46018", + "issue:46095", + "issue:46172", + "issue:46218", + "issue:46328" + ], + "soft_pairs": [ + "issue:43976|issue:44393", + "issue:44908|issue:45071", + "issue:44295|issue:45230", + "issue:44297|issue:45230", + "issue:44704|issue:44908", + "issue:41669|issue:44351", + "issue:44492|issue:46328", + "issue:41669|issue:45341", + "issue:43976|issue:45850", + "issue:44393|issue:45478", + "issue:43976|issue:46218", + "issue:44077|issue:44861", + "issue:43827|issue:44908", + "issue:31795|issue:33453", + "issue:44315|issue:44829", + "issue:43576|issue:44273", + "issue:44655|issue:44861", + "issue:44360|issue:45468", + "issue:45230|issue:45310", + "issue:44273|issue:44908", + "issue:43576|issue:46218", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:46018|issue:46172", + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:31515|issue:31795", + "issue:43704|issue:44485", + "issue:44485|issue:45468", + "issue:45850|issue:46218", + "issue:42371|issue:43704", + "issue:43502|issue:43519", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:43298|issue:46095", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:44261|issue:45468", + "issue:44351|issue:46218", + "issue:43352|issue:46218", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44492|issue:44998" + ], + "prepared_review_unit_hash": "360dc2b5de1027460ac5cc1603d312c50857f79b995ca013afb900c4d4850a06", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 43, + "serialized_chars": 39766, + "estimated_input_tokens": 9942, + "estimated_eval_tokens": 20140 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated closed issues with a few superficial overlaps around tokenizer/model-loading regressions in v5. I did not find any soft pair that safely looks like the same underlying bug/change, so I rejected all soft edges.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45310 is the most concrete, actionable loading regression in the set and the best anchor for the small v5/from_pretrained-related subgroup.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 has the clearest bug description, version scope, and likely reproducible model-loading failure, making it the best representative issue here.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Different problems: Python version compatibility vs Qwen3-VL output/vision bug." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Unrelated subsystems: LR scheduler kwarg handling vs PretrainedConfig type checking." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "One is a buffer access error, the other is a generic bug report with no clear matching failure mode." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch does not match the generic bug report enough to treat as the same bug." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor cached_file kwargs forwarding and inverse_sqrt scheduler kwargs are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star cleanup/perf issue is not the same as the HybridCache import failure." + }, + { + "left": "issue:44492", + "right": "issue:46328", + "accept": false, + "reason": "A cache-strategy typo and a padding performance bug are different issues." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Import-star usage cleanup is unrelated to a testing_utils bug." + }, + { + "left": "issue:43976", + "right": "issue:45850", + "accept": false, + "reason": "Both mention broad v5 breakage, but the concrete failures are different." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Different model families and failure modes; both are vision/model-specific bugs, not the same one." + }, + { + "left": "issue:43976", + "right": "issue:46218", + "accept": false, + "reason": "Python version support and Qwen3_5 import failure are not the same underlying bug." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "patchtsmixer post_init policy and tied-weight key handling are separate code paths." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs referencing removed pipeline APIs is unrelated to scheduler kwargs behavior." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion in model.forward is unrelated to tokenizer loading regression." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Different training/runtime behaviors: model_init Liger application vs flash_attention_3 degenerate training." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken transformers env command and lazy loading are different tools/features." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Pipeline save_pretrained support and tied-weight-key AttributeError are not the same bug." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer math and Gemma4AudioRelPositionalEncoding are unrelated." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "The first is too generic to establish that it is the same concrete Qwen3.5 MoE from_pretrained regression." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading and inverse_sqrt scheduler kwargs are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:46218", + "accept": false, + "reason": "Broken env subcommand and Qwen3_5 import failure are different failures." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Timestamp calculation in Qwen3VL Processor is not the same as GLM-5 RoPE implementation." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Processor timestamp math and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeek-V4 gating logic and a dtype/transpose expression bug are different concrete issues." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star cleanup is not the same as VRAM leakage in multi-threaded dataloading." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Model.forward documentation confusion and Llama 3.2 Vision model loading breakage are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Slow from_pretrained checkpoint loading and forward API documentation confusion are different issues." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leak in dataloader threads and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "Different model families and different numerical/positional-encoding problems." + }, + { + "left": "issue:45850", + "right": "issue:46218", + "accept": false, + "reason": "Generic v5 breakage and Qwen3_5 import failure are too broad to treat as the same bug." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 settings guidance and VRAM leakage are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "local_files_only network requests and timestamp calculation are different code paths." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp math and processor.save_pretrained missing files are different bugs." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Timestamp calculation and VRAM leakage are unrelated." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Audio-model continuous batching support and DeepSeek-V4 weight conversion are different concerns." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "GlmMoeDsaIndexer split return values and Gemma4AudioRelPositionalEncoding are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch in bark-small does not match Gemma-4 positional encoding." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "MLA q_a_layernorm epsilon precision issue is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "HybridCache import failure and Qwen3_5 import failure are different missing-symbol problems." + }, + { + "left": "issue:43352", + "right": "issue:46218", + "accept": false, + "reason": "Flash Attention 2.0 unsupported for NemotronHForCausalLM is unrelated to Qwen3_5 import errors." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Both are tokenizer v5 complaints, but one is general behavior drift and the other is a specific decoder-type mismatch." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwarg handling is unrelated to the placeholder 'Unemployment' issue." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "Cache-strategy typo is unrelated to the placeholder 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": 20, + "review_unit_index": 55, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-463", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:33453", + "issue:36296", + "issue:37934", + "issue:38617", + "issue:42936", + "issue:43240", + "issue:43298", + "issue:43525", + "issue:43572", + "issue:43618", + "issue:43668", + "issue:43717", + "issue:43813", + "issue:44031", + "issue:44230", + "issue:44623", + "issue:44861", + "issue:44938", + "issue:44998", + "issue:45290", + "issue:45310", + "issue:45850", + "issue:46018", + "issue:46144" + ], + "soft_pairs": [ + "issue:43618|issue:44998", + "issue:44998|issue:45850", + "issue:44938|issue:44998", + "issue:44861|issue:45290", + "issue:43525|issue:44861", + "issue:31515|issue:38617", + "issue:43572|issue:44861", + "issue:44230|issue:44623", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "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": "31655ebedf486722338388cb7c7f83b4d557a8c49aac22970a530f463811996a", + "component_budget": { + "node_count": 463, + "item_count": 463, + "soft_pair_count": 1704, + "serialized_chars": 883347, + "estimated_input_tokens": 220837, + "estimated_eval_tokens": 441930 + }, + "budget": { + "node_count": 26, + "item_count": 26, + "soft_pair_count": 25, + "serialized_chars": 21932, + "estimated_input_tokens": 5483, + "estimated_eval_tokens": 11222 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a noisy cluster of mostly unrelated issues that only share vague transformer-regression language. The soft similarity edges are not true duplicates; they span different models, subsystems, and failure modes. No pull requests are present.", + "confidence": 0.22, + "canonical_issue_reason": "Issue 45850 is the broadest umbrella complaint about post-5.0 breakage and is the closest thing to a cluster anchor, though it is still too generic to represent a real duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45850 is the most central/high-level issue in this set, but it should be treated as a coarse regression report rather than a duplicate target for the other concrete bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "Different topics entirely: CLIPOutput attentions vs an unrelated one-word issue." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "Unrelated issue titles; no shared bug or code path." + }, + { + "left": "issue:44938", + "right": "issue:44998", + "accept": false, + "reason": "Python 3.14 import/load failure is unrelated to the other issue." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Different code paths: tied-weight key handling vs chat template/tool-call crash." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Llama4Config pad token attribute regression is unrelated to tied-weight key crashes." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Checkpoint loading slowness and an ImportError are separate problems." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "StableLmConfig pad token attribute issue is not the same as tied-weight key handling." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 model support and missing processor save files are different issues." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLFlow reporting on failed training jobs is unrelated to from_pretrained performance." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Checkpoint load slowdown and tensor-parallel training bugs do not match." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "Both mention Qwen3.5, but one is FP8 support and the other is a from_pretrained regression; not the same fix." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Tokenizer loading regression and checkpoint loading slowdown are distinct." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Continuous batching for audio models is unrelated to FP8 Qwen support." + }, + { + "left": "issue:43717", + "right": "issue:46144", + "accept": false, + "reason": "Mamba init-weights behavior and RoFormer attention-interface usage are different model implementations." + }, + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2former ignore_value bug and DeepSeek-V4 gating issue are unrelated model-specific defects." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "Type-hint metadata and tokenizer regex warnings do not describe the same bug." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Loss kwargs propagation and Mask2former ignore_value handling are separate bugs." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss kwargs propagation is unrelated to RoFormer attention dispatch." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "No shared underlying issue; one is segmentation-loss logic, the other a tokenizer warning." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "Different model subsystems and failure modes." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Mask2former loss behavior and a quantization typo are unrelated." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Loss kwarg forwarding and a config type-hint issue are not duplicates." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "Incorrect type hint and typo in a quantization attribute are unrelated." + }, + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "RMS norm question for Llama4 is unrelated to ModernBERT type hints." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "No common bug; model loss logic vs config typing." + } + ] + }, + "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": 20, + "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": "6d70b3d2beccacc6eaf668728b81e99f027b2bfba7fb5aac82e3dc37f2e70f19", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2878, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2878, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The three issues are unrelated feature requests: truncation direction, ReduceLROnPlateau-like behavior, and Wav2Vec2 word time offsets. This cluster does not look like a true duplicate set.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:10519", + "right": "issue:11307", + "accept": false, + "reason": "Different topics and code paths: token truncation behavior vs Wav2Vec2 word boundary timestamps." + }, + { + "left": "issue:10519", + "right": "issue:11005", + "accept": false, + "reason": "One is about truncation direction, the other about scheduler functionality; no shared underlying bug or change." + }, + { + "left": "issue:11005", + "right": "issue:11307", + "accept": false, + "reason": "ReduceLROnPlateau-like scheduler behavior is unrelated to extracting time offsets in Wav2Vec2." + } + ] + }, + "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": 20, + "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": "279a3cab8c24c8e0fb2375c4fa5c28131dcf9784b77a37d573b4b89feb0f734e", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2478, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2478, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "No duplicate-worthy pair is evident here. The three issues cover different failure modes: tokenization mismatch, private-model auth/error messaging, and a generic model-loading failure. The soft similarities are superficial.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:16618 is the broadest user-facing loading problem, but the cluster is too heterogeneous to treat it as a true canonical issue.