diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index e8a04937418fe946bcdf3de46a06d85b791411d1..49604cf218af49174fb519977b0c58e081247e56 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": "20260607T060034Z", - "generated_at": "2026-06-07T06:08:53Z", + "snapshot_id": "20260607T120016Z", + "generated_at": "2026-06-07T12:08:25Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1148,114 +1148,6 @@ } ] }, - { - "cluster_id": "cluster-45183-4", - "summary": "Cluster of 2 issues and 2 PRs centered on issue #46463.", - "status": "open", - "confidence": 0.8, - "canonical_issue_number": 46463, - "canonical_pr_number": 46465, - "issue_numbers": [ - 45183, - 46463 - ], - "pr_numbers": [ - 46465, - 46470 - ], - "evidence_types": [ - "closing_reference", - "shared_issue_target", - "soft_similarity" - ], - "pr_comparisons": [ - { - "left_pr_number": 46465, - "right_pr_number": 46470, - "code_similarity": 0.63, - "size_similarity": 0.737, - "file_overlap": 0.333, - "area_overlap": 1.0, - "patch_similarity": 0.214, - "shared_filenames": [ - "src/transformers/models/mt5/modeling_mt5.py", - "src/transformers/models/t5/modeling_t5.py", - "src/transformers/models/umt5/modeling_umt5.py" - ], - "shared_file_areas": [ - { - "filename": "src/transformers/models/mt5/modeling_mt5.py", - "left_ranges": [ - [ - 654, - 660 - ], - [ - 664, - 670 - ] - ], - "right_ranges": [ - [ - 654, - 660 - ], - [ - 664, - 670 - ] - ] - }, - { - "filename": "src/transformers/models/t5/modeling_t5.py", - "left_ranges": [ - [ - 643, - 649 - ], - [ - 653, - 659 - ] - ], - "right_ranges": [ - [ - 643, - 649 - ], - [ - 653, - 659 - ] - ] - }, - { - "filename": "src/transformers/models/umt5/modeling_umt5.py", - "left_ranges": [ - [ - 612, - 618 - ], - [ - 622, - 628 - ] - ], - "right_ranges": [ - [ - 612, - 618 - ], - [ - 622, - 628 - ] - ] - } - ] - } - ] - }, { "cluster_id": "cluster-45363-3", "summary": "Cluster of 3 related pull requests linked by soft_similarity.", @@ -1892,7 +1784,7 @@ "cluster_id": "cluster-45561-3", "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", "status": "open", - "confidence": 0.8, + "confidence": 0.55, "canonical_issue_number": 45561, "canonical_pr_number": 45645, "issue_numbers": [ @@ -1903,9 +1795,7 @@ 45645 ], "evidence_types": [ - "closing_reference", - "shared_issue_target", - "soft_similarity" + "closing_reference" ], "pr_comparisons": [ { @@ -2080,6 +1970,216 @@ } ] }, + { + "cluster_id": "cluster-46463-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46463.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 46463, + "canonical_pr_number": 46465, + "issue_numbers": [ + 46463 + ], + "pr_numbers": [ + 46465, + 46470 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46465, + "right_pr_number": 46470, + "code_similarity": 0.63, + "size_similarity": 0.737, + "file_overlap": 0.333, + "area_overlap": 1.0, + "patch_similarity": 0.214, + "shared_filenames": [ + "src/transformers/models/mt5/modeling_mt5.py", + "src/transformers/models/t5/modeling_t5.py", + "src/transformers/models/umt5/modeling_umt5.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/mt5/modeling_mt5.py", + "left_ranges": [ + [ + 654, + 660 + ], + [ + 664, + 670 + ] + ], + "right_ranges": [ + [ + 654, + 660 + ], + [ + 664, + 670 + ] + ] + }, + { + "filename": "src/transformers/models/t5/modeling_t5.py", + "left_ranges": [ + [ + 643, + 649 + ], + [ + 653, + 659 + ] + ], + "right_ranges": [ + [ + 643, + 649 + ], + [ + 653, + 659 + ] + ] + }, + { + "filename": "src/transformers/models/umt5/modeling_umt5.py", + "left_ranges": [ + [ + 612, + 618 + ], + [ + 622, + 628 + ] + ], + "right_ranges": [ + [ + 612, + 618 + ], + [ + 622, + 628 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43656-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 43656, + "issue_numbers": [], + "pr_numbers": [ + 43656, + 43842 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 43656, + "right_pr_number": 43842, + "code_similarity": 0.405, + "size_similarity": 0.5, + "file_overlap": 1.0, + "area_overlap": 0.0, + "patch_similarity": 0.036, + "shared_filenames": [ + "src/transformers/cli/serve.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cli/serve.py", + "left_ranges": [ + [ + 11, + 18 + ], + [ + 30, + 36 + ], + [ + 315, + 323 + ], + [ + 665, + 671 + ], + [ + 931, + 937 + ], + [ + 1843, + 1849 + ], + [ + 1868, + 1874 + ] + ], + "right_ranges": [ + [ + 54, + 61 + ], + [ + 587, + 593 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44002-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44002, + "issue_numbers": [], + "pr_numbers": [ + 44002, + 44029 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44002, + "right_pr_number": 44029, + "code_similarity": 0.135, + "size_similarity": 0.43, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.329, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, { "cluster_id": "cluster-44018-2", "summary": "Cluster of 2 related pull requests linked by soft_similarity.", @@ -2356,16 +2456,7 @@ ] } ], - "duplicate_issues": [ - { - "cluster_id": "cluster-45183-4", - "canonical_issue_number": 46463, - "duplicate_issue_numbers": [ - 45183 - ], - "reason": "Issues in cluster-45183-4 are treated as duplicates because they share closing_reference, shared_issue_target, soft_similarity evidence." - } - ], + "duplicate_issues": [], "duplicate_prs": [ { "cluster_id": "cluster-41115-3", @@ -2394,6 +2485,15 @@ "target_issue_number": 43240, "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." }, + { + "cluster_id": "cluster-43656-2", + "canonical_pr_number": 43656, + "duplicate_pr_numbers": [ + 43842 + ], + "target_issue_number": null, + "reason": "PRs in cluster-43656-2 are treated as duplicates because they share soft_similarity evidence." + }, { "cluster_id": "cluster-43698-3", "canonical_pr_number": 43779, @@ -2420,6 +2520,15 @@ "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-44002-2", + "canonical_pr_number": 44002, + "duplicate_pr_numbers": [ + 44029 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44002-2 are treated as duplicates because they share soft_similarity evidence." + }, { "cluster_id": "cluster-44018-2", "canonical_pr_number": 44068, @@ -2438,15 +2547,6 @@ "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-45183-4", - "canonical_pr_number": 46465, - "duplicate_pr_numbers": [ - 46470 - ], - "target_issue_number": 46463, - "reason": "PRs in cluster-45183-4 are treated as duplicates because they converge on issue #46463 with closing_reference, shared_issue_target, soft_similarity evidence." - }, { "cluster_id": "cluster-45363-3", "canonical_pr_number": 46339, @@ -2464,7 +2564,7 @@ 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." + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference evidence." }, { "cluster_id": "cluster-45647-3", @@ -2474,13 +2574,22 @@ ], "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-46463-3", + "canonical_pr_number": 46465, + "duplicate_pr_numbers": [ + 46470 + ], + "target_issue_number": 46463, + "reason": "PRs in cluster-46463-3 are treated as duplicates because they converge on issue #46463 with closing_reference, shared_issue_target evidence." } ], "best_issue": { - "cluster_id": "cluster-45183-4", - "issue_number": 46463, - "reason": "Issue #46463 is the strongest global issue candidate because it is open, belongs to a cluster with 4 artifacts, and carries 0 discussion comments plus 6 inbound references.", - "score": 146.0 + "cluster_id": "cluster-41211-3", + "issue_number": 41211, + "reason": "Issue #41211 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 6 discussion comments plus 6 inbound references.", + "score": 148.0 }, "best_pr": { "cluster_id": "cluster-43979-11", diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 8fd1d1671b4679f750e12d66939a5cc132157618..212343d31ea5961508145c12dc8beb6f2096692a 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": "20260607T060034Z", - "generated_at": "2026-06-07T06:08:53Z", + "snapshot_id": "20260607T120016Z", + "generated_at": "2026-06-07T12:08:25Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -134,373 +134,367 @@ "issue:41762|issue:43596", "issue:39290|issue:41720" ], - "prepared_review_unit_hash": "9068cabf761cd78322c812403c5129fd644ad3b34c57ef2db9bf5570dfef3a8f", + "prepared_review_unit_hash": "8cb746d2b977866aed72a267fc6eca5ad93cf15a412264ceb38fcb9b8467a521", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 65, - "serialized_chars": 47170, - "estimated_input_tokens": 11793, - "estimated_eval_tokens": 23842 + "serialized_chars": 47165, + "estimated_input_tokens": 11792, + "estimated_eval_tokens": 23840 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most items are unrelated standalone bugs or docs problems. The only clear duplicate family is the DeepSpeed ZeRO-3 index-out-of-bounds reports around Bert/Gemma3, with 43596 as the best anchor.", - "confidence": 0.82, - "canonical_issue_reason": "Issue 43596 is the cleanest representative of the duplicate ZeRO-3 index-out-of-bounds family: it states the exact error and the DeepSpeed ZeRO-3/Bert loading context clearly.", + "summary": "This cluster is mostly loose similarity across unrelated model, docs, dtype, and distributed-training reports. The best-supported duplicate candidates are 43596/43638 (same ZeRO-3/Bert zero-size index error family) and 45859/45902 (same Qwen3.5 MoE rotary-embedding/offload bug).", + "confidence": 0.81, + "canonical_issue_reason": "issue:43596 is the likely canonical for the ZeRO-3/Bert zero-size index-error cluster: it is broader and better-scoped than 43638, which reads like a narrower Bert/non-pretrained variant of the same failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "43596 best captures the shared DeepSpeed ZeRO-3 loading failure and is the most concrete, reusable representative of the accepted duplicate set.", + "best_issue_reason": "issue:43596 is the best issue representative overall because it is the broadest clear fit for the ZeRO-3/Bert index-error family; this is a likely duplicate candidate, not a confirmed duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Different failures: missing Qwen3.5 config field vs chat/completions streaming error." + "reason": "Different bugs: Qwen3.5 config attribute error vs streaming chat/completions failure." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Different hardware/runtime paths: AMD GPU failure vs MPS histogram op on Int." + "reason": "Different hardware/runtime failures: AMD GPU issue vs MPS histogram on Int." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Different model/runtime bug: image preprocessor dtype mismatch vs GPT-OSS on AMD GPUs." + "reason": "Gemma3 dtype mismatch in preprocessing is unrelated to GPT-OSS AMD GPU failure." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Different concrete issue: autocast dtype mismatch vs GPT-OSS on AMD GPUs." + "reason": "Autocast dtype mismatch and AMD GPU failure are different code paths." }, { "left": "issue:45237", "right": "issue:46468", "accept": false, - "reason": "Different models and symptoms; not the same underlying bug." + "reason": "Unrelated: GPT-OSS AMD GPU bug vs Nemotron example breakage." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Different problem: AMD GPU execution vs invalid method call in GraniteMoEHybrid." + "reason": "Different problems: AMD GPU support vs invalid method call in GraniteMoEHybrid." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "Different code paths: DeepSpeed collective mismatch vs ONNX export failure." + "reason": "DeepSpeed collective mismatch and ModernBERT ONNX export are unrelated." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "Different model-specific export vs FP8 configuration failure." + "reason": "ModernBERT export issue vs Qwen3Moe FP8 failure; same broad area only." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "Different bugs: ONNX export error vs Qwen3 auto device mapping assert." + "reason": "ONNX export bug vs Qwen3 device-mapping CUDA assert." }, { "left": "issue:45901", "right": "issue:46015", "accept": false, - "reason": "Different issues: task crash vs outdated RAG docs." + "reason": "Table QA crash and outdated RAG examples are different issues." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are docs cleanups for v5 removals, but they target different removed content and are not the same change." + "reason": "Both are docs regressions, but they target different removed v5 references and are not the same concrete issue." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "Different issues: ONNX export failure vs GenerationMixin import regression." + "reason": "Export error vs GenerationMixin import error are unrelated." }, { "left": "issue:44918", "right": "issue:46165", "accept": false, - "reason": "Different bugs: TRL SFT embedding unpacking vs resume/trust_remote_code loading issue." + "reason": "TRL input-embeddings bug vs resume-training/trust_remote_code issue." }, { "left": "issue:44805", "right": "issue:45823", "accept": false, - "reason": "Different symptoms and subsystems: mask/index shape mismatch vs Gemma4 PLE device mismatch." + "reason": "Mask shape mismatch and Gemma4 PLE device mismatch are different failures." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Not the same bug; one is ZeRO-3 index error, the other is a mask shape mismatch." + "reason": "Bert ZeRO-3 index error is unrelated to tensor mask shape mismatch." }, { "left": "issue:43638", "right": "issue:46165", "accept": false, - "reason": "Different failures: ZeRO-3 Bert loading vs resume training with locally downloaded model." + "reason": "Different problems: Bert ZeRO-3 loading vs model resume/trust_remote_code." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Different errors; only one is the ZeRO-3 zero-size index issue family." + "reason": "ZeRO-3 Bert index error vs mask/index shape mismatch are not the same bug." }, { "left": "issue:44805", "right": "issue:46294", "accept": false, - "reason": "Different model path: mask shape mismatch vs Qwen3.5 video token/feature mismatch." + "reason": "Mask/tensor shape mismatch is unrelated to video feature/token mismatch." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Different issue: mask/index mismatch vs GPT-OSS tensor-parallel failure." + "reason": "Different failures: tensor shape mismatch vs GPT-OSS tensor-parallel issue." }, { "left": "issue:46165", "right": "issue:46468", "accept": false, - "reason": "Different bugs and contexts; no shared code-path evidence." + "reason": "Resume-training issue vs Nemotron example breakage are unrelated." }, { "left": "issue:44560", "right": "issue:46165", "accept": false, - "reason": "Different model/input handling bugs; not the same underlying change." + "reason": "Video StopIteration in Qwen3-vl-embedding is unrelated to resume training." }, { "left": "issue:43596", "right": "issue:46165", "accept": false, - "reason": "Different issues: ZeRO-3 Bert index error vs resume training/local model loading." + "reason": "Bert ZeRO-3 loading bug vs training-resume issue; not the same code path." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "Different subsystems: BarkProcessor voice preset vs ModernBERT ONNX export." + "reason": "BarkProcessor voice preset bug is unrelated to ModernBERT ONNX export." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Different DeepSpeed failures: collective mismatch vs ZeRO-3 Bert index error." + "reason": "Collective mismatch in Jamba/ZeRO-3 is different from Bert zero-size index error." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Both report the same ZeRO-3 index-0-out-of-bounds failure during model initialization/loading; same underlying bug family." + "reason": "Same underlying ZeRO-3/Bert zero-size index error; 43638 is the narrower non-pretrained Bert variant." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Different issues: Liger Kernel cross-entropy dispatch vs GPT-OSS tensor-parallel failure." + "reason": "Liger Kernel cross-entropy dispatch bug vs GPT-OSS TP issue." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Different bugs: detokenization regression vs LoRA merge/save/load output mismatch." + "reason": "Detokenization bug and LoRA merge/save mismatch are different issues." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "Different issues: Qwen NVFP4 runtime error vs LoRA save/load inconsistency." + "reason": "Qwen-3-8B-NVFP4 runtime issue is unrelated to LoRA save/reload inconsistency." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Different training/runtime mismatches; not the same concrete bug." + "reason": "FSDP checkpointing size mismatch is not the same as mask/tensor shape mismatch." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "Different DeepSpeed errors: weight dimensionality vs collective mismatch." + "reason": "Both mention DeepSpeed, but they are different model/error scenarios." }, { "left": "issue:45859", "right": "issue:45902", - "accept": false, - "reason": "Same module family, but different bugs: CPU offload incompatibility vs uninitialized inv_freq after materialization." + "accept": true, + "reason": "Same Qwen3.5 MoE rotary-embedding/offload bug around inv_freq materialization and device handling." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "Different bugs: ZeRO-3 Bert index error vs GPT-OSS tensor-parallel issue." + "reason": "Bert ZeRO-3 index error vs GPT-OSS MoE TP issue." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "Different failures: activation checkpointing/FSDP tensor recomputation vs trainer API signature mismatch." - }, - { - "left": "issue:43828", - "right": "issue:45161", - "accept": false, - "reason": "Different model/runtime bugs: Phi dtype mismatch vs GPT-OSS tensor parallel failure." + "reason": "Activation-checkpointing/FSDP bug vs CustomTrainer kwarg incompatibility." }, { - "left": "issue:44918", + "left": "issue:43596", "right": "issue:45161", "accept": false, - "reason": "Different failures in different training stacks; no shared underlying bug." + "reason": "Different bugs: Bert ZeRO-3 loading vs GPT-OSS tensor-parallel support." }, { "left": "issue:43716", "right": "issue:45507", "accept": false, - "reason": "Different model-specific bugs: Mistral-3 dtype mismatch vs GraniteMoEHybrid invalid method call." + "reason": "Gemma3 dtype mismatch is unrelated to GraniteMoEHybrid invalid method." }, { "left": "issue:45874", "right": "issue:46294", "accept": false, - "reason": "Different issues: reverse embedding lookup cost vs video token/feature mismatch." + "reason": "Expensive reverse-embedding lookup is unrelated to video token/template mismatch." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "Different models and error modes; not the same change or bug." + "reason": "NVFP4 runtime error and AMD GPU failure are different issues." }, { "left": "issue:39290", "right": "issue:45081", "accept": false, - "reason": "Different tokenizer/config regressions; unrelated model families and failures." + "reason": "Gemma3 config attribute regression vs Mistral tokenizer regex crash." }, { "left": "issue:45823", "right": "issue:45874", "accept": false, - "reason": "Different Gemma4 bugs: device-map mismatch vs expensive reverse embedding lookup." + "reason": "Gemma4 PLE device mismatch and reverse-embedding slowdown are unrelated." }, { "left": "issue:43425", "right": "issue:44292", "accept": false, - "reason": "Different incompatibilities: torch version support vs Qwen NVFP4 runtime error." + "reason": "Torch/transformers version incompatibility vs Qwen-3-8B-NVFP4 runtime error." }, { "left": "issue:34928", "right": "issue:41720", "accept": false, - "reason": "Different bugs: FSDP checkpointing recomputation vs Qwen3 device-mapping assert." + "reason": "FSDP checkpointing mismatch vs Qwen3 auto device-map CUDA assert." }, { - "left": "issue:43827", + "left": "issue:43828", "right": "issue:44512", "accept": false, - "reason": "Both are docs cleanup, but they cover different removed features and are not the same duplicate." + "reason": "Autocast dtype mismatch vs docs mention of removed run command." }, { "left": "issue:43828", "right": "issue:45685", "accept": false, - "reason": "Different model/runtime bugs: autocast dtype mismatch vs MPS histogram op limitation." + "reason": "Autocast dtype mismatch and MPS histogram issue are unrelated." }, { "left": "issue:44918", "right": "issue:45685", "accept": false, - "reason": "Different training/runtime problems with no shared code-path evidence." + "reason": "TRL input embeddings bug vs MPS histogram issue." }, { "left": "issue:43039", "right": "issue:43828", "accept": false, - "reason": "Different issues in different models; only broad dtype-related similarity." + "reason": "Liger Kernel cross-entropy bug vs autocast dtype mismatch." }, { "left": "issue:30277", "right": "issue:36331", "accept": false, - "reason": "Different bugs: DeepSpeed collective mismatch vs custom trainer signature breakage." + "reason": "DeepSpeed collective mismatch vs CustomTrainer compute_loss signature change." }, { "left": "issue:34928", "right": "issue:36010", "accept": false, - "reason": "Different issues: FSDP/activation checkpointing vs GenerationMixin import error." + "reason": "FSDP checkpointing issue vs GenerationMixin import error." }, { "left": "issue:45237", "right": "issue:46015", "accept": false, - "reason": "Different model/runtime bug vs docs issue." + "reason": "AMD GPU runtime issue vs outdated RAG docs/examples." }, { "left": "issue:43039", "right": "issue:44805", "accept": false, - "reason": "Different failures; same broad theme of runtime errors is not enough." + "reason": "Liger kernel cross-entropy dispatch bug is unrelated to mask shape mismatch." }, { "left": "issue:43866", "right": "issue:46402", "accept": false, - "reason": "Different model corruption/load failures with no shared underlying bug." + "reason": "Corrupted Ovis checkpoint is unrelated to Sam3Model weights not loading." }, { "left": "issue:45800", "right": "issue:46304", "accept": false, - "reason": "Both mention torch/transformers compatibility, but they are different version and environment regressions." + "reason": "Version incompatibility and USE_DISTRIBUTED=0 breakage are related only at a very broad level." }, { "left": "issue:43296", "right": "issue:46304", "accept": false, - "reason": "Different incompatibilities: vLLM/PaddleOCR-VL load failure vs USE_DISTRIBUTED=0 torch support regression." + "reason": "PaddleOCR-VL/vLLM load failure is unrelated to the distributed-support regression." }, { "left": "issue:41762", "right": "issue:43596", - "accept": true, - "reason": "Both describe the same ZeRO-3 index-0-out-of-bounds failure while loading models; same underlying bug family." + "accept": false, + "reason": "Both are ZeRO-3 index errors, but on different models and likely different code paths." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Different model/runtime bugs: Gemma3 config attribute missing vs Qwen3 device-mapping assert." + "reason": "Gemma3 config attribute regression vs Qwen3 auto device-mapping CUDA assert." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: the only clearly supported duplicate family is the DeepSpeed ZeRO-3 index-out-of-bounds loading bug around 43596/43638/41762, and the chosen anchor is reasonable. The soft-edge verdicts stay conservative; the rejected pairs are generally separated by model, subsystem, or error mode, and the accepted ZeRO-3 pairings are well supported by nearly identical titles and contexts." + "accept": false, + "feedback": "The overall clustering is conservative, but the rationale overstates evidence in a few places. In particular, calling issue:43596 the \u2018strongest confirmed duplicate pair\u2019 is too strong for a packet with only soft similarity evidence and no explicit duplicate link. Please downgrade that language to \u2018likely\u2019/\u2018best candidate\u2019 and avoid implying confirmation not present in the packet." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -611,292 +605,208 @@ "issue:43531|issue:45070", "issue:45020|issue:45800" ], - "prepared_review_unit_hash": "8fc7a60d906982a3f5a9a1f462e33520c59358c5ae4272109f9b5514d3c94907", + "prepared_review_unit_hash": "bdac03df1027d7083405d55fb0fe716ca5d59f8fd2b8227776beb56c7aaa4a4a", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 42, - "serialized_chars": 40101, - "estimated_input_tokens": 10026, - "estimated_eval_tokens": 20308 + "serialized_chars": 40096, + "estimated_input_tokens": 10024, + "estimated_eval_tokens": 20304 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are unrelated lookalikes; only the AutoTokenizer v5 regression reports (Granite-only vs broader multi-model report) appear to be the same underlying bug.", - "confidence": 0.91, - "canonical_issue_reason": "issue:45920 is the broadest, most representative description of the tokenizer regression, and it subsumes the narrower Granite-specific report in issue:45812.", + "summary": "This cluster is mostly heterogeneous regression reports, not one cohesive duplicate set. The only strong duplicate-like pair is the AutoTokenizer wrong-token-IDs regression across models.", + "confidence": 0.69, + "canonical_issue_reason": "Issue #45920 is the broadest and most complete report of the tokenizer ID regression, covering multiple affected model families and the same v5-era failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 best captures the underlying bug because it generalizes the wrong-token-ID regression across multiple model families rather than one model family.", + "best_issue_reason": "#45920 is the best umbrella issue: it is open, explicit about the shared tokenizer regression, and aggregates the widest model impact.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44589", "right": "issue:46304", "accept": false, - "reason": "Both are transformer compatibility failures, but one is a Float8 storage lookup error and the other is a distributed/USE_DISTRIBUTED=0 breakage; different bugs." + "reason": "Different failures: float8 storage lookup vs distributed support breakage. Same release family, but not the same bug." }, { "left": "issue:45823", "right": "issue:46294", "accept": false, - "reason": "Different model families and different failure modes: device mismatch in Gemma4 PLE versus video-token mismatch in Qwen3.5 apply_chat_template." + "reason": "Different models and code paths: Gemma4 device placement vs Qwen3.5 video chat-template token/feature mismatch." }, { "left": "issue:45237", "right": "issue:45901", "accept": false, - "reason": "No shared code-path signal; one is AMD GPU support for GPT-OSS, the other is a table-question-answering crash." + "reason": "AMD GPU execution issue vs table-question-answering crash; no concrete shared bug." }, { "left": "issue:43901", "right": "issue:44512", "accept": false, - "reason": "Both are docs-related, but they point to different deprecations: return_all_scores versus the removed transformers run command." + "reason": "Both are docs regressions, but they concern unrelated documentation updates and not one underlying code change." }, { "left": "issue:44841", "right": "issue:45823", "accept": false, - "reason": "Different models and subsystems: Voxtral processor failure versus Gemma4 device placement." + "reason": "Voxtral processor failure and Gemma4 device mismatch are distinct model-specific problems." }, { "left": "issue:43039", "right": "issue:44560", "accept": false, - "reason": "Liger cross_entropy dispatch and Qwen3.5 video StopIteration are unrelated." + "reason": "Liger Kernel cross-entropy behavior and Qwen3-vl StopIteration are unrelated code paths." }, { "left": "issue:43975", "right": "issue:46015", "accept": false, - "reason": "One is a detokenization regression; the other is outdated RAG documentation." + "reason": "Tokenizer detokenization bug and outdated RAG examples are unrelated." }, { "left": "issue:45812", "right": "issue:45920", "accept": true, - "reason": "Same underlying AutoTokenizer v5 regression: 45920 is the broader report of wrong token IDs across model families, while 45812 is the Granite-specific instance." + "reason": "Both report the same underlying AutoTokenizer v5 regression producing wrong token IDs across multiple model families." }, { "left": "issue:44863", "right": "issue:46015", "accept": false, - "reason": "NemotronH checkpoint loading and RAG docs are unrelated." + "reason": "NemotronH checkpoint-loading issue is unrelated to RAG documentation examples." }, { "left": "issue:45020", "right": "issue:46304", "accept": false, - "reason": "Remote_code breakage and USE_DISTRIBUTED=0 support are different compatibility issues." + "reason": "Broad remote_code compatibility breakage and USE_DISTRIBUTED=0 support are different regressions." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Checkpoint corruption in Ovis2 is not the same as NemotronH failing to load its checkpoints." + "reason": "A corrupted Ovis2 checkpoint is not the same as NemotronH implementation loading failures." }, { "left": "issue:44292", "right": "issue:46015", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime error is unrelated to RAG documentation." + "reason": "Qwen-3-8B-NVFP4 runtime error is unrelated to documentation examples." }, { "left": "issue:43872", "right": "issue:46304", "accept": false, - "reason": "bitsandbytes Int8Params constructor mismatch is a different incompatibility than distributed-disabled PyTorch support." + "reason": "bitsandbytes init argument incompatibility is unrelated to distributed build support." }, { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Both mention sliding_window, but one is Gemma3TextConfig missing sliding_window_pattern under vLLM and the other is a Qwen3-MoE sliding_window issue; not clearly the same bug." + "reason": "Both mention sliding_window, but one is a missing config attribute for vLLM/Gemma3 and the other is a Qwen3-MoE behavior issue." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Both are IndexErrors, but the tensor shape mismatch and the zero-sized tensor out-of-bounds failure are distinct." + "reason": "Both are IndexErrors, but the reported shapes/contexts differ enough that a shared root cause is not established." }, { "left": "issue:43792", "right": "issue:45901", "accept": false, - "reason": "Whisper loading/can't run is unrelated to table-question-answering crashes." + "reason": "Whisper model loading and table QA task crashes are unrelated." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "Different model-load failures: GLM-4.7-Flash unit-test loading versus corrupted Ovis2 checkpoint." + "reason": "A unit-test model loading failure is not the same as a corrupted checkpoint report." }, { "left": "issue:43425", "right": "issue:45070", "accept": false, - "reason": "Torch version incompatibility and pydantic PretrainedConfig field breakage are different issues." + "reason": "Torch version incompatibility and PretrainedConfig/pydantic regression are separate compatibility problems." }, { "left": "issue:43596", "right": "issue:45800", "accept": false, - "reason": "DeepSpeed ZeRO-3 Bert index error and torch 2.4.1/transformers 5.8.0 incompatibility are not the same bug." + "reason": "DeepSpeed ZeRO-3 BERT loading error and torch 2.4.1/v5.8.0 incompatibility are not the same bug." }, { "left": "issue:43975", "right": "issue:45901", "accept": false, - "reason": "Incorrect detokenization for DeepSeek coder and table-question-answering crashes are unrelated." + "reason": "Different subsystems: detokenization vs table QA runtime crash." }, { "left": "issue:43792", "right": "issue:45070", "accept": false, - "reason": "Whisper loading and pydantic config-field regression are unrelated." + "reason": "Whisper model loading and pydantic config breakage are unrelated." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Both are Qwen3.5 chat-related, but the titles indicate different failures and code paths: one crashes with transformers chat, the other fails in streaming chat/completions." + "reason": "Both involve Qwen3.5, but one is a chat crash and the other is a streaming API failure; different code paths." }, { "left": "issue:44292", "right": "issue:45901", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime failure is unrelated to table QA crashes." - }, - { - "left": "issue:39290", - "right": "issue:45823", - "accept": false, - "reason": "Gemma3 sliding_window_pattern missing and Gemma4 PLE device mismatch do not describe the same underlying bug." + "reason": "NVFP4 model execution error and table QA crash are unrelated." }, { "left": "issue:43404", "right": "issue:43866", "accept": false, - "reason": "Tied lm_head weights in Mistral3 and Ovis2 checkpoint corruption are unrelated." + "reason": "Tied-weight bug in Mistral3 is unrelated to an Ovis2 checkpoint corruption report." }, { "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Qwen3.5 streaming chat failure and an MPS histogram_mps int error are unrelated." - }, - { - "left": "issue:43541", - "right": "issue:45685", - "accept": false, - "reason": "Mixtral torch-dynamo grouped_mm tracing error and MPS histogram_mps failure are different code paths." + "reason": "Qwen3.5 streaming API failure and MPS histogram/int error are unrelated." }, { "left": "issue:43541", "right": "issue:43828", "accept": false, - "reason": "Both involve MoE and dtype issues, but grouped_mm tracing on Mixtral is not the same bug as autocast dtype mismatch on Phi-tiny-MoE." - }, - { - "left": "issue:44841", - "right": "issue:45084", - "accept": false, - "reason": "Processor failure for Voxtral and a non-template-node compilation error are unrelated." - }, - { - "left": "issue:36010", - "right": "issue:39290", - "accept": false, - "reason": "Importing GenerationMixin and Gemma3 sliding_window_pattern errors are unrelated compatibility breaks." - }, - { - "left": "issue:45800", - "right": "issue:45823", - "accept": false, - "reason": "Torch/transformers incompatibility and Gemma4 device mismatch are different failures." - }, - { - "left": "issue:41093", - "right": "issue:41762", - "accept": false, - "reason": "Both are IndexErrors under DeepSpeed, but the mask-shape mismatch and the zero-sized-tensor out-of-bounds error are distinct." - }, - { - "left": "issue:43296", - "right": "issue:43366", - "accept": false, - "reason": "PaddleOCR-VL loading in vLLM and GGUF gpt-oss support are unrelated model/loader issues." - }, - { - "left": "issue:43572", - "right": "issue:45070", - "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig is not the same as the PretrainedConfig pydantic field regression." + "reason": "Both are dtype/tracing-related, but they affect different models and different execution mechanisms; not one concrete bug." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "Both are shape-related errors, but the mask/index mismatch and TRL embedding unpacking failure are different." - }, - { - "left": "issue:44387", - "right": "issue:45005", - "accept": false, - "reason": "Int4 quantization OOM and tied-weights issues on translation models are unrelated." - }, - { - "left": "issue:43541", - "right": "issue:45161", - "accept": false, - "reason": "Mixtral grouped_mm tracing failure and GPT-OSS tensor-parallel-only issue are different." - }, - { - "left": "issue:43828", - "right": "issue:45464", - "accept": false, - "reason": "Phi-tiny-MoE autocast dtype mismatch and Qwen3.5 streaming API failure are unrelated." - }, - { - "left": "issue:44291", - "right": "issue:45070", - "accept": false, - "reason": "init_empty_weights/_is_hf_initialized TypeError and the pydantic PretrainedConfig regression are separate compatibility bugs." - }, - { - "left": "issue:43366", - "right": "issue:45464", - "accept": false, - "reason": "GGUF gpt-oss support and Qwen3.5 streaming chat failure are unrelated." - }, - { - "left": "issue:43531", - "right": "issue:45070", - "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and a pydantic config-field breakage are unrelated." + "reason": "Both are training-time tensor/mask issues, but the error shapes and implicated paths differ; too little evidence for one root cause." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "remote_code model breakage and torch 2.4.1/transformers 5.8.0 incompatibility are different issues." + "reason": "Remote_code breakage and torch/transformers version incompatibility are separate compatibility issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analysis is grounded in the packet and stays conservative on the soft-edge pairs. The 45812 \u2194 45920 linkage is well supported by the issue titles, and the broader canonicalization to 45920 is reasonable." + "feedback": "The analysis is well grounded in the packet and stays conservative on the soft-edge pairs. The choice of #45920 as the umbrella/canonical issue is supported by the broader, open report, and the only accepted duplicate pair (#45812/#45920) is plausibly the same AutoTokenizer v5 token-ID regression. No overstatement stands out." }, "evaluator_used": true, "retried": false, @@ -1001,14 +911,14 @@ "issue:43866|issue:44403", "issue:38175|issue:43054" ], - "prepared_review_unit_hash": "320f8e07d41bb85ccfb6179ee56a66c0c29b4aac90cdad438de0c6ea3d8f2878", + "prepared_review_unit_hash": "535b3f02d8e1476c6d996fbf5958398a63e018205eed12d94986283cc6f78487", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, @@ -1023,219 +933,219 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is still mostly a set of small duplicate-prone subgroups, with the strongest one centered on the Transformers v5 non-persistent-buffer regression. I tightened the custom-module pair: 45632/45698 is too under-specified to call a duplicate confidently, so it should stay rejected.", + "summary": "The cluster is mostly heterogeneous, but a few pairs are strong duplicates: the v5 non-persistent-buffer corruption reports, a Qwen3.5 video-token mismatch regression, and a trust_remote_code/custom-module resolution collision. Most other pairs only share broad subsystem or model-family similarity and should stay separate.", "confidence": 0.78, - "canonical_issue_reason": "issue:43950 is the strongest representative issue: it describes a concrete Transformers v5 regression with a clear mechanism (non-persistent buffer corruption) and cleanly covers the related buffer-materialization reports.", + "canonical_issue_reason": "issue:43950 is the clearest root-cause report for the non-persistent-buffer regression, and it lines up directly with the shorter duplicate descriptions in 43644 and 44534.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the best overall issue candidate because it is specific, reproducible, and has the clearest duplicate neighborhood among the listed issues.", + "best_issue_reason": "issue:43950 is the best canonical anchor because it is concrete, regression-focused, and names the exact failure mode and mechanism; it best represents the duplicate buffer-corruption subgroup in this cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43716", "right": "issue:46294", "accept": false, - "reason": "Both are multimodal input-mismatch reports, but they involve different models and different failure points, so they do not look like the same concrete bug." + "reason": "Different models and different multimodal failures; both are video/image input issues, but not the same concrete bug." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "These describe different failures: torch dynamo grouped_mm tracing versus a missing Float8 storage type." + "reason": "Unrelated runtime errors: torch dynamo grouped_mm tracing versus missing Float8 storage object." }, { "left": "issue:44479", "right": "issue:46294", "accept": true, - "reason": "Both point to the same Qwen video-input regression family: video feature/token handling and chat-template mismatch in Qwen3.5 models." + "reason": "Both describe the same Qwen3.5 video-input regression, with 46294 appearing to be a specific symptom of the broader token/feature mismatch in 44479." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Both are index/shape errors, but the titles suggest different code paths and symptoms; that is too broad to treat as one bug." + "reason": "Both are IndexErrors, but the failing code paths are different: mask-length mismatch versus ZeRO-3 zero-sized parameter access." }, { "left": "issue:43054", "right": "issue:43994", - "accept": true, - "reason": "Both are SigLIP2 model-quality reports with broadly the same bad-output symptom, so they plausibly reflect the same underlying model/processor issue." + "accept": false, + "reason": "Same model family, but the reported problems differ: degraded text embeddings versus nonsensical AutoModel/pipeline output." }, { "left": "issue:45759", "right": "issue:45902", "accept": false, - "reason": "These are different bugs: composite-config unwrapping versus uninitialized rotary embedding memory after materialization." + "reason": "Different bugs: multimodal config unwrapping versus uninitialized rotary embedding memory after meta materialization." }, { "left": "issue:45542", "right": "issue:45800", "accept": false, - "reason": "Unrelated dependency/backend problems; there is no shared concrete failure mode." + "reason": "No shared code-path bug; one is a missing TensorFlow backend check, the other is a torch/transformers version incompatibility." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "Different failures and different paths: FSDP weight-shape error versus TRL input-embedding unpacking." + "reason": "Both mention training/loading, but they are distinct failures on different models and different paths." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Only a broad overlap around multimodal/training; the model families and actual bugs differ." + "reason": "Different model families and different multimodal/training failure modes." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "No clear shared bug beyond both being model-behavior complaints." + "reason": "No meaningful overlap beyond being model-specific inference complaints." }, { "left": "issue:45632", "right": "issue:45698", - "accept": false, - "reason": "The titles only share a loose custom-module/loading theme; they do not clearly identify the same bug or a specific cache-collision mechanism, so the duplicate claim is too speculative." + "accept": true, + "reason": "Both point to the same trust_remote_code/custom-module resolution problem: cache/path collision causing the wrong custom module to be loaded." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Different issues: Jupyter/custom-model initialization regression versus flash-attn2 fallback gating." + "reason": "Unrelated issues: Jupyter/custom model initialization versus flash-attn2 fallback gating." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Both mention tied weights, but they appear to affect different model classes and likely require different fixes." + "reason": "Both involve tied weights, but they affect different model classes and appear to be separate bugs rather than one mergeable fix." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "Quantization/FineGrainedFP8 failure and flash-attn2 fallback gating are unrelated." + "reason": "Different failures: FineGrainedFP8Config on Qwen3Moe versus fallback logic blocked when fa2 is absent." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantized offloading behavior is not the same bug as tied-weights regression." + "reason": "Quantization offloading behavior is unrelated to tied-weight issues." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache crash on Qwen3.5 is unrelated to tied-weight handling." + "reason": "PagedAttentionCache linear_attention crash is a different code path than tied-weight handling." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading failure is a different problem from translation-model weight tying." + "reason": "Checkpoint-loading failure for NemotronH is not the same bug as tied weights in translation models." }, { "left": "issue:45183", "right": "issue:46463", - "accept": true, - "reason": "Both are input-validation/XOR-check issues that produce wrong or misleading errors, and 46463 looks like a model-specific instance of the broader validation bug." + "accept": false, + "reason": "Both are validation/error-message bugs, but they target different validation logic and are not clearly the same underlying defect." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "Completely different topics: DSA indexer discussion versus docs about a removed command." + "reason": "Completely unrelated: DSA indexer discussion versus outdated docs about a removed command." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "Different model/trainer path versus an invalid-method-call bug." + "reason": "Different model-specific failures with no shared concrete code-path bug." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "Both are documentation-adjacent, but they are not the same issue." + "reason": "Both are documentation/text issues, but about different docs and different fixes." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Different models and different failures; no clear shared bug." + "reason": "Different models and different runtime symptoms; no evidence of the same bug." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Unrelated model execution failure versus invalid method call." + "reason": "Unrelated model execution failure versus invalid-method call in a different model implementation." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Both describe the same non-persistent-buffer corruption regression in Transformers v5, just phrased differently." + "reason": "Same regression: Transformers v5 corrupts/overwrites non-persistent buffers during load/materialization." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "They are both SigLIP2-related, but one is a docs/example mismatch plus quantization failure while the other is an output-quality regression." + "reason": "SigLIP2 is the shared subject, but one is a documentation example/model-processor mismatch and the other is a model-quality complaint." }, { "left": "issue:43644", "right": "issue:43950", "accept": true, - "reason": "These are the same non-persistent-buffer v5 regression, with 43644 as the broader symptom report and 43950 as the specific manifestation." + "reason": "Direct duplicate descriptions of the same non-persistent-buffer junk regression in Transformers v5." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "Different issues: custom trainer API mismatch versus ZeRO-3 loading index error." + "reason": "Different failure modes: custom trainer signature mismatch versus DeepSpeed ZeRO-3 empty-tensor indexing." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Both involve dtype/training instability, but the model, mechanism, and failure mode differ." + "reason": "Different models and different numeric-instability/training failures; not the same concrete bug." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Different quantization/storage compatibility failures." + "reason": "Both are quantization/storage-related TypeErrors, but they involve different missing incompatibilities and different code paths." }, { "left": "issue:43950", "right": "issue:45902", - "accept": true, - "reason": "Both indicate the same non-persistent-buffer initialization/materialization problem; 45902 is a specific manifestation." + "accept": false, + "reason": "Both involve buffers/initialization, but the concrete bug is different: buffer corruption on load versus uninitialized rotary embedding memory." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality and version-dependent tokenization are different issues." + "reason": "Too broad and not the same bug; one is SigLIP2 embedding quality, the other is tokenization changing across versions." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Checkpoint corruption and noisy load logging are unrelated." + "reason": "Unrelated issues: checkpoint corruption versus loading noise/doc-style behavior." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Same model family, but the reported symptoms are not specific enough to confirm the same bug." + "reason": "Same model family, but the symptoms are different enough that they do not clearly look like one underlying bug." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The cluster summary is mostly grounded, but one soft-edge acceptance is too speculative: issue:45632 vs issue:45698 is only supported by a loose shared theme around custom-module loading, not by a clear title-level indication of the same bug or a specific cache-collision mechanism. That makes the duplicate claim stronger than the packet justifies. The non-persistent-buffer grouping around issue:43950 is reasonable, but the 45632/45698 verdict breaks the requested conservative bar." + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The accepted duplicate pairs are plausible from the titles alone: the non-persistent-buffer regression set (43644/43950/44534), the Qwen3.5 video-input mismatch set (44479/46294), and the trust_remote_code/custom-module loading collision set (45632/45698). The non-duplicate rejections are also appropriately cautious. I did not see a clear overstatement that would require rejection." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1334,20 +1244,20 @@ "issue:43792|issue:44220", "issue:36331|issue:41093" ], - "prepared_review_unit_hash": "5e1de58b5c108359241082186f46e25c0a759056ddcf0f755125ec8933bf9481", + "prepared_review_unit_hash": "48b2be9ed773463c93ab5bdf3dfa7139b77a3d58c01f2187153f2a6a8bf9a887", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36830, + "serialized_chars": 36831, "estimated_input_tokens": 9208, "estimated_eval_tokens": 18672 }, @@ -1356,204 +1266,204 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous cluster with several unrelated model-loading/tokenizer/perf reports. I only accept a few clear duplicate subgroups: the removed-v5 pipeline docs issue, the `_is_hf_initialized`/`init_empty_weights` loading regression, and the SigLIP2 implementation discrepancy.", - "confidence": 0.79, - "canonical_issue_reason": "issue:44291 is the clearest root-cause report of the shared `_is_hf_initialized`/`init_empty_weights` regression and is the best anchor among the true duplicates in this cluster.", + "summary": "This cluster is mostly a grab bag of unrelated Transformer bugs and docs regressions. I only accepted a few clear duplicate pairs; most soft links are same-subsystem but not the same concrete defect.", + "confidence": 0.89, + "canonical_issue_reason": "No single cluster-wide canonical issue: the items split into separate subgroups. If forced to anchor one bug, issue:44291 is the clearest runtime regression report, and issue:44509 is the clearest docs regression report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44291 is the best overall issue representative because it precisely states the failure mode, version range, and concrete loading path, making it the most useful dedupe target.", + "best_issue_reason": "issue:44291 is the strongest standalone issue here: it states the root cause clearly, is reproducible, and cleanly covers the same unexpected `_is_hf_initialized` regression seen in issue:43872.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Both are model-specific loading failures, but Voxtral error messaging and Molmo tied-weights behavior are different bugs." + "reason": "Different failures: tokenizer loading error messaging vs missing `all_tied_weights_keys` on Molmo." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Same model family, but one is a Flash Attention packed-sequence crash and the other is CPU-offload incompatibility in rotary embedding; different code paths." + "reason": "Both are Qwen3.5-MoE related, but one is a flash-attn packed-sequence crash and the other is a CPU-offload rotary embedding bug." }, { "left": "issue:45663", "right": "issue:45902", "accept": false, - "reason": "Different models and different failures: FSDP2 shared_kv_states KeyError vs uninitialized inv_freq after meta->to_empty(cuda)." + "reason": "Distinct model/code paths: Gemma4 FSDP2 shared-kv rebuilding vs Qwen3_5Moe rotary-memory initialization." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Both are memory/training complaints, but gradient checkpointing under LoRA and int4 CUDA reserved-memory OOM are distinct problems." + "reason": "Different problems: LoRA gradient-checkpointing ineffectiveness vs int4 quantization memory growth/OOM." }, { "left": "issue:43054", "right": "issue:43493", "accept": true, - "reason": "Both point to the same SigLIP2 implementation mismatch/regression that makes HF outputs diverge from the original JAX behavior and hurts text embeddings." + "reason": "Both describe the same SigLIP2 implementation discrepancy; one reports worse embeddings, the other names the HF-vs-JAX mismatch." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Different model load failures with no evidence of the same root cause." + "reason": "Unrelated loading issues for different models; one is a bad error message, the other is a model load failure." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Generic loading noise versus a specific ScandiBERT tokenizer/model load failure; not the same bug." + "reason": "Generic loading-noise complaint vs a specific ScandiBERT load failure." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Detokenization regression and `flash_attn` import error are unrelated." + "reason": "Different areas: tokenizer detokenization regression vs Python 3.13 flash_attn import-key handling." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "One is a general loading-noise complaint; the other is a specific model load failure for cjvt/sleng-bert." + "reason": "Same broad loading area, but not the same bug; one is log noise, the other is a specific model load failure." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "Both mention memory, but they describe different Qwen MoE training issues and different symptoms." + "reason": "Both involve MoE training, but checkpointing behavior and memory usage are separate defects." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "`AutoConfig` missing fields with `trust_remote_code=True` is not the same as the broader remote-code breakage report." + "reason": "Both mention remote_code, but one is missing fields from AutoConfig and the other is a broader loading regression." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Gemma3 ZeRO-3 load failure and Qwen3Moe FP8 failure are unrelated model/config bugs." + "reason": "Different model-specific failures: DeepSpeed ZeRO-3 load crash vs FineGrainedFP8Config failure." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Different models and error surfaces; no shared concrete bug." + "reason": "Different models and symptoms; Qwen3VL weight-only loading is not the same as GLM5 loading." }, { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "MPS histogram kernel limitation is unrelated to Gemma4 device-map mismatch during forward." + "reason": "Different device/runtime issues: MPS histogram support vs Gemma4 device-map forward mismatch." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Importing Qwen2.5-VL and bitsandbytes `_is_hf_initialized` are different failure modes." + "reason": "Completely different bugs: missing Qwen2.5-VL import vs `_is_hf_initialized` constructor incompatibility." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "GLM-4V loading failure and Molmo missing `all_tied_weights_keys` are unrelated model bugs." + "reason": "Different model-loading failures for different code paths; no shared concrete defect." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer-loading regression and flash-attn2 fallback gating are unrelated." + "reason": "Tokenizer loading regression vs flash-attn2 fallback being blocked by other checks." }, { "left": "issue:45020", "right": "issue:45823", "accept": false, - "reason": "Remote-code model breakage and Gemma4 device placement are different issues." + "reason": "Remote-code loading regression and Gemma4 PLE device-mapping mismatch are unrelated." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "`add-new-model-like` tokenizer mapping bug is not the same as the broader remote-code regression." + "reason": "`add-new-model-like` tokenizer mapping bug and remote_code regression are different issues." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "LoRA compatibility across HF versions and a tied-embeddings warning are not the same underlying bug." + "reason": "Qwen3 fine-tuning warning and hf4.x\u2192hf5.x LoRA incompatibility are not the same underlying bug." }, { "left": "issue:43828", "right": "issue:45823", "accept": false, - "reason": "Autocast dtype mismatch in Phi MoE and Gemma4 device mismatch are unrelated." + "reason": "Autocast dtype mismatch and device-map mismatch are separate runtime errors." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Mistral tokenizer regex crash and flash-attn2 fallback blocking are different subsystems." + "reason": "Tokenizer regex patch crash vs flash-attn2 fallback gating are different code paths." }, { "left": "issue:45715", "right": "issue:46315", "accept": false, - "reason": "Both are tokenizer performance regressions, but they affect different APIs and likely different hot paths." + "reason": "Both are tokenizer performance regressions, but they hit opposite conversion APIs and are not the same bug." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "Broad remote-code breakage and Kimi-K2.5 tokenizer regression are not the same concrete bug." + "reason": "Both involve model loading/remote code, but the concrete failures are different." }, { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "MobileLLM loading failure and flash-attn2 fallback gating are unrelated." + "reason": "MobileLLM key error and flash-attn2 fallback gating are unrelated." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "Tied lm_head weights in Mistral3 and non-persistent buffer corruption are different code-path bugs." + "reason": "Tied-weight handling in Mistral3 and non-persistent buffer corruption are different regressions." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Both are the same v5 docs cleanup: stale references to removed pipeline tasks after pipeline removals." + "reason": "Same docs regression: both complain that v5 docs still reference removed pipeline-based summarization/translation tasks." }, { "left": "issue:43872", "right": "issue:44291", "accept": true, - "reason": "These describe the same `_is_hf_initialized` TypeError during model loading with `init_empty_weights` in Transformers 5.x." + "reason": "Same root bug: both report the unexpected `_is_hf_initialized` argument during model initialization/loading." }, { "left": "issue:42915", "right": "issue:43872", "accept": false, - "reason": "Qwen3Moe FP8 failure and the bitsandbytes `_is_hf_initialized` TypeError are unrelated." + "reason": "Qwen3MoE FP8 failure and bitsandbytes `_is_hf_initialized` incompatibility are separate regressions." }, { "left": "issue:43792", "right": "issue:44220", "accept": false, - "reason": "Both are Whisper-related, but one is a generic run failure and the other is a specific fbank feature-extraction issue; not enough evidence they are the same bug." + "reason": "Whisper run failure and `_torch_extract_fbank_features()` issue are related at most by subsystem, not the same concrete bug." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Completely different errors: custom trainer signature mismatch versus tensor-mask shape mismatch." + "reason": "Different training/runtime errors: unexpected `num_items_in_batch` kwarg vs mask/tensor shape mismatch." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays mostly conservative. The accepted duplicate groups are supported by the issue titles: the v5 docs/pipeline cleanup pair, the `_is_hf_initialized` / `init_empty_weights` loading regression pair, and the SigLIP2 discrepancy pair. The rejected soft-edge pairs are phrased cautiously and do not overclaim a match without stronger evidence." + "feedback": "The analysis is well grounded in the packet and stays conservative on the soft-edge calls. The accepted pairs are plausibly the same concrete defect or docs regression, while the rejected pairs are mostly clearly distinct by title and failure mode. The cluster summary also accurately reflects that this is a mixed bag with only a few true duplicates." }, "evaluator_used": true, "retried": false, @@ -1656,20 +1566,20 @@ "issue:44560|issue:44805", "issue:38175|issue:43994" ], - "prepared_review_unit_hash": "14db6cf623108edddb7731073f8bad63aaf7273151b41e1ac42ff6d3996d11a7", + "prepared_review_unit_hash": "51234b80f8339ba1dc47ba9b04e9538542602add32579869b6ce12d39fbec1f9", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37154, + "serialized_chars": 37155, "estimated_input_tokens": 9289, "estimated_eval_tokens": 18834 }, @@ -1678,181 +1588,202 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose set of unrelated Transformers bug reports; several are about model-loading regressions, but the soft pairs are not convincing duplicates on titles alone. No PRs are present. The broadest issue here is the generic model-loading regression under Transformers 5.x.", - "confidence": 0.41, - "canonical_issue_reason": "issue:43957 is the broadest and most representative of the shared subtheme: Transformers 5.x model-loading failures across multiple models and loading modes.", + "summary": "Mostly unrelated issues with superficial title overlap. The only clear duplicate-style pair is 43645/43646 about the Transformers 5.0 custom-model initialization regression. The SigLIP2 pair 38175/43994 is only a plausible same-model relation from the titles, not a confirmed duplicate on the packet alone.", + "confidence": 0.79, + "canonical_issue_reason": "43645 is the broader, more self-contained report of the Transformers 5.0 custom-model initialization regression; 43646 is the narrower follow-up on the same failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43957 best captures the cluster's dominant model-loading regression theme and is the most useful umbrella issue for triage.", + "best_issue_reason": "43645 is the strongest canonical issue candidate because it cleanly describes the bug and already subsumes the narrower duplicate report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "Both involve Qwen/DeepSpeed training, but one is weight loading under accelerate+deepspeed and the other is gradient averaging with model_accepts_loss_kwargs; different bugs." + "reason": "Both involve DeepSpeed, but one is about weight conversion and the other about gradient averaging; different bugs." }, { "left": "issue:42947", "right": "issue:45923", "accept": false, - "reason": "Both touch training features, but gradient checkpointing ineffectiveness and a missing supports_gradient_checkpointing flag are not the same failure." + "reason": "LoRA gradient-checkpointing behavior and a missing gradient-checkpointing flag are only loosely related." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading problems, but they affect different models and describe different symptoms/root causes." + "reason": "Different models and different failures: bad Voxtral tokenizer messaging vs tokenizer loading failure for est-roberta." }, { "left": "issue:43940", "right": "issue:45313", "accept": false, - "reason": "Both are ZeRO-3 weight-loading failures, but the affected model families and reported missing components differ; not enough to call them duplicates." + "reason": "Both are Qwen/ZeRO-3 weight-loading issues, but the titles point to different model generations and different failure descriptions." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "Both are model-loading failures, but they concern different architectures and checkpoints." + "reason": "Both are model-loading problems, but the packet does not show the same model or the same root cause." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "These describe unrelated bugs: RLHF NaNs from 3D position_ids versus an AttributeError in flash_attention.py for sink-less models." + "reason": "Qwen3.5 NaNs from SDPA/position_ids and a flash-attention AttributeError are distinct code-path failures." }, { "left": "issue:41093", "right": "issue:44805", "accept": false, - "reason": "The error text is similar, but the mask/tensor shape mismatch could arise from many distinct code paths; no clear same-root-cause evidence." + "reason": "Both are mask-shape IndexErrors, but there is no evidence they are the same model or same code path." }, { "left": "issue:44479", "right": "issue:44560", "accept": false, - "reason": "Both involve Qwen video/VL paths, but one is a broad video-input regression and the other is a specific embedding StopIteration; too different to merge confidently." + "reason": "A broad video-input regression and a specific StopIteration in Qwen3-vl-embedding are not the same bug." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Different models and different loading symptoms; only a vague tokenizer-loading overlap." + "reason": "Both concern loading, but the symptoms and affected models are different." }, { "left": "issue:44863", "right": "issue:46402", "accept": false, - "reason": "Both are checkpoint-loading issues, but they affect different models and likely different loader code paths." + "reason": "Different model families and different loading symptoms; not enough to treat as one bug." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "Completely unrelated: DeepSpeed training weight shape error versus image processor segmentation-map handling." + "reason": "Unrelated training/runtime error versus image-processor segmentation-map failure." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "PagedAttentionCache/linear_attention crash and checkpoint-loading failure are unrelated." + "reason": "PagedAttentionCache linear_attention crash is unrelated to NemotronH checkpoint loading." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "No meaningful overlap beyond both being model-specific bugs; architecture support and autocast dtype mismatch are separate issues." + "reason": "GGUF gpt-oss support and a Phi-tiny-MoE dtype mismatch are unrelated changes." }, { "left": "issue:45325", "right": "issue:45902", "accept": false, - "reason": "Different subsystems entirely: Qwen2.5-VL rope indexing versus Qwen3.5-MoE rotary embedding materialization." + "reason": "Different bugs: rope-index scaling for Qwen2.5-VL versus uninitialized-memory after meta-to_empty materialization." }, { "left": "issue:45663", "right": "issue:45823", "accept": false, - "reason": "Same model family, but one is an FSDP2 shared-state KeyError and the other is a device-map forward mismatch; distinct bugs." + "reason": "Gemma4 FSDP2 shared_kv_states and Gemma4 PLE device mismatch are separate issues." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Unrelated Qwen3-MoE sliding-window bug versus gpt-oss GGUF support request." + "reason": "gpt-oss GGUF support and Qwen3-MoE sliding_window behavior do not match." }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "AudioFlamingo batching leak and Whisper empty-transcription generation failure are different code paths." + "reason": "AudioFlamingo3 batched leakage and Whisper empty-transcription handling are unrelated." }, { "left": "issue:43716", "right": "issue:44560", "accept": false, - "reason": "Mistral-3 image preprocessor dtype mismatch is unrelated to Qwen3-VL video input StopIteration." + "reason": "Mistral-3 dtype mismatch and Qwen3-vl video StopIteration are different code paths." }, { "left": "issue:43950", "right": "issue:45834", "accept": false, - "reason": "A from_pretrained buffer regression and a long-ocr index error are unrelated." + "reason": "Non-persistent buffer corruption is unrelated to Kosmos2.5 OCR index errors." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "ImportError for a missing exported class is not the same as meta-device model-loading breakage." + "reason": "Missing import of Qwen2_5_VLForConditionalGeneration is not the same as meta-device loading regressions." }, { "left": "issue:44560", "right": "issue:45685", "accept": false, - "reason": "Video-input StopIteration and an MPS histogram not implemented error are unrelated." + "reason": "Video-input regression and MPS histogram support are unrelated." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "GGUF architecture support and Float8 storage lookup failure are different bugs." + "reason": "gpt-oss support and missing Float8 storage are unrelated." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "FineGrainedFP8Config failure is a separate Qwen3-MoE quantization issue, not the generic meta-device loading regression." + "reason": "GenerationMixin import failure and meta-device loading breakage are different issues." }, { "left": "issue:43645", "right": "issue:43646", - "accept": false, - "reason": "Very close, but one is notebook-specific custom-model initialization while the other is a broader custom-model initialization regression; not enough evidence they are the same bug." + "accept": true, + "reason": "Same Transformers 5.0 custom-model initialization regression; 43646 is the narrower duplicate report." }, { "left": "issue:43550", "right": "issue:44910", "accept": false, - "reason": "torch.compile/SDPA failure on Bamba and Flash Attention illegal memory access on Qwen3.5 are distinct model/runtime issues." + "reason": "Different failures: torch.compile/SDPA on Bamba versus flash-attention illegal-memory-access on Qwen3.5." }, { "left": "issue:44560", - "right": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Both mention Qwen3.5/video/embedding themes, but the concrete failures are different." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py execution failure is unrelated to gpt-oss GGUF support." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "GenerationMixin import error and FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:44661", + "right": "issue:45823", "accept": false, - "reason": "Both mention index-type failures only superficially; the video StopIteration and mask/tensor shape mismatch are not the same bug." + "reason": "TOKENIZER_MAPPING_NAMES/add-new-model-like bug is unrelated to Gemma4 PLE device mismatch." }, { "left": "issue:38175", "right": "issue:43994", "accept": false, - "reason": "Both concern SigLIP2 outputs, but one is zero probabilities while the other is nonsensical AutoModel/pipeline results; not a clear duplicate." + "reason": "Same model family, but the packet only shows related symptoms (zero probabilities vs nonsensical results), not enough to confirm the same underlying bug." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, + "evaluator_result": { + "accept": false, + "feedback": "The review is mostly cautious, but it overstates evidence in a couple of places. In particular, treating issue:38175 / issue:43994 as a true duplicate-style pair is not well grounded from the packet alone: the titles indicate the same model family, but one is about zero probabilities and the other about nonsensical results, so this is at best a plausible relation, not a confirmed duplicate. The summary also says it found \"two true duplicate-style pairs,\" which is stronger than the available evidence supports. The 43645 / 43646 acceptance is plausible, but the overall conclusion should stay more conservative unless the issue bodies clearly establish duplication." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -1929,41 +1860,41 @@ "issue:42491|issue:43720", "issue:41720|issue:42915", "issue:44960|issue:44977", - "issue:44451|issue:44488", "issue:42915|issue:44756", + "issue:44451|issue:44488", "issue:45313|issue:45357", "issue:42491|issue:45276", "issue:45507|issue:46468", "issue:44488|issue:44991", "issue:43708|issue:44075", - "issue:39692|issue:43994", "issue:43644|issue:43645", + "issue:39692|issue:43994", "issue:45488|issue:45812", "issue:43856|issue:43873", "issue:43854|issue:43931", - "issue:43257|issue:43854", "issue:43856|issue:44387", + "issue:43257|issue:43854", "issue:45362|issue:45406", - "issue:42491|issue:45292", "issue:44960|issue:46050", + "issue:42491|issue:45292", "issue:42491|issue:43931", - "issue:45276|issue:45335", - "issue:44164|issue:44291" + "issue:44164|issue:44291", + "issue:45276|issue:45335" ], - "prepared_review_unit_hash": "2c036efdb8d8db50021061e50bb62599099d3b64da5f5fc18c45a280747eb914", + "prepared_review_unit_hash": "611b492fb3c971fef32030b0cb10f6e9fcb0275314ffc1a9ce89054eae79a71e", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36226, + "serialized_chars": 36225, "estimated_input_tokens": 9057, "estimated_eval_tokens": 18370 }, @@ -1972,174 +1903,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is still mostly heterogeneous, with a few small duplicate pockets. The clearest one is the Transformers v5 non-persistent-buffer regression around 43644/44534, which looks like the same bug reported in near-exactly similar terms; there is also a likely duplicate pocket around `resize_token_embeddings` propagation.", - "confidence": 0.77, - "canonical_issue_reason": "43644 is the cleanest canonical anchor because it sits at the center of the clearest duplicate pocket: 44534 describes the same Transformers v5 non-persistent-buffer junk regression in near-exactly similar terms.", + "summary": "Most pairs look distinct from the metadata alone. The strongest duplicate candidate is 43644/44534: they share the same core wording about Transformers v5 filling non-persistent buffers with junk, so they likely describe the same bug, but the packet does not prove it and should not be treated as confirmed. The rest appear to be separate issues across different models or code paths.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 43644 is the best canonical candidate because it is the clearest, most general title for the non-persistent-buffer junk regression, and 44534 closely mirrors that wording. Based on titles alone, this looks like a probable duplicate pair rather than a confirmed one.", "canonical_pr_reason": null, - "best_issue_reason": "43644 is the strongest representative issue in this set; it anchors the clearest duplicate pocket and is more central than the more model-specific one-off reports.", + "best_issue_reason": "Issue 43644 is the best representative issue overall because it most directly names the shared regression in a model-agnostic way and is the strongest metadata match for 44534.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43632", "right": "issue:45834", "accept": false, - "reason": "Different bugs: `_is_hf_initialized` flag regression vs a Kosmos2.5 long-input index error." + "reason": "Different bugs: HF init-flag regression vs. a Kosmos2.5 OCR mask/index error." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Different failure modes: ZeRO-3 index error on BertModel vs `add-new-model-like`/`TOKENIZER_MAPPING_NAMES` logic." + "reason": "Both involve loading/training machinery, but one is a ZeRO-3 BertModel indexing failure and the other is tokenizer-mapping/model-registration logic." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "Unrelated: zero3/BertModel index error vs tokenizer-mapping model registration bug." + "reason": "Different concrete failures: ZeRO-3/BertModel index error vs. a tokenizer-mapping constraint in model registration." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Both are config-validation issues, but for different models and different checks; not the same underlying bug." + "reason": "Both are config-related, but they affect different models and the failures are not the same code path." }, { "left": "issue:44661", "right": "issue:44805", "accept": false, - "reason": "Model-registration/tokenizer-mapping issue vs a mask-shape IndexError in a different code path." + "reason": "Tokenizer-mapping registration bug vs. a tensor/mask shape mismatch at inference time." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "Different problems: dequantized loading with device_map offload vs Strix Halo mmap OOM mitigation." + "reason": "Both mention loading/offload or memory, but the specific bugs are different: quantized model offload loading vs. mmap-driven OOM on Strix Halo." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "Same model family, but one is zero probabilities at inference and the other is docs-example/model-processor mismatch plus quantization failure." + "reason": "SigLIP2 output anomaly vs. documentation example errors; not the same underlying bug." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Same regression: both describe Transformers v5 filling non-persistent buffers with junk; the titles are near-exact variants of the same bug report." + "reason": "Very strong metadata match: the titles are near-paraphrases of the same Transformers v5 non-persistent-buffer junk regression, so this is the only likely duplicate pair in the packet." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Different issues: Qwen3 MoE LoRA compatibility across versions vs BitNet packed-weights not unpacked during accelerate loading." + "reason": "Different underlying problems: Qwen3 MoE LoRA compatibility vs. BitNet packed-weight loading with accelerate." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Both involve Qwen3 family, but one is a device-mapping cuda assert and the other is a FineGrainedFP8Config failure." + "reason": "Qwen3 auto device-map CUDA assert is unrelated to the Qwen3Moe FineGrainedFP8Config failure." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "Different models and code paths: GLM5 vs Qwen3.5 flash-attention generation." + "reason": "Different models and different symptoms: GLM5 issue vs. Qwen3.5 flash-attention generation problem." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "FP8 config failure and mmap/OOM behavior are distinct bugs." }, { "left": "issue:44451", "right": "issue:44488", "accept": false, - "reason": "Both are model-loading regressions, but they are for different model repos and there is no clear evidence of the same underlying bug." + "reason": "Both are model-loading failures, but the titles do not indicate the same concrete bug or shared code path." }, { "left": "issue:43708", "right": "issue:44075", "accept": false, - "reason": "Trainer checkpoint-resume step counting vs SGD optimizer arguments are unrelated." + "reason": "Trainer resume/max_steps math bug vs. optimizer SGD arguments not being used." }, { - "left": "issue:39692", - "right": "issue:43994", + "left": "issue:43644", + "right": "issue:43645", "accept": false, - "reason": "Both are SigLIP2-related, but one is docs/example breakage and the other is bad outputs from AutoModel/pipeline." + "reason": "Buffer initialization regression vs. Jupyter custom-model initialization issue; different failure modes." }, { - "left": "issue:43644", - "right": "issue:43645", + "left": "issue:39692", + "right": "issue:43994", "accept": false, - "reason": "Buffer-junk regression vs Jupyter custom-model initialization problem; different symptoms and code paths." + "reason": "Docs example errors vs. wrong outputs from AutoModel/pipeline; similar model family, different bug." }, { "left": "issue:45488", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions in v5, but they affect different model families and different tokenizer mechanisms." + "reason": "Different tokenizer regressions: ByteLevel/metaspace override in DeepSeek vs. Granite token-ID mismatch." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Different issues: Qwen3 MoE training memory inefficiency vs quantization offloading not working as expected." + "reason": "Training-memory inefficiency vs. quantization/offloading behavior; broad memory theme only." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Different loading failures for different model families; no shared concrete bug." + "reason": "Different load failures and different model families; no clear shared fix." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Accelerate+DeepSpeed Qwen3 MoE weight conversion issue vs GLM-4.7-Flash unit-test loading issue." + "reason": "MoE weight conversion with accelerate+deepspeed is unrelated to GLM-4.7-Flash test loading." }, { - "left": "issue:43856", - "right": "issue:44387", + "left": "issue:45362", + "right": "issue:45406", "accept": false, - "reason": "Both are memory-related, but one is training inefficiency and the other is int4 quantization causing OOM." + "reason": "Qwen3.5 chat crash and Gemma4 processor attribute error are distinct failures." }, { - "left": "issue:45362", - "right": "issue:45406", + "left": "issue:44960", + "right": "issue:46050", "accept": false, - "reason": "Qwen3.5 chat crash vs Gemma4Processor missing `_tokenizer`; unrelated." + "reason": "GLM5 issue vs. quantized save/load regression after sinq; different affected paths." }, { - "left": "issue:42491", - "right": "issue:45276", + "left": "issue:44488", + "right": "issue:44991", "accept": false, - "reason": "Qwen3 MoE LoRA/version-compatibility issue vs Gemma4 `resize_token_embeddings` behavior." + "reason": "Both are tokenizer-loading failures, but for different models and without evidence of a shared concrete bug." }, { - "left": "issue:44960", - "right": "issue:46050", + "left": "issue:44164", + "right": "issue:44291", "accept": false, - "reason": "GLM5 issue vs sinq quantization save/load breakage; different bugs." + "reason": "save/from_pretrained extra_state handling and init_empty_weights _is_hf_initialized are separate breakages." }, { - "left": "issue:42491", - "right": "issue:43931", + "left": "issue:43856", + "right": "issue:44387", "accept": false, - "reason": "Different model families and failure modes: LoRA/version compatibility vs Qwen3-VL shape mismatch." + "reason": "Training-memory issue vs. int4 quantization OOM; related theme only, not the same bug." }, { "left": "issue:45276", "right": "issue:45335", - "accept": true, - "reason": "Same underlying `resize_token_embeddings` regression: the resized embeddings are not propagated to model-specific output/decoder embeddings." + "accept": false, + "reason": "Both involve resize_token_embeddings, but on different architectures and different embedding/output wiring." }, { - "left": "issue:44164", - "right": "issue:44291", + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen3.5 regressions: ZeRO-3 language_model weight loading vs. incorrect visual encoder keys on save_pretrained." + }, + { + "left": "issue:45507", + "right": "issue:46468", "accept": false, - "reason": "Both involve save/load paths, but extra_state handling and `_is_hf_initialized` argument errors are different bugs." + "reason": "GraniteMoEHybrid invalid-method bug and Nemotron example breakage are unrelated." } ] }, "evaluator_result": { "accept": false, - "feedback": "The cluster-level read is mostly grounded and the soft-edge verdicts are conservative, but one key claim is a bit too strong for the packet: `43644`/`44534` look like a near-exact duplicate pair, not an exact title-level duplicate. Please soften that wording (e.g. \u201cnear-exact\u201d or \u201csame regression\u201d) and avoid implying stronger evidence than the titles alone show." + "feedback": "The cluster summary is a bit too strong for the packet. The 43644/44534 pair is the strongest candidate because they share the same title and high similarity, but the packet does not confirm it as a duplicate, so phrases like \u201cconfirmed duplicate,\u201d \u201cone true duplicate pair,\u201d and \u201calmost verbatim\u201d overstate the evidence. The rest of the soft-edge rejections are generally conservative enough, but the canonical/best-issue rationale should be downgraded to probabilistic language grounded only in the metadata." }, "evaluator_used": true, "retried": true, @@ -2228,13 +2171,13 @@ "issue:43957|issue:44756", "issue:43335|issue:45663", "issue:42947|issue:43381", - "issue:45276|issue:45292", "issue:45685|issue:46468", + "issue:45276|issue:45292", "issue:44863|issue:46468", "issue:44164|issue:44756", "issue:43854|issue:45216", - "issue:43381|issue:44928", "issue:44479|issue:44877", + "issue:43381|issue:44928", "issue:45684|issue:45698", "issue:45588|issue:45663", "issue:43646|issue:43708", @@ -2243,20 +2186,20 @@ "issue:45812|issue:46315", "issue:43828|issue:43957" ], - "prepared_review_unit_hash": "98ce9e45e6c4d6bce3381adec99bef82482259bf6d44ca88b005fe7a2c3c24e0", + "prepared_review_unit_hash": "acfff889b2fab1228730fd3224488a66852ca145a66055d79a2df27beb3363f2", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37466, + "serialized_chars": 37467, "estimated_input_tokens": 9367, "estimated_eval_tokens": 18990 }, @@ -2265,210 +2208,210 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab bag of unrelated model-specific bugs. Only one soft edge looks like a true duplicate: the resize_token_embeddings regression around Gemma4/output embeddings. No PRs are present.", - "confidence": 0.94, - "canonical_issue_reason": "issue:45292 is the cleanest canonical issue: it states the underlying resize_token_embeddings/output_embeddings regression in a general form, while issue:45276 is a model-specific manifestation of the same bug.", + "summary": "Mostly heterogeneous Transformers bug reports, with a couple of tight duplicate families: resize_token_embeddings not propagating to model-specific embeddings, and untied lm_head weights in multimodal generation models. I\u2019d use the generic resize_token_embeddings report as the main canonical issue.", + "confidence": 0.62, + "canonical_issue_reason": "issue:45292 is the broadest, most model-agnostic report in the resize_token_embeddings family and cleanly subsumes the Gemma4/T5Gemma variants.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45292 is the best representative because it is the broadest, clearest description of the only confirmed duplicate pair in the set.", + "best_issue_reason": "issue:45292 is the best umbrella issue because its title captures the shared core bug without being model-specific, making it the clearest duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Both are flash-attention crashes, but one is a Qwen3.5 packed-sequence/position_ids bug and the other is an s_aux=None AttributeError in sink-less models; different code paths." + "reason": "Both involve flash attention, but the failures are different: packed-sequence/position_ids misinterpretation vs s_aux=None handling on sink-less models." }, { "left": "issue:43404", "right": "issue:43454", - "accept": false, - "reason": "Both mention untied lm_head weights, but they affect different model implementations and likely different tying logic; not the same concrete bug." + "accept": true, + "reason": "Both report untied lm_head weights causing bad generation in multimodal model classes; this looks like the same missing weight-tying bug." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Qwen3 device-mapping CUDA assert and AudioFlamingo3 batched inference leakage are unrelated failures." + "reason": "Different symptoms and code paths: device-mapping CUDA assert on Qwen3 vs batched-inference token/embedding leakage in AudioFlamingo3." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Torch version incompatibility and meta-device loading breakage are different compatibility issues." + "reason": "One is a Torch 2.10 compatibility report, the other is a meta-device model-loading regression; not the same bug." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "A SigLIP2 docs example error and a GLM5 issue are unrelated." + "reason": "Doc-example errors for SigLIP2 and a separate GLM5 issue are unrelated." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Both are model-loading complaints, but the titles point to different models and likely different root causes." + "reason": "Different problems: tokenizer vocab-size mismatch vs a model-loading failure for cjvt/sleng-bert." }, { "left": "issue:42947", "right": "issue:45704", "accept": false, - "reason": "Gradient checkpointing with PEFT LoRA and a T5 Apex RMSNorm memory leak are unrelated." + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the T5/apex RMSNorm memory leak." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Both involve quantization/memory, but one is offloading behavior and the other is increased reserved memory/OOM under int4; not the same bug." + "reason": "Both mention quantization and memory, but one is offloading behavior and the other is reserved-memory/OOM under int4; not the same concrete bug." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "An import error for GenerationMixin is unrelated to Whisper return_language pipeline behavior." + "reason": "ImportError for GenerationMixin and Whisper pipeline return_language breakage are separate issues." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Gemma4 FSDP2 shared_kv_states and Qwen multimodal text_config unwrapping are different code paths." + "reason": "Gemma4 FSDP2 shared_kv_states KeyError and Qwen multimodal text_config unwrapping are different model-init problems." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "init_empty_weights argument handling and Strix Halo mmap OOM are unrelated." + "reason": "Unexpected _is_hf_initialized in init_empty_weights is unrelated to the Strix Halo mmap OOM report." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Gemma4 FSDP2 attention state rebuilding and Qwen3_5Moe CPU-offload incompatibility are different bugs." + "reason": "Different bugs: Gemma4 FSDP2 attention state rebuild vs Qwen3_5Moe rotary embedding CPU-offload incompatibility." }, { "left": "issue:44155", "right": "issue:45823", "accept": false, - "reason": "AudioFlamingo3 token leakage and Gemma4 device mismatch during auto device mapping are not the same issue." + "reason": "AudioFlamingo3 token leakage is not the same as Gemma4 PLE device mismatch with device_map=auto." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Int4 quantization OOM and mmap disablement for Strix Halo are unrelated." + "reason": "One is int4 quantization memory regression; the other is mmap/OOM on Strix Halo." }, { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Both are quantized-model loading problems, but they concern different models/formats and different failure modes." + "reason": "Both are loading/quantization issues, but they affect different models and different failure modes (load error vs fallback to bf16)." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading breakage and Strix Halo mmap OOM are separate issues." + "reason": "Meta-device loading regression and Strix Halo mmap/OOM are not the same underlying bug." }, { "left": "issue:43335", "right": "issue:45663", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and Gemma4 FSDP2 attention state bugs are unrelated." + "reason": "SwitchTransformers sparse-layer construction and Gemma4 FSDP2 attention state are unrelated." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Both mention gradient checkpointing, but one is ineffective LoRA checkpointing and the other is an eval-mode restriction; different problems." - }, - { - "left": "issue:45276", - "right": "issue:45292", - "accept": true, - "reason": "Same underlying resize_token_embeddings regression: the Gemma4-specific report and the generic output_embeddings report describe the same failure to propagate resized embeddings." + "reason": "Both mention gradient checkpointing, but one is ineffective LoRA checkpointing and the other is an eval-mode restriction." }, { "left": "issue:45685", "right": "issue:46468", "accept": false, - "reason": "A missing MPS histogram implementation and a broken Nemotron example are not the same bug." + "reason": "MPS int histogram limitation in MoE and the Nemotron example failure are different issues." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same resize_token_embeddings bug: the Gemma4-specific report is a narrower manifestation of the general output_embeddings propagation failure." }, { "left": "issue:44863", "right": "issue:46468", "accept": false, - "reason": "NemotronH checkpoint loading and a broken official example may be related in theme, but the titles do not indicate the same concrete code-path problem." + "reason": "NemotronH checkpoint-loading failure and a broken official Nemotron example are too broad to treat as the same concrete bug." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "save/from_pretrained extra_state handling is unrelated to Strix Halo mmap OOM." + "reason": "save/from_pretrained extra_state handling is unrelated to disabling mmap to avoid OOM." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "Unit-test model loading for GLM-4.7-Flash and Qwen3.5 save_pretrained regression are different issues." + "reason": "GLM-4.7-Flash load failure in tests and Qwen3.5 save_pretrained regression are separate model issues." }, { - "left": "issue:43381", - "right": "issue:44928", + "left": "issue:44479", + "right": "issue:44877", "accept": false, - "reason": "Eval-mode gradient checkpointing limitation and RLHF NaN explosion from Qwen3.5 position_ids are unrelated." + "reason": "Video-input regression for Qwen VL models and strict granite_speech config loading are unrelated." }, { - "left": "issue:44479", - "right": "issue:44877", + "left": "issue:43381", + "right": "issue:44928", "accept": false, - "reason": "A video-input regression and a strict-config loading failure are unrelated." + "reason": "Gradient checkpointing eval-mode prohibition is not the same as the Qwen3.5 RLHF NaN/SDPA fallback regression." }, { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Both involve custom-model save/load flows, but one is read-only permission propagation and the other is wrong module loading; not the same bug." + "reason": "Read-only permission propagation from save_pretrained and wrong custom-module loading after save_pretrained are different bugs." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Flash-attention s_aux handling and Gemma4 shared_kv_states rebuilding are distinct code paths." + "reason": "Flash-attention s_aux=None handling and Gemma4 FSDP2 shared_kv_states rebuilding are unrelated." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model initialization in Transformers 5.0.0 and Trainer max_steps recalculation on resume are unrelated." + "reason": "Custom model initialization in Transformers 5.0 and Trainer max_steps calculation on resume are different regression classes." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Notebook custom model initialization breakage and save/from_pretrained extra_state handling are different issues." + "reason": "Notebook-specific custom-model init breakage is not the same as save/from_pretrained extra_state handling." }, { "left": "issue:45292", "right": "issue:45335", - "accept": false, - "reason": "Both are resize_token_embeddings regressions, but the affected submodules and model families differ; not enough evidence they are the same code-path bug." + "accept": true, + "reason": "Same resize_token_embeddings propagation bug; t5gemma.decoder.embed_tokens is another model-specific target not being updated." }, { "left": "issue:45812", "right": "issue:46315", "accept": false, - "reason": "Wrong token IDs in Granite models and slower convert_tokens_to_ids in v5 are different tokenizer regressions." + "reason": "Wrong token IDs for Granite models and slow convert_tokens_to_ids with many added tokens are separate tokenizer problems." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "An autocast dtype mismatch in Phi-tiny-MoE is unrelated to meta-device loading failures." + "reason": "Autocast dtype mismatch on Phi-tiny-MoE and meta-device loading regressions are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analysis is grounded in the packet and stays conservative on the soft edges. The only accepted duplicate pair, issue:45276 vs issue:45292, is well supported by the titles and similarity, and the canonical/best-issue choices are reasonable. The note that no PRs are present is also consistent with the packet." + "feedback": "The summary stays grounded in the packet and the accepted soft pairs are conservative. The resize_token_embeddings family is the clearest canonical cluster, and the untied lm_head pair is also reasonably supported by the titles. I don\u2019t see an overstatement that would require rejection." }, "evaluator_used": true, "retried": false, @@ -2554,9 +2497,9 @@ "issue:44479|issue:44805", "issue:44589|issue:45084", "issue:43906|issue:46326", - "issue:43901|issue:44509", "issue:46050|issue:46326", "issue:43425|issue:44918", + "issue:43901|issue:44509", "issue:43994|issue:44079", "issue:36032|issue:43452", "issue:43299|issue:43931", @@ -2575,14 +2518,14 @@ "issue:43611|issue:43644", "issue:43582|issue:43632" ], - "prepared_review_unit_hash": "42de33a65d7260384164a5dc946d1e9afec9871c6f46ef9a3b7d7ef3874722a0", + "prepared_review_unit_hash": "e19ee05a3bcef4c609d105c04a88556d70a96a5d0512302213d7efd3da323f55", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, @@ -2597,222 +2540,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a heterogeneous mix of Transformers v5 regression reports, mostly around tokenizers, model loading, dtype/device mismatches, and a few docs/runtime issues. None of the soft pairs look like true duplicates of the same underlying bug or change.", - "confidence": 0.91, - "canonical_issue_reason": "If a single anchor is needed, \"transformers version changes the tokenization\" is the broadest version-regression report in the set, but it does not really subsume the many unrelated loading/runtime bugs around it.", + "summary": "This cluster is mostly a loose set of unrelated Transformers regression reports, many around v5/model-loading/tokenizer behavior but not true duplicates. No PRs are present; among the issues, the broadest and clearest representative is the base_model_prefix loading regression.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43611 is the most general, user-facing loading regression in the set and best matches the cluster\u2019s broad v5 breakage theme.", "canonical_pr_reason": null, - "best_issue_reason": "\"Transformers 5.0.0 breaks loading models with the `base_model_prefix` attribute\" is the cleanest concrete regression report and a reasonable representative issue, though still only one narrow slice of the cluster.", + "best_issue_reason": "issue:43611 is the strongest cluster representative: concrete, widely applicable, and closer to the shared model-loading regression pattern than the more model-specific tickets.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Both are tokenizer-related, but one is a vocab-size mismatch for a specific model and the other is a separate tokenizer loading failure for ScandiBERT." + "reason": "Both are tokenizer/model-loading complaints, but they involve different models and different failure modes; not the same bug." }, { "left": "issue:43425", "right": "issue:43828", "accept": false, - "reason": "Torch compatibility vs. a Phi-tiny-MoE autocast dtype mismatch; different failure modes and code paths." + "reason": "One is a general torch version incompatibility report, the other a Phi MoE autocast dtype mismatch; distinct problems." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Float8 storage lookup error is unrelated to the Strix Halo mmap/OOM workaround." + "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are unrelated runtime issues." }, { "left": "issue:44683", "right": "issue:45823", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 is a different bug from Gemma4 PLE device mismatch during forward." + "reason": "Torch 2.9 flex_attention compilation failure is not the same as Gemma4 PLE device-map mismatch." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper return_language pipeline regression is distinct from the broad tokenization-change report." + "reason": "Whisper return_language regression and a general tokenization change report are different behaviors and code paths." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "Base_model_prefix loading failure and Jupyter custom-model initialization are separate model-loading regressions." + "reason": "Model loading via base_model_prefix and custom model initialization in Jupyter are separate regressions." }, { "left": "issue:43425", "right": "issue:45823", "accept": false, - "reason": "Torch version incompatibility does not match the Gemma4 device_map=\"auto\" forward mismatch." + "reason": "Version incompatibility versus a Gemma4 forward device mismatch; not one underlying bug." }, { "left": "issue:43425", "right": "issue:43716", "accept": false, - "reason": "General torch incompatibility is unrelated to a multimodal image preprocessor/model dtype mismatch." + "reason": "These are both compatibility-style reports, but the affected paths and symptoms are different." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "PagedAttentionCache linear_attention crash and template-node compilation failure are different subsystems." + "reason": "PagedAttentionCache/Qwen3.5 crash is unrelated to the template-node compilation error." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "An isolated reproduction of another issue is not the same bug as the PagedAttentionCache crash." + "reason": "An isolated reproduction of another issue is not the same as the PagedAttentionCache linear_attention crash." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "torch.export with torch_compilable_check is a different path from Perceiver interpolation failing at non-default resolution." + "reason": "torch.export failures for torch_compilable_check and Perceiver non-default resolution failure are different bugs." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline behavior and MXFP4 quantization fallback are unrelated." + "reason": "Whisper pipeline regression and MXFP4 quantization fallback are unrelated." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "Qwen VL video input regression is not the same as a generic mask shape mismatch." + "reason": "A video-input regression across Qwen VL variants does not match the generic mask/index shape mismatch." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Float8 storage lookup failure and template compilation error have no common underlying bug." + "reason": "Missing Float8 storage and non-template-node compilation are unrelated." }, { "left": "issue:43906", "right": "issue:46326", "accept": false, - "reason": "Isolated reproduction of issue 38071 and mistral_common continue_final_message are different tokenizer behaviors." - }, - { - "left": "issue:43901", - "right": "issue:44509", - "accept": false, - "reason": "Both are docs-related, but they cover different removed/changed pipeline APIs." + "reason": "Reproduction of issue 38071 and mistral_common continue_final_message behavior are different tokenizer concerns." }, { "left": "issue:46050", "right": "issue:46326", "accept": false, - "reason": "Quantized save/load corruption and tokenizer continue_final_message are separate features." + "reason": "Quantized save/load breakage and a mistral_common message-format bug are not the same issue." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch 2.10 incompatibility is unrelated to TRL SFT embedding unpacking." + "reason": "Torch version incompatibility and Qwen3.5 SFT input-unpacking failure are distinct problems." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs-related, but they document different deprecated/changed pipeline behaviors." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "SigLIP2 bad outputs and ModelOutput key assignment with None are different bugs." + "reason": "SigLIP2 incorrect outputs and ModelOutput key assignment are separate functional bugs." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "T5 special-tokens method conflict and gguf_file from_pretrained breakage are unrelated tokenizer/model-loading failures." + "reason": "T5 tokenizer special-token conflict and gguf_file loading failures affect different code paths." }, { "left": "issue:43299", "right": "issue:43931", "accept": false, - "reason": "Both involve Qwen3-VL, but one is a loader break for MoE models and the other is a weights-shape mismatch." + "reason": "Both mention Qwen3-VL loading, but one is MoE/dev-version breakage and the other is a weight-shape mismatch; not the same root cause." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "CamemBERT MLM prediction regression is not the same as tokenizer loading failure for est-roberta." + "reason": "CamemBERT masked-LM prediction regression and EMBEDDIA tokenizer loading failure are unrelated." }, { "left": "issue:43596", "right": "issue:44683", "accept": false, - "reason": "DeepSpeed ZeRO3 init IndexError and compiled flex_attention torch>=2.9 failure are distinct problems." + "reason": "DeepSpeed zero3/BertModel index error is not the compiled flex_attention torch-version bug." }, { "left": "issue:44164", "right": "issue:45759", "accept": false, - "reason": "extra_state save/load handling and text_config unwrapping in from_config are separate model serialization bugs." + "reason": "extra_state save/load handling and composite Qwen config unwrapping are different serialization/configuration issues." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Tokenizer loading failure for est-roberta is unrelated to the Mistral regex patch crash." + "reason": "Two tokenizer-loading problems, but with different models and different failure mechanisms; do not dedupe." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "Qwen3 MoE memory inefficiency and disabling mmap on Strix Halo are different performance/memory issues." + "reason": "Qwen3 MoE training memory inefficiency and Strix Halo mmap OOM are not the same underlying problem." }, { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper pipeline regression and CamemBERT masked LM predictions are different model behaviors." + "reason": "Whisper pipeline behavior and CamemBERT masked-LM predictions are unrelated model-specific regressions." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "extra_state serialization and video input regression are not the same code-path." + "reason": "Serialization extra_state handling and Qwen VL video-input regression do not match." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "sleng-bert loading failure and Mistral regex patch crash are separate tokenizer issues." + "reason": "Loading a specific BERT model and the Mistral regex-patch tokenizer crash are different bugs." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "Tokenizer mapping-name failure and compiled flex_attention failure are unrelated." + "reason": "A model-listing/add-new-model-like failure is unrelated to the flex_attention compilation regression." }, { "left": "issue:44737", "right": "issue:45859", "accept": false, - "reason": "XLNet CPU arange placement bug and Qwen3_5Moe rotary embedding CPU-offload incompatibility are different device-placement issues." + "reason": "Both are device-placement related, but one is an XLNet CPU-arange inefficiency and the other is CPU-offload incompatibility in Qwen3_5Moe; not the same fix." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Two different model-loading regressions: sleng-bert vs physical-intelligence/fast." + "reason": "Two separate model-loading failures with different models and symptoms." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer _gpu_broadcast_one NoneType error is unrelated to Whisper return_language pipeline behavior." + "reason": "Trainer GPU broadcast NoneType error and Whisper return_language regression are unrelated." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Base_model_prefix loading and non-persistent buffer junk initialization are distinct v5 regressions." + "reason": "base_model_prefix loading breakage and non-persistent buffer junk-fill are distinct v5 regressions." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "Apple Silicon warmup TypeError and _is_hf_initialized flag breakage are different failures." + "reason": "AppleSilicon allocator warmup TypeError and the _is_hf_initialized flag breakage are different issues." } ] }, @@ -2895,12 +2838,12 @@ "issue:44451|issue:44987", "issue:44811|issue:45584", "issue:43541|issue:45084", - "issue:40444|issue:46326", "issue:42175|issue:46178", + "issue:40444|issue:46326", "issue:42915|issue:44530", "issue:43644|issue:43646", - "issue:44206|issue:44220", "issue:43452|issue:43482", + "issue:44206|issue:44220", "issue:43883|issue:44488", "issue:36032|issue:43645", "issue:43317|issue:43482", @@ -2919,233 +2862,236 @@ "issue:43994|issue:44610", "issue:43322|issue:43540" ], - "prepared_review_unit_hash": "66f60d1072ef47b2326028f61eeb0e5ce237c3732a5de3d18a10901a33e2a2d5", + "prepared_review_unit_hash": "7dc17ff6832b9bab97b108a477eae5c48b9cf08685a240a829f3941cea283639", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37421, - "estimated_input_tokens": 9356, - "estimated_eval_tokens": 18968 + "serialized_chars": 37416, + "estimated_input_tokens": 9354, + "estimated_eval_tokens": 18964 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly superficially similar Transformers v5/model-loading regressions, but they do not look like the same underlying bug. I would not merge any of the soft pairs as duplicates on title-level evidence alone.", - "confidence": 0.29, - "canonical_issue_reason": "issue:43646 is the broadest, most generic regression report in the set ('custom model initialization' in v5), so it is the best umbrella representative if a single issue must anchor the cluster.", + "summary": "This set is mostly heterogeneous: the soft pairs span unrelated tokenizer, model-loading, vision, audio, and backend issues. I only see one likely duplicate pair in the audio feature-extraction path; the rest look like false positives rather than the same underlying bug.", + "confidence": 0.43, + "canonical_issue_reason": "issue:44206 is the clearest concrete bug report in the only likely duplicate family: it names the exact failing audio feature-extraction path, version, and crash mode. The companion issue is more generic, so 44206 is the better anchor.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the most cluster-representative issue because it is broadly phrased, version-specific, and closer to an umbrella regression than the more model-specific reports.", + "best_issue_reason": "issue:44206 is the strongest canonical candidate because it identifies a specific regression (`LasrFeatureExtractor` / `_torch_extract_fbank_features` / unsupported `center` arg) rather than a broad symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46285", "right": "issue:46402", "accept": false, - "reason": "Different models and different failure modes: CLIPTextModel structure inconsistency vs Sam3 text encoder weights not loading." + "reason": "Different models and failures: CLIP text-model structure inconsistency vs Sam3 text-encoder weights not loading." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Mistral-3 image preprocessor dtype mismatch is unrelated to a tokenizer/template compilation error." + "reason": "Unrelated bugs: image preprocessor dtype mismatch vs Mistral regex patch/tokenizer attribute crash." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "High perplexity on a model/dataset benchmark is unrelated to runtime special-token post-processor updates." + "reason": "Perplexity regression on a specific model is not the same as runtime special-token/post-processor update behavior." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "Tokenizer method-name conflict and Qwen3Moe FP8 loading failure are different bugs." + "reason": "Tokenizer class conflict for T5 is unrelated to Qwen3Moe FP8 loading failure." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "AutoConfig trust_remote_code missing fields is unrelated to PagedAttentionCache group-type handling." + "reason": "AutoConfig missing fields under `trust_remote_code` is a different path from Qwen3.5 PagedAttentionCache group-type crashes." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "TrainerState token accounting bug is not the same as Qwen2.5-VL iterable dataset multi-image failure." + "reason": "Trainer token-count tracking and Qwen2.5-VL multi-image iterable dataset failures are separate problems." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Both are loading failures, but for different models with no evidence of the same root cause." + "reason": "Both are model-loading failures, but they affect different repos and likely different code paths; not enough to merge as one bug." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both involve Whisper, but one is batch_decode skip_special_tokens behavior and the other is empty-transcription generation after align_special_tokens." + "reason": "Both involve Whisper, but one is decode skip-special-tokens behavior and the other is empty-transcription generation after align_special_tokens." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Torch dynamo grouped_mm tracing error is unrelated to a template-node compilation TypeError." + "reason": "Torch dynamo grouped_mm tracing failure is unrelated to the non-template-node compile error." }, { - "left": "issue:40444", - "right": "issue:46326", + "left": "issue:42175", + "right": "issue:46178", "accept": false, - "reason": "Multimodal iterable-dataset image batching and mistral_common continue_final_message are unrelated." + "reason": "Package extra/backend installation coverage is different from an invalid BACKENDS_MAPPING entry for tensorflow_text." }, { - "left": "issue:42175", - "right": "issue:46178", + "left": "issue:40444", + "right": "issue:46326", "accept": false, - "reason": "Package extra/back-end installation coverage and BACKENDS_MAPPING validation are different backend-selection problems." + "reason": "Qwen2.5-VL iterable dataset multi-image handling is unrelated to mistral_common continue_final_message behavior." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "FP8 configuration failure in Qwen3Moe is not the same as linear_attention group handling in PagedAttentionCache." + "reason": "Different Qwen-family failures: FineGrainedFP8Config loading vs linear_attention group-type crash in PagedAttentionCache." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Both are v5 regressions, but one is non-persistent buffer serialization and the other is custom model initialization; not the same concrete bug." - }, - { - "left": "issue:44206", - "right": "issue:44220", - "accept": false, - "reason": "Unsupported center arg in LasrFeatureExtractor is different from a generic _torch_extract_fbank_features issue." + "reason": "Both mention Transformers v5 breakage, but one is about non-persistent buffers and the other about custom model initialization; not the same concrete bug." }, { "left": "issue:43452", "right": "issue:43482", "accept": false, - "reason": "GGUF-related loading issues are similar, but one is a general from_pretrained breakage and the other is a specific Qwen2.5-GGUF model report; not enough evidence of one bug." + "reason": "GGUF loading is the broad theme, but one is an API/path break around `gguf_file`, while the other is a specific Qwen2.5-GGUF v5 load failure." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": true, + "reason": "These point to the same audio feature-extraction failure path; the generic `_torch_extract_fbank_features()` report matches the unsupported-arg crash described in 44206." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Missing all_tied_weights_keys in Molmo and inability to load a specific BERT model are unrelated." + "reason": "Missing `all_tied_weights_keys` in Molmo is unrelated to a model-loading failure for cjvt/sleng-bert." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "Tokenizer special-method conflict and notebook custom-model initialization are different failures." + "reason": "Tokenizer special-token method conflict is unrelated to custom model initialization breaking in notebooks." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "Dequantized model device-map/offload loading is unrelated to GGUF loading regression." + "reason": "Device-map/offload loading failure is a different bug from GGUF loading regression in Transformers v5." }, { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Non-template-node compilation error and GraniteMoEHybrid invalid method call are different code paths." + "reason": "Compile-time template-node error is unrelated to GraniteMoEHybrid calling an invalid method." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Qwen3.5 hidden-state output bug and Qwen2.5-VL rope index scaling bug are unrelated model behaviors." + "reason": "Qwen3.5 hidden-state output bug and Qwen2.5-VL rope-index scaling bug affect different models and code paths." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Two separate model/tokenizer loading failures with no shared concrete bug evidence." + "reason": "Both are version-related loading failures, but for different models and different tokenizer/model code paths." }, { "left": "issue:44464", "right": "issue:45987", "accept": false, - "reason": "Chunked generation inconsistency under compiled forward is different from StaticCache sequence-length return type." + "reason": "Chunked-generation inconsistency under compiled forward is not the same as `StaticCache.get_seq_length()` returning the wrong type." }, { "left": "issue:43746", "right": "issue:46326", "accept": false, - "reason": "PEFT local-checkpoint loading for GraniteSpeech is unrelated to mistral_common continue_final_message behavior." + "reason": "PEFT local-checkpoint loading for GraniteSpeech is unrelated to mistral_common `continue_final_message` handling." }, { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Audio chat-template sample-rate defaulting and MIMI batch-padding-mask differences are separate audio processor bugs." + "reason": "Audio processor sampling-rate defaults and MIMI encoder padding-mask behavior are different issues." }, { "left": "issue:43645", "right": "issue:44534", "accept": false, - "reason": "Custom model init in Jupyter and non-persistent buffer junk values are different v5 regressions." + "reason": "Generic custom-model initialization breakage is not the same as non-persistent buffers being filled with junk." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Custom model initialization breakage and non-persistent buffer junk are not the same defect." + "reason": "Custom model initialization regression and junk non-persistent buffers are related only at a high level, not the same underlying bug." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Untied lm_head weights in one model family and device-dependent serialization inconsistency are related in theme but not clearly the same bug." + "reason": "Weight-tying failure in Mistral3 is different from device-dependent serialization inconsistency for `lm_head.weight`." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "Molmo tied-weights metadata and tokenizer loading for est-roberta are unrelated." + "reason": "Different components and failure modes: Molmo tied-weights attribute error vs tokenizer loading for est-roberta." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "General loading noise and a specific Mistral regex patch crash do not match as one bug." + "reason": "General loading noise is not the same as a Mistral tokenizer regex patch crash." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "FP8 model failure and template-node compilation TypeError are unrelated." + "reason": "Qwen3Moe FP8 loading and template-node compilation failures are unrelated." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "SigLIP2 nonsensical outputs and OmDet-Turbo processor/model input-size mismatch are different model/processor issues." + "reason": "Nonsensical model outputs and a processor/model input-size mismatch are distinct problems." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Llava Next loading segfault and Qwen3OmniMoe video-processing ValueError are different multimodal failures." + "reason": "Llava Next loading segfault and Qwen3OmniMoe video-processing ValueError are different code paths and symptoms." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is appropriately conservative and grounded in the packet. The only accepted soft pair, issue:44206 vs issue:44220, is supported by closely related audio feature-extraction titles, and the remaining rejects are framed as different models, code paths, or failure modes without overstating equivalence. The canonical choice and low confidence also fit the evidence." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3242,220 +3188,220 @@ "issue:44683|issue:45020", "issue:44079|issue:45663" ], - "prepared_review_unit_hash": "f20a3e4ab1d457b8603cdba2caff085a2b23f69bb7e03a11bb5d64aa0e7acc2a", + "prepared_review_unit_hash": "65371c118f57de265cdb65d75bab5f514022acc734533b0f9c61c7889b381b5e", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36150, - "estimated_input_tokens": 9038, - "estimated_eval_tokens": 18332 + "serialized_chars": 36153, + "estimated_input_tokens": 9039, + "estimated_eval_tokens": 18334 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is mostly a loose semantic cluster of unrelated issues with only a couple of real duplicate-looking pairs. The strongest true duplicates are the Qwen3.5 save_pretrained regression pair and the tokenizer special-token/post-processor regression pair; the rest overlap only at a broad subsystem level.", - "confidence": 0.76, - "canonical_issue_reason": "issue:45216 is the clearest representative: it is a concrete Qwen3.5 save_pretrained regression with a version boundary, and issue:45357 reads like a later follow-up of the same checkpoint-key problem.", + "summary": "This cluster is mostly a loose similarity set of unrelated Transformers issues across different models/subsystems. I found one plausible duplicate pair around the Qwen3.5 save_pretrained checkpoint regression; the rest should stay separate.", + "confidence": 0.87, + "canonical_issue_reason": "If forced to pick a representative issue, issue 45216 is the best canonical: it states the broader Qwen3.5 save_pretrained checkpoint regression, while 45357 reads like a narrower follow-up about the same broken serialization path.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45216 is the best single issue to anchor the cluster because it is specific, actionable, and has an obvious matching follow-up; most other items here are not true duplicates.", + "best_issue_reason": "Issue 45216 is the best global representative because it is broader, earlier, and more likely to capture the root save_pretrained regression for Qwen3.5; no other issue cleanly covers the rest of the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "Different models and different failure modes: batched-vs-single padding-mask mismatch vs DAC latents/STE mismatch." + "reason": "Different bugs: MIMI batching/padding-mask mismatch vs DAC from_latents forward/STE mismatch." }, { "left": "issue:43089", "right": "issue:45987", "accept": false, - "reason": "One is generation performance overhead, the other is a cache API return-type bug; same broad area, not the same defect." + "reason": "Unrelated symptoms: generation GPU-sync/dispatch overhead vs StaticCache.get_seq_length returning the wrong type." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "Different tokenizer-loading crashes for different models; no clear shared code-path bug." + "reason": "Both are tokenizer-loading failures, but the model and root cause differ: ScandiBERT load failure vs Mistral regex patch crash." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Both involve tokenizer loading, but one is a model loading failure and the other is a Mistral regex patch crash." + "reason": "Different failures: loading physical-intelligence/fast vs Mistral tokenizer regex/backend-tokenizer crash." }, { "left": "issue:44991", "right": "issue:46315", "accept": false, - "reason": "One is a tokenizer load failure, the other is a performance regression in slow-tokenizer lookup." + "reason": "One is a tokenizer load failure, the other is a performance regression in convert_tokens_to_ids." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "Unrelated issues: torch.export/torch_compilable_check versus image processor size mismatch." + "reason": "Different code paths: torch.export exportability with torch_compilable_check vs a processor/model input-size mismatch." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "Different subsystems and symptoms: fbank feature extraction vs compiled flex_attention on newer torch." + "reason": "Audio fbank feature extraction bug vs compiled flex_attention regression on newer torch; not the same bug." }, { "left": "issue:42915", "right": "issue:44291", "accept": false, - "reason": "Both are loading/training failures, but the causes differ: FP8 config handling vs unexpected _is_hf_initialized arg." + "reason": "Distinct initialization/runtime problems: FineGrainedFP8Config failure vs unexpected _is_hf_initialized argument in init_empty_weights." }, { "left": "issue:45216", "right": "issue:45357", "accept": true, - "reason": "Same Qwen3.5 save_pretrained regression family; both report incorrect checkpoint contents/keys and look mergeable into one fix." + "reason": "Same underlying save_pretrained serialization regression for Qwen3.5; 45357 looks like a narrower continuation about incorrect visual encoder keys." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "Different tokenizer bugs: AddedToken special-flag duplication vs a specific tokenizer load regression." + "reason": "AddedToken special-argument conflict is unrelated to loading physical-intelligence/fast." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "Different loading failures on different models; only superficial tokenizer/model-load similarity." + "reason": "Different tokenizer/model issues: T5 tokenizer special-token method conflict vs MobileLLM-125M key error on load." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "Different FP8-related failures: config init versus missing Float8 storage type." + "reason": "FP8 config failure and missing Float8 storage are separate runtime compatibility issues." }, { "left": "issue:44479", "right": "issue:45325", "accept": false, - "reason": "Related Qwen VL family, but one is a broad video-input regression and the other is a specific still-image position-id scaling bug." + "reason": "Related Qwen VL area, but not the same concrete bug: broad video-input regression across several models vs still-image temporal position scaling in Qwen2.5-VL." }, { "left": "issue:46209", "right": "issue:46326", "accept": false, - "reason": "Completely different problems: FP8 inference on B200 versus mistral_common continue_final_message behavior." + "reason": "Unrelated: FP8 inference on B200 vs continue_final_message behavior for mistral_common tokenizers." }, { "left": "issue:36032", "right": "issue:41720", "accept": false, - "reason": "Different tokenizer-loading error vs device-mapping cudaErrorAssert; no shared underlying bug." + "reason": "T5 tokenizer loading conflict is unrelated to Qwen3 auto device mapping cudaErrorAssert." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "Both are v5 regressions, but one is an _is_hf_initialized flag breakage and the other is an mmap/OOM platform issue." + "reason": "Different regressions: _is_hf_initialized handling in Transformers v5 vs mmap/OOM behavior on Strix Halo." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Both mention tied embeddings, but the concrete bugs differ: Mistral3 lm_head tying vs LoRA merge with extended vocabulary." + "reason": "Both involve tied embeddings, but the bugs differ: missing lm_head tying in Mistral3 vs LoRA merge collapse with extended vocabulary." }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Unrelated model/config failures; no shared code-path or symptom." + "reason": "Qwen3-MoE sliding_window bug and granite_speech strict-config loading are unrelated." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Same model family, but different defects: multi-image iterable dataset finetuning vs still-image temporal position-id scaling." + "reason": "Multi-image iterable finetuning failure vs Qwen2.5-VL still-image temporal position scaling; different symptoms and code paths." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL loading in vLLM and Qwen3-MoE sliding_window are unrelated." + "reason": "PaddleOCR-VL loading failure is unrelated to Qwen3-MoE sliding_window behavior." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "Different Transformers v5 regressions with different mechanics: base_model_prefix loading vs non-persistent buffer initialization." + "reason": "base_model_prefix loading breakage and non-persistent buffer junk are separate v5 regressions." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch version incompatibility and loading noise are not the same bug." + "reason": "Torch 2.10 incompatibility is not the same as noisy transformer loading logs." }, { "left": "issue:40990", "right": "issue:44811", "accept": false, - "reason": "Completely different issues: perplexity on GPT-OSS vs Whisper batch_decode special-token handling." + "reason": "Perplexity regression on gpt-oss-20b is unrelated to Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "Different model/runtime failures; no evidence of a shared root cause." + "reason": "Whisper runtime failure and missing Float8 storage type are different compatibility bugs." }, { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "Generic loading noise vs a specific tokenizer load regression are not duplicates." + "reason": "Loading noise vs tokenizer load failure; not the same issue." }, { "left": "issue:33666", "right": "issue:40444", "accept": false, - "reason": "Both are Qwen VL-related, but one is multi-GPU training and the other is iterable-dataset multi-image finetuning." + "reason": "Both are Qwen2-VL-family training issues, but one is multi-GPU training and the other is iterable-dataset multi-image finetuning; not the same bug." }, { "left": "issue:43632", "right": "issue:44534", "accept": false, - "reason": "Both are Transformers v5 regressions, but they affect different internals and would not be merged into one fix." + "reason": "_is_hf_initialized regression and non-persistent buffer junk are separate Transformers v5 problems." }, { "left": "issue:43421", "right": "issue:44568", - "accept": true, - "reason": "Same tokenizer special-token/post-processor class of bug; runtime special-token updates plausibly explain add_special_tokens failing to add BOS/EOS." + "accept": false, + "reason": "Runtime post-processor update support is not the same as add_special_tokens failing to add BOS/EOS for a specific tokenizer." }, { "left": "issue:44683", "right": "issue:45020", "accept": false, - "reason": "Different regression surfaces: compiled flex_attention on new torch vs remote_code model loading failures." + "reason": "Compiled flex_attention on torch>=2.9 and remote_code model loading failures are unrelated." }, { "left": "issue:44079", "right": "issue:45663", "accept": false, - "reason": "ModelOutput key assignment bug and Gemma4 FSDP2 attention failure are unrelated." + "reason": "ModelOutput key assignment bug and Gemma4 shared_kv_states FSDP2 error are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster summary is well grounded in the packet and stays appropriately conservative: most pairs are rejected for clear subsystem/symptom differences, and the two accepted pairs both have a credible shared defect family in the titles. The canonical issue choice around issue:45216 is also supported by the packet." + "feedback": "The summary is well grounded in the packet and stays appropriately conservative. The only accepted duplicate pair, issue:45216 and issue:45357, is clearly aligned by title and scope, and the remaining soft-edge rejections are mostly cautious distinctions across different models/subsystems. No obvious overstatement or mergeability problem." }, "evaluator_used": true, "retried": false, @@ -3540,8 +3486,8 @@ "issue:43818|issue:45357", "issue:44079|issue:44534", "issue:41720|issue:45081", - "issue:43377|issue:45491", "issue:43531|issue:44291", + "issue:43377|issue:45491", "issue:40444|issue:45381", "issue:44265|issue:45072", "issue:43644|issue:44079", @@ -3557,221 +3503,224 @@ "issue:42915|issue:43482", "issue:43701|issue:45325" ], - "prepared_review_unit_hash": "7ae36d4cb7bf9c9286487a4b4da20d86c0b98d4da1ed78c18701ef343df0df59", + "prepared_review_unit_hash": "1ac6c6af845b96a00a91ece34571672f10f7a39f268dc731cff80b0b12764b92", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36867, - "estimated_input_tokens": 9217, - "estimated_eval_tokens": 18690 + "serialized_chars": 36863, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a loose cluster of Transformers regression reports across many unrelated models/subsystems. The shared theme is broad v5 loading/init breakage, but the soft pairs do not look like safe duplicates.", - "confidence": 0.87, - "canonical_issue_reason": "If one issue must serve as the cluster umbrella, issue:43632 is the best representative because it describes a core Transformers v5 initialization/loading regression with broad downstream impact. Most other items are narrower model-specific bugs, so they should not be merged into it as duplicates.", + "summary": "This cluster is mostly a grab bag of unrelated transformer regressions. The only soft edge that looks like the same underlying bug is the Qwen2.5-VL position-id regression pair (45325/45381); the rest are too symptom- or model-specific to merge safely.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 45325 is the clearest root-cause report: it pinpoints the Qwen2.5-VL `get_rope_index` position-id scaling bug, which can reasonably subsume the closely related video-input report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43632 is the strongest global representative for this cluster because it is the most systemic v5 regression among the issues; it is broader than the model-specific reports and most likely to anchor related follow-up discussion.", + "best_issue_reason": "45325 is the best representative issue because it names the concrete failing code path and is the most general description of the accepted duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43381", "right": "issue:45923", "accept": false, - "reason": "Different problems: eval-mode gradient checkpointing behavior vs a missing supports_gradient_checkpointing flag on a trust_remote_code model." + "reason": "Different bugs: eval-mode gradient checkpointing use versus a missing `supports_gradient_checkpointing` flag." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "Different failure modes: empty-weights init argument handling vs float8 storage lookup during load." + "reason": "Unrelated TypeErrors in different loading paths (`init_empty_weights` vs float8 storage lookup)." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "Both affect sequence-classification training, but one is flash_attention_3 degeneracy and the other is num_labels=1 loss logic; not the same bug." + "reason": "Different training failures: FlashAttention-3 instability versus zero loss for `num_labels=1` classification." }, { "left": "issue:44462", "right": "issue:45812", "accept": false, - "reason": "Tokenizer repository-file selection vs Granite token-ID regression are distinct code paths and symptoms." + "reason": "Both are tokenizer regressions, but one ignores `tokenizer.json` while the other yields wrong Granite token IDs." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "GGUF/v5 model loading breakage vs CLIPTokenizer model_max_length constant are unrelated." + "reason": "Different failures: GGUF loading under v5 versus an absurd `model_max_length` on CLIPTokenizer." }, { "left": "issue:43482", "right": "issue:43525", "accept": false, - "reason": "Different model-specific load errors: GGUF loading vs Llama4Config missing pad_token_id." + "reason": "No shared code path; GGUF loading and `Llama4Config.pad_token_id` are separate issues." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL multi-image finetuning failures." + "reason": "Different multimodal problems: progressive generation with `past_key_values` versus multi-image iterable finetuning." }, { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "gpu+cpu offload/dequantized loading vs Qwen3 MoE training memory usage are different issues." + "reason": "Offload/device-map loading failure and Qwen3 MoE training memory use are not the same bug." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "AutoTokenizer/Voxtral error messaging vs Apple Silicon allocator warmup TypeError are unrelated." + "reason": "Bad tokenizer error messaging is unrelated to the Apple Silicon allocator warmup TypeError." }, { "left": "issue:42175", "right": "issue:45542", "accept": false, - "reason": "Both mention TensorFlow availability, but one is torch-extra dependency wiring and the other is tensorboard-without-TF runtime behavior; not clearly the same bug." + "reason": "Both mention TF/backend packaging, but they fail at different points and are not the same fix." }, { "left": "issue:43818", "right": "issue:46402", "accept": false, - "reason": "Video-LLaVA tower/temporal-attention problems vs Sam3 text-encoder weight loading are unrelated." + "reason": "Different model families and symptoms: Video-LLaVA weight/temporal-attention issues versus Sam3Model text-encoder loading." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Different model families and mechanisms: Video-LLaVA tower architecture vs Qwen3.5 save_pretrained visual encoder keys." + "reason": "Video-LLaVA architecture issues are unrelated to Qwen3.5 save_pretrained key serialization." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "ModelOutput key assignment bug vs non-persistent buffer junk fill are different regressions." + "reason": "ModelOutput key assignment and non-persistent buffer junk are distinct v5 regressions." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "Qwen3 auto device mapping cuda assert vs Mistral tokenizer regex patch AttributeError are unrelated." + "reason": "Qwen3 auto device-map CUDA assert is unrelated to Mistral tokenizer regex patching." }, { - "left": "issue:43377", - "right": "issue:45491", + "left": "issue:43531", + "right": "issue:44291", "accept": false, - "reason": "MIMI padding-mask batching bug vs Gemma3 mixed-length sliding-window NaNs are different model-path issues." + "reason": "Sliding-window handling in Qwen3-MoE is unrelated to the `_is_hf_initialized` init_empty_weights issue." }, { - "left": "issue:43531", - "right": "issue:44291", + "left": "issue:43377", + "right": "issue:45491", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior vs init_empty_weights/_is_hf_initialized loading regression are unrelated." + "reason": "Both involve batching/padding, but one is missing padding-mask support in MIMI and the other is NaNs from Gemma3 sliding-window all-padding cases." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen2.5-VL issues, but one is multi-image finetuning and the other is video vision_position_ids; not the same bug." + "reason": "Both are Qwen2.5-VL multimodal issues, but one is multi-image finetuning and the other is video position-id handling." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "torch.export/torch_compilable_check failure vs bfloat16 dtype mismatches are different code paths." + "reason": "Export with `torch_compilable_check` and bf16 dtype mismatches are different problems." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Non-persistent buffer serialization junk vs ModelOutput key assignment are unrelated regressions." + "reason": "One is a v5 non-persistent buffer regression; the other is `ModelOutput` key assignment behavior." }, { "left": "issue:45325", "right": "issue:45381", - "accept": false, - "reason": "Both concern Qwen2.5-VL position IDs, but one is still-image temporal scaling and the other is video input vision_position_ids; too different to merge." + "accept": true, + "reason": "Both point at the same Qwen2.5-VL vision/temporal position-id computation in 5.3.0; a single fix to `get_rope_index`/position-id handling could plausibly cover both still-image and video inputs." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM efficient loading vs ModelOutput key assignment are unrelated." + "reason": "FSDP CPU RAM-efficient loading and ModelOutput key assignment are unrelated." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both are v5 model-init regressions, but base_model_prefix loading and custom model initialization are different concrete failures." + "reason": "Both are broad v5 regressions, but one is `base_model_prefix` loading and the other custom model initialization; not the same concrete bug." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "TensorFlow backend packaging issue vs PaddleOCR-VL load failure are unrelated." + "reason": "Packaging/backend installation issue versus a PaddleOCR-VL loading regression." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Whisper loading failure vs meta-device loading regression are different bugs." + "reason": "Whisper runtime failure and meta-device model loading are separate issues." }, { "left": "issue:43994", "right": "issue:44898", "accept": false, - "reason": "Siglip2 bad outputs vs Perceiver non-default-resolution failure are unrelated." + "reason": "Different vision-model failures with different code paths and reproduction conditions." }, { "left": "issue:45325", "right": "issue:45834", "accept": false, - "reason": "Qwen2.5-VL rope-index bug vs Kosmos2.5 long-OCR index error are different models and failure mechanisms." + "reason": "Qwen2.5-VL position-id scaling and Kosmos2.5 long-OCR index errors are different bugs." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "AyaVision lm_head tying bug vs general lm_head.weight serialization inconsistency are related only at a high level, not the same underlying issue." + "reason": "Both mention `lm_head`, but one is missing weight tying and the other is inconsistent serialization depending on device." }, { "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "Core v5 initialization/loading regression vs bfloat16 dtype mismatch are unrelated." + "reason": "`_is_hf_initialized` breakage is unrelated to bf16 dtype mismatches." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure vs Qwen2.5 GGUF loading failure are unrelated." + "reason": "Different model/load failures: Qwen3MoE FP8 config versus Qwen2.5-GGUF loading under v5." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "resume_from_checkpoint key mismatch is unrelated to Qwen2.5-VL rope-index scaling." + "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL position-id scaling." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The result is grounded in the packet and the only accepted soft edge (45325/45381) is plausibly the same Qwen2.5-VL position-id bug. The remaining pair judgments are conservative and do not overclaim sameness from the available titles alone." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3839,8 +3788,8 @@ "soft_pairs": [ "issue:44291|issue:44877", "issue:44479|issue:45823", - "issue:43873|issue:45701", "issue:43606|issue:44265", + "issue:43873|issue:45701", "issue:43296|issue:45800", "issue:43262|issue:43746", "issue:40990|issue:45584", @@ -3851,8 +3800,8 @@ "issue:43881|issue:44488", "issue:43606|issue:43819", "issue:39692|issue:40444", - "issue:43454|issue:43550", "issue:41720|issue:45823", + "issue:43454|issue:43550", "issue:46305|issue:46315", "issue:43377|issue:44610", "issue:43454|issue:45072", @@ -3868,20 +3817,20 @@ "issue:43749|issue:44534", "issue:44991|issue:46291" ], - "prepared_review_unit_hash": "f96809f3838edb49f98640db51e1bd3badba6635df4fcc4d67482915fbb02662", + "prepared_review_unit_hash": "413c4918b048acf428dccef48ade10eec2c347dbceb87b5371269ce9ea5362b2", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36339, + "serialized_chars": 36338, "estimated_input_tokens": 9085, "estimated_eval_tokens": 18426 }, @@ -3890,180 +3839,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated Transformers/model-loading regressions that only look superficially similar. I rejected every soft pair because none describe the same underlying bug or a change that could plausibly be merged into one fix. If one issue must represent the cluster, the broad offloading/quantization regression is the best anchor.", - "confidence": 0.93, - "canonical_issue_reason": "issue:43873 is the broadest, most representative ticket in the set: it centers on offloading/quantization failure modes and has the strongest cross-reference footprint among the listed issues.", + "summary": "No soft pair looks like the same concrete bug/change; the list is a mix of unrelated model-loading, tokenizer, offload, and multimodal regression reports. If a single representative issue is needed, the broad offloading/quantization report is the best anchor.", + "confidence": 0.83, + "canonical_issue_reason": "issue:43873 is the broadest and most central offloading/quantization bug report in the set, so it is the best umbrella issue if one representative must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 best captures the recurring loading/offloading regression theme and is the most useful single issue to anchor this otherwise mixed cluster.", + "best_issue_reason": "issue:43873 has the clearest generic scope around offloading plus quantization, making it the most suitable cluster representative among these unrelated reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "Different bugs: unexpected `_is_hf_initialized` arg during init-empty-weights vs strict config rejection for granite_speech." + "reason": "Different failure modes: init_empty_weights/_is_hf_initialized vs strict config loading for granite_speech." }, { "left": "issue:44479", "right": "issue:45823", "accept": false, - "reason": "Different models and failures: video-input regression in Qwen VL variants vs Gemma4 PLE device mismatch in forward." + "reason": "Both are multimodal regressions, but one is a video-input regression and the other is a device-mismatch forward failure." }, { - "left": "issue:43873", - "right": "issue:45701", + "left": "issue:43606", + "right": "issue:44265", "accept": false, - "reason": "One is offloading/quantization behavior; the other is tokenizer behavior. Same broad repo area, not the same bug." + "reason": "Bark CPU offload device mismatch is unrelated to torch.export.export failing on torch_compilable_check." }, { - "left": "issue:43606", - "right": "issue:44265", + "left": "issue:43873", + "right": "issue:45701", "accept": false, - "reason": "CPU offload device mismatch for Bark-small is unrelated to torch.export failing on torch_compilable_check." + "reason": "Offloading/quantization bug vs tokenizer behavior change; not the same code-path problem." }, { "left": "issue:43296", "right": "issue:45800", "accept": false, - "reason": "PaddleOCR-VL load failure in vLLM is not the torch 2.4.1 / transformers 5.8 incompatibility issue." + "reason": "Different incompatibilities: PaddleOCR-VL loading in vLLM vs torch/transformers version mismatch." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio chat-template sample-rate default bug is unrelated to GraniteSpeech PEFT checkpoint loading." + "reason": "Audio chat-template sampling-rate default is unrelated to PEFT adapter checkpoint loading." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "High perplexity on gpt-oss-20b and Whisper empty-transcription generation are distinct behaviors." + "reason": "Perplexity regression on a language model is unrelated to Whisper empty-transcription generation failure." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "BarkProcessor voice_preset failure and progressive generation with inputs_embeds/past_key_values are different code paths." + "reason": "Voice preset handling in Bark is unrelated to progressive generation with inputs_embeds/past_key_values." }, { "left": "issue:44568", "right": "issue:45920", "accept": false, - "reason": "Both are tokenizer regressions, but one is a specific mdeberta special-tokens bug and the other is a broad wrong-ID regression across many models." + "reason": "Specific BOS/EOS insertion bug for one tokenizer is not the same as broad wrong-token-IDs regression across multiple models." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "Generation-cache behavior is unrelated to an A800 cudaErrorAssert from auto device mapping." + "reason": "Generation cache/input-embed issue vs auto device-mapping cudaErrorAssert are different problems." }, { "left": "issue:44186", "right": "issue:44265", "accept": false, - "reason": "LayoutLMv2Tokenizer padding/NER crash is not the same problem as export failures caused by torch_compilable_check." + "reason": "Tokenizer crash on NER/padding is unrelated to export failures involving torch_compilable_check." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "Different model-loading failures for different models; no shared concrete bug path." + "reason": "Model loading failure for glm-4v-9b vs tokenizer loading failure for cjvt/sleng-bert are unrelated." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Bark CPU offload device mismatch is unrelated to DAC.from_latents not matching forward due to missing STE." + "reason": "Both mention loading/latents, but one is CPU offload device mismatch and the other is DAC from_latents vs forward mismatch." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "Doc-example errors for SigLIP2 are not the same as Qwen2.5-VL finetuning failures with multi-image iterable datasets." + "reason": "SigLIP2 documentation errors are unrelated to Qwen2.5-VL finetuning with multi-image iterable datasets." }, { - "left": "issue:43454", - "right": "issue:43550", + "left": "issue:41720", + "right": "issue:45823", "accept": false, - "reason": "Weight-tying bug in AyaVisionConfig and torch.compile SDPA failure in Bamba are different concrete issues." + "reason": "Both are device-related, but one is auto device-map crash on A800 and the other is a forward-time device mismatch in Gemma4 PLE." }, { - "left": "issue:41720", - "right": "issue:45823", + "left": "issue:43454", + "right": "issue:43550", "accept": false, - "reason": "Both mention device_map=auto, but one is a CUDA assert on A800 and the other is a forward-time device mismatch in a different model." + "reason": "lm_head tying/garbage generation is unrelated to torch.compile SDPA failure." }, { "left": "issue:46305", "right": "issue:46315", "accept": false, - "reason": "Streaming mask off-by-one and tokenizer performance regression are unrelated." + "reason": "Streaming mask off-by-one is unrelated to tokenizer performance regression." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "MIMI padding-mask inconsistency is unrelated to OmDet-Turbo's processor/model input-size mismatch." + "reason": "Padding-mask batch inconsistency in MIMI is unrelated to OmDet-Turbo processor image-size mismatch." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "lm_head weight tying bug is unrelated to bfloat16 dtype mismatches in inference." + "reason": "Weight tying bug in AyaVision is unrelated to dtype mismatch in SwitchTransformers/TimmWrapperModel inference." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Dequantized model loading on gpu+cpu offload is related in theme to offloading, but it is a different concrete failure than the quantization/offloading bug in 43873." + "reason": "Similar offload/quantization theme, but one is dequantized-model loading on gpu+cpu offload and the other is a broader offloading regression; not enough evidence of one concrete bug." }, { "left": "issue:43257", "right": "issue:45923", "accept": false, - "reason": "MoE weight conversion under accelerate+deepspeed is unrelated to a missing supports_gradient_checkpointing flag." + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to a missing supports_gradient_checkpointing flag." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "Perplexity regression on gpt-oss-20b is not the same bug as gradient checkpointing being ineffective with PEFT LoRA." + "reason": "High perplexity on a model is unrelated to gradient checkpointing not being effective with LoRA." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "Missing padding-mask support in MIMI encoder is unrelated to Perceiver's non-default-resolution failure." + "reason": "MIMI encoder batching bug is unrelated to Perceiver image classification at non-default resolution." }, { "left": "issue:43322", "right": "issue:46050", "accept": false, - "reason": "Llava Next segfault on load and sinq quantized save/load breakage are different lifecycle failures." + "reason": "Llava Next segfault on load is unrelated to sinq quantized model save/load breakage." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig and failure to unwrap text_config in composite Qwen configs are different config-loading bugs." + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to AutoModelForCausalLM.from_config not unwrapping text_config." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "AyaVision lm_head tying bug is unrelated to BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:45325", + "right": "issue:46315", + "accept": false, + "reason": "Qwen2.5-VL rope-index regression is unrelated to slow convert_tokens_to_ids behavior." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "cudaErrorAssert under auto device mapping is not the same as a Llava Next loading segfault." }, { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "FSDP CPU RAM efficient loading is not the same as non-persistent buffers being filled with junk in v5." + "reason": "FSDP CPU RAM efficient loading is unrelated to non-persistent buffer initialization junk in v5." }, { "left": "issue:44991", "right": "issue:46291", "accept": false, - "reason": "Tokenizer loading failure for EMBEDDIA/est-roberta is unrelated to evaluate import failing because kernels needs version/revision." - }, - { - "left": "issue:45325", - "right": "issue:46315", - "accept": false, - "reason": "Qwen2.5-VL rope-index temporal scaling bug is unrelated to slow convert_tokens_to_ids behavior." + "reason": "Tokenizer loading regression for EMBEDDIA/est-roberta is unrelated to kernels v0.15 requiring version/revision on import." } ] }, @@ -4141,8 +4102,8 @@ "issue:44466|issue:45357", "issue:43526|issue:43883", "issue:43881|issue:44991", - "issue:40990|issue:45701", "issue:44387|issue:45704", + "issue:40990|issue:45701", "issue:36331|issue:42617", "issue:43065|issue:46402", "issue:44186|issue:45072", @@ -4161,28 +4122,28 @@ "issue:44610|issue:44898", "issue:44186|issue:44610", "issue:43295|issue:45701", - "issue:45632|issue:45684", "issue:42915|issue:43317", + "issue:45632|issue:45684", "issue:36010|issue:42617", "issue:43425|issue:46291", "issue:41720|issue:45161", "issue:43525|issue:46291", "issue:44671|issue:44743" ], - "prepared_review_unit_hash": "22a57311dcb485f5797b5f9ca0420f489f58b51ca69899b37bca26fd932a9f31", + "prepared_review_unit_hash": "6ea41a1f64abaf757a91a9b736e5c31da52186c8811aa7cb5cdeeeb7af02f907", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 32, - "serialized_chars": 36824, + "serialized_chars": 36821, "estimated_input_tokens": 9206, "estimated_eval_tokens": 18668 }, @@ -4191,204 +4152,204 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is a loose collection of distinct Transformers bugs and regressions, not one cohesive duplicate cluster. All soft-edge pairs look like false positives: they share broad themes (loading, offloading, serialization, model families) but not the same concrete code-path failure.", - "confidence": 0.96, - "canonical_issue_reason": "No single canonical issue stands out; the items span unrelated models, subsystems, and failure modes rather than one underlying bug.", + "summary": "These soft links are mostly keyword-similar but not true duplicates: they span different models, code paths, and failure modes (trainer, serving, quantization, tokenization, processors, loading). I would reject every soft edge.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue fits the cluster: the linked issues describe distinct bugs rather than one underlying defect.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue is a strong representative of the whole set; at most a few issues are broad, but none subsume the others or the pairwise topics.", + "best_issue_reason": "Issue 43873 is the most discussion-heavy and has inbound references, so it is the closest thing to a hub, but it is still too broad to serve as a true duplicate representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "Different problems: running 3d_parallel.py vs a Qwen3MoE FP8 model failure; no shared code path." + "reason": "Different failures in different contexts: 3d_parallel.py execution vs FineGrainedFP8Config loading." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Both involve Qwen3-family generation, but one is an auto-device-map CUDA assert and the other is flash-attention behavior." + "reason": "Both involve Qwen3/Qwen3.5, but one is device-map CUDA assert and the other is flash-attention generation behavior." }, { "left": "issue:44466", "right": "issue:45357", "accept": false, - "reason": "Both concern save/serialization regressions, but one is tied lm_head weight handling and the other is wrong visual encoder keys." + "reason": "Different bugs: device-dependent lm_head serialization vs incorrect visual encoder keys on save_pretrained." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "Unrelated issues: BeitImageProcessorFast label reduction vs Molmo tied-weights attribute missing." + "reason": "Unrelated components: BeitImageProcessorFast label reduction vs Molmo tied-weights attribute error." }, { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "Both are loading failures, but one is a model load and the other is a tokenizer load for a different model." + "reason": "Different models and subsystems: glm-4v loading vs tokenizer loading for est-roberta." }, { - "left": "issue:40990", - "right": "issue:45701", + "left": "issue:44387", + "right": "issue:45704", "accept": false, - "reason": "Perplexity on GPT-OSS is not the same bug as a general tokenization change across versions." + "reason": "Different root causes: int4 quantization memory growth vs apex RMSNorm memory leak." }, { - "left": "issue:44387", - "right": "issue:45704", + "left": "issue:40990", + "right": "issue:45701", "accept": false, - "reason": "Different memory issues: int4 quantization OOM vs an apex RMSNorm memory leak." + "reason": "Perplexity regression on one model/dataset is not the same bug as a tokenization behavior change." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "Trainer compute_loss signature mismatch is unrelated to a 3d_parallel.py runtime failure." + "reason": "Trainer compute_loss signature breakage is unrelated to running 3d_parallel.py." }, { "left": "issue:43065", "right": "issue:46402", "accept": false, - "reason": "Same broad Sam3 area, but dummy Conv2d in a decoder and missing text-encoder weights are different bugs." + "reason": "Sam3 pixel decoder dummy conv2d is a different issue from Sam3 text encoder weights not loading." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "Tokenizer crash on NER/padding is unrelated to dtype mismatches in inference." + "reason": "Tokenizer padding/truncation crash and bfloat16 dtype mismatch are different code paths." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Compute_loss API breakage is unrelated to Qwen3MoE FP8 model loading/execution." + "reason": "CustomTrainer API breakage is unrelated to Qwen3Moe FP8 loading failure." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Video processor batched-shape bug is not the same as tokenization changing across versions." + "reason": "Video processor batch-shape bug is unrelated to tokenization changes." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config loading for granite_speech is a different failure from tokenizer loading for EMBEDDIA/est-roberta." + "reason": "Strict config loading for granite_speech is not the same as tokenizer loading failure." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Both mention offload/memory, but quantized offloading and disabling mmap on Strix Halo are different code paths." + "reason": "Quantization/offloading behavior and mmap OOM on Strix Halo are different problems." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "torch.compile with SDPA on Bamba is unrelated to packed-weight unpacking during accelerate loading." + "reason": "torch.compile/SDPA failure in Bamba is unrelated to packed-weight unpacking during accelerate loading." }, { "left": "issue:43577", "right": "issue:43708", "accept": false, - "reason": "BLIP2 dtype propagation is unrelated to Trainer resume_from_checkpoint max_steps calculation." + "reason": "BLIP2 dtype propagation and Trainer resume max_steps calculation are unrelated." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "Both touch save_pretrained/tied-weight behavior, but Mistral3 lm_head tying and Qwen3.5 visual encoder keys are distinct bugs." + "reason": "Tied lm_head weights in Mistral3 are a different defect from Qwen3.5 visual encoder key serialization." }, { "left": "issue:44008", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma multimodal issues, but a variable name collision in Gemma3n forward is not the same as Gemma4 PLE device mismatch." + "reason": "Gemma3n forward variable collision and Gemma4 PLE device mismatch are distinct bugs." }, { "left": "issue:42175", "right": "issue:46291", "accept": false, - "reason": "Packaging/backend inclusion is unrelated to kernels v0.15 requiring version or revision metadata." + "reason": "Missing TensorFlow backend in torch extras is unrelated to kernels v0.15 version/revision requirements." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Qwen3 auto-device-map CUDA assert is not the same as a 3d_parallel.py execution problem." + "reason": "Different models and symptoms: auto device-map CUDA assert vs 3d_parallel.py execution failure." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Continuous-batching multimodal crash and Gemma4Processor missing _tokenizer are different serve-time failures." + "reason": "Both involve serve, but the crashes come from different causes: string `.to` misuse vs missing `_tokenizer`." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Different device/runtime bugs: Bamba SDPA compile failure vs bark-small CPU offload device mismatch." + "reason": "Different device/runtime issues: torch.compile SDPA failure vs CPU-offload device mismatch." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Both are image-size related, but processor output size mismatch and Perceiver interpolation failure are separate issues." + "reason": "Both are processor/image-size issues, but they affect different models and preprocessing behavior." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "Tokenizer crashes and image processor resolution bugs do not share the same underlying defect." + "reason": "Tokenizer crash in LayoutLMv2 and OmDet-Turbo image-size mismatch are unrelated." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "A regression around processor.tokenizer/images is unrelated to a broad tokenization behavior change." + "reason": "Processor.tokenizer regression is not the same as version-dependent tokenization changes." }, { - "left": "issue:45632", - "right": "issue:45684", + "left": "issue:42915", + "right": "issue:43317", "accept": false, - "reason": "Both involve local model files/caching, but cache-path collisions and read-only permission propagation are different problems." + "reason": "Qwen3Moe FP8 loading and dequantized device_map offload failures are different bugs." }, { - "left": "issue:42915", - "right": "issue:43317", + "left": "issue:45632", + "right": "issue:45684", "accept": false, - "reason": "Qwen3MoE FP8 failure is unrelated to loading a dequantized model with gpu+cpu offload." + "reason": "Cache-path collision and read-only permission propagation are unrelated save/load issues." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "GenerationMixin import breakage is unrelated to running 3d_parallel.py." + "reason": "Importing GenerationMixin and running 3d_parallel.py are unrelated failures." }, { "left": "issue:43425", "right": "issue:46291", "accept": false, - "reason": "Torch version incompatibility is not the same as kernels metadata requirements during import." + "reason": "Torch 2.10 incompatibility is unrelated to kernels version/revision handling." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Different parallelism failures: auto device mapping on Qwen3 vs only-TP not working for GPT-OSS MoE." + "reason": "Auto device-mapping CUDA assert is not the same as tensor-parallel support failure in GPT-OSS MoE." }, { "left": "issue:43525", "right": "issue:46291", "accept": false, - "reason": "Missing pad_token_id on Llama4Config is unrelated to kernels version/revision import requirements." + "reason": "Missing pad_token_id in Llama4Config is unrelated to kernels dependency versioning." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "CamemBERT masked-LM prediction regression and Qwen3.5 recurrent-state reset are distinct model bugs." + "reason": "CamemBERT MLM prediction regression and Qwen3.5 recurrent-state reset are different model-specific bugs." } ] }, @@ -4467,8 +4428,8 @@ "issue:44186|issue:44361", "issue:43421|issue:45812", "issue:42915|issue:43322", - "issue:43526|issue:43881", "issue:43540|issue:43782", + "issue:43526|issue:43881", "issue:36331|issue:41720", "issue:43316|issue:44913", "issue:29159|issue:36032", @@ -4477,216 +4438,216 @@ "issue:43550|issue:43653", "issue:43606|issue:43720", "issue:30064|issue:34634", - "issue:45406|issue:45464", "issue:43873|issue:45704", + "issue:45406|issue:45464", "issue:44530|issue:46291", "issue:41720|issue:43746", "issue:44387|issue:45701", "issue:45357|issue:46402", "issue:43606|issue:43653", "issue:43531|issue:45759", - "issue:44492|issue:44509", "issue:42490|issue:43329", "issue:34634|issue:41720", + "issue:44492|issue:44509", "issue:34567|issue:35707", "issue:44625|issue:45479" ], - "prepared_review_unit_hash": "27b4c38c93fd645c183408a82e3ff8335b16f14d2d3dd4ac40206d1efad7c9ad", + "prepared_review_unit_hash": "dd94a886ad4eea7cc87d3708e98357cdd3a279bc0e3f92765785adac94389c04", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36307, - "estimated_input_tokens": 9077, - "estimated_eval_tokens": 18410 + "serialized_chars": 36311, + "estimated_input_tokens": 9078, + "estimated_eval_tokens": 18412 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a loose collection of unrelated issues; none of the soft-edge pairs look like the same underlying bug or change, so all should remain separate. The most representative issue is the broad Granite tokenizer regression because it is concrete, high-impact, and has the strongest evidence of scope.", - "confidence": 0.61, - "canonical_issue_reason": "issue:45812 is the strongest representative: it describes a concrete, broad regression affecting all Granite models, with clear symptoms and inbound references.", + "summary": "No soft pair looks like the same underlying bug/change. The items span distinct models, tokenizers, processors, configs, and runtime paths, so they should remain separate issues.", + "confidence": 0.97, + "canonical_issue_reason": "issue:43873 is the most central/high-activity report in this set (highest discussion plus multiple inbound references) and is the broadest bug description among the issues.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the best overall issue in this set because it is specific, high-impact, and well-scoped compared with the other mostly isolated bug reports.", + "best_issue_reason": "issue:43873 is the best representative issue because it has the strongest signal of impact and follow-on references, even though this set does not form a true duplicate cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45081", "right": "issue:45812", "accept": false, - "reason": "Different tokenizer bugs in different model families; one is a Mistral regex/backend_tokenizer crash, the other is Granite token IDs regression." + "reason": "Different tokenizer families and failure modes: Mistral regex patching vs Granite token IDs regression." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Unrelated model/config issues: Qwen3-MoE sliding window behavior vs StableLmConfig pad_token_idx missing after v5 update." + "reason": "Unrelated model/config bugs: Qwen3-MoE sliding window behavior vs StableLm pad_token_idx after upgrade." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "Both are tokenizer-related, but they affect different tokenizers, task flows, and failure modes; not the same bug." + "reason": "Both tokenizer-related, but they concern different classes and errors: LayoutLMv2 NER/padding crash vs MLuke AttributeError." }, { "left": "issue:43421", "right": "issue:45812", "accept": false, - "reason": "Runtime special-token/post-processor sync is a different problem from Granite AutoTokenizer producing wrong IDs." + "reason": "Runtime post-processor update bug is distinct from Granite AutoTokenizer generating wrong token IDs." }, { "left": "issue:42915", "right": "issue:43322", "accept": false, - "reason": "Different model-loading failures with different symptoms and causes; no shared code-path evidence." + "reason": "Different domains: Qwen3Moe FP8 config failure vs Llava Next segmentation fault on load." }, { - "left": "issue:43526", - "right": "issue:43881", + "left": "issue:43540", + "right": "issue:43782", "accept": false, - "reason": "Image label reduction bug vs unrelated GLM loading failure." + "reason": "Different Qwen multimodal paths and symptoms: video input ValueError vs weight_only load error." }, { - "left": "issue:43540", - "right": "issue:43782", + "left": "issue:43526", + "right": "issue:43881", "accept": false, - "reason": "Different Qwen multimodal issues: video input processing vs from_pretrained weight_only loading." + "reason": "Beit image-processor label reduction bug is unrelated to glm-4v-9b loading failure." }, { "left": "issue:36331", "right": "issue:41720", "accept": false, - "reason": "Trainer compute_loss signature mismatch is unrelated to Qwen3 auto device mapping CUDA assertion." + "reason": "Trainer API signature mismatch is unrelated to Qwen3 auto device mapping cudaErrorAssert." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Both are config/API issues, but they concern different models and different fields; not the same underlying defect." + "reason": "Both are config issues, but different fields/models: Gemma3TextConfig API discrepancy vs GPTNeoX rotary_pct reload regression." }, { "left": "issue:29159", "right": "issue:36032", "accept": false, - "reason": "General tokenizer inconsistency vs a specific T5Tokenizer load-time AttributeError; different bugs." + "reason": "Tokenizer inconsistency vs T5Tokenizer method-name conflict are different bugs with different root causes." }, { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "MacOS bus error with a CLIP community model is unrelated to progressive generation with inputs_embeds/past_key_values." + "reason": "MacOS CLIP bus error is unrelated to progressive generation with inputs_embeds/past_key_values." }, { "left": "issue:43388", "right": "issue:45059", "accept": false, - "reason": "Different subsystems: gather_for_metrics batching vs SAM3 text/bbox behavior." + "reason": "Metrics batch-label dropping and SAM3 text/bbox behavior are different code paths and symptoms." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "torch.compile/SDPA failure in Bamba is unrelated to BigBirdTokenizer special-token registration." + "reason": "torch.compile SDPA failure for Bamba is unrelated to BigBirdTokenizer special-token registration." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "Both involve loading/offload, but one is a CPU offload device mismatch and the other is unpacking packed BitNet weights during accelerate loading." + "reason": "Both involve loading/offloading, but they are separate bugs: CPU-offload device mismatch vs packed-weight unpacking during accelerate loading." }, { "left": "issue:30064", "right": "issue:34634", "accept": false, - "reason": "Image processor segmentation-map handling and BarkProcessor voice_preset are unrelated." + "reason": "Image segmentation-map processing bug is unrelated to BarkProcessor voice_preset handling." }, { - "left": "issue:45406", - "right": "issue:45464", + "left": "issue:43873", + "right": "issue:45704", "accept": false, - "reason": "Both are serving/inference issues, but they affect different models and failure paths." + "reason": "Quantization/offloading behavior is a different bug from T5's apex FusedRMSNorm memory leak." }, { - "left": "issue:43873", - "right": "issue:45704", + "left": "issue:45406", + "right": "issue:45464", "accept": false, - "reason": "Quantization offloading behavior and a T5 apex RMSNorm memory leak are unrelated." + "reason": "Different API surfaces and failures: Gemma4Processor missing _tokenizer vs Qwen3.5 streaming chat/completions failure." }, { "left": "issue:44530", "right": "issue:46291", "accept": false, - "reason": "PagedAttentionCache linear_attention crash and kernels version/revision import failure are unrelated." + "reason": "PagedAttentionCache linear_attention crash is unrelated to kernels package version/revision import failure." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "CUDA assertion with Qwen3 auto device mapping is a different code path from GraniteSpeech local checkpoint loading with PEFT adapters." + "reason": "Qwen3 auto device mapping cuda error is unrelated to GraniteSpeech PEFT local checkpoint loading." }, { "left": "issue:44387", "right": "issue:45701", "accept": false, - "reason": "Int4 quantization OOM in Transformers 5.x is unrelated to tokenization changes across versions." + "reason": "CUDA memory growth under int4 quantization is unrelated to a broad tokenization-version change report." }, { "left": "issue:45357", "right": "issue:46402", "accept": false, - "reason": "Both are multimodal/model-loading issues, but one is incorrect visual encoder key saving and the other is missing text encoder weight loading." + "reason": "Both are multimodal save/load issues, but one is wrong visual encoder keys on save and the other is missing text encoder weights on load." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "Device mismatch during CPU offload is unrelated to BigBirdTokenizer special-token registration." + "reason": "Different tokenizer/model subsystems and unrelated failure modes." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and unwrapping text_config for composite Qwen multimodal configs are different defects." - }, - { - "left": "issue:44492", - "right": "issue:44509", - "accept": false, - "reason": "A documentation typo is not the same as stale docs still mentioning removed pipeline tasks." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to AutoModelForCausalLM from_config text_config unwrapping for Qwen multimodal configs." }, { "left": "issue:42490", "right": "issue:43329", "accept": false, - "reason": "Local-path import warning and undefined video-branch variables in _get_num_multimodal_tokens are unrelated." + "reason": "Trailing-slash import warning is unrelated to multimodal token counting video-branch undefined names." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping CUDA assert are unrelated." + "reason": "BarkProcessor voice_preset bug is unrelated to Qwen3 device-mapping cuda assertions." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategy docs/code is not the same as stale docs listing removed pipeline tasks." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "TrainerState token counting and progressive generation with cached embeddings are unrelated." + "reason": "TrainerState token counter update bug is unrelated to progressive generation with embeddings and cache reuse." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Qwen3.5 num_labels propagation and a cross-model zero-loss classification bug are different configuration/loss issues." + "reason": "Qwen3.5 num_labels propagation is a config wiring bug, while the other is a classification loss degeneracy across models." } ] }, @@ -4768,15 +4729,15 @@ "issue:44077|issue:44625", "issue:44062|issue:44451", "issue:45588|issue:45759", - "issue:44368|issue:45357", "issue:43421|issue:43927", + "issue:44368|issue:45357", "issue:44987|issue:45042", "issue:39692|issue:44610", "issue:44918|issue:44936", "issue:43782|issue:45381", "issue:43746|issue:46050", - "issue:43381|issue:43844", "issue:43421|issue:43653", + "issue:43381|issue:43844", "issue:42886|issue:44843", "issue:45704|issue:45923", "issue:46032|issue:46463", @@ -4791,223 +4752,223 @@ "issue:39692|issue:43873", "issue:43596|issue:44220" ], - "prepared_review_unit_hash": "ba5233ee3da20cdf95e8aa4a7b0df4addcb7aa2300cfce4db6e8b51e33749a5d", + "prepared_review_unit_hash": "d0a423acdfe6f5fd1717a0e707e3d7cc9e2dd59c3d949df2f7e4544f67971ce3", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 37110, - "estimated_input_tokens": 9278, - "estimated_eval_tokens": 18812 + "serialized_chars": 37114, + "estimated_input_tokens": 9279, + "estimated_eval_tokens": 18814 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated issue reports; only one pair is a clear duplicate: the two Qwen2_5_VLProcessor.apply_chat_template batched-input/padding=False crashes. The rest are different bugs in different models/code paths and should not be merged as duplicates.", - "confidence": 0.89, - "canonical_issue_reason": "issue:44545 is the best canonical issue for the only true duplicate pair: it matches issue:44514 exactly, and it has slightly more discussion/inbound activity, making it the better representative report.", + "summary": "The cluster is mostly heterogeneous, with two clear duplicate pairs standing out: 44514/44545 and 42886/44843. The Mistral-regex pair 45081/45356 should be rejected as related tokenizer regressions, not proven the same bug.", + "confidence": 0.69, + "canonical_issue_reason": "If one issue must anchor the cluster, 44545 is the best local canonical because it is an exact duplicate of 44514 and has slightly more inbound context; the set is otherwise mixed, so this is only a local choice.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44545 is the strongest issue to keep overall because it is the exact duplicate pair representative and has the most activity among the matched reports.", + "best_issue_reason": "44545 is the strongest representative issue in the cluster because it sits in the clearest exact-duplicate pair (44514/44545) and is well-supported by the packet.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading regressions, but they affect different models and likely different code paths; not the same underlying bug." + "reason": "Both are tokenizer-loading problems, but they affect different models and the packet does not show the same underlying bug." }, { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Exact same crash description for Qwen2_5_VLProcessor.apply_chat_template with batched input and padding=False; these are duplicates." + "reason": "Near-identical title and symptom; these appear to be the same batched apply_chat_template crash." }, { "left": "issue:43531", "right": "issue:44589", "accept": false, - "reason": "Completely different failures: Qwen3-MoE sliding_window behavior vs a Float8 storage TypeError." + "reason": "Different failures: Qwen3-MoE sliding-window behavior vs a Float8 storage TypeError." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Different subsystems and symptoms: Perceiver image resolution handling vs dtype mismatch in bfloat16 inference." + "reason": "Different subsystems and symptoms: Perceiver resize/interpolation vs dtype mismatch in inference." }, { "left": "issue:45759", "right": "issue:45859", "accept": false, - "reason": "Both involve Qwen-family models, but one is from_config/text_config unwrapping and the other is CPU-offload rotary embedding incompatibility; different bugs." + "reason": "Multimodal config unwrapping and CPU-offload incompatibility are different code-path problems." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "No shared underlying bug; one is a post_init contract issue, the other is num_labels propagation in Qwen3.5." + "reason": "A post_init restriction on patchtsmixer is unrelated to Qwen3.5 num_labels propagation." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "Different tokenizer problems with no concrete overlap: AddedToken keyword collision vs model/tokenizer load failure for ScandiBERT." + "reason": "An AddedToken keyword collision is unrelated to a model-specific tokenizer load failure." }, { "left": "issue:45588", "right": "issue:45759", "accept": false, - "reason": "Different code paths and failure modes: flash_attention sink-less handling vs Qwen composite config unwrapping." + "reason": "Flash-attention sink handling is unrelated to Qwen multimodal config unwrapping." }, { - "left": "issue:44368", - "right": "issue:45357", + "left": "issue:43421", + "right": "issue:43927", "accept": false, - "reason": "A warning about tie_word_embeddings is unrelated to the incorrect visual encoder keys saved in Qwen3.5." + "reason": "Both are tokenizer-related, but runtime special-token post-processing is not the same bug as save/load loss of custom token IDs." }, { - "left": "issue:43421", - "right": "issue:43927", + "left": "issue:44368", + "right": "issue:45357", "accept": false, - "reason": "Both involve token/config metadata, but one is runtime post-processor updating and the other is save/load token ID loss causing IndexError; not the same bug." + "reason": "A training warning about tie_word_embeddings does not match a visual-encoder save_pretrained regression." }, { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Different regressions: model loading failure for physical-intelligence/fast vs PIL image processors incorrectly requiring torchvision." + "reason": "A model-loading regression and a PIL image processor dependency issue are different problems." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "Both mention vision/model mismatches, but one is a SigLIP2 doc/example plus quantization issue and the other is an OmDet-Turbo processor size mismatch." + "reason": "Both mention processor/model mismatches, but the concrete failures are different models and different symptoms." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Different trainer failures: Qwen3.5 embedding unpacking in TRL SFT vs evaluate() failing after train()." + "reason": "TRl embedding unpacking failure is unrelated to trainer.evaluate() failing after trainer.train()." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Different Qwen vision-language bugs: weight_only load error vs incorrect video vision_position_ids." + "reason": "Different Qwen-VL bugs: weight_only load failure vs wrong vision_position_ids in video input." }, { "left": "issue:43746", "right": "issue:46050", "accept": false, - "reason": "Local PEFT checkpoint loading failure is unrelated to the sinq quantization save/load regression." + "reason": "PEFT local checkpoint loading and sinq quantized save/load corruption are different issues." }, { - "left": "issue:43381", - "right": "issue:43844", + "left": "issue:43421", + "right": "issue:43653", "accept": false, - "reason": "Gradient checkpointing in eval mode is a separate behavior from exploding gradients with HfDeepSpeedConfig + ZeRO-3." + "reason": "Both involve tokenizer state, but the concrete bugs differ: runtime post-processor updates vs BigBird mask-token registration." }, { - "left": "issue:43421", - "right": "issue:43653", + "left": "issue:43381", + "right": "issue:43844", "accept": false, - "reason": "Both concern special-token handling, but one is post-processor updates and the other is BigBirdTokenizer mask token registration; too different to merge." + "reason": "Gradient checkpointing in eval mode is unrelated to ZeRO-3 gradient growth." }, { "left": "issue:42886", "right": "issue:44843", - "accept": false, - "reason": "Both involve HF_HUB_OFFLINE tokenizer failures, but 44843 is specifically the mistral regex/model_info call; 42886 is broader and not clearly the same concrete bug." + "accept": true, + "reason": "Same offline tokenizer-loading regression; 44843 identifies the concrete _patch_mistral_regex model_info() call that breaks HF_HUB_OFFLINE." }, { "left": "issue:45704", "right": "issue:45923", "accept": false, - "reason": "A T5 RMSNorm memory-leak issue is unrelated to a missing supports_gradient_checkpointing flag on Nemotron-3-Nano-Omni." + "reason": "A T5 RMSNorm memory leak is unrelated to a missing gradient-checkpointing flag." }, { "left": "issue:46032", "right": "issue:46463", "accept": false, - "reason": "Mamba2Mixer use_cache correctness and T5 decoder-input validation error messaging are unrelated." + "reason": "Mamba2 use_cache correctness and a T5 decoder-input error message are not the same issue." }, { "left": "issue:45704", "right": "issue:45859", "accept": false, - "reason": "Different problems: apex RMSNorm memory leak vs Qwen3_5MoeTextRotaryEmbedding CPU offload incompatibility." + "reason": "A normalization memory leak is unrelated to Qwen3_5Moe CPU-offload behavior." }, { "left": "issue:41720", "right": "issue:45704", "accept": false, - "reason": "Qwen3 auto device mapping cudaErrorAssert is unrelated to a T5 apex RMSNorm memory leak." + "reason": "Auto device mapping cuda asserts and a T5 RMSNorm leak have no common underlying change." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "TokenizersBackend post-processor updates and LayoutLMv2Tokenizer NER/padding crashes are different tokenizer bugs." + "reason": "Different tokenizer failures: special-token post-processing vs LayoutLMv2 NER padding/truncation crashes." }, { "left": "issue:45663", "right": "issue:45923", "accept": false, - "reason": "Gemma-4 FSDP2 shared_kv_states handling is unrelated to Nemotron gradient-checkpointing flag metadata." + "reason": "Gemma-4 FSDP2 shared_kv_states bug is unrelated to a missing capability flag on Nemotron-3-Nano-Omni." }, { "left": "issue:43531", "right": "issue:43645", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is a model/runtime issue, not the Jupyter custom-model initialization regression." + "reason": "Qwen3-MoE sliding-window behavior is unrelated to Jupyter custom-model initialization in Transformers 5.0.0." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "Different vision-language/model bugs: SigLIP2 nonsensical outputs vs CamemBERT masked LM prediction regression." + "reason": "Different model regressions with different symptoms: SigLIP2 pipeline nonsense vs CamemBERT masked-LM prediction errors." }, { "left": "issue:45081", "right": "issue:45356", "accept": false, - "reason": "Both touch tokenizer regex logic, but 45081 is an AttributeError on backend_tokenizer and 45356 is a codec/warning regression for Kimi-K2.5." + "reason": "They share the Mistral-regex area, but the packet does not prove the same underlying bug rather than two related tokenizer regressions." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "PEFT local checkpoint loading and quantization/offloading behavior are not the same failure mode." + "reason": "Local PEFT checkpoint loading and quantization/offloading behavior are distinct issues." }, { "left": "issue:39692", "right": "issue:43873", "accept": false, - "reason": "SigLIP2 documentation/quantization errors are unrelated to quantization offloading behavior." + "reason": "SigLIP2 docs/model-processor mismatch plus quantization failure is not the same as general offloading with quantization." }, { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "Deepspeed ZeRO-3/BertModel index errors and _torch_extract_fbank_features() are unrelated." + "reason": "A ZeRO-3/BertModel indexing error is unrelated to _torch_extract_fbank_features()." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: the only clearly duplicated pair is the two Qwen2_5_VLProcessor.apply_chat_template batched-input/padding=False reports, and the canonical choice of issue:44545 is supported by the slightly higher activity in the packet. The other soft-edge rejections are conservative and do not overstate duplicate evidence based on the titles/data provided." + "accept": false, + "feedback": "Most of the conservative false/true calls are grounded, but one accepted edge looks overstated: issue:45081 vs issue:45356 share the general _patch_mistral_regex area, yet the packet does not show they are the same underlying bug rather than two related tokenizer regressions. Also, the canonical-issue rationale for 44545 leans on \u2018clearer report,\u2019 which is not directly supported by the packet. The cluster otherwise looks mixed, so I\u2019d reject on precision grounds." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -5082,8 +5043,8 @@ "issue:43653|issue:43720", "issue:44220|issue:44661", "issue:43550|issue:45072", - "issue:43618|issue:43697", "issue:43295|issue:45834", + "issue:43618|issue:43697", "issue:44062|issue:44488", "issue:43550|issue:43819", "issue:43540|issue:44008", @@ -5092,8 +5053,8 @@ "issue:45183|issue:45479", "issue:43322|issue:45685", "issue:43531|issue:43643", - "issue:43335|issue:43756", "issue:44843|issue:45081", + "issue:43335|issue:43756", "issue:43388|issue:46032", "issue:43716|issue:44008", "issue:43643|issue:44877", @@ -5103,202 +5064,202 @@ "issue:43950|issue:45987", "issue:44821|issue:45084" ], - "prepared_review_unit_hash": "141e4b756d2cd82c7dc1c733232c6e4bf2f5325105e5839858b9f44bc99344e2", + "prepared_review_unit_hash": "e5386a338e95e8f253d297c84940dfb5c2f8c727d1b1f2e7db5259e51ee46cf1", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35902, - "estimated_input_tokens": 8976, - "estimated_eval_tokens": 18208 + "serialized_chars": 35898, + "estimated_input_tokens": 8975, + "estimated_eval_tokens": 18206 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a heterogeneous set of mostly unrelated issues. I do not see a true duplicate cluster or a single clear canonical issue/PR. All soft-edge pairs look like false positives based on different root causes, models, or code paths.", + "summary": "The cluster is highly heterogeneous: it mixes unrelated tokenizer, multimodal, config, inference, and training bugs across different models and code paths. I do not see a single canonical issue or PR that would safely represent the whole set.", "confidence": 0.97, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No issue cleanly subsumes the others; the items span unrelated regressions and model-specific bugs. If forced to pick a representative, none would be reliable enough for canonicalization.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45356", "right": "issue:45907", "accept": false, - "reason": "Different subsystems and failures: tokenizer regex/regression vs Hub file listing forwarding. No shared underlying bug." + "reason": "Different subsystems: tokenizer regression vs Hub file-listing revision forwarding. No shared underlying bug." }, { "left": "issue:42915", "right": "issue:43746", "accept": false, - "reason": "Different models and causes: Qwen3Moe FP8 config failure vs GraniteSpeech PEFT/local checkpoint loading. Not the same change or code path." + "reason": "Different models and failures: Qwen3Moe FP8 config vs GraniteSpeech PEFT checkpoint loading." }, { "left": "issue:43540", "right": "issue:46294", "accept": false, - "reason": "Both involve Qwen video input, but the models and failure modes differ; not enough evidence of one underlying bug." + "reason": "Both are Qwen video issues, but one is a video-processing ValueError and the other is chat-template feature/token mismatch; separate defects." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Image segmentation map processing vs GenerationMixin import error are unrelated." + "reason": "Unrelated: image processor segmentation-map handling vs GenerationMixin import error." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "BitNet accelerate loading/unpacking issue vs DAC forward/from_latents mismatch. Different components and fixes." + "reason": "Different code paths: accelerate loading/unpacked BitNet weights vs DAC forward/from_latents STE mismatch." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "Tokenizer special-token decode bug vs BitNet weight unpacking bug; unrelated code paths." + "reason": "Tokenizer special-token registration issue vs BitNet weight unpacking bug; unrelated." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Audio feature extraction issue vs tokenizer mapping/model registration problem. Not the same bug." + "reason": "Audio feature extraction bug vs tokenizer mapping lookup failure; no same underlying change." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "Different model families and symptoms: torch.compile/SDPA on Bamba vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + "reason": "Both are model inference bugs, but in different models and mechanisms: SDPA/torch.compile vs bf16 dtype mismatch." }, { - "left": "issue:43618", - "right": "issue:43697", + "left": "issue:43295", + "right": "issue:45834", "accept": false, - "reason": "CLIP attentions assignment regression vs RTDetrV2 output difference in v5; both are output regressions but not the same code path." + "reason": "Both are multimodal regressions, but one is processor.tokenizer/image-to-tokenizer breakage and the other is OCR long-input index error." }, { - "left": "issue:43295", - "right": "issue:45834", + "left": "issue:43618", + "right": "issue:43697", "accept": false, - "reason": "Processor/tokenizer regression vs Kosmos2.5 OCR index error. Different model behaviors and likely unrelated fixes." + "reason": "Different regressions: CLIP output attentions field vs RTDetrV2 output changes." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "AddedToken argument duplication vs tokenizer loading failure for a specific model. Similar area, but not the same underlying defect." + "reason": "One is AddedToken constructor misuse; the other is loading a specific model/tokenizer failure. Symptoms and fixes differ." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Separate model-specific correctness problems; no shared failure mechanism beyond both being inference bugs." + "reason": "Unrelated model bugs: SDPA compile failure vs DAC latent reconstruction mismatch." }, { "left": "issue:43540", "right": "issue:44008", "accept": false, - "reason": "Qwen video-input processing error vs Gemma3n variable name collision in forward. Different models and code paths." + "reason": "Different multimodal paths: Qwen3OmniMoe video processing vs Gemma3n audio mask variable collision." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "LayoutLMv2 tokenizer padding/NER crash vs Perceiver image resolution/interpolation failure. Unrelated." + "reason": "Tokenizer padding/NER crash vs image-classification resolution/interpolation bug; no common root cause." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "Tokenizer special-token handling vs DAC latent decoding mismatch. No shared root cause." + "reason": "BigBird special-token registration is unrelated to DAC latent/STE behavior." }, { "left": "issue:45183", "right": "issue:45479", "accept": false, - "reason": "Input validation error messages vs zero-loss classification regression. Different semantics and fixes." + "reason": "Both touch sequence classification, but one is input-validation messaging and the other is zero-loss behavior; distinct bugs." }, { "left": "issue:43322", "right": "issue:45685", "accept": false, - "reason": "Llava Next loading segfault vs MPS histogram kernel not implemented. Completely different problems." + "reason": "Completely different areas: Llava Next load crash vs MPS histogram implementation in MoE." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE vs AutoConfig trust_remote_code field loss. Not the same bug." + "reason": "Qwen3-MoE sliding-window issue vs AutoConfig trust_remote_code field-loss regression; unrelated." }, { - "left": "issue:43335", - "right": "issue:43756", + "left": "issue:44843", + "right": "issue:45081", "accept": false, - "reason": "SwitchTransformers sparse-layer construction vs Smollm3 RoPE-layer dropping. Different model config issues." + "reason": "Both involve _patch_mistral_regex, but one is offline model_info access and the other is backend_tokenizer AttributeError; separate failure modes." }, { - "left": "issue:44843", - "right": "issue:45081", + "left": "issue:43335", + "right": "issue:43756", "accept": false, - "reason": "Both touch _patch_mistral_regex, but one is offline model_info behavior and the other is backend_tokenizer AttributeError. Related area, not the same concrete bug." + "reason": "SwitchTransformers sparse-layer config bug vs Smollm3 RoPE-layer-count regression; different model internals." }, { "left": "issue:43388", "right": "issue:46032", "accept": false, - "reason": "Metrics gathering last-batch truncation vs Mamba2 use_cache correctness. Unrelated runtime paths." + "reason": "Evaluation gather/truncation bug vs Mamba2 cache correctness bug; not the same issue." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Image preprocessor/model dtype mismatch vs Gemma3n audio mask attribute collision. Different model components." + "reason": "Different models and mismatches: Mistral image-preprocessor dtype vs Gemma3n audio mask variable collision." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "Both involve config loading, but one is missing fields with trust_remote_code and the other is strict config rejection for granite_speech. Not the same defect." + "reason": "Both are config-loading regressions, but one is missing fields from trust_remote_code and the other is strict config rejection for granite_speech." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Image processor segmentation-map failure vs CustomTrainer compute_loss signature mismatch are unrelated." + "reason": "Image processor segmentation maps vs CustomTrainer.compute_loss signature change; unrelated." }, { "left": "issue:41720", "right": "issue:44991", "accept": false, - "reason": "Qwen3 auto device mapping CUDA assert vs tokenizer loading for EMBEDDIA/est-roberta. Different subsystems." + "reason": "Qwen3 auto device-map CUDA assert vs tokenizer loading failure for a different model; no shared bug." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Both are v5 output regressions, but different models and likely different root causes. Too broad to merge." + "reason": "Both are v5 output regressions, but they affect different models and likely different code paths." }, { "left": "issue:43950", "right": "issue:45987", "accept": false, - "reason": "from_pretrained buffer corruption vs StaticCache return-type contract issue. Different APIs and code paths." + "reason": "from_pretrained buffer restoration bug vs StaticCache return-type contract bug; unrelated." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "AutoImageProcessor URL loading vs template compilation error. No meaningful overlap." + "reason": "AutoImageProcessor URL loading vs Jinja/template compilation error; different subsystems." } ] }, @@ -5386,8 +5347,8 @@ "issue:46294|issue:46463", "issue:34634|issue:36010", "issue:44805|issue:44936", - "issue:44485|issue:44509", "issue:28218|issue:29159", + "issue:44485|issue:44509", "issue:43577|issue:43582", "issue:44561|issue:45020", "issue:45987|issue:46192", @@ -5402,20 +5363,20 @@ "issue:45372|issue:46326", "issue:43582|issue:44857" ], - "prepared_review_unit_hash": "9e02754b085a858eb3247ec6922944e3945ec6e797af914ad203da077ebdb43c", + "prepared_review_unit_hash": "75328e0aaf8484a3e7736861c176582eaf177cfe4b7f97c1bf2bc571f77c4a3d", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36614, + "serialized_chars": 36616, "estimated_input_tokens": 9154, "estimated_eval_tokens": 18564 }, @@ -5424,192 +5385,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a collection of distinct Transformers issues. There is one clear duplicate pair around tokenizer added-token spacing / slow-vs-fast inconsistency, but the other soft pairs are only loosely related by broad subsystem or version changes and should stay separate.", - "confidence": 0.86, - "canonical_issue_reason": "issue:28218 is the clearest concrete tokenizer bug report and is earlier than the paired inconsistency report, so it works well as the canonical issue for that duplicate pair.", + "summary": "This cluster is mostly heterogeneous: most pairs are clearly unrelated bugs across different models, tokenizers, loaders, and docs. Only a couple of tokenizer/pipeline-related pairs look plausibly like duplicates or the same upstream change.", + "confidence": 0.76, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the strongest global issue candidate because it is broad, high-signal, and captures a major recent-version regression theme affecting remote_code models.", + "best_issue_reason": "No single issue cleanly represents the whole cluster. If forced to pick a local representative, issue:43825 is the most concrete of the near-matches, but the cluster is too mixed to treat it as a true canonical.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43577", "right": "issue:43818", "accept": false, - "reason": "Different models and different failures: dtype loading bug vs missing video temporal attention/weight tying." + "reason": "Different bugs and subsystems: BLIP2 dtype handling vs Video-LLaVA weight/temporal-attention issues." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Both involve tokenization/processor behavior, but one is runtime special-token post-processing and the other is Whisper batch decode skip_special_tokens handling." + "reason": "Different tokenizer behaviors: runtime special-token/post-processor sync vs Whisper batch_decode skip_special_tokens." }, { "left": "issue:39692", "right": "issue:45834", "accept": false, - "reason": "Unrelated problems: SigLIP2 docs/example errors versus Kosmos2.5 long-OCR index error." + "reason": "Docs/example problems for SigLIP2 vs a Kosmos2.5 OCR index error; not the same bug." }, { "left": "issue:43540", "right": "issue:45381", "accept": false, - "reason": "Both are video-input issues, but they concern different Qwen models and different code paths." + "reason": "Both are video-input issues, but they affect different Qwen models and likely different code paths." }, { "left": "issue:45584", "right": "issue:45834", "accept": false, - "reason": "Whisper empty-transcription generation bug is unrelated to Kosmos2.5 OCR indexing." + "reason": "Whisper empty-transcription failure after align_special_tokens is unrelated to Kosmos2.5 OCR indexing." }, { "left": "issue:44186", "right": "issue:44442", "accept": false, - "reason": "Tokenizer-related, but one is LayoutLMv2 NER padding/truncation and the other is AutoTokenizer loading FastSpeech2ConformerTokenizer." + "reason": "LayoutLMv2 tokenizer NER/padding crash is unrelated to AutoTokenizer loading FastSpeech2ConformerTokenizer." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Different Qwen VL issues: one is from_pretrained weight_only loading, the other is flash-attention generation behavior." + "reason": "Qwen3VL weight_only loading error vs Qwen3.5 flash-attention generation bug; different failure modes." }, { "left": "issue:43644", "right": "issue:43749", "accept": false, - "reason": "Both concern loading, but one is non-persistent buffer initialization and the other is FSDP CPU RAM efficient loading." + "reason": "Non-persistent buffer corruption and FSDP CPU RAM efficient loading are different loading-time bugs." }, { "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "Different failure modes and models: CPU offload device mismatch vs bfloat16 dtype mismatch." + "reason": "CPU offload device mismatch in Bark differs from dtype mismatches in SwitchTransformers/TimmWrapperModel." }, { "left": "issue:44479", "right": "issue:45834", "accept": false, - "reason": "Qwen video regression and Kosmos2.5 OCR index error are unrelated." + "reason": "Qwen video-input regression is unrelated to Kosmos2.5 long-OCR index errors." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "One is a v5 error-message wording issue; the other is a SigLIP2 model/pipeline correctness bug." + "reason": "Pipeline-task messaging/docs around v5 removals vs SigLIP2 example/model mismatch; not the same issue." }, { "left": "issue:46294", "right": "issue:46463", "accept": false, - "reason": "Both are validation errors, but they affect different models and different validations." + "reason": "Qwen3.5 video chat-template mismatch and T5 decoder validation messaging are unrelated." }, { "left": "issue:34634", "right": "issue:36010", "accept": false, - "reason": "BarkProcessor voice_preset and GenerationMixin import error are unrelated." + "reason": "BarkProcessor voice_preset failure is unrelated to the GenerationMixin import error." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "The first is a shape-mismatch IndexError; the second is a trainer lifecycle failure after train()." - }, - { - "left": "issue:44485", - "right": "issue:44509", - "accept": false, - "reason": "GLM-5 RoPE implementation discussion and docs still mentioning removed pipelines are not the same bug." + "reason": "Mask/tensor shape mismatch is unrelated to trainer.evaluate() failing after train()." }, { "left": "issue:28218", "right": "issue:29159", "accept": true, - "reason": "These describe the same tokenizer bug: added-token spacing / inconsistency between slow and fast tokenizer behavior." + "reason": "Both point to the same tokenizer inconsistency around added tokens / slow-vs-fast behavior, with the extra-space symptom fitting the broader mismatch." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE implementation discussion is unrelated to docs still mentioning removed pipeline tasks." }, { "left": "issue:43577", "right": "issue:43582", "accept": false, - "reason": "Different runtime problems: Blip2 dtype state vs Apple Silicon caching_allocator_warmup TypeError." + "reason": "BLIP2 dtype retention and Apple Silicon allocator warmup TypeError are unrelated." }, { "left": "issue:44561", "right": "issue:45020", "accept": false, - "reason": "The latter is a broad umbrella for remote_code breakage; the former is one specific cause, so they are not clearly the same issue." + "reason": "The former is a specific missing-helper regression; the latter is a broad umbrella for remote_code breakages, not the same concrete bug." }, { "left": "issue:45987", "right": "issue:46192", "accept": false, - "reason": "StaticCache return type bug and return_tensors keyword handling are unrelated." + "reason": "StaticCache seq-length contract issue is unrelated to return_tensors being ignored when text_kwargs is passed." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "Packed-weight accelerate loading failure is unrelated to bfloat16 dtype mismatches." + "reason": "BitNet packed-weight unpacking during accelerate loading is unrelated to dtype mismatches in other models." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "OmDet-Turbo image-size mismatch is unrelated to dtype mismatch issues." + "reason": "OmDet-Turbo processor/model input-size mismatch is unrelated to bfloat16 dtype mismatches." }, { "left": "issue:44671", "right": "issue:45701", "accept": false, - "reason": "CamemBERT masked-LM prediction regression and tokenization version changes are too different to treat as duplicates." + "reason": "CamemBERT masked-LM regression is not the same as a general tokenization-version-change report." }, { "left": "issue:43825", "right": "issue:43827", - "accept": false, - "reason": "Both mention v5 pipeline removals, but one is an error message bug and the other is stale documentation." + "accept": true, + "reason": "Both concern the v5 pipeline-removal fallout: one is the runtime error/help text, the other is stale docs still referencing removed pipeline tasks." }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "Both involve Qwen VL, but one is a loading error and the other is a RoPE position-id scaling bug." + "reason": "Qwen3VL weight_only loading error and Qwen2.5-VL rope-index scaling are different issues." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "Both are model-loading regressions, but they hit different APIs and failure points." + "reason": "FSDP CPU RAM loading failure and init_empty_weights _is_hf_initialized TypeError are distinct model-loading regressions." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Different AttributeError failures in different components: Mistral regex patching vs Gemma4Processor tokenization state." + "reason": "Mistral tokenizer regex patch crash is unrelated to Gemma4Processor missing _tokenizer in transformers serve." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "TensorFlow backend packaging and Mistral tokenizer patching are unrelated." + "reason": "Missing TensorFlow in a torch install extra is unrelated to the Mistral tokenizer backend crash." }, { "left": "issue:45372", "right": "issue:46326", "accept": false, - "reason": "Both involve mistral_common, but one is a missing import during Gemma 4 loading and the other is a continue_final_message tokenizer behavior bug." + "reason": "Gemma4 processor import failure due to ReasoningEffort is unrelated to continue_final_message behavior for mistral_common tokenizers." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "Apple Silicon warmup TypeError and CUDA float16 AMP crash are unrelated hardware/runtime issues." + "reason": "Apple Silicon allocator warmup TypeError is unrelated to LwDetrImageLoss AMP/CUDA crashes." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet, and the soft-edge decisions are appropriately conservative overall. The accepted tokenizer pair is plausibly mergeable as a single bugfix, and the other pairs are reasonably kept separate based on distinct models, APIs, or failure modes." + "feedback": "The summary is grounded in the packet: the cluster is indeed mostly heterogeneous, and the two accepted pairs are conservative matches based on the titles alone. I don\u2019t see any overstatement of evidence in the soft-edge verdicts." }, "evaluator_used": true, "retried": false, @@ -5688,8 +5649,8 @@ "issue:43653|issue:44912", "issue:44060|issue:44368", "issue:45084|issue:45538", - "issue:43526|issue:44554", "issue:43742|issue:45020", + "issue:43526|issue:44554", "issue:43329|issue:45600", "issue:44589|issue:45042", "issue:43452|issue:44991", @@ -5699,8 +5660,8 @@ "issue:43550|issue:44912", "issue:43819|issue:43927", "issue:44387|issue:44617", - "issue:43257|issue:43329", "issue:36010|issue:43645", + "issue:43257|issue:43329", "issue:43720|issue:44610", "issue:43854|issue:44977", "issue:38175|issue:44610", @@ -5711,214 +5672,214 @@ "issue:45459|issue:45803", "issue:43262|issue:44811" ], - "prepared_review_unit_hash": "f1a50878cad5b333e726b7683bb80d22fb9af26f5885d7a6037de8f6bd8c2b21", + "prepared_review_unit_hash": "6901391969a54bcc8855f9fe4bb95bca5b429c6ad772fe16e031d5ec9fad5e75", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36831, - "estimated_input_tokens": 9208, - "estimated_eval_tokens": 18672 + "serialized_chars": 36825, + "estimated_input_tokens": 9207, + "estimated_eval_tokens": 18670 }, "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 Transformers issues across loading, quantization, processors, training, cache, and model-specific regressions. I found no pair that is clearly the same underlying bug/change, so no canonical issue or PR emerges.", - "confidence": 0.92, - "canonical_issue_reason": "No single canonical issue: the items cover distinct failures in different subsystems and model families, with only superficial thematic overlap.", + "summary": "This cluster is heterogeneous: the soft pairs are mostly false positives spanning unrelated regressions in loading, audio, vision, training, cache, and quantization. I would not merge any of the suggested pairs. If one issue has to anchor triage, the broad remote_code regression issue is the closest umbrella, but it does not truly duplicate the rest.", + "confidence": 0.43, + "canonical_issue_reason": "issue:45020 is the broadest umbrella in the set: it describes a general regression where recent Transformers versions break remote_code models, which is the most plausible triage sink among otherwise unrelated bugs.", "canonical_pr_reason": null, - "best_issue_reason": "No best issue candidate because there is no coherent duplicate cluster to anchor on; the closest matches are still different bugs.", + "best_issue_reason": "issue:45020 is the most representative high-level regression issue in this cluster; it is broad enough to serve as an anchor for triage, even though it is not a true duplicate of the other items.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43906", "right": "issue:46209", "accept": false, - "reason": "Unrelated issues: one is a reproduction of an upstream correctness issue, the other is FP8 inference failure on B200 hardware." + "reason": "Both touch FP8/model loading, but one is an isolated reproduction and the other is a B200 inference failure; different bugs and code paths." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Different code paths: audio chat-template sampling-rate defaulting vs DAC latent reconstruction mismatch." + "reason": "Audio chat-template sampling-rate defaulting and DAC latent/forward mismatch are unrelated defects." }, { "left": "issue:45003", "right": "issue:46285", "accept": false, - "reason": "Different problems: unsafe sys.modules access in modeling_utils vs CLIP text-model structure inconsistency." + "reason": "Unsafe sys.modules access is unrelated to the CLIP text-model structure inconsistency." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "Completely different areas: BarkProcessor voice preset vs trainer compute_loss signature change." + "reason": "Bark voice_preset and CustomTrainer compute_loss signature breakage are completely different areas." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "Different bugs: BeitImageProcessorFast reduce_labels behavior vs RTDetrV2 output mismatch in v5." + "reason": "BEiT label reduction and RTDetrV2 output divergence are different model/processor bugs." }, { "left": "issue:43653", "right": "issue:44912", "accept": false, - "reason": "Unrelated: BigBirdTokenizer special-token registration vs GGUF loading fallback/quantization behavior." + "reason": "BigBirdTokenizer special-token registration and MXFP4 quantization loading are unrelated." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "Similar theme of tied-weight warnings, but different models and symptoms; not clearly the same concrete bug." + "reason": "Both mention tied-weights warnings, but they involve different Qwen models and likely different root causes." }, { "left": "issue:45084", "right": "issue:45538", "accept": false, - "reason": "Different failures: template compilation error vs CLIPTokenizer max_length constant misuse." + "reason": "Template-node compilation error and CLIPTokenizer max_length constant are unrelated." }, { - "left": "issue:43526", - "right": "issue:44554", + "left": "issue:43742", + "right": "issue:45020", "accept": false, - "reason": "No overlap beyond being model/library issues; one is image-label reduction, the other an MPS attention correctness bug." + "reason": "MobileLLM key error is a specific loading failure, while remote_code breakage is a broad version regression; not the same bug." }, { - "left": "issue:43742", - "right": "issue:45020", + "left": "issue:43526", + "right": "issue:44554", "accept": false, - "reason": "One is a model load KeyError, the other is a broad remote_code regression; too different to merge." + "reason": "BEiT label reduction and MPS attention correctness are unrelated." }, { "left": "issue:43329", "right": "issue:45600", "accept": false, - "reason": "Different bugs in different areas: multimodal token counting video branch vs removed Sam3LiteText auto-mapping references." + "reason": "Undefined video-token helper usage and stale auto_mappings config references are distinct issues." }, { "left": "issue:44589", "right": "issue:45042", "accept": false, - "reason": "Unrelated: missing Float8 storage handling vs unnecessary torchvision requirement in PIL processors." + "reason": "Float8 storage lookup failure is unrelated to PIL backend processors requiring torchvision." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "Different loading regressions affecting different artifacts and APIs; not the same underlying bug." + "reason": "gguf_file handling for tokenizer/model loading is different from the est-roberta tokenizer regression." }, { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "Different training failures: zero3/non-pretrained BERT index error vs post-train evaluate failure." + "reason": "DeepSpeed ZeRO3 index error during training and evaluate-after-train failure are different trainer issues." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "No relation: placeholder issue text vs a concrete reproduction of an upstream issue." + "reason": "The placeholder issue title has no concrete technical overlap with the FP8 reproduction issue." }, { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "Different subsystems: local checkpoint loading with PEFT adapters vs Whisper batch_decode skip_special_tokens behavior." + "reason": "PEFT local checkpoint loading and Whisper batch_decode skip_special_tokens handling are unrelated." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "Different model/runtime failures: torch.compile with SDPA vs MXFP4 quantization fallback on load." + "reason": "Bamba torch.compile+SDPA failure is unrelated to MXFP4 quantization fallback." }, { "left": "issue:43819", "right": "issue:43927", "accept": false, - "reason": "Different bugs: DAC STE mismatch vs DiaConfig losing custom token IDs after save/load." + "reason": "DAC latent reconstruction mismatch and DiaConfig save/load token-ID loss are different defects." }, { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "Both mention OOM, but causes differ: int4 quantization memory regression vs Sam3Video model OOM." + "reason": "Both are OOM-related symptoms, but one is an int4 memory regression and the other is a Sam3Video-specific exhaustion issue." }, { - "left": "issue:43257", - "right": "issue:43329", + "left": "issue:36010", + "right": "issue:43645", "accept": false, - "reason": "Different model families and failures: Qwen3 MoE weight conversion vs multimodal token counting video-branch bug." + "reason": "GenerationMixin import breakage and Jupyter custom-model initialization failure are different release regressions." }, { - "left": "issue:36010", - "right": "issue:43645", + "left": "issue:43257", + "right": "issue:43329", "accept": false, - "reason": "Both are upgrade regressions, but one is a GenerationMixin import breakage and the other is custom-model initialization in notebooks." + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed and multimodal token counting video-branch bugs are unrelated." }, { "left": "issue:43720", "right": "issue:44610", "accept": false, - "reason": "Different issues: BitNet packed-weight loading vs OmDet-Turbo processor/model input-size mismatch." + "reason": "BitNet packed-weight loading and OmDet-Turbo processor size mismatch are separate problems." }, { "left": "issue:43854", "right": "issue:44977", "accept": false, - "reason": "Different model-specific generation/load failures with different triggers and code paths." + "reason": "GLM-4.7-Flash test loading failure and Qwen3.5 flash-attention generation bug do not share a concrete code path." }, { "left": "issue:38175", "right": "issue:44610", "accept": false, - "reason": "Different symptoms and components: SigLIP2 zero probabilities vs OmDet-Turbo processor producing wrong resolution." + "reason": "Unexpected zero probabilities in SigLIP2 and OmDet-Turbo processor resolution mismatch are unrelated." }, { "left": "issue:36010", "right": "issue:36331", "accept": false, - "reason": "Unrelated upgrade regressions: import error for GenerationMixin vs trainer compute_loss signature mismatch." + "reason": "GenerationMixin import regression and CustomTrainer compute_loss signature mismatch are unrelated." }, { "left": "issue:46421", "right": "issue:46424", "accept": false, - "reason": "Related cache area, but one is lazy-init recompilation with chunked prefill and the other is max_cache_length handling for quantized cache only." + "reason": "Both concern cache behavior, but one is about lazy initialization causing recompilation and the other about max_cache_length semantics; not the same bug." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Different domains: Qwen3-MoE sliding_window bug vs PIL image processors requiring torchvision." + "reason": "Qwen3-MoE sliding_window issue and torchvision dependency regression are unrelated." }, { "left": "issue:39692", "right": "issue:44912", "accept": false, - "reason": "Documentation/example errors for SigLIP2 are unrelated to GGUF/MXFP4 loading behavior." + "reason": "SigLIP2 documentation/example errors and MXFP4 quantization loading are different classes of problems." }, { "left": "issue:45459", "right": "issue:45803", "accept": false, - "reason": "Both involve exception handling, but they are distinct locations and failure modes; not the same bug." + "reason": "Tokenizer protobuf error handling and Fuyu bare-except swallowing errors are separate exception-handling bugs." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Both touch audio processing, but one is sampling-rate defaulting and the other is batch_decode skip_special_tokens handling." + "reason": "Audio processor chat-template sampling-rate default and Whisper batch_decode skip_special_tokens are different processor behaviors." } ] }, @@ -6000,8 +5961,8 @@ "issue:34567|issue:40990", "issue:29159|issue:34567", "issue:44568|issue:45356", - "issue:43749|issue:44756", "issue:43606|issue:45859", + "issue:43749|issue:44756", "issue:43257|issue:43931", "issue:42915|issue:43742", "issue:44291|issue:45042", @@ -6011,9 +5972,9 @@ "issue:43697|issue:45357", "issue:43927|issue:44912", "issue:42915|issue:45538", - "issue:44568|issue:45488", "issue:43452|issue:43582", "issue:33357|issue:34567", + "issue:44568|issue:45488", "issue:42175|issue:43824", "issue:41720|issue:43540", "issue:44062|issue:44991", @@ -6021,20 +5982,20 @@ "issue:44479|issue:45381", "issue:28282|issue:34689" ], - "prepared_review_unit_hash": "e8338dbbd40803f2f5ee199031a2f1925abb6b1dd0c24a57743d7847ab65e086", + "prepared_review_unit_hash": "020b54f38b82471e73a85432321c971106df420d75c2e9d8aefd671803c1befd", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35931, + "serialized_chars": 35929, "estimated_input_tokens": 8983, "estimated_eval_tokens": 18222 }, @@ -6043,193 +6004,196 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a grab-bag of mostly unrelated Transformers issues: tokenizer regressions, multimodal/video bugs, model-loading failures, and offload/quantization problems. From titles alone, none of the soft pairs is clearly the same concrete bug with enough confidence to merge.", - "confidence": 0.39, - "canonical_issue_reason": "issue:45488 is the clearest concrete regression report in the set: it names a specific tokenizer override and a visible downstream breakage, while many others are either broader umbrellas or different subsystems entirely.", + "summary": "Mostly unrelated issue reports; only one soft pair looks like the same video-input regression in the Qwen VL family. Everything else is too model-/path-specific to be treated as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "issue:44479 is the best umbrella issue: it describes the broader v5.3.0 video-input regression across multiple Qwen VL/MoE variants, and issue:45381 appears to be a narrower instance of that same bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45488 is the best representative issue overall because it is specific, reproducible, and high impact, whereas most other items in the cluster are either unrelated or too vague to serve as a canonical duplicate target.", + "best_issue_reason": "issue:44479 is the most representative issue because it captures the shared regression across several affected models rather than one model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42490", "right": "issue:43257", "accept": false, - "reason": "Import-warning path handling and Qwen3 MoE weight conversion are unrelated bugs." + "reason": "Different areas: local-path import warning vs Qwen3 MOE weight conversion." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Deepspeed weight conversion failure vs CUDA OOM are different problems." + "reason": "Qwen3 MOE loading/conversion bug vs GPT-oss OOM; different failure mode." }, { "left": "issue:43329", "right": "issue:44479", "accept": false, - "reason": "Both are video-input related, but one is a missing helper in token counting and the other is a broader regression; not enough to prove the same code-path bug." + "reason": "Both are video-related, but one is undefined helpers in token counting and the other is a broader video-input regression." }, { "left": "issue:43454", "right": "issue:43577", "accept": false, - "reason": "Weight tying for AyaVision and Blip2 dtype casting are different model-loading issues." + "reason": "AyaVision weight tying vs BLIP2 dtype casting; unrelated code paths." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "Both touch config loading, but one is missing fields from trust_remote_code and the other is about unwrapping text_config in from_config; different bugs." + "reason": "AutoConfig missing fields with trust_remote_code is not the same as AutoModelForCausalLM.from_config not unwrapping text_config." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Video token counting helper failure and batched video processor shape mismatch are related area-wise but not the same defect." + "reason": "Undefined video-branch helpers vs batched video processor shape bug; different bugs." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "Trainer token-count tracking and perplexity on GPT-oss are unrelated." + "reason": "Trainer token counting metric issue vs high perplexity on a model; no shared root cause." }, { "left": "issue:29159", "right": "issue:34567", "accept": false, - "reason": "Tokenizer inconsistency and TrainerState token-count updates are different subsystems." + "reason": "Tokenizer slow/fast inconsistency vs TrainerState token counter not updating." }, { "left": "issue:44568", "right": "issue:45356", "accept": false, - "reason": "Both are tokenizer regressions, but they describe different concrete failures and model families." + "reason": "Both tokenizer regressions, but for different models and different tokenizer behaviors." }, { - "left": "issue:43749", - "right": "issue:44756", + "left": "issue:43606", + "right": "issue:45859", "accept": false, - "reason": "FSDP RAM-efficient loading and mmap OOM on Strix Halo are different memory/loading issues." + "reason": "Both involve CPU offload, but different models and likely different failing code paths." }, { - "left": "issue:43606", - "right": "issue:45859", + "left": "issue:43749", + "right": "issue:44756", "accept": false, - "reason": "Both involve CPU offload, but the models and failure modes differ; not clearly the same bug." + "reason": "FSDP CPU RAM-efficient loading vs mmap OOM on Strix Halo; different problems." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Qwen3 MoE conversion and Qwen3-VL weight-shape mismatch are different loading failures." + "reason": "Qwen3 MOE weight conversion vs Qwen3-VL shape mismatch; same family only at a high level." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Fine-grained FP8 config failure and MobileLLM key error are unrelated." + "reason": "Qwen3 MOE FP8 config failure vs MobileLLM key error; unrelated." }, { "left": "issue:44291", "right": "issue:45042", "accept": false, - "reason": "init_empty_weights TypeError and PIL-backend torchvision dependency regression are unrelated." + "reason": "init_empty_weights argument handling vs image processor torchvision requirement; unrelated." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "Qwen3 MoE FP8 failure and AutoImageProcessor URL loading are different problems." + "reason": "Qwen3 MOE FP8 loading vs AutoImageProcessor URL loading; no shared bug." }, { "left": "issue:44987", "right": "issue:45538", "accept": false, - "reason": "Loading physical-intelligence/fast and CLIPTokenizer max_length are unrelated tokenizer/model-loading issues." + "reason": "Model-loading regression for one repo vs CLIPTokenizer max-length bug; unrelated." }, { "left": "issue:44849", "right": "issue:45701", "accept": false, - "reason": "Qwen3.5 hidden-states behavior and a generic tokenization-change report are not the same concrete bug." + "reason": "Qwen3.5 hidden-states bug vs a broad tokenization-version complaint; not the same concrete defect." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "RTDetrV2 output mismatch and Qwen3.5 save_pretrained key corruption are different issues." + "reason": "RTDetrV2 output differences vs Qwen3.5 save_pretrained key regression; different subsystems." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "Saving/loading DiaConfig token IDs and MXFP4 load fallback are unrelated." + "reason": "DiaConfig token ID persistence vs GPT-oss MXFP4 load fallback; unrelated." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Qwen3 MoE FP8 failure and CLIPTokenizer max_length are unrelated." - }, - { - "left": "issue:44568", - "right": "issue:45488", - "accept": false, - "reason": "Missing BOS/EOS insertion and LlamaTokenizer pre-tokenizer override are different tokenizer regressions." + "reason": "Qwen3 MOE FP8 loading vs CLIPTokenizer model_max_length; not the same issue." }, { "left": "issue:43452", "right": "issue:43582", "accept": false, - "reason": "gguf AutoTokenizer/AutoModel loading and AppleSilicon allocator TypeError are unrelated." + "reason": "gguf_file loading bug vs Apple Silicon allocator TypeError; different paths." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "MacOS bus error on a CLIP model and TrainerState token counting are unrelated." + "reason": "MacOS bus error on CLIP model vs TrainerState token counter update; unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Tokenizer regressions, but different root causes and model families; not mergeable into one fix." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Missing TensorFlow backend in a torch install and missing Qwen2_5_VL import are unrelated." + "reason": "Packaging/backend dependency issue vs missing Qwen2.5-VL symbol import; unrelated." }, { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Auto device-map CUDA assert and Qwen3OmniMoe video ValueError are different failure modes." + "reason": "Auto device-mapping cuda assert vs video processing ValueError in Qwen3OmniMoe; different failures." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "AddedToken special-arg conflict and tokenizer loading failure for est-roberta are not the same bug." + "reason": "AddedToken keyword duplication vs tokenizer loading for a specific model; different bugs." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "Dequantized model offload loading failure and CLIPTokenizer max_length are unrelated." + "reason": "Dequantized model offload failure vs tokenizer max-length constant bug; unrelated." }, { "left": "issue:44479", "right": "issue:45381", - "accept": false, - "reason": "Both are Qwen VLM video regressions, but one is a broad multi-model regression and the other is a specific vision_position_ids issue; titles do not prove the same root cause." + "accept": true, + "reason": "45381 is a specific report of the same Qwen2.5-VL video-input regression described broadly in 44479; same code path and likely same fix." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "PyTorch ImportError and Llama 3.2 Vision model-loading breakage are unrelated." + "reason": "PyTorch missing import error vs Llama 3.2 Vision model-loading breakage; unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: issue:44479 is clearly the broader v5.3.0 video-input regression report, and issue:45381 reads like a narrower report of the same Qwen VL video-position-id bug. The other soft-edge rejections are conservative based on the issue titles, with no obvious overclaiming." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -6308,12 +6272,12 @@ "issue:44737|issue:46463", "issue:42175|issue:44488", "issue:43577|issue:45072", - "issue:43257|issue:43756", - "issue:43906|issue:46468", "issue:43493|issue:43994", + "issue:43906|issue:46468", + "issue:43257|issue:43756", "issue:44291|issue:44821", - "issue:34567|issue:36331", "issue:43526|issue:45479", + "issue:34567|issue:36331", "issue:44062|issue:45081", "issue:43582|issue:43782", "issue:27957|issue:28282", @@ -6325,208 +6289,208 @@ "issue:43756|issue:45325", "issue:45823|issue:46463" ], - "prepared_review_unit_hash": "f1ebc9ce38c8aa8b0275cfe16b908b020e191d8368cd7b873376938c5080276d", + "prepared_review_unit_hash": "1882a07dfdbed8eba0c9b543a42bd3206753d28100034f63f5e50ebd2bce3722", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36322, - "estimated_input_tokens": 9081, - "estimated_eval_tokens": 18418 + "serialized_chars": 36319, + "estimated_input_tokens": 9080, + "estimated_eval_tokens": 18416 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are unrelated bug reports across different models or subsystems. The only likely duplicate is the SigLIP2 implementation mismatch vs. the report of nonsensical outputs/pipeline failure.", - "confidence": 0.68, - "canonical_issue_reason": "issue:43493 is the clearest root-cause report in the only likely duplicate pair; issue:43994 reads like the downstream symptom report.", + "summary": "This cluster is mostly a mix of unrelated Transformers bug reports across different models, features, and code paths. The only plausible duplicate pair is the two SigLIP2 reports; the rest are too model-specific or symptom-only to safely merge.", + "confidence": 0.84, + "canonical_issue_reason": "No single canonical fits the whole cluster cleanly; if one must be chosen, issue 43493 is the best root-cause report because it describes the underlying SigLIP2 implementation discrepancy more directly than the user-facing symptom in 43994.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43493 is the best representative issue because it states the underlying model discrepancy rather than only the observed bad outputs.", + "best_issue_reason": "Issue 43493 is the strongest standalone issue here: concrete, technically specific, and closest to the shared underlying defect in the only plausible duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error and an unreleased PEFT version bump are unrelated." + "reason": "Different problems: DeepSpeed ZeRO3 deque failure vs an unrelated PEFT version bump." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Both mention video inputs, but they target different Qwen model families and different failure modes." + "reason": "Both mention video inputs, but they target different Qwen model families and the titles point to separate regressions." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Different models and different errors; not the same bug." + "reason": "Same general area, but different model families and different failure modes (ValueError vs StopIteration)." }, { "left": "issue:44987", "right": "issue:45907", "accept": false, - "reason": "CI forwarding `revision` and a Transformers load failure are unrelated." + "reason": "One is a model load failure, the other is a repo-listing API regression; not the same underlying bug." }, { "left": "issue:24643", "right": "issue:29159", "accept": false, - "reason": "DeepSpeed training weight-shape error and tokenizer inconsistency are different problems." + "reason": "DeepSpeed training error and tokenizer inconsistency are unrelated code paths." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "MobileLLM key error and `init_empty_weights` argument handling do not look like the same code-path bug." + "reason": "Both are loading-time errors, but one is a key lookup for a specific model and the other is an init_empty_weights compatibility issue." }, { "left": "issue:28282", "right": "issue:29127", "accept": false, - "reason": "PyTorch missing import error vs. LayoutLMv3 box-validation messaging are unrelated." + "reason": "ImportError about missing PyTorch is unrelated to a LayoutLMv3 error-message clarity issue." }, { "left": "issue:42915", "right": "issue:43749", "accept": false, - "reason": "Qwen3 MoE FP8 loading and FSDP CPU RAM efficient loading are different issues." + "reason": "Different subsystems: FineGrainedFP8Config on Qwen3Moe vs FSDP CPU RAM efficient loading." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "Different models and symptoms; one is zero probabilities, the other is a Perceiver image-classification failure." + "reason": "Different models and different symptoms; no clear shared code-path bug." }, { "left": "issue:44737", "right": "issue:46463", "accept": false, - "reason": "XLNet CPU tensor creation bug and T5 decoder-input error message are unrelated." + "reason": "XLNet device placement bug and T5 decoder-input validation messaging are unrelated." }, { "left": "issue:42175", "right": "issue:44488", "accept": false, - "reason": "Backend install extras and a model loading failure are unrelated." + "reason": "Packaging/backend selection issue vs a model-loading failure; not the same bug." }, { "left": "issue:43577", "right": "issue:45072", "accept": false, - "reason": "Both are dtype-related, but they affect different models and likely different implementation paths." + "reason": "Both involve dtype, but they affect different models and likely different conversion paths." }, { - "left": "issue:43257", - "right": "issue:43756", - "accept": false, - "reason": "Qwen3 MoE weight conversion and Smollm3 RoPE-layer behavior are unrelated." + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "These both point to the same SigLIP2 defect: HF output quality/behavior diverging from the expected JAX behavior for the same model." }, { "left": "issue:43906", "right": "issue:46468", "accept": false, - "reason": "An isolated reproduction of another issue and a broken Nemotron example are not clearly the same bug." + "reason": "One is a reproduction of an older issue; the other is a broken example. No evidence they share the same root cause." }, { - "left": "issue:43493", - "right": "issue:43994", - "accept": true, - "reason": "Both report the same SigLIP2 output correctness problem; one frames it as an HF-vs-JAX discrepancy, the other as nonsensical AutoModel/pipeline outputs." + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to Smollm3 RoPE layer dropping." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "`init_empty_weights` initialization and loading AutoImageProcessor from URL are different failure modes." + "reason": "init_empty_weights compatibility and loading AutoImageProcessor from URL are different failure modes." }, { - "left": "issue:34567", - "right": "issue:36331", + "left": "issue:43526", + "right": "issue:45479", "accept": false, - "reason": "Trainer token-count tracking and custom `compute_loss` signature incompatibility are unrelated." + "reason": "A fast image-processor label reduction bug is not the same as the sequence-classification zero-loss issue." }, { - "left": "issue:43526", - "right": "issue:45479", + "left": "issue:34567", + "right": "issue:36331", "accept": false, - "reason": "BeitImageProcessorFast label reduction and sequence-classification loss degeneracy are different bugs." + "reason": "TrainerState token counting and CustomTrainer compute_loss signature mismatch are separate trainer issues." }, { "left": "issue:44062", "right": "issue:45081", "accept": false, - "reason": "AddedToken keyword conflict and Mistral regex patch crashing on missing backend tokenizer are unrelated." + "reason": "AddedToken keyword collision and Mistral regex patching crash are unrelated tokenizer bugs." }, { "left": "issue:43582", "right": "issue:43782", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and Qwen3VL weight-only loading error are different issues." + "reason": "Apple Silicon allocator warmup TypeError and Qwen3VL weight_only loading error are different code paths." }, { "left": "issue:27957", "right": "issue:28282", "accept": false, - "reason": "XLM-R with Flash Attention 2 and missing PyTorch import are unrelated." + "reason": "XLM-RoBERTa Flash Attention 2 and missing PyTorch ImportError are unrelated." }, { "left": "issue:43404", "right": "issue:45987", "accept": false, - "reason": "Mistral3 lm_head tying and `StaticCache.get_seq_length()` return type are unrelated." + "reason": "A tied-weights bug in Mistral3 and a StaticCache return-type bug are not the same underlying issue." }, { "left": "issue:44737", "right": "issue:45588", "accept": false, - "reason": "XLNet positional encoding on CPU and flash-attention sink handling are different code paths." + "reason": "XLNet CPU tensor allocation and flash_attention sink-less AttributeError are unrelated." }, { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "A Janus test failure and a Phi-4 multimodal load failure are not the same bug." + "reason": "Janus test failure and Phi-4 multimodal loading failure are different models and different bugs." }, { "left": "issue:43756", "right": "issue:46399", "accept": false, - "reason": "Smollm3 RoPE-layer dropping and Qwen3.5 kernel registration error are unrelated." + "reason": "RoPE-layer dropping in Smollm3 and missing apply_rotary_pos_emb decorator in Qwen3.5 are unrelated." }, { "left": "issue:43906", "right": "issue:45381", "accept": false, - "reason": "A generic reproduction issue and a specific Qwen2.5-VL vision_position_ids regression are not clearly duplicates." + "reason": "A repro of an older issue and a qwen2.5-vl video-position-id bug are not the same bug." }, { "left": "issue:43756", "right": "issue:45325", "accept": false, - "reason": "Smollm3 RoPE-layer behavior and Qwen2.5-VL temporal position scaling are unrelated." + "reason": "Different Qwen2.5-VL position/rope problems; not enough evidence they are the same fix." }, { "left": "issue:45823", "right": "issue:46463", "accept": false, - "reason": "Gemma4 device mismatch and T5 decoder-input validation messaging are unrelated." + "reason": "Gemma4 device mismatch and T5 decoder validation messaging are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays mostly conservative. The only accepted soft pair, issue:43493 vs issue:43994, is plausibly the same SigLIP2 bug based on the shared model/title context. The other soft-edge rejections are reasonably cautious given the differing models, subsystems, or failure modes." + "feedback": "The cluster summary is well grounded in the packet, and the lone accepted soft pair (43493/43994) is a conservative, defensible duplicate judgment. The other pairwise rejections are also appropriately cautious and do not overstate evidence." }, "evaluator_used": true, "retried": false, @@ -6598,8 +6562,8 @@ ], "soft_pairs": [ "issue:45042|issue:45081", - "issue:43452|issue:44488", "issue:43742|issue:44589", + "issue:43452|issue:44488", "issue:45357|issue:45920", "issue:43531|issue:44821", "issue:45405|issue:45600", @@ -6623,190 +6587,190 @@ "issue:44361|issue:44442", "issue:43295|issue:45356" ], - "prepared_review_unit_hash": "cff54a995fd2ac0ce582ff07dfb4ec8129f121374a2669a447f46bb1c1a3df9a", + "prepared_review_unit_hash": "2ae273d47742d6eeada63284c0e19ea7fd7e8ec75b628116495fa96398210539", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 25, - "serialized_chars": 35418, - "estimated_input_tokens": 8855, - "estimated_eval_tokens": 17966 + "serialized_chars": 35423, + "estimated_input_tokens": 8856, + "estimated_eval_tokens": 17968 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous cluster of unrelated issues. The only plausible duplicate thread is the tied-weights regression in v5 around serialization/loading for translation models.", + "summary": "The cluster is mostly unrelated issues across different models and code paths. I only see one plausible duplicate pair: the v5 tied-weights regression reports (44466/45005).", "confidence": 0.74, - "canonical_issue_reason": "Issue 45005 is the broadest representative of the only credible duplicate thread: tied-weights serialization/loading regressions in v5. It subsumes the more specific device-dependent lm_head.weight report in 44466.", + "canonical_issue_reason": "issue:44466 is the most concrete tied-weights report: it names the exact lm_head.weight serialization symptom, the v5 regression context, and the device-dependent behavior.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45005 has the clearest, broadest scope among the related items and is the best anchor for the tied-weights regression thread.", + "best_issue_reason": "issue:44466 is the best representative issue because it is the most specific, actionable description of the tied-weights bug and has the clearest regression framing.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Different failures: torchvision dependency in PIL image processors vs Mistral regex patch crash during tokenizer loading." + "reason": "Different subsystems and failures: image processor torchvision dependency vs Mistral regex tokenizer crash." }, { - "left": "issue:43452", - "right": "issue:44488", + "left": "issue:43742", + "right": "issue:44589", "accept": false, - "reason": "Both are loading-related, but one is a gguf_file regression and the other is a specific model load failure; not the same underlying bug." + "reason": "Both are loading errors, but they describe different concrete failures and likely different root causes." }, { - "left": "issue:43742", - "right": "issue:44589", + "left": "issue:43452", + "right": "issue:44488", "accept": false, - "reason": "Unrelated load-time errors: KeyError for MobileLLM loading vs Float8 storage lookup/type handling." + "reason": "One is a gguf_file/from_pretrained regression; the other is a specific model-loading failure. No clear same bug." }, { "left": "issue:45357", "right": "issue:45920", "accept": false, - "reason": "Different model families and different operations: incorrect visual encoder keys on save vs wrong token IDs on AutoTokenizer." + "reason": "Qwen3.5 save_pretrained key serialization bug vs broad AutoTokenizer wrong-token-IDs regression; different code paths." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Qwen3-MoE sliding-window logic is unrelated to AutoImageProcessor URL loading." + "reason": "Qwen3-MoE sliding_window behavior and AutoImageProcessor URL loading are unrelated." }, { "left": "issue:45405", "right": "issue:45600", "accept": false, - "reason": "One is a dependency version bump not yet on PyPI; the other is stale auto_mappings references causing CI breakage." + "reason": "PEFT version-pin issue and removed-config CI breakage are unrelated maintenance problems." }, { "left": "issue:44913", "right": "issue:46121", "accept": false, - "reason": "Both touch config/rope settings, but one is silent default reversion on reload and the other is a TypeError during validation conversion." + "reason": "Rotary_pct reload defaulting and rope-validation TypeError are different config bugs." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "BitNet packed-weight loading and Perceiver positional interpolation are unrelated bugs." + "reason": "BitNet packed-weight loading bug vs Perceiver interpolation failure; different models and mechanisms." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both involve config fields, but one is generic trust_remote_code config loss and the other is a specific Qwen vision config field dropped by strict parsing." + "reason": "Both mention config fields, but one is generic trust_remote_code missing fields and the other is a specific strict-decorated config field drop." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "Different regressions on different models: RTDetrV2 output drift vs Qwen3.5 save_pretrained key mapping." + "reason": "RTDetr output mismatch and Qwen3.5 save_pretrained corruption are distinct regressions." }, { "left": "issue:45200", "right": "issue:46463", "accept": false, - "reason": "Gemma 4 mm_token_type_ids behavior is unrelated to T5 decoder input validation error messaging." + "reason": "Gemma 4 mm_token_type_ids handling and T5 decoder-input validation messaging are unrelated." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed training RuntimeError and LayoutLMv3 error-message clarity are unrelated." + "reason": "DeepSpeed weight shape RuntimeError vs LayoutLMv3 error-message clarity; different issues entirely." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "Qwen3.5 num_labels propagation and granite_speech strict config loading are different config bugs." + "reason": "Qwen3.5 num_labels propagation and granite_speech strict config loading are different config-loading bugs." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "Gemma3 ZeRO-3 loading crash and trainer.evaluate after train are not the same issue." + "reason": "ZeRO-3 model-load index error and trainer.evaluate-after-train failure do not share the same code path." }, { "left": "issue:44493", "right": "issue:46050", "accept": false, - "reason": "Unexpected position-id keys and sinq quantization save/load breakage are different serialization/loading problems." + "reason": "Unexpected position-id keys and quantized save/load breakage are different symptoms." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen2.5-VL, but one is batched chat template padding and the other is video vision_position_ids; not the same bug." + "reason": "One is a batched apply_chat_template crash; the other is wrong vision_position_ids in video input. Different bugs." }, { "left": "issue:44466", "right": "issue:45005", "accept": true, - "reason": "These are the same tied-weights regression thread in v5: 44466 reports device-dependent lm_head.weight serialization, while 45005 describes the broader translation-model tied-weights failure." + "reason": "Both describe v5 tied-weights regressions in translation-style models and plausibly stem from the same save/load serialization path." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Both involve AutoTokenizer/from_pretrained, but the root causes differ: gguf_file loading breakage vs an unconditional model_info call in the Mistral regex patch." + "reason": "gguf_file/from_pretrained breakage and unconditional model_info() in offline mode are distinct tokenizer-loading bugs." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC latent decoding mismatch and Whisper batch_decode skip_special_tokens handling are unrelated." + "reason": "DAC latent reconstruction mismatch and Whisper batch_decode skip_special_tokens handling are unrelated." }, { "left": "issue:30064", "right": "issue:33666", "accept": false, - "reason": "Void segmentation-map processing and Qwen2-VL multi-GPU training are different subsystems and failures." + "reason": "Void segmentation-map processing and Qwen2-VL multi-GPU training are unrelated." }, { "left": "issue:43688", "right": "issue:43749", "accept": false, - "reason": "Auxiliary-loss normalization and FSDP CPU RAM efficient loading are distinct training behaviors." + "reason": "Auxiliary-loss normalization and FSDP CPU-efficient loading are different training/loading issues." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "PyTorch import detection and NoneType broadcast in Trainer are unrelated errors." + "reason": "PyTorch import detection failure and NoneType GPU broadcast error are unrelated." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "ImportError for missing PyTorch and MacOS bus error on CLIP are unrelated." + "reason": "Missing PyTorch import and MacOS bus error on CLIP are unrelated." }, { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer failures on different models; no shared underlying bug is evident." + "reason": "Different tokenizer classes and failures; no shared underlying bug is evident." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "Both are regressions, but one is processor/tokenizer API breakage and the other is Kimi-K2.5 codec handling plus warning text." + "reason": "Processor/tokenizer API regression and Kimi-K2.5 codec handling regression are different bugs." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative overall. The only accepted soft pair, 44466/45005, is plausibly the same tied-weights regression thread, and the other pair verdicts avoid overstating duplicates." + "feedback": "The analysis is grounded in the packet and stays conservative. The only accepted soft pair, 44466/45005, is plausibly the same tied-weights serialization regression; the rest are reasonably treated as unrelated based on their titles and described failure modes. The summary does not overstate the evidence." }, "evaluator_used": true, "retried": false, @@ -6880,24 +6844,24 @@ "issue:45372|issue:46170", "issue:43653|issue:44361", "issue:44936|issue:45823", - "issue:44484|issue:44704", "issue:38175|issue:43295", + "issue:44484|issue:44704", "issue:43452|issue:43577", "issue:45542|issue:46178", "issue:24643|issue:27957", - "issue:43653|issue:44568", "issue:29159|issue:33357", "issue:43577|issue:43697", + "issue:43653|issue:44568", "issue:44843|issue:44991", "issue:44464|issue:44743", "issue:43646|issue:43749", - "issue:43688|issue:45440", - "issue:44514|issue:45290", "issue:43531|issue:43742", - "issue:43756|issue:46402", + "issue:44514|issue:45290", + "issue:43688|issue:45440", + "issue:43653|issue:43927", "issue:43688|issue:45704", "issue:45491|issue:46002", - "issue:43653|issue:43927", + "issue:43756|issue:46402", "issue:43550|issue:43927", "issue:43688|issue:45701", "issue:43550|issue:44361", @@ -6907,206 +6871,215 @@ "issue:32090|issue:34567", "issue:41720|issue:44464" ], - "prepared_review_unit_hash": "5d343f24406ca68960f6d4a74a2e0a391ee26de14306eb52a356c3656f13b2e4", + "prepared_review_unit_hash": "ccdfd091da38a4874a1c748820d9c2297e5a1ce13eb1e275288bcd97025054df", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36427, - "estimated_input_tokens": 9107, - "estimated_eval_tokens": 18470 + "serialized_chars": 36432, + "estimated_input_tokens": 9108, + "estimated_eval_tokens": 18472 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issue pairs across tokenizer, loading, training, and backend areas. The only likely duplicate is the `apply_chat_template` crash pair (44514/45290), which appears to describe the same assistant/tool-call tokenization failure from titles alone.", - "confidence": 0.86, - "canonical_issue_reason": "44514 is the earlier issue and its title matches the broader `apply_chat_template` crash area best. 45290 looks like a probable later duplicate, but with only titles/metadata available I\u2019d phrase this as a likely duplicate rather than a confirmed same-crash report.", + "summary": "This cluster is highly heterogeneous: it mixes unrelated tokenizer, trainer, multimodal, MoE, config, and loading regressions. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": "If a representative issue is required, #43646 is the broadest v5-regression report in the set; otherwise there is no true canonical duplicate for this cluster.", "canonical_pr_reason": null, - "best_issue_reason": "44514 is the best canonical issue because it is earlier and the title is the clearest broad match for the shared `apply_chat_template` failure. 45290 is the more specific follow-up report, so it fits as the duplicate.", + "best_issue_reason": "#43646 is the closest umbrella representative because it describes a broad Transformers 5.0 regression affecting custom model initialization, which is more general than the other very specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45372", "right": "issue:46170", "accept": false, - "reason": "Different failure modes: Gemma4 processor import regression vs Windows/ROCm torch.distributed import/type error." + "reason": "Different failure domains: Gemma processor loading/import dependency vs Windows ROCm torch.distributed import/type error." }, { "left": "issue:43653", "right": "issue:44361", "accept": false, - "reason": "Both tokenizer-related, but the titles point to different bugs and different models." + "reason": "Both are tokenizer bugs, but they affect different tokenizers and different concrete failures." }, { "left": "issue:44936", "right": "issue:45823", "accept": false, - "reason": "Trainer evaluate-after-train behavior is unrelated to a Gemma4 device-mismatch forward error." + "reason": "Trainer lifecycle failure vs Gemma4 device-map forward mismatch; unrelated code paths." }, { - "left": "issue:44484", - "right": "issue:44704", + "left": "issue:38175", + "right": "issue:43295", "accept": false, - "reason": "`max_shard_size` save_pretrained behavior is unrelated to AutoProcessor kwargs propagation." + "reason": "Different models and symptoms: zero probabilities in SigLIP2 vs processor.tokenizer/image-handling regression." }, { - "left": "issue:38175", - "right": "issue:43295", + "left": "issue:44484", + "right": "issue:44704", "accept": false, - "reason": "SigLIP2 zero probabilities and a custom model processor/tokenizer regression are different issues." + "reason": "A save_pretrained shard-size question is unrelated to AutoProcessor kwargs forwarding." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "`gguf_file` loading breakage is unrelated to BLIP2 dtype staying float32." + "reason": "GGUF loading breakage is not the same bug as Blip2 dtype remaining float32." }, { "left": "issue:45542", "right": "issue:46178", "accept": false, - "reason": "Both involve backend/TensorFlow wording, but the concrete errors are different." + "reason": "Both mention backend/TF issues, but they are different backend resolution failures with different triggers." }, { "left": "issue:24643", "right": "issue:27957", "accept": false, - "reason": "DeepSpeed weight-dimension training error is unrelated to XLMRoberta Flash Attention 2 support." - }, - { - "left": "issue:43653", - "right": "issue:44568", - "accept": false, - "reason": "Different tokenizer special-token bugs on different models; not the same defect." + "reason": "DeepSpeed weight-shape runtime error is unrelated to XLMRoberta Flash Attention 2." }, { "left": "issue:29159", "right": "issue:33357", "accept": false, - "reason": "Tokenizer consistency issue and CLIP bus error on macOS are unrelated." + "reason": "Tokenizer inconsistency and MacOS bus error on CLIP are different bugs." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "BLIP2 dtype retention and RTDetrV2 output divergence are different code-path problems." + "reason": "Model dtype loading bug vs RTDetrV2 output divergence are distinct issues." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Different tokenizer special-token behaviors on different models; not the same bug." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Offline hub access in `_patch_mistral_regex` is unrelated to tokenizer loading for EMBEDDIA/est-roberta." + "reason": "Offline model_info call in tokenizer patch is unrelated to est-roberta tokenizer loading failure." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Compiled chunked generation inconsistency is not the same as Qwen3.5 recurrent-state reset with cache." + "reason": "Both involve generation state, but one is compiled-forward inconsistency and the other is cache reset logic in a different model." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Custom model initialization breakage in v5 is unrelated to FSDP CPU RAM efficient loading." + "reason": "Custom model initialization regression is unrelated to FSDP CPU RAM efficient loading." }, { - "left": "issue:43688", - "right": "issue:45440", + "left": "issue:43531", + "right": "issue:43742", "accept": false, - "reason": "Both are MoE-related, but the auxiliary-loss normalization bug and DeepseekV3MoE divergence are different problems." + "reason": "Sliding-window behavior in Qwen3-MoE is not the same as a key error when loading MobileLLM-125M." }, { "left": "issue:44514", "right": "issue:45290", - "accept": true, - "reason": "Very similar `apply_chat_template` crash wording, both centered on batched/assistant/tool-call handling and empty content; likely the same underlying bug." + "accept": false, + "reason": "Same API surface, but different crashes: batched padding=False vs assistant tool-call messages with no content." }, { - "left": "issue:43531", - "right": "issue:43742", + "left": "issue:43688", + "right": "issue:45440", "accept": false, - "reason": "Qwen3-MoE sliding-window issue is unrelated to a MobileLLM-125M key error on load." + "reason": "Auxiliary-loss normalization in MoE models is different from DeepSeekV3MoE implementation divergence." }, { - "left": "issue:43756", - "right": "issue:46402", + "left": "issue:43653", + "right": "issue:43927", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch is unrelated to Sam3Model text-encoder weights not loading." + "reason": "Tokenizer special-token registration bug and DiaConfig custom token-ID loss are unrelated." }, { "left": "issue:43688", "right": "issue:45704", "accept": false, - "reason": "MoE auxiliary-loss normalization is unrelated to T5 using apex.FusedRMSNorm and leaking memory." + "reason": "MoE auxiliary-loss normalization is unrelated to a T5 apex RMSNorm memory leak." }, { "left": "issue:45491", "right": "issue:46002", "accept": false, - "reason": "Different NaN symptoms with different triggers and model contexts." + "reason": "Both mention NaNs, but the model, locus, and root cause are different." }, { - "left": "issue:43653", + "left": "issue:43756", + "right": "issue:46402", + "accept": false, + "reason": "RoPE-layer dropping in Smollm3 is unrelated to Sam3 text encoder weights not loading." + }, + { + "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "BigBird special-token decode bug is unrelated to DiaConfig losing custom token IDs after save/load." + "reason": "torch.compile/SDPA failure is not the same bug as custom token IDs being lost on save/load." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Auxiliary-loss normalization is unrelated to broad tokenization changes." }, { "left": "issue:43550", "right": "issue:44361", "accept": false, - "reason": "torch.compile + SDPA failure for Bamba-9B-v2 is unrelated to MLukeTokenizer task AttributeError." + "reason": "Compiler/SDPA model failure and tokenizer AttributeError are different code paths." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Video-input regression across Qwen VL models is unrelated to Qwen3.5 num_labels propagation." + "reason": "Video-input regression across Qwen variants is distinct from num_labels propagation in Qwen3.5 config." }, { "left": "issue:43643", "right": "issue:44561", "accept": false, - "reason": "Both involve `trust_remote_code`, but the concrete failures are different." + "reason": "Missing fields from trust_remote_code AutoConfig is different from the removed is_torch_fx_available regression." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Compiled chunked-generation inconsistency is unrelated to Qwen3.5 flash-attention generation failure." + "reason": "Compiled-forward chunked generation inconsistency is not the same as flash-attention generation failure." }, { "left": "issue:32090", "right": "issue:34567", "accept": false, - "reason": "Trainer broadcast TypeError is unrelated to `num_input_tokens_seen` not updating." + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are unrelated trainer bugs." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "Qwen3 auto device-map cuda assert is unrelated to compiled forward chunked-generation inconsistency." + "reason": "Auto device mapping cudaErrorAssert on A800 is a different failure than chunked generation inconsistency under compilation." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The soft-edge rejections are mostly conservative, but the canonical duplicate call is a bit overstated for the evidence available. In the packet we can see only the titles/metadata, not the bodies, so claims like \u201cmore specific reproduction details\u201d and \u201csame crash\u201d are stronger than what the packet directly supports. The safer phrasing would be that 44514 is the earlier, likely related report, with 45290 as a probable duplicate based on similar apply_chat_template crash wording." - }, - "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 }, @@ -7173,21 +7146,21 @@ "issue:46365" ], "soft_pairs": [ - "issue:43504|issue:43720", "issue:44977|issue:46209", + "issue:43504|issue:43720", "issue:43504|issue:43550", "issue:32090|issue:33666", "issue:43504|issue:44610", - "issue:43577|issue:44554", "issue:33666|issue:36010", + "issue:43577|issue:44554", "issue:44466|issue:46050", "issue:43742|issue:44821", "issue:43577|issue:45588", - "issue:44554|issue:46365", "issue:43720|issue:43927", "issue:43262|issue:43295", - "issue:43606|issue:43927", + "issue:44554|issue:46365", "issue:43452|issue:44451", + "issue:43606|issue:43927", "issue:45823|issue:46050", "issue:44849|issue:46209", "issue:33666|issue:36331", @@ -7203,18 +7176,18 @@ "issue:44849|issue:45357", "issue:44568|issue:45812", "issue:44488|issue:44843", - "issue:43493|issue:43844", "issue:39692|issue:43295", + "issue:43493|issue:43844", "issue:43278|issue:43844" ], - "prepared_review_unit_hash": "bba9661e6a027b8b268d4565b4a54b609f010308ff93d2a3f9469b90b091c836", + "prepared_review_unit_hash": "8577d56ef3334bb1a421eae8a43c9d3e307adc7a6ca6ef6331357cdbf447c423", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, @@ -7229,210 +7202,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "All candidate edges are superficial text-similarity matches. They span unrelated models, configs, and failure modes, so there is no true duplicate cluster here.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "No soft-edge pair looks like a true duplicate; most are unrelated bugs across different models, processors, or runtime paths. If a single issue must anchor the cluster, `issue:45812` is the strongest standalone regression, but the set is not one coherent duplicate group.", + "confidence": 0.87, + "canonical_issue_reason": "`issue:45812` is the clearest high-impact core regression here: broad `AutoTokenizer` token-id breakage across Granite models with multiple inbound references.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "`issue:45812` is the best representative issue because it is concrete, user-facing, and broader in impact than the other standalone reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43504", - "right": "issue:43720", + "left": "issue:44977", + "right": "issue:46209", "accept": false, - "reason": "Different failures: BEiT legacy preset field vs BitNet packed-weight loading." + "reason": "Different bugs: Qwen3.5 flash-attention generation vs FP8 inference failure on B200." }, { - "left": "issue:44977", - "right": "issue:46209", + "left": "issue:43504", + "right": "issue:43720", "accept": false, - "reason": "Different models and paths: Qwen3.5 flash-attention generation vs FP8 B200 inference." + "reason": "Different code paths: legacy-field model loading vs BitNet packed-weight unpacking during accelerate load." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "Legacy field load error is unrelated to Bamba torch.compile SDPA crash." + "reason": "Unrelated failures: BEiT preset loading vs Bamba torch.compile SDPA issue." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Trainer gpu_broadcast TypeError is unrelated to Qwen2-VL multi-GPU training." + "reason": "Both involve training, but the error types and affected models are different." }, { "left": "issue:43504", "right": "issue:44610", "accept": false, - "reason": "Model preset load bug is unrelated to processor image-size mismatch." + "reason": "Legacy config-field load bug is unrelated to OmDet-Turbo processor image-size mismatch." }, { - "left": "issue:43577", - "right": "issue:44554", + "left": "issue:33666", + "right": "issue:36010", "accept": false, - "reason": "Blip2 dtype retention bug differs from MPS attention correctness issue." + "reason": "Different failures: multi-GPU training vs import-time `GenerationMixin` regression." }, { - "left": "issue:33666", - "right": "issue:36010", + "left": "issue:43577", + "right": "issue:44554", "accept": false, - "reason": "Multi-GPU training issue is unrelated to GenerationMixin import failure." + "reason": "Different bugs: BLIP2 dtype casting vs MPS attention correctness when head dims differ." }, { "left": "issue:44466", "right": "issue:46050", "accept": false, - "reason": "Tied-weight serialization bug differs from sinq quantized save/load breakage." + "reason": "Both are save/load related, but one is tied-weight serialization and the other is quantization persistence." }, { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "MobileLLM key error is unrelated to AutoImageProcessor-from-URL loading." + "reason": "Loading a causal LM vs loading an image processor from URL are separate problems." }, { "left": "issue:43577", "right": "issue:45588", "accept": false, - "reason": "Blip2 dtype casting is unrelated to flash_attention.py s_aux=None crash." - }, - { - "left": "issue:44554", - "right": "issue:46365", - "accept": false, - "reason": "MPS attention correctness issue is unrelated to Trainer import conflict." + "reason": "BLIP2 dtype mismatch is unrelated to flash-attention crashing on `s_aux=None`." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "BitNet packed-weight load bug is unrelated to DiaConfig token-id persistence." + "reason": "Packed-weight loading bug and DiaConfig token-id persistence bug are different underlying issues." }, { "left": "issue:43262", "right": "issue:43295", "accept": false, - "reason": "Audio sampling-rate default is unrelated to processor.tokenizer/image-handling regression." + "reason": "Audio chat-template sampling-rate defaulting and processor/tokenizer API regression are distinct." }, { - "left": "issue:43606", - "right": "issue:43927", + "left": "issue:44554", + "right": "issue:46365", "accept": false, - "reason": "CPU-offload device mismatch is unrelated to DiaConfig save/load token IDs." + "reason": "MPS attention correctness bug is unrelated to `Trainer` import failing with datasets installed." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "GGUF from_pretrained breakage is unrelated to ScandiBERT model loading." + "reason": "Generic `from_pretrained`/gguf breakage is not the same as the ScandiBERT load failure." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to config token-id loss after save/load." }, { "left": "issue:45823", "right": "issue:46050", "accept": false, - "reason": "Gemma4 device-mapped forward mismatch differs from sinq save/load corruption." + "reason": "Different issues: Gemma4 device placement during forward vs sinq quantization persistence." }, { "left": "issue:44849", "right": "issue:46209", "accept": false, - "reason": "Qwen3.5 hidden-state output bug is unrelated to FP8 B200 inference." + "reason": "Qwen3.5 hidden-states bug is separate from FP8 inference on B200." }, { "left": "issue:33666", "right": "issue:36331", "accept": false, - "reason": "Multi-GPU training issue is unrelated to custom compute_loss signature change." + "reason": "Distributed training setup issue and custom-trainer signature mismatch are not the same bug." }, { "left": "issue:43994", "right": "issue:46192", "accept": false, - "reason": "SigLIP2 incorrect results differ from return_tensors being ignored." + "reason": "SigLIP2 correctness regression and `return_tensors` handling bug are unrelated." }, { "left": "issue:43278", "right": "issue:43688", "accept": false, - "reason": "BF16/FP32 eval dtype drift is unrelated to aux-loss normalization." + "reason": "Embedding dtype drift in eval is not the same as auxiliary-loss normalization." }, { "left": "issue:43720", "right": "issue:44361", "accept": false, - "reason": "BitNet accelerate loading bug is unrelated to MLukeTokenizer task AttributeError." + "reason": "BitNet packed-weight loading and MLukeTokenizer task AttributeError are different failures." }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Different configs and fields: Llama4 pad_token_id vs StableLm pad_token_idx." + "reason": "Similar theme, but different configs and attributes (`pad_token_id` vs `pad_token_idx`) and no shared concrete code path." }, { "left": "issue:44625", "right": "issue:46002", "accept": false, - "reason": "Qwen3.5 num_labels propagation is unrelated to classification-head NaNs." + "reason": "`num_labels` propagation bug is distinct from NaNs in classification heads." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer extra-space bug is unrelated to MacOS CLIP bus error." + "reason": "Tokenizer added-space bug is unrelated to the MacOS CLIP bus error." }, { "left": "issue:43493", "right": "issue:43825", "accept": false, - "reason": "SigLIP2 JAX/HF discrepancy is unrelated to pipeline translation-support message." + "reason": "SigLIP2 implementation discrepancy and pipeline error-message regression are different issues." }, { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Phi-4 multimodal load failure is unrelated to an unreleased PEFT version bump." + "reason": "Model loading failure and unreleased dependency version bump are unrelated." }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "Docs-example model/processor errors are not the same as core SigLIP2 implementation mismatch." + "reason": "Both mention SigLIP2, but one is a doc example error set and the other is an implementation discrepancy." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Qwen3.5 hidden-state bug differs from save_pretrained visual-key regression." + "reason": "Hidden-states bug and incorrect visual-encoder key serialization are separate regressions." }, { "left": "issue:44568", "right": "issue:45812", "accept": false, - "reason": "mDeBERTa special-token regression is unrelated to Granite tokenizer IDs." + "reason": "Different tokenizer regressions affecting different models and code paths." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "Model loading failure for sleng-bert is unrelated to offline model_info patch." - }, - { - "left": "issue:43493", - "right": "issue:43844", - "accept": false, - "reason": "SigLIP2 implementation mismatch is unrelated to ZeRO-3 gradient explosion." - }, - { - "left": "issue:39692", - "right": "issue:43295", - "accept": false, - "reason": "SigLIP2 docs example failure is unrelated to processor.tokenizer regression." + "reason": "Model load failure for a specific model is not the same as offline-mode breakage in `_patch_mistral_regex`." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Embedding dtype drift is unrelated to ZeRO-3 gradient inflation." + "reason": "Evaluation dtype mismatch is unrelated to ZeRO-3 gradient growth." } ] }, @@ -7512,18 +7473,18 @@ "issue:44464|issue:46424", "issue:33357|issue:33666", "issue:43334|issue:43572", - "issue:43257|issue:43299", "issue:43824|issue:44589", + "issue:43257|issue:43299", "issue:44589|issue:45520", - "issue:43504|issue:43606", "issue:43927|issue:44361", + "issue:43504|issue:43606", "issue:43334|issue:45759", "issue:43278|issue:43701", "issue:45229|issue:45823", "issue:43575|issue:45405", - "issue:44493|issue:46285", "issue:43572|issue:44291", "issue:44792|issue:45405", + "issue:44493|issue:46285", "issue:43504|issue:43653", "issue:43606|issue:44112", "issue:44464|issue:45584", @@ -7531,220 +7492,223 @@ "issue:43873|issue:46050", "issue:45588|issue:46050", "issue:43065|issue:46369", - "issue:44423|issue:44734", "issue:29942|issue:30333", - "issue:43844|issue:44964", - "issue:28218|issue:34567" + "issue:44423|issue:44734", + "issue:28218|issue:34567", + "issue:43844|issue:44964" ], - "prepared_review_unit_hash": "30318fcbecbab39d828df9fcf32ae92888c925c3e08046fd8ef17ad688882040", + "prepared_review_unit_hash": "3504df3302ee0c284cd2e8d18db9432d0df88cc5328e48eb90f8c62e714bef12", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36339, - "estimated_input_tokens": 9085, - "estimated_eval_tokens": 18426 + "serialized_chars": 36341, + "estimated_input_tokens": 9086, + "estimated_eval_tokens": 18428 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct bug reports across different model families and subsystems. The soft links mostly reflect broad topical similarity (Qwen, Gemma, loading, serving, quantization) rather than the same underlying defect, so none should be merged as duplicates.", - "confidence": 0.94, - "canonical_issue_reason": "No single issue is a clear canonical duplicate target; the items span unrelated failure modes and model families, and the soft pairs do not point to one shared underlying bug.", + "summary": "Most candidate pairs are clearly unrelated and should not be merged. The only plausible duplicate pair is the Qwen3 multimodal loading/config bug pair (43334 and 45759), where the second issue describes the underlying config-unwrapping problem more directly.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45759 is the best canonical issue for the one genuine duplicate pair because it states the root cause (`from_config` not unwrapping `text_config`) more directly than the symptom report in issue:43334.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue stands out because the cluster does not represent one concrete bug/change; it is better treated as multiple separate issues.", + "best_issue_reason": "issue:45759 is the strongest representative overall: it is specific about the failing code path and broad enough to cover the related Qwen3-VL loading symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Gemma4 text-only mm_token_type_ids handling is unrelated to a Gemma4 31B multi-GPU CUDA OOM." + "reason": "Different bugs: defaulting mm_token_type_ids in text-only fine-tuning vs multi-GPU inference CUDA OOM." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Both involve Qwen multimodal/attention paths, but one is a video-processing ValueError and the other is a flash-attention generation issue; different code paths and symptoms." + "reason": "Different failure modes and code paths: video input processing ValueError vs flash-attention generation behavior." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "One is a custom model initialization regression in v5, the other is inconsistent lm_head serialization by device; not the same bug." + "reason": "Unrelated: custom model initialization breakage vs inconsistent lm_head.weight serialization depending on device." }, { "left": "issue:44464", "right": "issue:46424", "accept": false, - "reason": "Chunked generation with compiled forward is unrelated to max_cache_length only applying to quantized cache." + "reason": "Different cache issues: compiled forward chunked generation vs max_cache_length only honored for quantized cache." }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "A MacOS bus error on CLIP is unrelated to Qwen2-VL multi-GPU training." + "reason": "Different models and subsystems: CLIP bus error on MacOS vs Qwen2-VL multi-GPU training." }, { "left": "issue:43334", "right": "issue:43572", "accept": false, - "reason": "Both are config-loading regressions after v5, but they affect different models and different missing fields; not the same underlying defect." + "reason": "Both are config-related, but they affect different models and different missing fields; not the same bug." }, { - "left": "issue:43257", - "right": "issue:43299", + "left": "issue:43824", + "right": "issue:44589", "accept": false, - "reason": "Both mention Qwen3 MoE, but one is accelerate/deepspeed weight conversion and the other is model loading in dev; too different to merge." + "reason": "Different errors and causes: missing import for a model class vs missing Float8 storage type." }, { - "left": "issue:43824", - "right": "issue:44589", + "left": "issue:43257", + "right": "issue:43299", "accept": false, - "reason": "Missing Qwen2.5-VL import is unrelated to the Float8 storage lookup error." + "reason": "Same family name, but different concrete bugs and loading paths: accelerate+deepspeed weight conversion vs Qwen3VL MoE load breakage." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage resolution and flash_attn import on Python 3.13 are unrelated failures." + "reason": "Different runtime/environment issues: Float8 storage lookup vs flash_attn import key error on Python 3.13." }, { - "left": "issue:43504", - "right": "issue:43606", + "left": "issue:43927", + "right": "issue:44361", "accept": false, - "reason": "Legacy BEiT preset loading and a CPU offload device mismatch in Bark are different bugs." + "reason": "Different token/config problems affecting different classes and symptoms; not the same underlying defect." }, { - "left": "issue:43927", - "right": "issue:44361", + "left": "issue:43504", + "right": "issue:43606", "accept": false, - "reason": "DiaConfig losing custom token IDs after save/load is unrelated to MLukeTokenizer task-time AttributeError." + "reason": "Different areas entirely: BEiT legacy field loading vs Bark CPU offload device mismatch." }, { "left": "issue:43334", "right": "issue:45759", - "accept": false, - "reason": "Both touch Qwen multimodal config handling, but one is a missing pad_token_id on text config and the other is from_config failing to unwrap text_config; related area, not the same concrete bug." + "accept": true, + "reason": "Same underlying Qwen3 multimodal loading bug: 43334 is the symptom, 45759 identifies the config-unwrapping root cause in the same code path." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "TrainerState token accounting and resume_from_checkpoint key mismatch are different training-state bugs." + "reason": "Different issues: dtype drift between train/eval vs resume_from_checkpoint key mismatch." }, { "left": "issue:45229", "right": "issue:45823", "accept": false, - "reason": "Gemma4 inference OOM and Gemma4 PLE device mismatch share a model family but not the same underlying issue." + "reason": "Both involve Gemma4 and devices, but one is CUDA OOM and the other is a device mismatch; not the same bug." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "Tensor-parallel OOM when loading Qwen2-57B is unrelated to the PEFT version bump." - }, - { - "left": "issue:44493", - "right": "issue:46285", - "accept": false, - "reason": "Unexpected position-id keys and CLIP text model structure changes are unrelated." + "reason": "Completely unrelated: tensor-parallel OOM for a model load vs a dependency version bump." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "A missing StableLmConfig field after v5 is not the same as init_empty_weights passing an unexpected _is_hf_initialized argument." + "reason": "Different compatibility problems: missing pad_token_idx vs unexpected _is_hf_initialized argument." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "A Janus image-generation test failure is unrelated to the PEFT dependency version issue." + "reason": "Different topics: a Janus image-generation test failure vs a PEFT minimum-version bump." + }, + { + "left": "issue:44493", + "right": "issue:46285", + "accept": false, + "reason": "Not the same bug: unexpected position-id keys across many models vs CLIP text model structure change." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "BEiT legacy field loading and BigBirdTokenizer special-token registration are different tokenizer/model-loading bugs." + "reason": "Different models and defects: BEiT legacy field loading vs BigBirdTokenizer special-token registration." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "A real CPU-offload device mismatch in Bark is not the same as a stale GraniteSpeech CI test override." + "reason": "Different issues: real CPU-offload device mismatch vs stale GraniteSpeech CI test." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Compiled-forward chunked generation is unrelated to Whisper failing on empty transcription after special-token alignment." + "reason": "Different cache/generation bugs with different triggers and symptoms." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 nonsensical outputs and unexpected position-id keys do not share a concrete code path." + "reason": "Different problems: SigLIP2 output quality issue vs position-id key regressions." }, { "left": "issue:43873", "right": "issue:46050", "accept": false, - "reason": "Quantization offloading behavior and sinq save/load breakage are distinct quantization bugs." + "reason": "Both mention memory/offloading/quantization concerns, but the concrete failures are different and not mergeable as one fix." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "Flash attention crashing on sink-less models is unrelated to sinq serialization/load failure." + "reason": "Different bugs: flash_attention sink-less model crash vs sinq save/load corruption." }, { "left": "issue:43065", "right": "issue:46369", "accept": false, - "reason": "A dummy Conv2d in Sam3PixelDecoder is not the same as memory overflow in sam3-litetext." - }, - { - "left": "issue:44423", - "right": "issue:44734", - "accept": false, - "reason": "Serve continuous-batching crashing on multimodal inputs and /v1/responses KV-cache continuation indexing are different server bugs." + "reason": "Different Sam3 topics: dummy Conv2d implementation vs memory overflow in sam3-litetext." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "Flash Attention 2 test failures and MLFlow reporting after failed training jobs are unrelated." + "reason": "Completely unrelated CI and MLflow reporting failures." }, { - "left": "issue:43844", - "right": "issue:44964", + "left": "issue:44423", + "right": "issue:44734", "accept": false, - "reason": "ZeRO-3 gradient explosion with HfDeepSpeedConfig is unrelated to Phi-4-multimodal load failure." + "reason": "Both are transformers serve crashes, but they hit different endpoints and have different root causes." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Tokenizer spacing after added tokens has nothing to do with TrainerState token counters." + "reason": "Tokenizer spacing bug vs TrainerState token-count update bug; no shared underlying defect." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "Different training/loader issues: ZeRO-3 gradient growth vs latest-transformers model loading failure." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The only accepted duplicate pair, 43334/45759, is plausibly the same Qwen3 multimodal loading/config bug, and the rejected pairs are explained at a level that matches the visible titles without overclaiming." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -7815,8 +7779,8 @@ "issue:45406|issue:46121", "issue:45084|issue:45520", "issue:44554|issue:45446", - "issue:15354|issue:38175", "issue:43582|issue:44977", + "issue:15354|issue:38175", "issue:43540|issue:45200", "issue:45072|issue:45198", "issue:45200|issue:45823", @@ -7840,94 +7804,94 @@ "issue:37428|issue:43824", "issue:43950|issue:46285" ], - "prepared_review_unit_hash": "03ccc337ef3ed00afc937cfa4efbf1673b47b0dad1544289a18de65eb8139921", + "prepared_review_unit_hash": "fe0ef5677431d33bbb91e03702472eb6eba3a1631232d5a0a30f78d58e11be20", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36116, - "estimated_input_tokens": 9029, - "estimated_eval_tokens": 18314 + "serialized_chars": 36117, + "estimated_input_tokens": 9030, + "estimated_eval_tokens": 18316 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set does not form a coherent duplicate cluster: the items span unrelated regressions in tokenizers, generation, multimodal processors, import paths, and training/runtime backends. None of the soft pairs looks like the same underlying bug/change, so there is no strong canonical issue or PR here.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "These items are not duplicates of one another; they span unrelated bugs across tokenizers, generation, model loading, video processing, quantization, and backend/import errors. The soft-similarity matches are mostly keyword overlap rather than the same underlying defect.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue fits this set: the issues describe different code paths and failure modes, so selecting one would misrepresent the cluster.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "No best issue candidate is suitable as a cluster representative because the set is heterogeneous and lacks a shared underlying bug/change.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42490", "right": "issue:43696", "accept": false, - "reason": "Local-path import warning vs CUDA OOM on GPT-oss-20b; unrelated failure modes." + "reason": "Local-path import warning vs CUDA out-of-memory on GPT-oss; unrelated failures." }, { "left": "issue:29159", "right": "issue:32090", "accept": false, - "reason": "Slow/fast tokenizer inconsistency vs Trainer _gpu_broadcast_one NoneType error; different code paths." + "reason": "Tokenizer behavior mismatch vs Trainer GPU broadcast NoneType error; different subsystems." }, { "left": "issue:45406", "right": "issue:46121", "accept": false, - "reason": "Gemma4Processor _tokenizer attribute error vs rope validation TypeError; unrelated bugs." + "reason": "Gemma4Processor `_tokenizer` attribute crash vs rope-validation list type error; unrelated." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Template node compilation error vs flash_attn import KeyError; no shared underlying issue." + "reason": "Template-node compile error vs flash_attn import backend KeyError; different issues." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "MPS attention correctness bug vs import version-check bug in flex_attention; too different." + "reason": "MPS attention correctness bug vs PyTorch version check bug in flex_attention; not the same bug." }, { - "left": "issue:15354", - "right": "issue:38175", + "left": "issue:43582", + "right": "issue:44977", "accept": false, - "reason": "TorchScript GeneratorExp export failure vs SigLIP2 zero probabilities; unrelated model issues." + "reason": "Apple Silicon warmup TypeError vs Qwen3.5 flash-attention generation issue; unrelated." }, { - "left": "issue:43582", - "right": "issue:44977", + "left": "issue:15354", + "right": "issue:38175", "accept": false, - "reason": "Apple Silicon caching allocator TypeError vs Qwen3.5 flash-attention generation bug; unrelated." + "reason": "TorchScript GeneratorExp export problem vs unexpected zero probabilities in SigLIP2; unrelated." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Video-processing ValueError in Qwen3OmniMoe vs Gemma4 text-only mm_token_type_ids defaulting issue." + "reason": "Video input processing ValueError vs Gemma4 multimodal token-type defaulting; different code paths." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "bfloat16 dtype mismatch in SwitchTransformers/TimmWrapperModel vs Wav2Vec2 save/tokenization failure." + "reason": "bfloat16 dtype mismatch in inference vs Wav2Vec2 save/tokenization failure; unrelated." }, { "left": "issue:45200", "right": "issue:45823", "accept": false, - "reason": "Both Gemma4-related, but one is missing mm_token_type_ids defaults and the other is a device_map forward mismatch." + "reason": "Missing mm_token_type_ids default vs PLE device mismatch in forward; not the same defect." }, { "left": "issue:43493", @@ -7939,43 +7903,43 @@ "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both touch apply_chat_template, but one is zero assistant masks for multimodal inputs and the other is a crash on tool-call messages." + "reason": "Both involve chat templates, but one is zero assistant masks for multimodal inputs and the other is tool-call/no-content crash; different bugs." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Missing flash support import vs meta-device loading regression; different subsystems and symptoms." + "reason": "Flash-attention import symbol error vs meta-device loading regression; unrelated loading issues." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "Generation kwargs sync issue after sync_gpus vs Qwen3.5 hidden-states bug; unrelated." + "reason": "Generation kwargs update after sync_gpus vs Qwen3.5 output_hidden_states bug; different generation paths." }, { "left": "issue:44186", "right": "issue:46192", "accept": false, - "reason": "LayoutLMv2Tokenizer crash on NER/batching vs return_tensors being ignored with text_kwargs; different tokenizer behaviors." + "reason": "LayoutLMv2Tokenizer crash on NER/padding vs return_tensors being ignored when text_kwargs is passed; different tokenizer bugs." }, { "left": "issue:45823", "right": "issue:46192", "accept": false, - "reason": "Gemma4 device mismatch vs return_tensors precedence bug; not the same underlying problem." + "reason": "Gemma4 device mismatch in forward vs return_tensors argument handling; unrelated." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Processor.tokenizer regression vs video processor batch-shape bug; both processor-related but distinct." + "reason": "Processor.tokenizer regression in custom code vs batched video processor shape bug; different models and failures." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "Processor/tokenizer regression vs OmDet-Turbo processor/model size mismatch; different concrete bugs." + "reason": "Processor/tokenizer regression vs OmDet-Turbo processor size mismatch; unrelated." }, { "left": "issue:45381", @@ -7987,61 +7951,61 @@ "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Processor/tokenizer regression vs offloading/quantization behavior; no shared code-path." + "reason": "Processor/tokenizer regression vs quantization offloading behavior; not the same issue." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "MPS attention correctness issue vs Kimi tokenizer codec-handling regression; unrelated." + "reason": "MPS attention correctness issue vs Kimi-K2.5 tokenizer regression; unrelated." }, { "left": "issue:39692", "right": "issue:46192", "accept": false, - "reason": "SigLIP2 docs/example errors vs return_tensors silently ignored; different scope and fix." + "reason": "SigLIP2 documentation/example errors vs return_tensors being ignored; docs and API behavior mismatch, not the same bug." }, { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "StableLmConfig pad_token_idx regression vs model-loading failure for physical-intelligence/fast; unrelated." + "reason": "Missing pad_token_idx in StableLmConfig vs model loading failure for physical-intelligence/fast; unrelated." }, { "left": "issue:45520", "right": "issue:46178", "accept": false, - "reason": "flash_attn import KeyError vs tensorflow_text backend mapping error; both import issues but not same bug." + "reason": "flash_attn import KeyError on Python 3.13 vs tensorflow_text backend mapping error; different import paths." }, { "left": "issue:43572", "right": "issue:43957", "accept": false, - "reason": "Config pad_token_idx issue vs meta-device loading regression; different code paths." + "reason": "StableLmConfig pad_token_idx regression vs meta-device model loading breakage; unrelated." }, { "left": "issue:43550", "right": "issue:43844", "accept": false, - "reason": "torch.compile/SDPA failure vs ZeRO-3 gradient inflation; unrelated training/runtime problems." + "reason": "Bamba torch.compile SDPA bug vs gradient inflation under ZeRO-3; different training/runtime problems." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "State dict serialization inconsistency vs LoRA merge collapse with extended vocab; different root causes." + "reason": "Serialization of tied lm_head weights vs LoRA merge collapse with extended vocab; superficially about weights, but different failure modes." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "flash_support import error vs missing Qwen2_5_VLForConditionalGeneration export; unrelated import problems." + "reason": "Different import failures: flash_attention_utils symbol missing vs Qwen2.5-VL class not exported." }, { "left": "issue:43950", "right": "issue:46285", "accept": false, - "reason": "Non-persistent buffer corruption on from_pretrained vs CLIP model structure inconsistency; different regression classes." + "reason": "Non-persistent buffer corruption during from_pretrained vs CLIP text model structure inconsistency; unrelated regressions." } ] }, @@ -8116,20 +8080,20 @@ ], "soft_pairs": [ "issue:44898|issue:45412", - "issue:45200|issue:45325", "issue:43450|issue:43575", + "issue:45200|issue:45325", "issue:42490|issue:43493", "issue:43334|issue:43525", "issue:43388|issue:43526", - "issue:44792|issue:45563", "issue:42915|issue:43824", + "issue:44792|issue:45563", "issue:43824|issue:43975", "issue:38175|issue:43493", "issue:28218|issue:35141", "issue:41720|issue:46209", "issue:43278|issue:43856", - "issue:43653|issue:44112", "issue:43526|issue:46192", + "issue:43653|issue:44112", "issue:44442|issue:45198", "issue:43262|issue:43450", "issue:38175|issue:45412", @@ -8144,20 +8108,20 @@ "issue:43827|issue:43901", "issue:43819|issue:45083" ], - "prepared_review_unit_hash": "899264c323800d92c835b1cc583a3e20ba7761889572b6005d7d1c6e4272b91a", + "prepared_review_unit_hash": "b6660799a45ced439a88a72f5f1783221efa2c7c1a1b4744e9310611e4b9f7b3", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36297, + "serialized_chars": 36300, "estimated_input_tokens": 9075, "estimated_eval_tokens": 18406 }, @@ -8166,180 +8130,180 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is not a duplicate cluster: the issues span many unrelated regressions across tokenization, chat templates, vision/audio processors, model config loading, memory, and CI/platform-specific failures.", - "confidence": 0.99, - "canonical_issue_reason": "No single canonical issue fits here; the items are heterogeneous and do not describe the same underlying bug.", - "canonical_pr_reason": "No pull requests were provided in the cluster.", - "best_issue_reason": "No global best issue: none of the issues is a representative superset of the others, and merging them would conflate unrelated bugs.", - "best_pr_reason": "No pull requests were provided in the cluster.", + "summary": "All candidate pairs look like near-topic neighbors rather than true duplicates; they share broad subsystems (tokenizers, multimodal chat templates, model loading, docs) but not the same concrete bug or fix path. I would reject every soft edge. For a representative issue, `issue:44521` is the strongest candidate because it is active, specific, and has the most evidence around a concrete multimodal `apply_chat_template` failure.", + "confidence": 0.72, + "canonical_issue_reason": "`issue:44521` is the most representative active bug report in the set: it has a concrete repro, multiple inbound refs, and a focused multimodal `apply_chat_template` failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "`issue:44521` is the best standalone issue to keep as the primary reference because it is open, well-scoped, and clearly actionable.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44898", "right": "issue:45412", "accept": false, - "reason": "Different bugs: Perceiver interpolation on non-default resolution vs RT-DETR model memory not releasing on deletion." + "reason": "Different bugs: Perceiver resolution handling vs RT-DETR model memory release; no shared code path." }, { - "left": "issue:45200", - "right": "issue:45325", + "left": "issue:43450", + "right": "issue:43575", "accept": false, - "reason": "Different code paths: Gemma 4 multimodal token IDs defaulting vs Qwen2.5-VL RoPE scaling of still-image positions." + "reason": "Video processor batch-shape bug vs Qwen2 load OOM; unrelated symptoms and fixes." }, { - "left": "issue:43450", - "right": "issue:43575", + "left": "issue:45200", + "right": "issue:45325", "accept": false, - "reason": "Batched video processor output-shape bug is unrelated to tensor-parallel OOM when loading a large Qwen model." + "reason": "Gemma 4 `mm_token_type_ids` defaulting vs Qwen2.5-VL rope indexing; both multimodal, but different code paths." }, { "left": "issue:42490", "right": "issue:43493", "accept": false, - "reason": "Trailing-slash import warning is unrelated to a SigLIP2 implementation discrepancy against JAX." + "reason": "Trailing-slash local path warning vs SigLIP2 HF/JAX discrepancy; unrelated issues." }, { "left": "issue:43334", "right": "issue:43525", "accept": false, - "reason": "Both mention missing pad_token_id, but they affect different model configs and loading paths (Qwen3-VL vs Llama4)." + "reason": "Both mention missing `pad_token_id`, but they affect different model configs and likely need different fixes." }, { "left": "issue:43388", "right": "issue:43526", "accept": false, - "reason": "Different utilities: gather_for_metrics tuple handling vs BeitImageProcessorFast reduce_labels truncation." + "reason": "Metric-gathering tuple truncation vs BeitImageProcessorFast label reduction; not the same underlying bug." }, { - "left": "issue:44792", - "right": "issue:45563", + "left": "issue:42915", + "right": "issue:43824", "accept": false, - "reason": "A Janus image-generation test failure is not the same as a stale paged-generate warning about num_return_sequences." + "reason": "Qwen3Moe FP8 failure vs Qwen2.5-VL import error; unrelated model-loading problems." }, { - "left": "issue:42915", - "right": "issue:43824", + "left": "issue:44792", + "right": "issue:45563", "accept": false, - "reason": "Qwen3Moe FP8Config failure and Qwen2.5-VL import error are separate model/setup problems." + "reason": "Janus image-generation test failure vs stale `num_return_sequences` warning; different generate-path issues." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "Importing a missing class is unrelated to a detokenization regression in DeepSeek coder." + "reason": "Importing Qwen2.5-VL class vs DeepSeek detokenization regression; not the same bug." }, { "left": "issue:38175", "right": "issue:43493", "accept": false, - "reason": "Both involve SigLIP2, but one is zero probabilities on a checkpoint and the other is an HF-vs-JAX implementation mismatch; not clearly the same bug." + "reason": "Both are SigLIP2-related, but one is zero probabilities and the other is implementation mismatch vs JAX; too broad to merge." }, { "left": "issue:28218", "right": "issue:35141", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization during post_init." + "reason": "Tokenizer spacing after added token vs embedding reinit on resize; unrelated." }, { "left": "issue:41720", "right": "issue:46209", "accept": false, - "reason": "Different failures: auto device mapping CUDA assert on A800 vs FP8 inference incompatibility on B200." + "reason": "Qwen3 auto device-map CUDA assert vs FP8 inference failure on B200; distinct failures." }, { "left": "issue:43278", "right": "issue:43856", "accept": false, - "reason": "Embedding dtype drift in evaluation is unrelated to Qwen3 MoE training memory inefficiency." + "reason": "Eval dtype regression vs MoE training memory usage; different layers of the stack." }, { - "left": "issue:43653", - "right": "issue:44112", + "left": "issue:43526", + "right": "issue:46192", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug is unrelated to a GraniteSpeech CI test stale device override." + "reason": "BeitImageProcessorFast label reduction bug vs `return_tensors` being ignored; unrelated APIs." }, { - "left": "issue:43526", - "right": "issue:46192", + "left": "issue:43653", + "right": "issue:44112", "accept": false, - "reason": "reduce_labels truncating label arrays is unrelated to return_tensors being ignored when text_kwargs is passed." + "reason": "BigBirdTokenizer special-token registration vs GraniteSpeech CI test staleness; not the same code-path problem." }, { "left": "issue:44442", "right": "issue:45198", "accept": false, - "reason": "AutoTokenizer loading FastSpeech2ConformerTokenizer and Wav2Vec2 save/tokenization failure are different tokenizer bugs." + "reason": "Different tokenizer/model loading failures for different architectures; no single fix would cover both." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting is unrelated to batched video processor shape handling." + "reason": "Audio chat-template sampling-rate default vs video batching shape bug; unrelated processors." }, { "left": "issue:38175", "right": "issue:45412", "accept": false, - "reason": "SigLIP2 zero-probabilities behavior and RT-DETR GC memory retention are unrelated." + "reason": "SigLIP2 output issue vs RT-DETR memory leak; completely different bugs." }, { "left": "issue:45520", "right": "issue:46170", "accept": false, - "reason": "Python 3.13 flash_attn import key error is unrelated to Windows ROCm torch.distributed import/type errors." + "reason": "Python 3.13 `flash_attn` import handling vs Windows ROCm `torch.distributed` dependency; unrelated import issues." }, { "left": "issue:45290", "right": "issue:46326", "accept": false, - "reason": "Both touch chat templates, but the failure modes differ: tool-call messages with no content vs mistral_common continue_final_message behavior." + "reason": "Both touch `apply_chat_template`, but one is tool-call/no-content crashes and the other is `continue_final_message` on mistral_common; not mergeable as one fix." }, { "left": "issue:45759", "right": "issue:46285", "accept": false, - "reason": "Composite Qwen multimodal config unwrapping and CLIPTextModel structure inconsistency are different model-family issues." + "reason": "Composite Qwen config unwrap bug vs CLIP text-model structure inconsistency; different models and code paths." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "OmDet-Turbo processor output size mismatch is unrelated to LwDetrImageLoss crashing under float16 AMP." + "reason": "Processor input-size mismatch vs float16 AMP crash in a loss module; unrelated." }, { "left": "issue:44561", "right": "issue:46234", "accept": false, - "reason": "Missing is_torch_fx_available after v5 removal and missing is_falcon_mamba_ssm_available are separate import-utils symbols." + "reason": "Removal of `is_torch_fx_available` breaking remote-code models vs missing `is_falcon_mamba_ssm_available`; different import-utils gaps." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "DAC.from_latents vs forward mismatch has no relation to a GraniteSpeech stale CI test." + "reason": "DAC latent/forward mismatch vs GraniteSpeech CI test failure; unrelated." }, { "left": "issue:44514", "right": "issue:44521", "accept": false, - "reason": "Both are apply_chat_template bugs, but one crashes on batched padding=False input while the other produces all-zero assistant masks for multimodal inputs." + "reason": "Both involve `apply_chat_template`, but one crashes on batched padding=False input and the other returns all-zero assistant masks; different failure modes." }, { "left": "issue:44206", "right": "issue:45584", "accept": false, - "reason": "LasrFeatureExtractor center-arg crash is unrelated to Whisper empty transcription after align_special_tokens." + "reason": "Unsupported `center` arg in LasrFeatureExtractor vs Whisper empty transcription after align_special_tokens; unrelated." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "These are separate documentation regressions for different pipelines and deprecated arguments, not one underlying issue." + "reason": "Both are docs regressions, but they refer to different pipeline docs and different deprecated behavior; not the same change." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "DAC latent decoding mismatch is unrelated to qwen3_omni_moe helper length behavior." + "reason": "DAC consistency bug vs `_get_feat_extract_output_lengths` behavior in qwen3_omni_moe; unrelated." } ] }, @@ -8442,218 +8406,224 @@ "issue:39692|issue:44493", "issue:43335|issue:45593", "issue:34689|issue:35141", - "issue:44360|issue:44561", "issue:42490|issue:43854", - "issue:43606|issue:44857" + "issue:43606|issue:44857", + "issue:44360|issue:44561" ], - "prepared_review_unit_hash": "86a90dcfdfde90959904399155e60b3da0a6980053afb2c1a1fad4c1d0e312bc", + "prepared_review_unit_hash": "90563a2ac5c0aa685dd70203b7e661839a6933ca17516ede6dea7a10b2e80504", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 33, - "serialized_chars": 37208, - "estimated_input_tokens": 9302, - "estimated_eval_tokens": 18860 + "serialized_chars": 37210, + "estimated_input_tokens": 9303, + "estimated_eval_tokens": 18862 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly unrelated issue-to-issue similarity candidates. Only one pair is a clear duplicate: the repeated Qwen3.5 MoE `from_pretrained` regression report. No PRs are present.", - "confidence": 0.97, - "canonical_issue_reason": "issue:45310 is the earlier, more complete report of the Qwen3.5 MoE `from_pretrained` regression, and issue:45478 is the later duplicate with the same failure mode.", + "summary": "This set is mostly a collection of unrelated issue reports with only one clear duplicate pair: the repeated Qwen3.5 MoE `from_pretrained` error (issue 45310 / 45478). The soft links are generally too broad to merge.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45310 is the earliest and more detailed report of the only clear duplicate pair in this set; issue:45478 repeats the same Qwen3.5 MoE `from_pretrained` failure.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the strongest canonical issue because it precisely captures the concrete regression, is earlier, and subsumes the duplicate report.", + "best_issue_reason": "issue:45310 is the strongest canonical issue candidate because it matches the only exact duplicate and is the earlier report. The remaining issues are unrelated enough that none is a better global representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "Different models and failure modes: AMP/CUDA loss crash vs missing auxiliary losses when denoising is off." + "reason": "Different models and failure modes: AMP/CUDA loss crash vs missing auxiliary losses in D-FINE." }, { "left": "issue:43335", "right": "issue:43550", "accept": false, - "reason": "Unrelated symptoms: sparse-layer config bug vs torch.compile/SDPA failure in Bamba." + "reason": "Both mention model behavior, but the bugs are unrelated: SwitchTransformers config vs Bamba torch.compile/SDPA." }, { "left": "issue:43611", "right": "issue:46402", "accept": false, - "reason": "Both involve loading, but the concrete breakages differ: `base_model_prefix` loading vs Sam3Model text encoder weights not loaded." + "reason": "Both are loading-related, but one is a generic v5 `base_model_prefix` regression and the other is Sam3Model text-encoder weights not loading." }, { "left": "issue:43606", "right": "issue:45083", "accept": false, - "reason": "Different subsystems and bugs: CPU offload device mismatch vs feature-length helper behavior." + "reason": "Device-offload mismatch in Bark vs a helper-function output-length issue in qwen3_omni_moe; no shared code-path." }, { "left": "issue:37428", "right": "issue:43425", "accept": false, - "reason": "A missing internal import is not the same as general Torch 2.10 incompatibility." + "reason": "Import error for a missing flash-attention helper vs torch 2.10 compatibility; too broad to be the same bug." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Flash Attention 2 test failures are unrelated to the Trainer `_gpu_broadcast_one` NoneType error." + "reason": "Flash Attention 2 test failures and a Trainer broadcast TypeError are different failures with no concrete overlap." }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Different code paths: Pegasus output changes vs Qwen3 modular recurrent state reset." + "reason": "Different models and mechanisms: Pegasus output changes vs Qwen3 recurrent state reset with cache." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "Config/sparse-layer bug is unrelated to tokenizer special-token registration." + "reason": "SwitchTransformers sparse-layer config bug vs BigBirdTokenizer special-token registration bug." }, { "left": "issue:45200", "right": "issue:45588", "accept": false, - "reason": "Different models and mechanics: Gemma4 text-only `mm_token_type_ids` defaulting vs flash-attention `s_aux=None` crash." + "reason": "Gemma4 mm_token_type_ids defaulting issue vs flash_attention.py crashing on `s_aux=None`; unrelated." }, { "left": "issue:45127", "right": "issue:45245", "accept": false, - "reason": "LoRA/tied-embedding collapse is unrelated to the category-cardinality limit." + "reason": "LoRA/extended-vocab collapse and category-count limits are unrelated model-training constraints." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "Different failures: SDPA/torch.compile model issue vs stale CI device override test." + "reason": "Both involve CI/model execution, but one is Bamba SDPA compile failure and the other is a GraniteSpeech stale test." }, { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "Python 3.13 import/`torch.jit.script` parsing error is unrelated to the qwen3_omni_moe feature-length helper." + "reason": "Python 3.13 import/`torch.jit.script` parsing problem vs qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:37428", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but they concern different missing symbols and different regressions." + "reason": "Both are import errors, but they concern different missing symbols and different root causes." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "DAC latent/forward mismatch is unrelated to unexpected position-id keys during loading." + "reason": "DAC latent/forward mismatch is a different bug from unexpected `position_id` keys." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "Different models and bugs: SwitchTransformers sparse-layer creation vs Beit legacy-field loading." + "reason": "SwitchTransformers sparse-layer creation and Beit legacy-field loading are separate config/loading issues." }, { "left": "issue:43278", "right": "issue:43825", "accept": false, - "reason": "Training/evaluate dtype drift is unrelated to the pipeline translation-task error message." + "reason": "Embedding dtype regression and pipeline translation messaging are unrelated v5 issues." }, { "left": "issue:44206", "right": "issue:45823", "accept": false, - "reason": "Unsupported `center` arg in a feature extractor is unrelated to Gemma4 PLE device mismatch." + "reason": "Unsupported `center` arg in LasrFeatureExtractor vs Gemma4 PLE device mismatch; different code paths." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "LayoutLMv3 error messaging and Trainer broadcast TypeError are different bugs." + "reason": "Clarifying an error message in LayoutLMv3 is not the same bug as `_gpu_broadcast_one` rejecting NoneType." }, { "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "Doc example issues are unrelated to BeitImageProcessorFast label reduction behavior." + "reason": "Documentation example errors vs BeitImageProcessorFast `reduce_labels` behavior; not the same underlying change." }, { "left": "issue:45588", "right": "issue:45859", "accept": false, - "reason": "Flash-attention sink handling and Qwen3_5Moe rotary CPU-offload compatibility are distinct problems." + "reason": "Both touch flash/offload-related areas, but one is `s_aux=None` handling in flash_attention.py and the other is rotary embedding CPU-offload incompatibility." }, { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Chat-template tool-call crash and Qwen2.5-VL video position-id bug are different code paths." + "reason": "Chat-template tool-call crash and Qwen2.5-VL video `vision_position_ids` are distinct failures." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "SwitchTransformers config bug is unrelated to qwen3_omni_moe output-length helper behavior." + "reason": "A SwitchTransformers sparse-layer config bug is unrelated to qwen3_omni_moe feature-length output." }, { "left": "issue:43720", "right": "issue:44112", "accept": false, - "reason": "BitNet packed-weight loading bug is unrelated to a stale CI device override test." + "reason": "BitNet packed-weight loading and GraniteSpeech CI stale override tests do not share a concrete code path." }, { "left": "issue:45459", "right": "issue:46234", "accept": false, - "reason": "Tokenizer protobuf error handling and missing import-utils symbol are not the same regression." + "reason": "Both involve import utilities, but one hides tokenizer errors without protobuf and the other is a missing availability helper." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "Pegasus v4/v5 output differences are unrelated to DeepSeek tokenizer regression." + "reason": "Different tokenizer/model regressions; similar v5 timing is not enough to merge." }, { "left": "issue:45310", "right": "issue:45478", "accept": true, - "reason": "Same Qwen3.5 MoE `from_pretrained` failure on transformers>=5.4.0; the later issue repeats the same bug report." + "reason": "These are the same Qwen3.5 MoE `from_pretrained` regression reported twice with matching titles and symptom." }, { "left": "issue:43335", "right": "issue:44112", "accept": false, - "reason": "Sparse-layer configuration and GraniteSpeech CI device override are unrelated." + "reason": "SwitchTransformers config and GraniteSpeech CI issues are unrelated." }, { - "left": "issue:39692", - "right": "issue:44493", + "left": "issue:42490", + "right": "issue:43854", "accept": false, - "reason": "SigLIP2 doc example errors are unrelated to unexpected position-id keys." + "reason": "Local-path trailing-slash warning and GLM-4.7-Flash unit-test loading failure are different loading bugs." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "Device mismatch during CPU offload is unrelated to LwDetrImageLoss AMP crash." + "reason": "CPU offload device mismatch in Bark vs float16 AMP crash in LwDetrImageLoss; no shared underlying bug." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer architecture comment and removal of `is_torch_fx_available` affecting trust_remote_code are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The assessment is grounded in the packet: 45310 and 45478 have the same title and are a plausible duplicate pair, with 45310 earlier. The rest of the soft-edge rejections are conservative and stay within the evidence available from titles/metadata. No PR-pair mergeability issue applies because the packet contains issues only." + "feedback": "The summary is well grounded in the packet: issue 45310/45478 is the only clearly duplicated pair by title and symptom, and the soft-edge rejections stay appropriately conservative. I didn\u2019t see any unsupported overstatement or an implausibly asserted mergeable pair." }, "evaluator_used": true, "retried": false, @@ -8698,8 +8668,8 @@ "issue:43746", "issue:43873", "issue:43927", - "issue:43976", "issue:44448", + "issue:44466", "issue:44493", "issue:44561", "issue:44568", @@ -8712,7 +8682,6 @@ "issue:45083", "issue:45229", "issue:45372", - "issue:45440", "issue:45491", "issue:45563", "issue:45588", @@ -8725,10 +8694,10 @@ ], "soft_pairs": [ "issue:45920|issue:46192", - "issue:38175|issue:44493", "issue:44857|issue:45072", - "issue:40444|issue:43526", + "issue:38175|issue:44493", "issue:28218|issue:32090", + "issue:40444|issue:43526", "issue:44448|issue:44977", "issue:43701|issue:45083", "issue:44977|issue:45229", @@ -8742,8 +8711,8 @@ "issue:44561|issue:44779", "issue:10000|issue:15354", "issue:37428|issue:44987", - "issue:43278|issue:43329", "issue:27957|issue:34689", + "issue:43278|issue:43329", "issue:43873|issue:45229", "issue:44857|issue:45588", "issue:41720|issue:45229", @@ -8752,210 +8721,210 @@ "issue:43278|issue:43381", "issue:45372|issue:45588", "issue:45588|issue:45823", - "issue:43976|issue:45440" + "issue:44448|issue:44466" ], - "prepared_review_unit_hash": "89cbee519de7c51b551509db93d5316b3d58f01c3622deef09ff3fa97103355a", + "prepared_review_unit_hash": "05164501a961d68bd153844bbeacc4a381cde8e39d93a96db2f154afcb2ee06d", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36537, - "estimated_input_tokens": 9135, - "estimated_eval_tokens": 18526 + "serialized_chars": 36048, + "estimated_input_tokens": 9012, + "estimated_eval_tokens": 18280 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: tokenizer regressions, flash-attention issues, loading/import failures, dtype/device-mapping bugs, and unrelated training/runtime errors. There is no shared underlying bug or code path, so none of the soft pairs should be merged.", - "confidence": 0.96, - "canonical_issue_reason": "No canonical issue: the items span unrelated subsystems and failure modes across different models, versions, and workflows.", + "summary": "This is not a duplicate cluster: the items span many unrelated bugs, regressions, and model-specific failures. The soft-similarity pairs mostly share buzzwords or subsystems, but they describe different code paths and symptoms, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the set is heterogeneous and the candidate pairs are only textually similar, not the same underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue for this cluster; the closest thing to a hub is 45920, but it only covers one tokenizer-regression slice of the set.", + "best_issue_reason": "#45920 is the strongest standalone issue to use as a representative report: it describes a broad, reproducible tokenizer regression affecting multiple models and has inbound references.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45920", "right": "issue:46192", "accept": false, - "reason": "Tokenizer regression vs. a separate `return_tensors`/`text_kwargs` handling bug; different API surfaces and failure modes." - }, - { - "left": "issue:38175", - "right": "issue:44493", - "accept": false, - "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to the position-id key mismatch reported here." + "reason": "Tokenizer ID regression across many models vs. `return_tensors` being ignored when `text_kwargs` is passed; different API behaviors and failure modes." }, { "left": "issue:44857", "right": "issue:45072", "accept": false, - "reason": "Both involve precision, but one is a CUDA AMP crash in LwDetrImageLoss and the other is a bfloat16 inference dtype mismatch in different models." + "reason": "`LwDetrImageLoss` fp16/CUDA crash vs dtype mismatches in `SwitchTransformers`/`TimmWrapperModel` bfloat16 inference; unrelated modules." }, { - "left": "issue:40444", - "right": "issue:43526", + "left": "issue:38175", + "right": "issue:44493", "accept": false, - "reason": "Qwen2.5-VL multi-image finetuning failure and BeitImageProcessorFast label reduction are unrelated bugs." + "reason": "`siglip2` zero probabilities vs unexpected position-id key warnings across models; different symptom and likely cause." }, { "left": "issue:28218", "right": "issue:32090", "accept": false, - "reason": "Tokenizer extra-space behavior and Trainer `_gpu_broadcast_one` NoneType error are completely different problems." + "reason": "Tokenizer extra-space bug vs Trainer `_gpu_broadcast_one` NoneType error; unrelated code paths." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "IterableDataset multi-image finetuning failure vs `BeitImageProcessorFast.reduce_labels` returning one label; different subsystems." }, { "left": "issue:44448", "right": "issue:44977", "accept": false, - "reason": "Pegasus v4/v5 output drift is not the same issue as Qwen3.5 flash-attention generation failure." + "reason": "Pegasus v4/v5 output difference vs Qwen3.5 flash-attention generation crash; different model families and bugs." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "Checkpoint key mismatch during resume is unrelated to the qwen3_omni_moe helper-function bug." + "reason": "Checkpoint key mismatch on resume vs qwen3_omni_moe feature-length helper bug; unrelated behaviors." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Flash-attention generation bug vs. Gemma4 multi-GPU OOM; different subsystems and symptoms." + "reason": "Flash-attention generation failure vs Gemma4 multi-GPU CUDA OOM; different failure modes and code paths." }, { "left": "issue:43746", "right": "issue:46209", "accept": false, - "reason": "PEFT local-checkpoint loading in GraniteSpeech is unrelated to FP8 inference failing on B200." + "reason": "PEFT local checkpoint load failure vs FP8 inference on B200; no shared underlying defect." }, { "left": "issue:32090", "right": "issue:33357", "accept": false, - "reason": "NoneType GPU broadcast error and MacOS bus error on community CLIP are different runtime issues." + "reason": "`_gpu_broadcast_one` NoneType TypeError vs MacOS bus error with a CLIP model; different platforms and root causes." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Flash Attention 2 test failures do not match the embedding reinitialization bug in `post_init`." + "reason": "Flash Attention 2 test failures vs token-embedding resize reinitializing output embeddings; separate bugs." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "torch.compile/SDPA failure in Bamba-9B-v2 is unrelated to the qwen3_omni_moe helper-function problem." + "reason": "Bamba torch.compile/SDPA failure vs qwen3_omni_moe helper-function bug; unrelated model/runtime issues." }, { "left": "issue:44855", "right": "issue:46399", "accept": false, - "reason": "Python 3.13 import parsing failure in DebertaV2 is a different code path from Qwen3.5 rotary registration missing a decorator." + "reason": "Python 3.13 `@torch.jit.script` parsing/IndentationError vs missing decorator causing Qwen3.5 rotary registration error; superficially script-related but not the same bug." }, { "left": "issue:42490", "right": "issue:43575", "accept": false, - "reason": "Trailing-slash import warning and Qwen2-57B tp OOM are unrelated loading behaviors." + "reason": "Import warning for local paths with trailing slash vs OOM loading `Qwen2-57B-A14B-Instruct` with tensor parallelism; different problem class." }, { "left": "issue:43927", "right": "issue:45698", "accept": false, - "reason": "Both concern save/load, but one loses custom token IDs in config and the other loads the wrong custom module after save_pretrained." + "reason": "Custom token IDs lost on save/load vs wrong custom module loaded after `save_pretrained`; both involve serialization, but they affect different mechanisms." }, { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "Removal of `is_torch_fx_available` breaking trust_remote_code is unrelated to DeepSeek tokenizer regression." + "reason": "Removal of `is_torch_fx_available` breaking `trust_remote_code` vs Deepseek tokenizer regression in v5; unrelated API and tokenizer issues." }, { "left": "issue:10000", "right": "issue:15354", "accept": false, - "reason": "German DistilBertModel issue and ViT GeneratorExp torch.jit.script export failure are unrelated." + "reason": "German DistilBertModel issue vs GeneratorExp not supported by `torch.jit.script` while exporting ViT; no common underlying bug." }, { "left": "issue:37428", "right": "issue:44987", "accept": false, - "reason": "Missing `_flash_supports_window_size` import and a model-load failure for `physical-intelligence/fast` are not the same bug." + "reason": "Missing `_flash_supports_window_size` import vs failure loading `physical-intelligence/fast`; different import/runtime causes." }, { - "left": "issue:43278", - "right": "issue:43329", + "left": "issue:27957", + "right": "issue:34689", "accept": false, - "reason": "BF16-to-FP32 embedding dtype drift is unrelated to the undefined video-branch variables in `_get_num_multimodal_tokens`." + "reason": "XLMRoberta Flash Attention 2 support vs model loading broken for Llama 3.2 90B Vision Instruct; different models and symptoms." }, { - "left": "issue:27957", - "right": "issue:34689", + "left": "issue:43278", + "right": "issue:43329", "accept": false, - "reason": "XLMRoberta Flash Attention 2 compatibility and Llama 3.2 vision model loading are different model-specific failures." + "reason": "Embedding dtype changes from BF16 to FP32 in eval vs undefined helper names in video token counting; unrelated bugs." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Quantization offloading not working and Gemma4 multi-GPU OOM both involve memory, but they are distinct problems and code paths." + "reason": "Quantization offloading not working as expected vs Gemma4 multi-GPU OOM; both memory-related, but not the same defect." }, { "left": "issue:44857", "right": "issue:45588", "accept": false, - "reason": "A float16 AMP crash in LwDetrImageLoss is unrelated to an `s_aux=None` crash in flash_attention integration." + "reason": "`LwDetrImageLoss` fp16/CUDA crash vs `flash_attention.py` crashing on `s_aux=None`; distinct code paths." }, { "left": "issue:41720", "right": "issue:45229", "accept": false, - "reason": "Qwen3 auto device mapping assert and Gemma4 multi-GPU OOM are separate device/memory bugs." + "reason": "Qwen3 auto device mapping `cudaErrorAssert` vs Gemma4 multi-GPU OOM; different model-specific failures." }, { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Tokenizer special-token handling bug is unrelated to Gemma3 NaN embeddings from sliding-window attention." + "reason": "Tokenizer failing to add BOS/EOS with `add_special_tokens=True` vs Gemma3 NaN embeddings from sliding-window attention; tokenizer vs attention bug." }, { "left": "issue:44977", "right": "issue:45563", "accept": false, - "reason": "Qwen3.5 flash-attention generation failure and a stale paged-generate warning are not the same issue." + "reason": "Flash-attention generation crash vs stale `num_return_sequences` warning in paged generate; not the same bug." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Embedding dtype change during evaluate is unrelated to gradient checkpointing being disallowed in eval mode." + "reason": "Embedding dtype regression vs gradient checkpointing being disallowed in eval mode; unrelated behavior." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma 4 processor loading ImportError from `mistral_common` is unrelated to the flash_attention `s_aux=None` crash." + "reason": "`ReasoningEffort` import failure for Gemma 4 processor loading vs `flash_attention.py` AttributeError on `s_aux=None`; different subsystems." }, { "left": "issue:45588", "right": "issue:45823", "accept": false, - "reason": "Flash-attention crash and Gemma4 PLE device mismatch are different failure modes in different layers." + "reason": "`flash_attention.py` `s_aux=None` crash vs Gemma4 PLE device mismatch with `device_map='auto'`; distinct bugs." }, { - "left": "issue:43976", - "right": "issue:45440", + "left": "issue:44448", + "right": "issue:44466", "accept": false, - "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to DeepseekV3MoE behavioral divergence." + "reason": "Pegasus v4/v5 output differences vs device-dependent `lm_head.weight` serialization; both version-related, but not the same concrete issue." } ] }, @@ -9002,14 +8971,14 @@ "issue:43701", "issue:43756", "issue:43873", + "issue:43976", "issue:43994", "issue:44186", "issue:44206", "issue:44265", "issue:44361", - "issue:44448", + "issue:44442", "issue:44462", - "issue:44466", "issue:44479", "issue:44743", "issue:44871", @@ -9029,9 +8998,9 @@ "issue:46369" ], "soft_pairs": [ - "issue:44448|issue:44466", - "issue:43412|issue:45588", + "issue:43976|issue:45440", "issue:45372|issue:45406", + "issue:43412|issue:45588", "issue:25251|issue:32090", "issue:45198|issue:45803", "issue:43653|issue:44743", @@ -9041,9 +9010,9 @@ "issue:44361|issue:45198", "issue:44186|issue:45198", "issue:33357|issue:39692", + "issue:29127|issue:34567", "issue:43065|issue:45412", "issue:44206|issue:44479", - "issue:29127|issue:34567", "issue:43065|issue:43994", "issue:43611|issue:46121", "issue:44871|issue:44977", @@ -9051,216 +9020,223 @@ "issue:45459|issue:45712", "issue:42915|issue:42994", "issue:43278|issue:43575", - "issue:29942|issue:30064", "issue:45412|issue:46369", + "issue:29942|issue:30064", "issue:45372|issue:45459", "issue:43994|issue:45412", "issue:44265|issue:45198", "issue:43381|issue:43701", + "issue:43653|issue:44442", "issue:43756|issue:45440" ], - "prepared_review_unit_hash": "6dc869823af1d8b26d6c8c237db65ebadabf70553b4b20f0d1a4b3421d0c97f9", + "prepared_review_unit_hash": "2cb5259acb03ecfbc896aad9a09b3bb16291ae191eadd5e0d329d3b3bb325cdd", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36429, - "estimated_input_tokens": 9108, - "estimated_eval_tokens": 18472 + "soft_pair_count": 30, + "serialized_chars": 36616, + "estimated_input_tokens": 9154, + "estimated_eval_tokens": 18564 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the soft pairs mostly share broad themes (tokenizers, quantization, memory, vision models) but not the same underlying bug or fix path. I would not merge any of the soft pairs as duplicates.", - "confidence": 0.97, - "canonical_issue_reason": "Issue 45412 is the strongest standalone triage candidate: it has a concrete memory-leak symptom, user impact, and multiple references. That said, the set is not a true duplicate cluster, so this is only a practical anchor, not a duplicate canonical.", + "summary": "These are mostly unrelated issue candidates. Several share broad themes like Gemma, tokenizer, quantization, or memory behavior, but none of the soft pairs look like the same concrete bug/change.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 43873 is the broadest and most active representative in the set (open, multiple inbound refs, substantial discussion) even though there is no clear duplicate issue to canonize against.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45412 is the best representative issue because it is well-scoped, actionable, and high-impact compared with the others. It is the clearest single bug report, even though it is not actually duplicated by the rest.", + "best_issue_reason": "43873 is the strongest global representative because it has the most discussion and a broadly actionable runtime/quantization scope.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44448", - "right": "issue:44466", + "left": "issue:43976", + "right": "issue:45440", "accept": false, - "reason": "Both are v5-related regressions, but one is generation output drift and the other is serialization of tied weights; different code paths and fixes." + "reason": "Different problems: Python/version compatibility vs a DeepSeekV3MoE implementation divergence." }, { - "left": "issue:43412", - "right": "issue:45588", + "left": "issue:45372", + "right": "issue:45406", "accept": false, - "reason": "Different models and failure sites: Gemma3n export vs flash-attention runtime AttributeError. Not the same bug." + "reason": "Both involve Gemma 4, but one is an ImportError during processor loading and the other is a serve-time AttributeError for a missing tokenizer field." }, { - "left": "issue:45372", - "right": "issue:45406", + "left": "issue:43412", + "right": "issue:45588", "accept": false, - "reason": "Both involve Gemma 4, but one is an import/dependency failure and the other is a processor init attribute error. Distinct issues." + "reason": "Unrelated code paths: executorch export vs flash_attention integration crashing on sink-less models." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Completely unrelated: pipeline top_k shape change versus Trainer GPU broadcast TypeError." + "reason": "Pipeline top_k nesting bug is unrelated to Trainer NoneType broadcast failures." }, { "left": "issue:45198", "right": "issue:45803", "accept": false, - "reason": "Different components and root causes: Wav2Vec2 save/tokenization failure vs a bare-except in Fuyu tensor conversion." + "reason": "Both touch input processing, but one is save/tokenization failure in Wav2Vec2 and the other is a bare-except bug swallowing errors in Fuyu tensor conversion." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "Both are tokenizer/model bugs, but the first is special-token registration and the second is recurrent-state cache handling. Not mergeable into one fix." + "reason": "BigBird tokenizer special-token decode bug is unrelated to Qwen recurrent-state caching behavior." }, { "left": "issue:45245", "right": "issue:45563", "accept": false, - "reason": "Unrelated behaviors: category-count runtime error versus a stale generate warning." + "reason": "Categories-limit RuntimeError and stale generate warning are different issues with no shared failure path." }, { "left": "issue:32090", "right": "issue:33290", "accept": false, - "reason": "Different training stack problems: NoneType broadcast in Trainer vs DeepSpeed/Adafactor OOM." + "reason": "Trainer broadcast NoneType error is not the same as Deepspeed Adafactor OOM." }, { "left": "issue:42490", "right": "issue:43873", "accept": false, - "reason": "One is a local-path import warning; the other is quantization/offloading behavior. Same broad area, different bug." + "reason": "Local-path import warning is unrelated to quantization/offloading behavior." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "Different tokenizers/models and different failure modes. No shared underlying bug." + "reason": "Different tokenizers and different failure modes; no shared code bug." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 tokenizer task/padding crash is unrelated to the Wav2Vec2 save/tokenization issue." + "reason": "LayoutLMv2 NER/padding crash is unrelated to Wav2Vec2 save/tokenization failures." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "Both are vision-model problems, but one is a MacOS bus error and the other is a documentation/example mismatch plus quantization failure." + "reason": "MacOS bus error in CLIP loading is not the same as a SigLIP2 doc example/model-processor mismatch." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and TrainerState token counting are unrelated." }, { "left": "issue:43065", "right": "issue:45412", "accept": false, - "reason": "Different models and symptoms: dummy Conv2d placeholder vs memory not being released on delete." + "reason": "Dummy Conv2d in Sam3PixelDecoder is unrelated to RT-DETR memory release/GC behavior." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Unsupported feature-arg crash in a feature extractor versus a video-input regression in Qwen VL models. Different fixes." - }, - { - "left": "issue:29127", - "right": "issue:34567", - "accept": false, - "reason": "Unrelated: LayoutLMv3 error-message clarity versus TrainerState token counter not updating." + "reason": "Unsupported center arg in LasrFeatureExtractor is unrelated to the Qwen video regression." }, { "left": "issue:43065", "right": "issue:43994", "accept": false, - "reason": "Sam3 pixel-decoder placeholder layer and SigLIP2 inference nonsense are unrelated issues." + "reason": "Dummy layer cleanup and SigLIP2 nonsense outputs are different bugs." }, { "left": "issue:43611", "right": "issue:46121", "accept": false, - "reason": "Base-model-prefix loading regression and rope-validation TypeError are different model-loading/config bugs." + "reason": "base_model_prefix loading regression is unrelated to rope-validation list handling." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "Different models and symptoms: eos_token_id config mismatch versus flash-attention generation regression." + "reason": "Gemma eos_token_id mismatch and Qwen3.5 flash-attention generation failure are different model-specific issues." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Tokenizer decoder-type regression and AutoTokenizer ignoring tokenizer.json are separate loading/serialization issues." + "reason": "Tokenizer decoder-type regression in v5 is not the same as AutoTokenizer ignoring tokenizer.json." }, { "left": "issue:45459", "right": "issue:45712", "accept": false, - "reason": "Both concern fallback/error handling, but one is tokenizer decode exception masking and the other is leftover dummy PT classes. Not the same bug." + "reason": "Tokenizer error masking due to missing protobuf is unrelated to leftover dummy classes in dummy_pt_objects.py." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "Quantization-related, but one is FP8 config failure on model load and the other is quantized model saving failure. Different code paths." + "reason": "One is FP8 config loading for Qwen3Moe; the other is saving quantized models. Different lifecycle stages and failures." }, { "left": "issue:43278", "right": "issue:43575", "accept": false, - "reason": "Embedding dtype drift in eval and tensor-parallel OOM are unrelated." + "reason": "Embedding dtype drift in eval is unrelated to tensor-parallel OOM on Qwen2-57B-A14B-Instruct." }, { - "left": "issue:29942", - "right": "issue:30064", + "left": "issue:45412", + "right": "issue:46369", "accept": false, - "reason": "Flash Attention 2 test failures and void segmentation-map processing are different subsystems with no shared root cause." + "reason": "Both are memory-related, but RT-DETR GC leakage and Sam3-litetext memory overflow are not the same bug." }, { - "left": "issue:45412", - "right": "issue:46369", + "left": "issue:29942", + "right": "issue:30064", "accept": false, - "reason": "Both mention memory, but one is a GC leak in RT-DETR and the other is an overflow in sam3-litetext. Different bugs." + "reason": "FlashAttention 2 test failures are unrelated to void segmentation-map processing." }, { "left": "issue:45372", "right": "issue:45459", "accept": false, - "reason": "Gemma 4 loading import error and protobuf-related tokenizer error masking are separate issues." + "reason": "Gemma processor import failure and protobuf-related tokenizer error masking are separate issues." }, { "left": "issue:43994", "right": "issue:45412", "accept": false, - "reason": "SigLIP2 wrong outputs and RT-DETR memory leak are unrelated." + "reason": "SigLIP2 inference nonsense and RT-DETR memory retention are unrelated." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "torch.export failure on torch_compilable_check is a model-export issue, not the same as Wav2Vec2 save/tokenization failure." + "reason": "torch.export failure with torch_compilable_check is unrelated to Wav2Vec2 tokenizer/save_pretrained bugs." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient checkpointing eval-mode restriction and resume-from-checkpoint key mismatch are distinct training-state bugs." + "reason": "Gradient checkpointing in eval mode and resume_from_checkpoint key mismatch are different training-control issues." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "BigBird special-token decode bug is unrelated to AutoTokenizer failing to load a FastSpeech2ConformerTokenizer." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Both are model-implementation mismatches, but for different architectures and different concrete defects; not one fix." + "reason": "Smollm3 RoPE-layer mismatch is unrelated to DeepSeekV3MoE divergence." } ] }, @@ -9301,7 +9277,6 @@ "issue:43526", "issue:43531", "issue:43575", - "issue:43653", "issue:43701", "issue:43761", "issue:43824", @@ -9334,7 +9309,6 @@ "issue:46399" ], "soft_pairs": [ - "issue:43653|issue:44442", "issue:43421|issue:45491", "issue:27957|issue:29942", "issue:29127|issue:33357", @@ -9361,211 +9335,205 @@ "issue:44743|issue:44811", "issue:27957|issue:33290", "issue:43452|issue:45588", - "issue:44062|issue:45907", - "issue:33290|issue:34689" + "issue:33290|issue:34689", + "issue:44062|issue:45907" ], - "prepared_review_unit_hash": "4a8debc529e5f4460d701457db861bdc4083523425fcffd661e546b3bf9646c7", + "prepared_review_unit_hash": "1efeff33695d7ff2f23d4867eccd36d323767c7622ec6ca8e8f07aa62242f873", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36332, - "estimated_input_tokens": 9083, - "estimated_eval_tokens": 18422 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35435, + "estimated_input_tokens": 8859, + "estimated_eval_tokens": 17974 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a noisy, heterogeneous issue cluster with no clear duplicate group. The closest umbrella report is the generic Flash Attention 2 regression, but most other items are unrelated model, tokenizer, audio, or loading bugs.", - "confidence": 0.95, - "canonical_issue_reason": "Issue 29942 is the broadest and most generic report here (\u201cFailing Flash Attention 2 tests\u201d), so it is the closest fit as a representative issue, even though the cluster overall is not a strong duplicate set.", + "summary": "This cluster is highly heterogeneous: the soft pairs mostly connect unrelated bugs that only share broad Transformers vocabulary (models, tokenizers, attention, loading, OOM, etc.). I rejected every soft edge because the artifacts do not describe the same underlying defect or fixable change.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "29942 is the best single issue to stand in for the cluster because it has the widest scope among the mostly unrelated reports. It is not a true cluster center, just the least-specific representative.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43653", - "right": "issue:44442", - "accept": false, - "reason": "Both are tokenizer-related, but one is a BigBird special-token decode bug and the other is an AutoTokenizer loading failure for a different tokenizer; not the same bug." - }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Different code paths: runtime special-token/post-processor updates vs Gemma3 sliding-window NaN embeddings on GPU." + "reason": "Runtime special-token post-processing bug vs Gemma3 NaN embeddings under sliding-window/all-padding; different subsystems and failure modes." }, { "left": "issue:27957", "right": "issue:29942", "accept": false, - "reason": "Both mention Flash Attention 2, but one is XLM-RoBERTa model support and the other is failing tests; the symptoms and likely fixes are different." + "reason": "One is XLM-RoBERTa Flash Attention 2 support, the other is failing FA2 tests; related area but not the same bug." }, { "left": "issue:29127", "right": "issue:33357", "accept": false, - "reason": "LayoutLMv3 error-message clarity and a CLIP MacOS bus error are unrelated." + "reason": "LayoutLMv3 box-validation clarity issue vs MacOS bus error loading a CLIP model; unrelated." }, { "left": "issue:45440", "right": "issue:45593", "accept": false, - "reason": "DeepSeekV3MoE divergence and D-FINE auxiliary-loss behavior are different models and different bugs." + "reason": "DeepSeekV3 implementation divergence vs D-FINE auxiliary loss behavior; different models and defects." }, { "left": "issue:43994", "right": "issue:46369", "accept": false, - "reason": "SigLIP2 nonsensical outputs and SAM3 LiteText memory overflow are unrelated inference issues." + "reason": "SigLIP2 incorrect outputs vs SAM3 LiteText memory overflow; not the same code path." }, { "left": "issue:43526", "right": "issue:45834", "accept": false, - "reason": "BEiT image label reduction and Kosmos2.5 long-OCR indexing are different preprocessing/model bugs." + "reason": "BeitImageProcessorFast reduce_labels bug vs Kosmos2.5 long OCR index error; different processors and symptoms." }, { "left": "issue:43824", "right": "issue:44279", "accept": false, - "reason": "Qwen2.5-VL import failure and a generic dependency issue are not the same concrete problem." + "reason": "Specific Qwen2.5-VL import error vs a generic dependency issue; no shared concrete bug." }, { "left": "issue:43526", "right": "issue:44186", "accept": false, - "reason": "Both are tokenizer/image-processing related, but one is a BEiT fast image processor label bug and the other is a LayoutLMv2 tokenizer crash; different failure modes and fixes." + "reason": "BeitImageProcessorFast label reduction vs LayoutLMv2Tokenizer NER/padding crash; unrelated tokenizer/processor bugs." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "Error-message clarity in LayoutLMv3 and embedding reinitialization after resizing are unrelated." + "reason": "LayoutLMv3 error message concern vs token-embedding resize/post_init reinitialization bug; different functionality." }, { "left": "issue:44625", "right": "issue:45584", "accept": false, - "reason": "Qwen3.5 num_labels propagation and Whisper empty-transcription generation failure are different subsystems." + "reason": "Qwen3.5 num_labels propagation bug vs Whisper empty-transcription generation failure; different models and paths." }, { "left": "issue:43575", "right": "issue:43701", "accept": false, - "reason": "Qwen2-57B tensor-parallel OOM and checkpoint key mismatch are different bugs." + "reason": "Tensor-parallel OOM when loading Qwen2-57B vs resume_from_checkpoint key mismatch; unrelated training/loading problems." }, { "left": "issue:44442", "right": "issue:45072", "accept": false, - "reason": "FastSpeech2Conformer tokenizer loading and dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated." + "reason": "Tokenizer loading failure vs dtype mismatch in bf16 inference; not the same bug." }, { "left": "issue:42490", "right": "issue:43381", "accept": false, - "reason": "Local-path slash import warning and gradient-checkpointing eval-mode restriction are not the same issue." + "reason": "Import warning for local path with trailing slash vs gradient checkpointing in eval mode; unrelated." }, { "left": "issue:43575", "right": "issue:44387", "accept": false, - "reason": "Both involve OOM, but one is a specific model load with tensor parallelism and the other is a broader int4 quantization memory regression." + "reason": "Both mention memory, but one is TP load OOM for a specific Qwen model and the other is increased reserved memory under int4 quantization; different causes." }, { "left": "issue:42915", "right": "issue:43931", "accept": false, - "reason": "Qwen3Moe FP8 config failure and Qwen3-VL weight-shape mismatch are separate model-loading bugs." + "reason": "Qwen3Moe FP8 config failure vs Qwen3-VL weight-shape mismatch; distinct model/config issues." }, { "left": "issue:43761", "right": "issue:45859", "accept": false, - "reason": "CLIPVision hidden_states regression and Qwen3.5 MoE CPU-offload incompatibility are unrelated." + "reason": "CLIPVisionModel hidden_states regression vs Qwen3_5Moe rotary embedding CPU offload incompatibility; unrelated." }, { "left": "issue:43526", "right": "issue:44857", "accept": false, - "reason": "BEiT label reduction and LwDETR AMP/CUDA crash are unrelated image-model bugs." + "reason": "BeitImageProcessorFast label reduction vs LwDetrImageLoss AMP/CUDA crash; different components." }, { "left": "issue:46121", "right": "issue:46399", "accept": false, - "reason": "Rope-validation TypeError and Qwen3.5 kernel registration/decorator failure are different concrete code-path problems." + "reason": "Rope-validation TypeError from JSON-loaded list vs missing decorator for apply_rotary_pos_emb registration; unrelated implementation bugs." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Qwen3-VL pad_token_id loading failure and Qwen3-MoE sliding-window behavior are different model/config issues." + "reason": "Missing pad_token_id in Qwen3-VL text config vs sliding_window issue in Qwen3-MoE; different config fields and failures." }, { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Qwen3.5 flash-attention generation failure and the 2^24 category limit error are unrelated." + "reason": "Flash-attention generation bug for Qwen3.5 vs category-count RuntimeError; no shared code path." }, { "left": "issue:44464", "right": "issue:45491", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency and Gemma3 sliding-window NaNs are different generation bugs." + "reason": "Compiled-forward chunked generation inconsistency vs Gemma3 NaN embeddings with sliding-window attention; distinct bugs." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Audio chat-template sample-rate defaulting and a LASR feature extractor argument crash are unrelated audio-processing issues." + "reason": "Audio chat-template sample-rate default bug vs LasrFeatureExtractor unsupported center argument; unrelated audio-processing issues." }, { "left": "issue:27957", "right": "issue:29127", "accept": false, - "reason": "Flash Attention 2 support for XLM-RoBERTa and a LayoutLMv3 error-message problem do not share the same underlying bug." + "reason": "Flash Attention 2 support for XLM-RoBERTa vs LayoutLMv3 error-message clarity for box inputs; different models and concerns." }, { "left": "issue:44743", "right": "issue:44811", "accept": false, - "reason": "Qwen3.5 recurrent state reset and Whisper batch_decode skip_special_tokens behavior are unrelated." + "reason": "Qwen3.5 recurrent-state reset bug in modular code vs Whisper processor batch_decode skip_special_tokens issue; unrelated model/processor behavior." }, { "left": "issue:27957", "right": "issue:33290", "accept": false, - "reason": "Flash Attention 2 model support and Adafactor/deepspeed OOM are unrelated." + "reason": "FA2 support issue vs Adafactor/deepspeed OOM; no overlap beyond generic training/inference context." }, { "left": "issue:43452", "right": "issue:45588", "accept": false, - "reason": "gguf_file loading breakage and a flash_attention.py None-handling crash are different code paths." + "reason": "gguf_file breaks model/tokenizer loading vs flash_attention.py crash on s_aux=None; different failure sites and causes." }, { - "left": "issue:44062", - "right": "issue:45907", + "left": "issue:33290", + "right": "issue:34689", "accept": false, - "reason": "AddedToken special-arg duplication and list_repo_files revision forwarding are unrelated API bugs." + "reason": "Adafactor OOM in deepspeed vs Llama 3.2 Vision model-loading breakage; unrelated." }, { - "left": "issue:33290", - "right": "issue:34689", + "left": "issue:44062", + "right": "issue:45907", "accept": false, - "reason": "Adafactor/deepspeed OOM and Llama 3.2 Vision model loading breakage are not the same issue." + "reason": "AddedToken special-arg duplication vs list_repo_files missing revision forwarding; different APIs and bugs." } ] }, @@ -9646,9 +9614,9 @@ "issue:45003|issue:45440", "issue:43723|issue:45479", "issue:42490|issue:44743", - "issue:43761|issue:44079", "issue:43316|issue:43525", "issue:42175|issue:46234", + "issue:43761|issue:44079", "issue:34689|issue:43611", "issue:43720|issue:44442", "issue:43232|issue:43388", @@ -9667,136 +9635,136 @@ "issue:44855|issue:45137", "issue:44190|issue:44496" ], - "prepared_review_unit_hash": "f39f07340dd07e2311ca468cf846f59c5882df6599b786fdc395dc09bd48f4f7", + "prepared_review_unit_hash": "c46279eb56890f3738842754a18e983ba353523fd8b68df001fa214aa1275d5d", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35769, - "estimated_input_tokens": 8943, - "estimated_eval_tokens": 18142 + "serialized_chars": 35766, + "estimated_input_tokens": 8942, + "estimated_eval_tokens": 18140 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is a loose collection of unrelated Transformers issues; the soft pairs mostly share broad subsystem vocabulary but not the same underlying bug or fix. I rejected all soft edges.", - "confidence": 0.89, - "canonical_issue_reason": "No single canonical duplicate stands out because the issues span unrelated regressions, model-loading failures, config mismatches, generation bugs, and docs-only problems.", + "summary": "This cluster is a heterogeneous set of unrelated Transformers issues with only superficial overlap in themes like loading, generation, and config handling. I found no true duplicates among the soft edges, and there is no clear single canonical issue or PR for the cluster.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 45698 is the broadest and most currently active loading/saving bug in the set, with multiple inbound references and substantial discussion, so it is the best representative issue if one must be chosen.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Both mention custom model init/post_init behavior, but 44077 is a model-specific policy issue while 43646 is a broader v5 initialization regression; not the same bug." + "reason": "Both mention model initialization/post_init in v5, but one is a broad custom-model init regression and the other is a patchtsmixer-specific post_init policy change." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "Adafactor+DeepSpeed OOM is unrelated to token-embedding resize/post_init reinitialization." + "reason": "OOM with Adafactor+DeepSpeed is unrelated to embedding resize/post_init reinitialization." }, { "left": "issue:33290", "right": "issue:34567", "accept": false, - "reason": "OOM in training and a stale TrainerState token counter are different code paths and symptoms." + "reason": "Optimizer memory blowup and TrainerState token counter not updating are separate bugs." }, { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Gradient checkpointing in eval mode and quantization offloading are separate runtime behaviors." + "reason": "Gradient checkpointing in eval mode and quantization offloading are different execution paths and failure modes." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Tokenizer loading failure and Perceiver image interpolation failure are unrelated." + "reason": "Tokenizer-loading failure and Perceiver image-classification resolution failure are unrelated subsystems." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "Unsafe sys.modules access in modeling_utils is not the same as DeepseekV3MoE divergence from remote implementation." + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to DeepseekV3MoE implementation divergence." }, { "left": "issue:43723", "right": "issue:45479", "accept": false, - "reason": "AutoTokenizer loading in v5 and degenerate single-label loss are unrelated bugs." + "reason": "AutoTokenizer v5 loading regression and single-label classification zero-loss bug do not share the same underlying code path." }, { "left": "issue:42490", "right": "issue:44743", "accept": false, - "reason": "Local-path slash import warning is unrelated to recurrent state reset during cached generation." - }, - { - "left": "issue:43761", - "right": "issue:44079", - "accept": false, - "reason": "CLIPVision hidden_states regression and ModelOutput key assignment are related only at a very high level; the specific failure modes and likely fixes differ." + "reason": "Trailing-slash local model loading warning is unrelated to recurrent-state reset in Qwen3.5 cache handling." }, { "left": "issue:43316", "right": "issue:43525", "accept": false, - "reason": "Gemma3TextConfig API mismatch and missing pad_token_id on Llama4Config are different config regressions." + "reason": "Both involve config APIs, but one is a Gemma3TextConfig mismatch and the other is a missing pad_token_id on Llama4Config." }, { "left": "issue:42175", "right": "issue:46234", "accept": false, - "reason": "Backend packaging omission and a missing import_utils availability helper are unrelated." + "reason": "Torch-only install backend inclusion and a missing import_utils availability helper are different packaging/import issues." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression is model-specific; ModelOutput key assignment bug is a generic container issue." }, { "left": "issue:34689", "right": "issue:43611", "accept": false, - "reason": "Llama 3.2 model loading regression is not the same as the v5 base_model_prefix loading breakage." + "reason": "Both are loading regressions, but they concern different releases and different causes (Llama 3.2 vision load vs base_model_prefix handling)." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "BitNet packed-weight loading and tokenizer loading are different subsystems." + "reason": "BitNet packed-weight loading failure is unrelated to a tokenizer class import/load failure." }, { "