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Tokenization output mismatch with TensorFlow code is a different problem from failing to load bert-base-uncased; no shared concrete code-path bug is evident." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Bad error messaging for private-model download without login is an authentication/error-handling issue, not the same as a generic model-load failure." + } + ] + }, + "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": 20, + "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": "38e586304b66c31765c0ff4020f9734724fd0d4b67333806669638e0dcc2fa82", + "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 not duplicates: one requests Flash Attention 2 support for more architectures, the other requests GGUF support for more architectures. They share the same \u201ccommunity contribution\u201d framing but target different code paths and feature areas.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue applies here; the two issues are separate feature requests (Flash Attention 2 vs GGUF support).", + "canonical_pr_reason": null, + "best_issue_reason": "Neither is a duplicate of the other. If a representative issue is needed for the Flash Attention 2 topic, issue 26350 is the more established thread and has substantially more discussion.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different underlying requests: Flash Attention 2 architecture support vs GGUF architecture support. Similar wording and subsystem scope are not enough to treat them as the same bug/change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 5, + "component_count": 20, + "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-27088-8", + "nodes": [ + "issue:27088", + "issue:28936", + "issue:38435", + "issue:43834", + "issue:44365", + "issue:44370", + "issue:46222", + "issue:46226" + ], + "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:38435|issue:46222", + "issue:38435|issue:46226", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:43834|issue:46222", + "issue:43834|issue:46226", + "issue:44365|issue:44370", + "issue:44365|issue:46222", + "issue:44365|issue:46226", + "issue:44370|issue:46222", + "issue:44370|issue:46226", + "issue:46222|issue:46226" + ], + "prepared_review_unit_hash": "db5cb3e904618b82bdf7735a502e2b72577929c2762540d4e6e60e861d9b2272", + "component_budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13528, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13528, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mixed: the Turkish, Spanish, and Romanian issues are distinct language-specific translation requests, while the placeholder-title issues form one duplicate subgroup and the two Thai-titled issues form another exact duplicate pair.", + "confidence": 0.76, + "canonical_issue_reason": "issue:43834 is the best anchor for the largest clear duplicate subgroup: the three generic placeholder-title translation issues are near-identical, and this one is the earliest of that set. The language-specific requests are separate issues, so the full cluster is not one single duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43834 is the most useful representative of the duplicate placeholder subgroup; the only other clear duplicate is the exact Thai pair. No single issue cleanly represents the entire mixed cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Different language-specific translation request; boilerplate similarity alone is not enough." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "Different language-specific translation request; not the same underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Turkish request vs generic placeholder-title issue; too little to treat as the same issue." + }, + { + "left": "issue:27088", + "right": "issue:38435", + "accept": false, + "reason": "Different target languages and request contexts." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "Different target language; these are separate translation requests." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "Spanish request is distinct from the generic placeholder-titled issues." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "Different target language/request, despite similar template text." + }, + { + "left": "issue:28936", + "right": "issue:38435", + "accept": false, + "reason": "Spanish vs Romanian translation request; not the same issue." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Different language-specific request; similarity is only template-level." + }, + { + "left": "issue:38435", + "right": "issue:44370", + "accept": false, + "reason": "Romanian request is distinct from the placeholder-title issues." + }, + { + "left": "issue:38435", + "right": "issue:44365", + "accept": false, + "reason": "Romanian request vs generic placeholder-titled issue; not a duplicate." + }, + { + "left": "issue:38435", + "right": "issue:43834", + "accept": false, + "reason": "Different language-specific request; only the boilerplate overlaps." + }, + { + "left": "issue:38435", + "right": "issue:46222", + "accept": false, + "reason": "Romanian translation request and Thai-titled issue do not look like the same underlying request." + }, + { + "left": "issue:38435", + "right": "issue:46226", + "accept": false, + "reason": "Romanian translation request and Thai-titled issue do not look like the same underlying request." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": true, + "reason": "Same generic placeholder-title translation request with near-identical body; looks like a duplicate." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": true, + "reason": "Same generic placeholder-title translation request with near-identical body; looks like a duplicate." + }, + { + "left": "issue:43834", + "right": "issue:46222", + "accept": false, + "reason": "Different title/script and likely different request context; boilerplate similarity alone is insufficient." + }, + { + "left": "issue:43834", + "right": "issue:46226", + "accept": false, + "reason": "Different title/script and likely different request context; not enough evidence for a duplicate." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": true, + "reason": "Same generic placeholder-title translation request with near-identical body; looks like a duplicate." + }, + { + "left": "issue:44365", + "right": "issue:46222", + "accept": false, + "reason": "Different title/script and likely different request context." + }, + { + "left": "issue:44365", + "right": "issue:46226", + "accept": false, + "reason": "Different title/script and likely different request context." + }, + { + "left": "issue:44370", + "right": "issue:46222", + "accept": false, + "reason": "Different title/script and likely different request context." + }, + { + "left": "issue:44370", + "right": "issue:46226", + "accept": false, + "reason": "Different title/script and likely different request context." + }, + { + "left": "issue:46222", + "right": "issue:46226", + "accept": true, + "reason": "Exact same Thai-titled issue with identical body shape; this looks like a straightforward duplicate." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The mixed-cluster summary matches the packet: Turkish/Spanish/Romanian are clearly separate language-tagged requests, the three placeholder-title issues form a tight duplicate subgroup, and the two Thai-titled issues are an exact duplicate pair by title/body-length pattern. The soft-edge decisions stay cautious and do not overclaim beyond the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 20, + "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": "1809bfe7d2db5b243266308a25dd9e89f78249501be2f03d1324a3816789295a", + "component_budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21661, + "estimated_input_tokens": 5416, + "estimated_eval_tokens": 11088 + }, + "budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21661, + "estimated_input_tokens": 5416, + "estimated_eval_tokens": 11088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a loose collection of model-request issues with overlapping wording, but they target different models, families, and modalities. I found no true duplicate pair and no PRs in the set.", + "confidence": 0.98, + "canonical_issue_reason": "issue:42738 is the most umbrella-like representative in the group; it\u2019s generic and broad compared with the many specific model-request tickets, so it works best as a cluster anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42738 is the best issue candidate because it is the least model-specific and most likely to summarize the broader theme of the cluster, even though the items are not duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Different model requests: EuroBert vs DEIMv2; same template, not same underlying change." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert config addition is unrelated to Qwen3-TTS support." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert model config and RF-DETR model support are different models and code paths." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are separate model additions." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "Different audio/vision model requests; no shared concrete bug or change." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "RF-DETR and a weight-sparse transformer request are unrelated model implementations." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR vs Qwen3-ASR are different models and modalities." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "RF-DETR model support is unrelated to Qwen3-TTS support." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are different model families and changes." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT addition is not the same as adding the weight-sparse transformer model." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT and LFM2.5 Audio are unrelated model requests." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are distinct model additions." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are separate model requests despite similar naming." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support is unrelated to Qwen3-ASR support." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 and the weight-sparse transformer request target different code paths." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 architecture and ModernVBERT are different model implementations." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 and Voxtral TTS support are different model families." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated additions." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT models and Qwen3-ASR are different model targets." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are not the same change." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT models are more specific than the broad BERT-like RoPE request; not duplicates." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "BERT-like RoPE support and Voxtral TTS support are unrelated model requests." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "TTS and ASR requests for Qwen3 are related by family but still distinct support work." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-TTS support and Voxtral TTS support are different model-specific changes." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and implementations." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-TTS support is unrelated to the weight-sparse transformer request." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio are separate audio model requests." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio are distinct speech/audio model requests." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "Qwen3-ASR and EUPE are different model families and tasks." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR support is unrelated to the sparse transformer model request." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio and SAM3-LiteText are different model additions." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "OpenAI weight-sparse transformer and EUPE are unrelated architectures." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral TTS are distinct speech model support requests." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral-4B-TTS are unrelated model additions." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-ASR and DeepSeek-OCR2 are different modality/model requests." + }, + { + "left": "issue:45322", + "right": "issue:46133", + "accept": false, + "reason": "EUPE and TIPSv2 are separate vision-language model requests; same broad area, not the same change." + } + ] + }, + "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": 20, + "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-38740-2", + "nodes": [ + "issue:38740", + "issue:46443" + ], + "soft_pairs": [ + "issue:38740|issue:46443" + ], + "prepared_review_unit_hash": "abd9eec779977db39f047fee02065cfee0208a4393b2c31c9341c1d96525be8a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1505, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1505, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are unrelated: one is a closed docs request about adding Pruna as an optimization framework, and the other is an open technical issue about Vision RoPE refactoring. No duplicate or same-change relationship is evident.", + "confidence": 0.98, + "canonical_issue_reason": "If a representative issue is needed, 46443 is the better anchor because it is the open technical bug/feature discussion; 38740 is a separate closed documentation request.", + "canonical_pr_reason": null, + "best_issue_reason": "46443 is the best issue to keep as the cluster representative because it is the substantive open code-related issue, whereas 38740 is a closed docs-only item.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38740", + "right": "issue:46443", + "accept": false, + "reason": "These describe different topics and different artifact types of work: a docs request for Pruna versus a Vision RoPE refactor. They do not look like 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": 20, + "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": "27e0706e1d263b1917a29471235194bc690c815049d97956cbcd760a7db7bbfb", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are not duplicates: one is about block decoding behavior, the other is about adding DeepSeek v3.2 model support. The only link is weak semantic similarity, with no shared targets, files, or explicit references.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical representative for the other; they describe different features/problems and should remain separate.", + "best_pr_reason": "No PRs are present.", + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "Different underlying requests: block decoding mechanics vs model/version support. No shared code path, filenames, targets, or explicit linkage." + } + ] + }, + "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": 20, + "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": "4fe11e143217722151147f55411568f05ca25c06ea378d7b6a68547da6db9992", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1583, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1583, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are both about custom trainers, but they describe different problems: one is about `remove_unused_column` customizability in `Trainer`, the other is about FLOPS estimation handling for custom trainers. They should not be treated as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 43598 is the cleaner representative of the cluster because it has the more concrete Trainer customization focus and a clearer code-path target.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43598 is the best issue to keep as a representative example of the cluster, but not because it duplicates 43600; it is simply the more direct and actionable Trainer customization report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "Related theme, but different underlying behavior: `remove_unused_column` customization versus FLOPS estimation for custom trainers. Not the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 10, + "component_count": 20, + "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": "ec57effae4bd49b7794d11f1798968ebe291576c5cb9afe2c7acbc01ef550920", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related only at a very high level (Qwen model support), but they describe different gaps: one is about registering Qwen3-Omni in AutoModel/AutoModelForConditionalGeneration, the other is about adding AutoModelForSequenceClassification support for Qwen3.5.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #43684 is the cleaner representative because it targets a concrete registration gap in the core auto-model mappings; #44405 is a separate task for a different config and a different task head.", + "canonical_pr_reason": null, + "best_issue_reason": "#43684 is the best issue to keep as the representative item since it is broader and more directly about model registration, whereas #44405 is a distinct sequence-classification support request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different underlying requests: Qwen3-Omni auto-model registration vs Qwen3.5 sequence-classification support. They affect different model families and API paths, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 20, + "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": "e82584b54ea234ccfd57dc23605fb5c26b37d6c2fd3a528d185477da47dca653", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All four issues are extremely sparse and mostly consist of generic or identical titles with no meaningful body evidence. The cluster looks noisy rather than clearly duplicate-backed, so I would not merge any pair on the current data.", + "confidence": 0.36, + "canonical_issue_reason": "Issue 43859 is the only one with any discussion activity, so it is the least empty anchor in an otherwise low-signal cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "43859 is the best available representative because it has the most activity and shares the same generic title pattern as the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "No shared bug description or target is available; the similarity is only superficial from short titles/body snippets." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Insufficient evidence that these describe the same underlying problem; only weak title similarity is present." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "The available metadata does not show a common concrete bug, just generic title overlap." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": false, + "reason": "Identical generic titles alone are not enough to confirm a duplicate without body or target evidence." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "No concrete issue details are present to justify treating these as the same report." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "The pair lacks any substantive evidence beyond broad title similarity, so duplicate status is not supported." + } + ] + }, + "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": 20, + "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": "793e4744f25e0f8627164b3db25be0139b7253e09f8e5de68abd3ad7a5954f2e", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are not genuine duplicates of one bug/change; they look like unrelated spam or low-signal issues with only superficial title similarity. No PRs are present.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:44371 is the most generic placeholder-style item, but the cluster does not form a true duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are spam-like placeholders, but there is no evidence they describe the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Shared spammy format only; no concrete bug or change matches." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is a generic spam placeholder and the other is a documentation-suggestion title; unrelated." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Generic spam vs. a specific-looking phrase about Google LLC; no same underlying problem." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "Documentation enhancement and spam placeholder are not the same issue." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Only superficial similarity in low-information titles; no shared problem statement." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Both are spam placeholders, but there is no substantive evidence they refer to the same report." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Different titles and no technical context tying them to one defect." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "No shared code path or explicit linkage; reject as duplicate." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 20, + "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": "5ba7cfb17f4a64a167d899bf28d52005fe1800e26495556bf20b42fa6a6cd8a1", + "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": "These are duplicate issue reports with the same title and very similar content, describing the same energy-reduction proposal. Issue #45145 is the better canonical choice because it was created earlier and already has an inbound reference.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 is slightly preferable as the canonical issue because it was created earlier and already has an inbound reference, while #45151 appears to be the later duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best issue to keep as the primary tracker: same topic as #45151, earlier creation time, and more established discussion footprint.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Same title and same apparent report; this looks like a duplicate issue for the same underlying request." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The canonical-issue choice is grounded in the packet: #45145 was created earlier and has an inbound reference, while #45151 is later with no inbound references. The duplicate-issue verdict is also supported by the identical titles and soft-similarity signal. No clear overstatement beyond the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 20, + "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": 15, + "component_count": 20, + "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": "efe260d6eab4a9a8c569b6df8a6a46a68b072c426cdcc879cdb1f67c8a27ee01", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1672, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1672, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These look like two different 4-bit inference bugs: one is a Gemma4/audio-module dtype issue triggered by `torch.finfo()` on `uint8` weights, the other is a `BitsAndBytesConfig` regression where `llm_int8_skip_modules` drops the default `lm_head` exclusion and causes an assertion failure. They share the same broad area but not the same underlying failure mode.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 45674 is the better canonical issue for this cluster because it states a concrete configuration regression with a clear failure path; issue 45672 appears to be a separate, model-specific dtype error rather than the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45674 is the strongest representative because it describes a general, actionable 4-bit inference bug in `BitsAndBytesConfig`, whereas 45672 is narrower and tied to Gemma4 audio modules.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Both mention 4-bit inference, but the failures are different: `torch.finfo()` on `uint8` weights in audio modules versus a skipped-module exclusion regression causing an assertion. 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": 16, + "component_count": 20, + "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": "45076df33bf696a60dc81435a103fea2000b4f9d25e88eaf4bf7ff2927377758", + "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": "Both items look like unrelated, low-signal closed issues with only personal-name/date titles and no body or technical context. They should not be clustered as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "The titles refer to different names and different dates, with no shared bug report, feature request, or technical context. This is not 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": 17, + "component_count": 20, + "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": "e01f4268c4e9230bd9b5e3273bd1e26b15d59bf83b615ad0cf6f85f4ba42a884", + "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 not duplicates; they look like unrelated closed issues with only superficial title similarity from personal names/dates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "Different named subjects and dates, no shared body, targets, filenames, or code-related context; not 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": 18, + "component_count": 20, + "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-45810-2", + "nodes": [ + "issue:45810", + "issue:46295" + ], + "soft_pairs": [ + "issue:45810|issue:46295" + ], + "prepared_review_unit_hash": "e1de639e7eb24e8fd7a944d87d8ccf712e29514279e5f411267d45176540ebf7", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1541, + "estimated_input_tokens": 386, + "estimated_eval_tokens": 1028 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1541, + "estimated_input_tokens": 386, + "estimated_eval_tokens": 1028 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are unrelated: one requests a new Qwen3_5 token-classification model class, while the other requests FP8 training support. The soft similarity signal is not enough to treat them as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45810", + "right": "issue:46295", + "accept": false, + "reason": "Different underlying requests and code paths: model-class addition vs FP8 training support." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 19, + "component_count": 20, + "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": "434fb9c7093308f13bcb025257b5585ece98e96f9f154486699caea864deb6ef", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5728, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5728, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The set most likely splits into two near-duplicate issue pairs: 45870/45871 and 45872/45873. Issue 45869 looks separate and only loosely related by broad financial theme. The evidence supports duplicate triage, but not a claim of exact duplication.", + "confidence": 0.87, + "canonical_issue_reason": "45870 is the clearest representative for the example-request pair because 45871 matches it extremely closely in title and body length; this is strong near-duplicate evidence, though not enough to prove an exact duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "45870 is a good anchor for deduping one pair, but there is no single canonical issue for all five: 45872/45873 form a separate near-duplicate pair, and 45869 is distinct.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Different request types: a dataset bundle offer versus an example/tutorial request." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Different request types: a dataset bundle offer versus an example/tutorial request." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Very likely duplicate: identical title and matching body length/overall request wording." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Related financial/DataForge theme, but the requests are different: an example request versus a scraping-service request." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Related financial/DataForge theme, but the requests are different: an example request versus a scraping-service request." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Related theme only; the request wording points to different needs, so there is not enough evidence of the same underlying issue." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Related theme only; the request wording points to different needs, so there is not enough evidence of the same underlying issue." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Very likely duplicate: identical title and matching body length/overall request wording." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The clustering is too strong in a few places. The packet only provides soft_similarity plus matching titles/body lengths, which supports near-duplicate or very likely duplicate, but not a fully grounded claim of \"exact duplicate.\" Also, phrases like \"different code paths\" are unsupported because there is no diff/code evidence. The overall split into two duplicate pairs plus one separate item is plausible, but the soft-edge verdicts and summary should be more conservative." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 20, + "component_count": 20, + "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-46439-2", + "nodes": [ + "issue:46439", + "issue:46441" + ], + "soft_pairs": [ + "issue:46439|issue:46441" + ], + "prepared_review_unit_hash": "cba103cba046a56ae73fe5d5d0f80d4a829b633c661c46e90c36c74b64223420", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are related in topic (`cache_implementation=\"static\"` and linear-attention behavior), but the packet only provides soft similarity and no direct evidence that they are the same bug. Issue 46439 is the broader, more general report, while 46441 looks like a possibly related but distinct symptom report.", + "confidence": 0.47, + "canonical_issue_reason": "Issue 46439 is the better canonical issue only in a tentative sense: it describes the broader static-cache / linear-attention failure mode, whereas 46441 is a narrower Qwen3.5 `KeyError` symptom. The evidence here is not strong enough to prove they share the same root cause.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 46439 is the best issue candidate because it is more general and mechanism-focused, making it the most suitable representative if these reports are later confirmed related.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46439", + "right": "issue:46441", + "accept": false, + "reason": "The evidence is too weak to treat these as the same underlying bug. They share a broad static-cache / linear-attention theme, but the packet only shows title-level similarity and no direct linkage tying the shape corruption report to the Qwen3.5 `KeyError`." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The clustering is a bit too strong for the evidence in the packet. We only have title-level similarity plus a soft-similarity signal; there is no direct linkage showing issue 46441 is a symptom of the same bug as 46439. The summary and canonicalization language go beyond what\u2019s grounded here by asserting a shared underlying root cause rather than a plausible relation. A more conservative verdict would keep this as a tentative similarity, not an accepted same-bug pair." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 35, + "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:46267" + ], + "soft_pairs": [ + "pull_request:39895|pull_request:43424", + "pull_request:41855|pull_request:43823", + "pull_request:41855|pull_request:43424", + "pull_request:42668|pull_request:43823", + "pull_request:43251|pull_request:43823", + "pull_request:43251|pull_request:44827", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:43995", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:46267", + "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": "3f8a803f257d71da3b1473f409bb106d7a3f91d2ea7e8ae340b98bfd912a6bcf", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 38090, + "estimated_input_tokens": 9523, + "estimated_eval_tokens": 19302 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 38090, + "estimated_input_tokens": 9523, + "estimated_eval_tokens": 19302 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly unrelated open PRs spanning model additions, tokenizer utilities, docs, and infrastructure. The only true duplicate-looking pair is the two `fixed_cross_entropy` PRs (#43251 and #43254), which patch the same loss helper for the same issue. Everything else should stay separate.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43251 is the best canonical PR for the one real duplicate pair: it directly fixes `fixed_cross_entropy` in `loss_utils.py`, has the explicit issue target #43240, and is the more clearly established implementation of that change.", + "best_issue_reason": null, + "best_pr_reason": "PR #43251 is the strongest representative in this set because it is a concrete, narrowly scoped code-path fix with an explicit linked issue, and it matches the only clearly duplicate companion PR.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "Unrelated PRs: VideoPrism model integration vs ExecuTorch dynamic-shape export test. No shared underlying bug or change." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Different areas: Mistral Common tokenizer methods vs adding MobileLLM. Same broad tokenizer/model codebase, but not the same fix." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer API completion and ExecuTorch export testing are unrelated." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor auto-loading robustness and new MobileLLM model support are different changes." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Loss helper kwargs fix vs adding a new model package; no shared code-path problem." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "`fixed_cross_entropy` kwargs handling is unrelated to Mistral4 test/model updates." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "ExecuTorch export testing and Mistral4 fixes are unrelated." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "No same bug/change: export tests vs new model implementation." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "New MobileLLM model support is unrelated to `cls.config` resolution ordering." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "MobileLLM addition vs Falcon output-collection refactor are distinct changes." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon model refactor and model-creation skill content are unrelated." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon standardized output refactor and Mistral4 test fixes do not address the same concrete issue." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test/model work and `layrnorm` renaming across CLIP-like models are different fixes." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46267", + "accept": false, + "reason": "Mistral4 fixes and chat-response parsing updates are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 test/model work and TDT loss kernel integration are separate changes." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 test/model work and `cls.config` resolution ordering are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Touching `modeling_auto.py` is not enough to merge these: Mistral4 model support and model-creation skill are not the same change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "MobileLLM model support and the model-creation skill are unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs fix the same `fixed_cross_entropy` helper in `loss_utils.py` for the same underlying issue (#43240); the only differences are implementation style and API forwarding details." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet. The only clearly duplicate-looking pair is the two `fixed_cross_entropy` PRs (#43251 and #43254), which both target the same helper in `loss_utils.py` and share issue #43240, so treating them as mergeable into one bugfix PR is reasonable. The other pair verdicts stay conservative and do not overstate duplication." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 35, + "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-40546-3", + "nodes": [ + "pull_request:40546", + "pull_request:42781", + "pull_request:46331" + ], + "soft_pairs": [ + "pull_request:40546|pull_request:42781", + "pull_request:42781|pull_request:46331" + ], + "prepared_review_unit_hash": "b451aabbe9bd255d9a961c0edecd7e8ad2167ac6c7df83e99f4a660b052ffa3b", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10869, + "estimated_input_tokens": 2718, + "estimated_eval_tokens": 5692 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10869, + "estimated_input_tokens": 2718, + "estimated_eval_tokens": 5692 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are not duplicates: one adds core VibeVoice support, one adds a realtime VibeVoice variant, and one adds Parakeet-RNNT. They share some auto-registration and speech-model plumbing, but they do not fix the same concrete change.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 40546 is the best representative of the cluster because it is the main VibeVoice implementation and has the broadest, most foundational scope among the VibeVoice-related changes.", + "best_issue_reason": null, + "best_pr_reason": "PR 40546 is the strongest standalone candidate for merging because it introduces the core model support rather than a narrower follow-up variant or an unrelated ASR model.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Both are VibeVoice-related, but 40546 is the base VibeVoice implementation while 42781 adds a realtime variant with additional tokenizer/feature-extractor work. They are related but not the same underlying change and would not plausibly be merged as one PR." + }, + { + "left": "pull_request:42781", + "right": "pull_request:46331", + "accept": false, + "reason": "These target different model families: VibeVoice Realtime versus Parakeet-RNNT. Shared auto files are broad framework plumbing, not evidence of the same code-path fix or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 3, + "component_count": 35, + "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": "5bb6396d8249f92deaee4f521d0b6d2592a43403d72478f1bf600d6105c80a63", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are unrelated feature PRs: one adds TimesFM forecasting with covariates, the other adds DINOv3 image classification support. The soft similarity is superficial and should not be merged as a duplicate cluster.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: the PRs target different models, different code paths, and different user-facing features, so there is no single underlying change here.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a good global representative for the cluster because they are unrelated and do not describe the same bug or change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Reject: PR 40755 adds TimesFM covariates support across TimesFM docs/modeling/tests, while PR 41224 adds DINOv3 image classification support and updates DINOv3 docs/auto/modeling/tests. They touch different subsystems and fix different problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 4, + "component_count": 35, + "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": 35, + "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": "84c622cc01ce12eebfeea87300075161914cbbfe7a1a4b90da10c0b13d09da3f", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10436, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10436, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three items are separate model-addition PRs with similar boilerplate changes to docs and auto-registrations, but they introduce different models and codepaths. They are not duplicates or mergeable into one PR.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45630 is the strongest cluster representative because it has the highest review activity and the most complete-looking implementation surface among the three, but it is still a distinct model addition rather than a duplicate of the others.", + "best_issue_reason": null, + "best_pr_reason": "PR #45630 is the best single representative of the cluster for triage because it is the most active and most developed PR in the set; however, it should not be treated as the same change as the other two.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "DeepSeek OCR and Isaac are different model additions with different filenames, configs, and test suites. The overlap is just standard Transformers boilerplate for registering a new model." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Isaac and Kimi2-6 add different models and related modules. Shared edits to auto-mapping and docs are generic new-model integration work, not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 6, + "component_count": 35, + "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": "9c65b5f2b6f9e1f3e1de3b259482e1f7cd1d282100d21f340b3805949f1cbe8a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two separate GitHub Actions maintenance PRs: one broad Node 24 compatibility upgrade across many workflow actions, and one smaller upgrade focused on docker/build-push-action. They overlap in theme and some files, but not in the concrete change or code path.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 42975 is the broader representative: it updates many workflow actions and clearly states the Node 24 compatibility goal, making it a better cluster anchor than the narrower Docker action bump.", + "best_issue_reason": null, + "best_pr_reason": "PR 42975 is the best single representative because it covers the wider set of workflow action upgrades and has the clearest maintenance intent. PR 42976 is narrower and specific to Docker build action versioning.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "Do not merge: both are workflow dependency upgrades, but they change different actions for different purposes. 42975 targets checkout/cache/upload-artifact and other GitHub actions for Node 24 compatibility; 42976 upgrades docker/build-push-action in build workflows. They are related maintenance PRs, not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 35, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42978-4", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45490" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43451|pull_request:45490" + ], + "prepared_review_unit_hash": "72a72eade7e8515b66b9bbbc62b65447e36ffa5b9a48bfdd0a15c19a34017a1d", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14253, + "estimated_input_tokens": 3564, + "estimated_eval_tokens": 7384 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14253, + "estimated_input_tokens": 3564, + "estimated_eval_tokens": 7384 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are four separate model-addition pull requests (ViT NEPA, Molmo, Molmo2, CTSM). They share framework plumbing and docs patterns, but they do not fix the same code-path problem or represent mergeable duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No true duplicate PR exists here; if one must be used as the cluster representative, 43451 (Molmo2) is the most substantial and review-intensive model-addition PR, with the broadest implementation surface.", + "best_issue_reason": null, + "best_pr_reason": "43451 is the best representative only in the sense of being the most complete/active PR in the group. It is not a better duplicate than the others because all four are distinct model integrations.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different models and different integrations (ViT NEPA vs Molmo); shared auto/docs files are generic model-registration plumbing, not the same underlying change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "ViT NEPA and Molmo2 are unrelated model additions with different code paths and filenames; they are not mergeable into one fix/change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "ViT NEPA and CTSM are distinct model ports with no shared bug or concrete implementation target beyond standard registration/docs updates." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are separate model families; overlap is limited to common auto/docs/test scaffolding, so they are not duplicates." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo and CTSM address different models and different preprocessing/modeling code paths; same subsystem alone is not enough to merge them." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are unrelated model additions with only generic shared registration files, not the same concrete 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": 8, + "component_count": 35, + "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:46380" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43291|pull_request:46380", + "pull_request:43665|pull_request:44815", + "pull_request:43665|pull_request:46380", + "pull_request:44815|pull_request:46380" + ], + "prepared_review_unit_hash": "7cf21e7e07cb87eaad3b830f338746f76d76645f08439829e2243c4584949327", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 11145, + "estimated_input_tokens": 2787, + "estimated_eval_tokens": 5830 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 11145, + "estimated_input_tokens": 2787, + "estimated_eval_tokens": 5830 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "No duplicate cluster here: the four PRs touch unrelated areas (Whisper tokenization, CLIP/ViT test files, FP8 dequant/loading, and generation watermarking). All soft links should be rejected.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No true canonical PR emerges because the PRs are about different bugs and code paths; none are mergeable as one fix.", + "best_issue_reason": null, + "best_pr_reason": "PR 44815 is the strongest standalone code fix, with a concrete runtime/quantization change and supporting tests, but it is still unrelated to the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Different targets and problems: Whisper decoding/tokenizer code vs CLIP/ViT test edits; not the same underlying bug." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper tokenizer decoding fix is unrelated to FP8 dequant/model-loading changes in a different subsystem." + }, + { + "left": "pull_request:43291", + "right": "pull_request:46380", + "accept": false, + "reason": "Whisper tokenization vs generation watermarking/config validation are separate code paths and bugs." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "Test-file assertions in CLIP/ViT are not the same issue as FP8 dequant/loading logic; no shared concrete bug." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46380", + "accept": false, + "reason": "These are unrelated test edits versus generation/watermarking behavior; they do not fix the same problem." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46380", + "accept": false, + "reason": "FP8 quantization/loading fix and watermarking config validation are distinct runtime issues with different 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": 9, + "component_count": 35, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43488-2", + "nodes": [ + "pull_request:43488", + "pull_request:43532" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:43532" + ], + "prepared_review_unit_hash": "1bc29dd3ff7aed61a6bc4be4e922e4bcb0900d38b86751958dd2b66fec235e4b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4562, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4562, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two pull requests are both clearly non-production test/bot exercises, but they are not duplicates of the same underlying change: one tweaks packaging/version metadata and a Vit test, while the other rewires CI/workflow and testing utilities plus a Vit test. The only overlap is a shared test filename, which is not enough to merge them as one change.", + "confidence": 0.9, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43532 is the better canonical representative for the cluster because it has the broader and more distinctive bot/CI-oriented diff, touching multiple workflow and utility files rather than mostly packaging metadata.", + "best_issue_reason": null, + "best_pr_reason": "PR #43532 is the strongest single representative PR in this set: its title and diff better summarize the cluster\u2019s bot/demo nature and it spans the largest, most identifiable change surface.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Reject: the diffs address different changes. #43488 is about setup/version/dependency metadata plus a Vit test, while #43532 changes CI workflows, testing utilities, and notification code. Shared test-file overlap does not indicate the same 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": 10, + "component_count": 35, + "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": "9324f395da004cb57f2271acf65f92689bed1296ab86550bfb1319fdc4b48cbb", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6083, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6083, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related but distinct PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share a documentation/pipeline-registration pattern and some infrastructure, but they target different tasks and different model/test surfaces, so they should not be deduplicated.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43613 is the broader and more central change: it adds the visual segmentation pipeline with wider model/test coverage across SAM/EdgeTam/SAM2/SAM3 tracker, making it the more representative PR of this cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 43613 is the best representative because it has the larger, more complete implementation and test footprint. PR 43612 is narrower and focuses on a different segmentation task, so it is not the better cluster anchor.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Different underlying features/tasks: concept segmentation vs visual segmentation. Shared docs/registry plumbing 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": 35, + "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": "e080d2feb53b2923eb7ee59158fb69bcfb7bddd5571b047f0304244226ff6655", + "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` that both add small extension points to `Trainer`, but they target different code paths and solve different problems: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering for PEFT/custom trainers. Not a duplicate pair.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 43663 is the better representative of the cluster because it introduces a reusable Trainer hook with broader API impact; PR 43651 is a narrower fix for loss scaling behavior.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Different underlying changes: gradient-accumulation loss normalization vs. input-signature column filtering. They touch the same file, but not the same bug or mergeable fix path." + } + ] + }, + "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": 35, + "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": "b19de155753931771c328cd21432807f129908f505520351def2d9fad5a82283", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "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 server file but fix different NameError/typing problems: one addresses optional pydantic/TypeAdapter import handling, the other addresses annotation evaluation issues in serve.py more broadly. They should not be treated as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43842 is the more specific and concrete fix: it directly targets the TypeAdapter NameError when pydantic is absent, with a narrow import/annotation guard in the affected code path.", + "best_issue_reason": null, + "best_pr_reason": "PR 43842 is the best representative of the cluster because its title, issue target, and diff all align on one well-defined bug and a minimal fix. PR 43656 looks like a separate typing cleanup/runtime-annotation issue rather than the same underlying problem.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Different underlying bugs: 43656 changes type annotations/future annotations in serve.py, while 43842 specifically guards TypeAdapter for optional pydantic. Same file, but not the same concrete fix or 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": 13, + "component_count": 35, + "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": 35, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44024", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44066|pull_request:44071", + "pull_request:44018|pull_request:44068", + "pull_request:44066|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44019|pull_request:44722", + "pull_request:44086|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44116", + "pull_request:44068|pull_request:44722", + "pull_request:44018|pull_request:44116", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44085", + "pull_request:44018|pull_request:44086", + "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:44161", + "pull_request:44017|pull_request:44722", + "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:44073|pull_request:44161", + "pull_request:44059|pull_request:44071", + "pull_request:44056|pull_request:44101", + "pull_request:44018|pull_request:44059", + "pull_request:44059|pull_request:44068", + "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: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", + "pull_request:44024|pull_request:44076" + ], + "prepared_review_unit_hash": "77ec5ad138f50c1204a3243aace597dc003b3a33a62bd2afb960c9699d166bcb", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84496, + "estimated_input_tokens": 21124, + "estimated_eval_tokens": 42504 + }, + "budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 72, + "serialized_chars": 76930, + "estimated_input_tokens": 19233, + "estimated_eval_tokens": 38722 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a shared #43979 output-tracing refactor set, but it is not limited to GPT-Neo/GPT-J. The packet contains several true mergeable subclusters, including the GPT-J/CodeGen line centered on #44722 and the FocalNet/ImageGPT pair (44024/44076), plus a few other issue-linked output-tracing pairs; many remaining edges are only theme-similar and should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "#43979 is the common umbrella tracker repeatedly linked by these refactor PRs and is the clearest issue-level anchor for the cluster.", + "canonical_pr_reason": "PR #44722 is the strongest canonical PR candidate because it is issue-linked and directly overlaps with both GPT-J/CodeGen refactor PRs on the same concrete code path.", + "best_issue_reason": "No separate issue artifact is needed here; #43979 is the best global issue anchor for the whole refactor cluster.", + "best_pr_reason": "PR #44722 is the best overall representative because it sits at the center of the GPT-J/CodeGen subcluster and has the cleanest direct duplicate links.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different model families; the shared issue is only a broad refactor umbrella." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Different backbones and different concrete forward paths; same tracking issue is not enough." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": false, + "reason": "Same broader GPT-J area, but 44066 also rewrites CodeGen and the PRs are not the same concrete change." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "Different vision model code paths; they share only the standardized output-tracing pattern." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "Unrelated model families and code paths; only the refactor theme overlaps." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen vs EfficientNet are different concrete changes." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "Different model implementations; both are output-tracing refactors, but not the same bug/change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen and MGP-STR are unrelated code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "Different model stacks; same decorator migration is too broad to merge as one PR." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": false, + "reason": "Same GPT-Neo family, but the diffs are not the same concrete rewrite and do not look mergeable as one PR." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen and GPT-Neo are distinct model code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are different concrete changes." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet and GPT-J/CodeGen are unrelated implementations." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen do not share the same code path or bug." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet and GPT-J are unrelated model implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are different models with different output-handling code." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are separate implementations with no shared code path." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are unrelated model implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are different model code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are different model files and forward paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated concrete changes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "Different model code paths; same output-tracing style is not enough for duplication." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44086", + "accept": false, + "reason": "MPT and MGP-STR are separate implementations with separate output handling." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 and MGP-STR do not share the same code path or bug." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J and MGP-STR are unrelated; only the output-tracing initiative overlaps." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert are different model code paths." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT2 and GPT-J/CodeGen are distinct model implementations and cannot be merged as one PR." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different concrete changes despite similar decorator migration." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "SegFormer and LongT5 are unrelated model code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "SegFormer and GPT-J/CodeGen are not the same underlying bug/change." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "SegFormer and VisualBert are separate implementations; same refactor pattern only." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and SegFormer are different backbone code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are unrelated model implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are different model families with different output-handling code." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet and LongT5 are unrelated model implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44161", + "accept": false, + "reason": "GPT-Neo and LongT5 are different concrete changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are separate model implementations with no shared code path." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert and VisualBert are different model code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are unrelated models." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are different implementations and not a duplicate fix." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44129", + "accept": false, + "reason": "RegNet and SpeechT5 are unrelated concrete changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBert and GPT-Neo are different model implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "FNet/CVT and LongT5 are unrelated code paths despite the shared output-tracing initiative." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "FNet/CVT and SegFormer are different model implementations." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert and DeBERTa V2 are distinct code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa V2 and TextNet are unrelated model implementations." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "Both target the same GPT-J/CodeGen output-tracing rewrite, with overlapping files and the same concrete change." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Both modify the GPT-J output-tracing path in modeling_gptj.py, so they look like the same underlying change." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": true, + "reason": "Both are issue-linked standardized output-tracing refactors in the same umbrella initiative, and the packet treats them as mergeable." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": true, + "reason": "Both are standardized output-tracing refactors under the same issue, with the same concrete refactor pattern across the model code." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": true, + "reason": "Both are issue-linked output-tracing refactors that apply the same standardized capture pattern." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": true, + "reason": "Both are issue-linked output-tracing refactors in the same initiative and are close enough to be merged as one change." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44154", + "accept": true, + "reason": "Both implement the same standardized output-collection refactor pattern on compatible model code paths." + }, + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": true, + "reason": "Both are the same standardized output-tracing refactor applied to image-model encoder code, so they are plausibly mergeable into one PR." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is not fully grounded in the packet: it says only the GPT-Neo/GPT-J pairs look like true duplicates, but the packet also marks pull_request:44024 and pull_request:44076 as a true mergeable pair. That makes the summary too strong. The accepted soft-edge verdicts that are present are mostly fine, but the narrative overstates how exclusive the GPT-Neo/GPT-J duplication is." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 35, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "8aa08d8eee1c022c7e562169f98083749ab13fd0401ecbff32566135143804f3", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84496, + "estimated_input_tokens": 21124, + "estimated_eval_tokens": 42504 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 5, + "serialized_chars": 13066, + "estimated_input_tokens": 3267, + "estimated_eval_tokens": 6790 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a collection of similar-looking but model-specific PRs that all refactor output tracing/capture in different architectures. They are not the same underlying change, so the soft duplicate links should be rejected.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44129 is the most concrete anchor because it has an explicit issue target and a substantial SpeechT5-specific output-tracing refactor, but it is still not a duplicate of the others.", + "best_issue_reason": null, + "best_pr_reason": "PR 44129 is the best representative of the cluster because it is the most fully specified and ties to an explicit issue, while the others are smaller per-model refactors.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "Both are output-tracing refactors, but they touch different model code paths (VisualBert vs TextNet) and are independent per-model migrations, not one mergeable fix." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "MPNet and VisualBert are separate model implementations; these are parallel refactors of the same framework pattern, not the same underlying bug or change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "Different models and different code paths; similarity is only in adopting the same tracing utility, which is too broad to count as a duplicate." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated model implementations, and the PRs refactor output handling in different forward paths rather than fixing one shared defect." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "These affect different architectures and different output-tracing code; they are not the same concrete 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": 15, + "component_count": 35, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43998-13", + "nodes": [ + "pull_request:43998", + "pull_request:43999", + "pull_request:44000", + "pull_request:44001", + "pull_request:44002", + "pull_request:44003", + "pull_request:44004", + "pull_request:44025", + "pull_request:44026", + "pull_request:44027", + "pull_request:44028", + "pull_request:44029", + "pull_request:44030" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44028", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030" + ], + "prepared_review_unit_hash": "a1e406186f2e2e2f0ba25661a1689e74e7b6b34c1f2768d08c618f5ac3de4927", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40174, + "estimated_input_tokens": 10044, + "estimated_eval_tokens": 20344 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40174, + "estimated_input_tokens": 10044, + "estimated_eval_tokens": 20344 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs share a generic \"refactor output tracing\" theme, but they touch different model implementations and different forward/output code paths, so they are not duplicates of one another.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: each item is a separate model-specific refactor, not the same underlying code change.", + "best_issue_reason": null, + "best_pr_reason": "No PR is globally best for duplicate triage here; the cluster is a set of parallel refactors across different architectures, not one mergeable change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families and forward paths; same refactor theme only." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different architectures and files; not the same underlying change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Different model code paths; only the wording of the refactor is similar." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model families (timm_backbone vs mamba/falcon_mamba); separate fixes." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model implementations and output plumbing; not mergeable as one PR." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "Mobilenet_v1 and UnivNet are unrelated model paths; same refactor pattern only." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model backbones and forward signatures; not the same change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "MobileNetV1 and Depth Anything are separate model changes, not duplicates." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Different architectures and files; only a shared output-tracing refactor theme." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "Different model code paths; not the same bug or feature change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Vision-text dual encoder and UnivNet are unrelated implementations." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models and output paths; same naming pattern only." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model families and code paths; not one underlying change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Vision-text dual encoder and Depth Anything are separate model refactors." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "UnivNet and UperNet are unrelated model implementations." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and forward methods; not the same fix." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families; no shared concrete code-path problem." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "UperNet and RWKV are unrelated code paths; only the refactor theme matches." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model families and distinct output handling." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "Mamba/FalconMamba refactor vs SuperPoint refactor; separate changes." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Different model code paths (mamba vs codegen); not the same underlying issue." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Timm backbone and CodeGen are unrelated implementations." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "UnivNet and CodeGen fix different forward/output paths." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "UperNet and CodeGen are separate model refactors." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Vision-text dual encoder and CodeGen are different architectures." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Depth Anything and VisionEncoderDecoder are different model families." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Depth Anything and SpeechEncoderDecoder are unrelated code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "Depth Anything and SuperPoint are separate model-specific refactors." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Depth Anything and RWKV are unrelated implementations." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "Depth Anything and DPR touch different model code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "VisionEncoderDecoder and RWKV are unrelated model changes." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "VisionEncoderDecoder and SpeechEncoderDecoder are different code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "Different model families; not the same concrete bug/change." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "VisionEncoderDecoder and DPR are separate refactors." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "SpeechEncoderDecoder and SuperPoint are unrelated model implementations." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families and output paths." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "SpeechEncoderDecoder and DPR are separate changes." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "SuperPoint and RWKV are different code paths; same refactor style only." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "SuperPoint and DPR are unrelated model changes." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "RWKV and DPR are different model families; not mergeable as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 16, + "component_count": 35, + "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-44662-2", + "nodes": [ + "pull_request:44662", + "pull_request:45401" + ], + "soft_pairs": [ + "pull_request:44662|pull_request:45401" + ], + "prepared_review_unit_hash": "fd7a8f449d550442a2145d659c12fba948a24af60377cc13b37f293835146b67", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6550, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6550, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are unrelated model-addition PRs: one adds PenguinVL, the other adds Voxtral TTS. They only overlap on generic transformer integration files and docs infrastructure, not on the same underlying code path or fix.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44662 is the more complete and mature standalone addition: it is not draft, has much broader implementation coverage, and more review activity. PR 45401 is a separate draft PR scoped to a different model family and explicit issue.", + "best_issue_reason": null, + "best_pr_reason": "PR 44662 is the stronger representative PR in this cluster because it is a fuller implementation with more discussion; PR 45401 is a distinct Voxtral TTS support change and should not be merged with it.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different models and different functionality (vision-language vs text-to-speech). Shared auto-registration/docs changes are generic and do not indicate the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 35, + "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-44979-4", + "nodes": [ + "pull_request:44979", + "pull_request:45363", + "pull_request:45666", + "pull_request:46339" + ], + "soft_pairs": [ + "pull_request:44979|pull_request:45363", + "pull_request:45363|pull_request:46339", + "pull_request:45666|pull_request:46339" + ], + "prepared_review_unit_hash": "479dd366bd16587335484f727771d6c7a04d898225b71f8c26ec3102aa9f32e9", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 3, + "serialized_chars": 9633, + "estimated_input_tokens": 2409, + "estimated_eval_tokens": 5074 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 3, + "serialized_chars": 9633, + "estimated_input_tokens": 2409, + "estimated_eval_tokens": 5074 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is centered on the KernelConfig / n-to-1 kernel fusion feature line. PR 46339 looks like the strongest representative because it is the latest and most comprehensive iteration of that implementation, while PR 44979 is a separate foundational API PR and not a duplicate of the kernel-fusion PRs.", + "confidence": 0.79, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46339 is the best canonical PR because it is the most complete and mature iteration of the KernelConfig-based n-to-1 fusion work, with broader implementation and tests than the earlier drafts.", + "best_issue_reason": null, + "best_pr_reason": "PR 46339 best represents the cluster\u2019s concrete feature set: it continues the same KernelConfig fusion path as the earlier PRs and adds the broadest, most refined implementation changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "44979 is a new Module Fusion API; 45363 is a KernelConfig-based n-to-1 fusion PR. They are related, but not the same concrete change and would not be merged as one duplicate PR." + }, + { + "left": "pull_request:45363", + "right": "pull_request:46339", + "accept": true, + "reason": "Both target the same KernelConfig-based n-to-1 kernel fusion code path, touch overlapping kernel-loading files, and 46339 reads as a later refactor/extension of the same feature rather than a separate change." + }, + { + "left": "pull_request:45666", + "right": "pull_request:46339", + "accept": true, + "reason": "These are clearly the same feature family: both implement extended n-to-1 fusion via KernelConfig with substantial overlap in the same loading/integration files, so they could plausibly be unified into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The rejection of 44979 vs 45363 is well supported because they are different concrete changes (Module Fusion API vs KernelConfig-based n-to-1 fusion). The accepted soft pairs are also plausible: 45363 and 46339, and 45666 and 46339, share the same KernelConfig/n-to-1 fusion line and overlapping integration files, so treating them as mergeable iterations of one feature is reasonable. The canonical PR choice of 46339 is also supported by its later timestamp and broader shared implementation surface, though the wording could stay a bit more tentative." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 35, + "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": "eb56340009b2f194d26ffa5049b6e5a90dafbf30c504ab64c815c1b118a39b30", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests address the same AttributeError in `_patch_mistral_regex` by switching from `backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer`; `45317` is the more complete version because it adds tests and more explicit coverage.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "`pull_request:45317` is the best canonical PR because it fixes the same concrete bug as `45086` while also adding test coverage and clearer validation of `fix_mistral_regex=True`.", + "best_issue_reason": null, + "best_pr_reason": "`pull_request:45317` is the strongest PR overall: same code-path fix, broader validation, and extra tests make it the most complete and merge-ready representation of the change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying bug in `_patch_mistral_regex` on the same file and same fix; `45317` looks like a superset of `45086` with tests, so these are duplicate PRs." + } + ] + }, + "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": 35, + "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": "bd55a0034dc37f41aae1197f72296ff5d85884c87ea14635e10c7ca4788cbe69", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related NemotronH refactors, but they target different changes: one is a conversion-mapping/renaming update, the other is a broader model split and documentation update for NemotronH dense/sparse variants. They should not be treated as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45569 is the more complete and self-contained NemotronH change, covering the new dense/sparse model structure, auto mappings, docs, and tests. PR #45115 is a narrower preprocessing/mapping refactor that fits inside the larger split.", + "best_issue_reason": null, + "best_pr_reason": "PR #45569 is the best representative because it implements the broader, user-visible architectural change and includes the downstream plumbing needed for the new model variants.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Different underlying changes: #45115 only adds conversion-key renaming for NemotronH/NemotronH Dense checkpoints, while #45569 restructures the model into nemotron_h_dense and nemotron_h_sparse packages with docs and tests. They are related, but not the same fix and not plausibly 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": 20, + "component_count": 35, + "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": "31a8e902364fb933fc3b64a60e2643b6dd1c3fb0ac95d17b232c8f86973250b8", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5465, + "estimated_input_tokens": 1367, + "estimated_eval_tokens": 2990 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5465, + "estimated_input_tokens": 1367, + "estimated_eval_tokens": 2990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two related but distinct PRs: one is a broad heterogeneous-model support umbrella, the other is a narrower per-layer configuration PR. They share infrastructure files, but they are not the same change.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:45332 is the broader umbrella change: it includes heterogeneous modeling, cache, masking, and tests, while 45333 is a narrower config-only foundation that appears to feed into it.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:45332 is the best representative of the cluster because it captures the full heterogeneous-model feature set rather than just the configuration plumbing.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Reject: 45333 is a subset/foundation for per-layer config, while 45332 adds broader modeling/cache/masking support. They are related, but not the same concrete change and would not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 21, + "component_count": 35, + "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": "9cf887b3a0d4595f7f6f90529397857b274d1d3ca2b8393951495d2081420db1", + "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": "Both PRs are typing-related, but they are different changes: 45415 is a broad repo-wide type-checking sweep, while 45425 adds specific typing protocols/helpers around modeling_utils and related utilities. The shared files appear incidental to the overlap in typing work, not evidence of the same fix.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45415 is the broader umbrella change: it enables type checking across many transformers modules and adjusts pyproject type-ignore handling, so it best represents the cluster theme.", + "best_issue_reason": null, + "best_pr_reason": "PR 45425 is the more self-contained and concrete typing change, focused on specific protocols and helper types rather than a large sweeping type-checking rollout.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Reject: these are not the same underlying change. 45415 is a broad typing pass across many files, while 45425 adds targeted typing protocols/helpers and related model utilities; overlapping filenames do not make them 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": 22, + "component_count": 35, + "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-45513-2", + "nodes": [ + "pull_request:45513", + "pull_request:46084" + ], + "soft_pairs": [ + "pull_request:45513|pull_request:46084" + ], + "prepared_review_unit_hash": "2e855ee2d6c0f72a7ae8818b2d214c40111134ba6c6d7c22bf9727ac53b117f1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4858, + "estimated_input_tokens": 1215, + "estimated_eval_tokens": 2686 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4858, + "estimated_input_tokens": 1215, + "estimated_eval_tokens": 2686 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate cache-related fixes in different model families: one for GDN linear attention multi-token cached forward in Qwen/Olmo hybrid models, and one for Mamba2 use_cache with seq_len > 1. They are thematically similar but not the same concrete bug or change.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 46084 is the clearest standalone fix: it has an explicit issue reference, a narrowly defined Mamba2 cache bug, and a small, focused patch. PR 45513 is broader and spans multiple model implementations, but it is still a different bug.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45513", + "right": "pull_request:46084", + "accept": false, + "reason": "Both concern cached forward paths with seq_len > 1, but they target different architectures and code paths (Qwen/Olmo hybrid GDN linear attention vs Mamba2). Similar theme is not enough to treat them as duplicates or one merged 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": 23, + "component_count": 35, + "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": "101fb8a5e18d3d2c8dcd0c2655c3785da13d9bd0c7f4df0515d23906f91ab715", + "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 two related but distinct EP-loading fixes. They share a file and the broad EP theme, but one targets serialization/loading under accelerate launch and the other targets FSDP2/rank-0 broadcast overwriting expert weights. They should not be merged as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45662 is the more concrete fix: it addresses a specific corruption path where experts are overwritten by rank-0 broadcast during EP + FSDP2 loading, with a direct code-path change in core loading and MoE handling.", + "best_issue_reason": null, + "best_pr_reason": "PR #45662 is the best representative of the cluster because it describes a precise failure mode and the diff aligns with that bug; #45548 is a separate EP/ZeRO-3 configuration/loading issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Different underlying bugs and code paths: #45548 is about EP + DeepSpeed ZeRO-3 loading via accelerate launch and config serialization, while #45662 fixes EP + FSDP2 experts being overwritten by rank-0 broadcast." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 35, + "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": "7efb89899e0c59ebfcaf8bf024a6fbd9d3ca8042d0c8ebc7cdf2d3ad15cfb0ba", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same xdist-related captured_info artifact collision problem under issue #45561; PR #45645 is the broader, more complete fix and appears to subsume #45639.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45645 is the best canonical PR because it covers the full artifact-collision path: it makes the writer xdist-safe and also updates CI/log collection to handle multiple captured_info files, whereas #45639 only fixes the writer side.", + "best_issue_reason": null, + "best_pr_reason": "PR #45645 is the strongest candidate overall since it addresses both the root cause and the downstream log-consumption changes needed for the same xdist artifact issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Same underlying captured_info xdist collision bug. #45639 fixes the output-file write path; #45645 fixes that path too and also updates the workflow to consume multiple captured_info files, so they are plausibly part of one merged change set." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 25, + "component_count": 35, + "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-45670-3", + "nodes": [ + "pull_request:45670", + "pull_request:46320", + "pull_request:46401" + ], + "soft_pairs": [ + "pull_request:45670|pull_request:46320", + "pull_request:45670|pull_request:46401", + "pull_request:46320|pull_request:46401" + ], + "prepared_review_unit_hash": "29017647b96bf3dcbbd7316548066a6b68240a5d192e96109eefdd174d244f60", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7580, + "estimated_input_tokens": 1895, + "estimated_eval_tokens": 4046 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7580, + "estimated_input_tokens": 1895, + "estimated_eval_tokens": 4046 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are not duplicates: two are separate attn_implementation dispatch additions for different model families, and the third is an unrelated AutoModel registry tweak. They share a broad theme but not the same underlying code-path change.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46320 is the best canonical representative of the cluster theme because it is a focused implementation of attn_implementation dispatch for a specific model family, with model code, conversion mapping, and tests touched.", + "best_issue_reason": null, + "best_pr_reason": "PR #46320 is the strongest standalone PR here: it directly changes the model code path for dispatch support and includes validation. PR #46401 is a separate model-family implementation, and PR #45670 is unrelated to attn dispatch.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45670", + "right": "pull_request:46320", + "accept": false, + "reason": "Different changes: #45670 only adjusts AutoModel registry entries for glmasr/dia, while #46320 implements attn_implementation dispatch for RecurrentGemma." + }, + { + "left": "pull_request:45670", + "right": "pull_request:46401", + "accept": false, + "reason": "No shared underlying fix: #45670 is an AutoModel mapping nit, whereas #46401 adds attn_implementation dispatch support across BLIP-2/InstructBLIP variants." + }, + { + "left": "pull_request:46320", + "right": "pull_request:46401", + "accept": false, + "reason": "Both are about attn_implementation support, but they target different concrete model families and code paths (RecurrentGemma vs BLIP/InstructBLIP), so they are not 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": 26, + "component_count": 35, + "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-45720-2", + "nodes": [ + "pull_request:45720", + "pull_request:46259" + ], + "soft_pairs": [ + "pull_request:45720|pull_request:46259" + ], + "prepared_review_unit_hash": "e80f00b9f3889e25c36b3834764f58a37bd2c1c8cf3ca1e23ec367a0c0946dd4", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7189, + "estimated_input_tokens": 1798, + "estimated_eval_tokens": 3852 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7189, + "estimated_input_tokens": 1798, + "estimated_eval_tokens": 3852 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated linter/chore PRs: one enforces TRF003 around model forward return-tuple behavior, the other adds TRF018 about image/video processor boolean flags. They overlap only in broad modeling-related file coverage and should not be merged as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45720 is the better representative of the cluster because it makes a concrete behavioral rule change across model forward methods, whereas #46259 is a separate processor-rule/docs addition.", + "best_issue_reason": null, + "best_pr_reason": "PR #45720 is the stronger standalone PR for triage: it targets a specific code-path change in multiple model implementations and has a clearer execution-focused scope than the modeling-rule documentation/update in #46259.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45720", + "right": "pull_request:46259", + "accept": false, + "reason": "Reject: these implement different rules for different subsystems. #45720 changes model forward methods to support can_return_tuple/return_dict behavior; #46259 adds a rule about processor do_* flags being consumed during preprocess. Shared model filenames and linter/chore framing are too broad to indicate the same underlying 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": 27, + "component_count": 35, + "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": 28, + "component_count": 35, + "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-45932-2", + "nodes": [ + "pull_request:45932", + "pull_request:46100" + ], + "soft_pairs": [ + "pull_request:45932|pull_request:46100" + ], + "prepared_review_unit_hash": "1fdeba9d55e2f5e080e81b8de00f7ee88b6d220c649960d3fe1166a7ededc312", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4301, + "estimated_input_tokens": 1076, + "estimated_eval_tokens": 2408 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4301, + "estimated_input_tokens": 1076, + "estimated_eval_tokens": 2408 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two unrelated prank/diagnostic PRs: 45932 is a multi-file CI/model/config change, while 46100 is a ViT test-only PR that injects failing assertions. They do not fix the same underlying issue.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45932 is the more substantive representative: it changes workflow and several model/config files. PR 46100 is a trivial test-only PR that intentionally breaks tests, so it is not a canonical change.", + "best_issue_reason": null, + "best_pr_reason": "PR 45932 is the closest thing to a real code change and spans the actual affected files; 46100 is clearly a test sabotage PR and not suitable as a representative artifact.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "Reject: 45932 modifies CI secrets and model/config code; 46100 only adds failing assertions in ViT tests. They are not the same bug or a plausible single merged 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": 29, + "component_count": 35, + "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-46073-3", + "nodes": [ + "pull_request:46073", + "pull_request:46154", + "pull_request:46196" + ], + "soft_pairs": [ + "pull_request:46073|pull_request:46154", + "pull_request:46073|pull_request:46196", + "pull_request:46154|pull_request:46196" + ], + "prepared_review_unit_hash": "ad2a39bc61e8c8c865807cb26a2af1bf40960908cbd3ea7740a610c6f2ea2823", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 8082, + "estimated_input_tokens": 2021, + "estimated_eval_tokens": 4298 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 8082, + "estimated_input_tokens": 2021, + "estimated_eval_tokens": 4298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three pull requests are related by the same feature theme (SDPA dispatch), but they target different model implementations and code paths, so they should not be treated as duplicates of one another.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46196 is the broadest representative of the cluster, covering multiple model families and the largest change set around SDPA dispatch support.", + "best_issue_reason": null, + "best_pr_reason": "PR 46196 is the strongest single representative because it spans the widest scope and is most central to the shared SDPA-dispatch theme.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46073", + "right": "pull_request:46154", + "accept": false, + "reason": "Same high-level feature area, but CTRL and T5 are separate model implementations with distinct attention code paths; these are not the same underlying fix and would not plausibly merge into one PR." + }, + { + "left": "pull_request:46073", + "right": "pull_request:46196", + "accept": false, + "reason": "They both add SDPA support, but one is for CTRL while the other covers wav2vec2_conformer/bert and SeamlessM4T variants; different concrete model paths, so not duplicates." + }, + { + "left": "pull_request:46154", + "right": "pull_request:46196", + "accept": false, + "reason": "T5 SDPA dispatch is a separate model-specific change from the audio/seq2seq model set in 46196; same theme, but not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 30, + "component_count": 35, + "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-46126-2", + "nodes": [ + "pull_request:46126", + "pull_request:46269" + ], + "soft_pairs": [ + "pull_request:46126|pull_request:46269" + ], + "prepared_review_unit_hash": "a96cd855ac275d1569de9217eb64667ad57ecb835eaa6fac44d2670e8ba76818", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5571, + "estimated_input_tokens": 1393, + "estimated_eval_tokens": 3042 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5571, + "estimated_input_tokens": 1393, + "estimated_eval_tokens": 3042 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are not duplicates: one PR is a focused EP/trainer integration change, while the other is a broad distributed branch-base umbrella touching many unrelated distributed areas. The overlap in shared files is too coarse to indicate the same concrete fix.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46126 is the more focused, concrete change centered on EP + Trainer integration; PR 46269 is a broad branch-base PR with many distributed subsystems and is not a clean duplicate anchor.", + "best_issue_reason": null, + "best_pr_reason": "PR 46126 is the best representative of the shared overlap because it targets a specific code-path change, whereas PR 46269 is too broad and mixed-scope to serve as a precise canonical PR.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46126", + "right": "pull_request:46269", + "accept": false, + "reason": "Reject: they do not appear to fix the same underlying bug/change. The shared files are in broad distributed infrastructure, but 46126 is a targeted EP/trainer integration PR and 46269 is a large distributed branch-base PR spanning many unrelated areas." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 31, + "component_count": 35, + "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-46138-6", + "nodes": [ + "pull_request:46138", + "pull_request:46432", + "pull_request:46501", + "pull_request:46502", + "pull_request:46503", + "pull_request:46504" + ], + "soft_pairs": [ + "pull_request:46138|pull_request:46504", + "pull_request:46138|pull_request:46432", + "pull_request:46138|pull_request:46503", + "pull_request:46138|pull_request:46501", + "pull_request:46138|pull_request:46502", + "pull_request:46432|pull_request:46503", + "pull_request:46432|pull_request:46501", + "pull_request:46432|pull_request:46502", + "pull_request:46432|pull_request:46504", + "pull_request:46501|pull_request:46503", + "pull_request:46501|pull_request:46502", + "pull_request:46501|pull_request:46504", + "pull_request:46502|pull_request:46503", + "pull_request:46502|pull_request:46504", + "pull_request:46503|pull_request:46504" + ], + "prepared_review_unit_hash": "52d168666bc44aae1c1509a804c30f7339330bafde167a88face08e9516a575b", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 15, + "serialized_chars": 14228, + "estimated_input_tokens": 3557, + "estimated_eval_tokens": 7370 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 15, + "serialized_chars": 14228, + "estimated_input_tokens": 3557, + "estimated_eval_tokens": 7370 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are all unrelated workflow-maintenance changes; the similarity is mostly the generic 'chore/update workflow' pattern, not the same underlying fix or feature.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46504 is the cleanest single-purpose change in the set: it fixes shell-safe commit-message handling in build-ci-docker-images and is narrowly scoped.", + "best_issue_reason": null, + "best_pr_reason": "PR 46504 is the best representative of the cluster because it addresses one concrete CI bug with a self-contained, plausibly mergeable change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46138", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflows and different bugs: stdin-safe message passing in self-comment CI vs commit-message parsing in build-ci-docker-images." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46432", + "accept": false, + "reason": "These touch separate CI workflows with unrelated goals; one changes message piping, the other hardens SSH runner handling and env cleanup." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46503", + "accept": false, + "reason": "No shared concrete code-path: self-comment CI message handling is unrelated to benchmark_v2 triggering logic." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46501", + "accept": false, + "reason": "Different workflow files and different concerns; these are both chore edits but not the same change or bug." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46502", + "accept": false, + "reason": "Self-comment CI command input handling is unrelated to assign-reviewers checkout behavior." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46503", + "accept": false, + "reason": "SSH runner security/environment changes and benchmark dispatch logic are separate workflow paths with no single fix to combine." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46501", + "accept": false, + "reason": "These modify different workflows for different reasons; one updates SSH runner hardening, the other changes anti-slop permissions." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46502", + "accept": false, + "reason": "No same underlying bug: SSH runner secrets handling does not match assign-reviewers base-ref checkout." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46504", + "accept": false, + "reason": "Different CI workflows and distinct shell concerns; not the same code-path problem." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46503", + "accept": false, + "reason": "Anti-slop permission scoping and benchmark workflow dispatch are unrelated workflow changes." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46502", + "accept": false, + "reason": "Both are workflow maintenance, but they fix different jobs in different files with no shared defect." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46504", + "accept": false, + "reason": "Changing anti-slop job permissions is unrelated to build-ci-docker-images command parsing." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46503", + "accept": false, + "reason": "Assign-reviewers checkout ref behavior is unrelated to benchmark_v2 event gating." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflow files and unrelated fixes; not plausibly one merged PR." + }, + { + "left": "pull_request:46503", + "right": "pull_request:46504", + "accept": false, + "reason": "Benchmark_v2 dispatch support and build-ci-docker-images shell quoting address different workflows and bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 32, + "component_count": 35, + "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-46152-2", + "nodes": [ + "pull_request:46152", + "pull_request:46207" + ], + "soft_pairs": [ + "pull_request:46152|pull_request:46207" + ], + "prepared_review_unit_hash": "c330232a3454f187dd643e40e2bda06d9955ee2d805c7e68e45376915da0996a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4275, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4275, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs share a high-level theme of avoiding direct file-open source inspection, but they target different code paths and bugs: one in import_utils module-source resolution, the other in modeling_utils attention-implementation detection.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46207 is the better representative of the cluster because it has the clearer, self-contained code-path fix and accompanying test coverage, while still matching the shared source-inspection theme.", + "best_issue_reason": null, + "best_pr_reason": "PR #46207 is the strongest standalone PR for this cluster: it fixes a concrete heuristic in _can_set_attn_implementation and includes tests; #46152 is a separate utility change in a different module.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46152", + "right": "pull_request:46207", + "accept": false, + "reason": "Reject: they are not the same underlying change. #46152 rewrites import_utils path handling for self-source resolution, while #46207 changes modeling_utils's attention-implementation support check. Similar pattern, different bug and code path." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 33, + "component_count": 35, + "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-46173-3", + "nodes": [ + "pull_request:46173", + "pull_request:46436", + "pull_request:46474" + ], + "soft_pairs": [ + "pull_request:46436|pull_request:46474", + "pull_request:46173|pull_request:46474" + ], + "prepared_review_unit_hash": "e27c88da5177d2c02deea31816182e226ff098990ec23b74aff7aa80d73da519", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5921, + "estimated_input_tokens": 1481, + "estimated_eval_tokens": 3218 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5921, + "estimated_input_tokens": 1481, + "estimated_eval_tokens": 3218 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs are near-exact duplicates fixing the same StaticLayer `get_seq_length` type annotation in `cache_utils.py`; the third PR is unrelated and changes static cache length handling in generation.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46173 is the best canonical representative: it has the same patch as #46474, includes the explicit issue link (#45987), and has more discussion/context.", + "best_issue_reason": null, + "best_pr_reason": "PR #46173 is the strongest overall PR in the cluster because it directly matches the underlying change, is clearly scoped to the bug, and is better documented than the later duplicate.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46436", + "right": "pull_request:46474", + "accept": false, + "reason": "Different changes: #46436 adjusts static cache max length selection in generation, while #46474 only updates the `StaticLayer.get_seq_length` type hint. Same subsystem, but not the same underlying fix." + }, + { + "left": "pull_request:46173", + "right": "pull_request:46474", + "accept": true, + "reason": "Same code change in `src/transformers/cache_utils.py`, same issue target (#45987), and the diffs are effectively identical. These are duplicate PRs for the same fix." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The verdicts are grounded in the packet. #46173 and #46474 have identical diffs in src/transformers/cache_utils.py and share issue target #45987, so treating them as duplicate PRs is justified. #46436 changes a different file and a different fix in generation/utils, so rejecting it as the same underlying fix is conservative. The summary is also consistent with the evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 34, + "component_count": 35, + "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-46293-2", + "nodes": [ + "pull_request:46293", + "pull_request:46478" + ], + "soft_pairs": [ + "pull_request:46293|pull_request:46478" + ], + "prepared_review_unit_hash": "7bd39ae48d659fdee338d62b65f52e0c5fa39368b02b874770cb97e7816d0b13", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4443, + "estimated_input_tokens": 1111, + "estimated_eval_tokens": 2478 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4443, + "estimated_input_tokens": 1111, + "estimated_eval_tokens": 2478 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs target the same hub-kernels area and share issue #46291, but they implement different fixes: one updates kernel repo/version metadata for kernels>=0.15.0, while the other pins LayerRepository revisions to main and adds broader config/test changes. They are related, but not the same duplicate change.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46478 is the better canonical PR because it is broader and more complete: it updates multiple hub-kernel references, touches kernel config, and includes test coverage, making it the stronger representative of the fix set.", + "best_issue_reason": null, + "best_pr_reason": "PR #46478 is the best overall PR in the cluster because it addresses the problem more comprehensively and is more suitable as the merge target for the shared hub-kernel compatibility work.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46293", + "right": "pull_request:46478", + "accept": false, + "reason": "Both mention the same issue and file, but the concrete changes differ: #46293 adjusts repo/version compatibility metadata, while #46478 pins revisions to main and extends config/tests. They are related fixes, not the same PR-worthy code 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": 35, + "component_count": 35, + "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-46376-2", + "nodes": [ + "pull_request:46376", + "pull_request:46417" + ], + "soft_pairs": [ + "pull_request:46376|pull_request:46417" + ], + "prepared_review_unit_hash": "0d196d39d760fb23881e231f917e55438aeb55209250ee7868e6b1b99168b07b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5959, + "estimated_input_tokens": 1490, + "estimated_eval_tokens": 3236 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5959, + "estimated_input_tokens": 1490, + "estimated_eval_tokens": 3236 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are unrelated: one is a draft infrastructure change for continuous batching multimodal support, and the other is a full model addition for HunYuan VL. They should not be clustered as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46417 is the cleaner standalone representative of this set because it is a complete, non-draft model addition; PR 46376 is a draft, broader infrastructure change and not the same underlying work.", + "best_issue_reason": null, + "best_pr_reason": "PR 46417 is the best PR in this cluster for triage since it is self-contained, non-draft, and materially more complete than the unrelated draft PR 46376.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46376", + "right": "pull_request:46417", + "accept": false, + "reason": "Different changes and code paths: continuous batching multimodal infrastructure vs. adding the HunYuan VL model. They do not look like the same underlying fix or feature." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/manifest.json b/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..58491c6596dcc96d03c08d8207390bc5377b0a03 --- /dev/null +++ b/snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260609t120115z", + "artifacts": { + "hybrid": "snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-06-09T12:12:39Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260609T120115Z", + "variant": "hybrid" +} diff --git a/snapshots/20260609T120115Z/manifest.json b/snapshots/20260609T120115Z/manifest.json index 17c985d2d28ad52f554597360e19eabc802ecb49..2dfc25b9420bbd9d8d3c53775171310c2df89881 100644 --- a/snapshots/20260609T120115Z/manifest.json +++ b/snapshots/20260609T120115Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 1519 }, "extracted_at": "2026-06-09T12:01:15Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260609t120115z", + "runs": { + "hybrid-model-20260609t120115z": { + "analysis_id": "hybrid-model-20260609t120115z", + "artifacts": { + "hybrid": "snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260609T120115Z/analysis-runs/hybrid-model-20260609t120115z/manifest.json", + "model": null, + "published_at": "2026-06-09T12:12:39Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260609T120115Z", "watermark": {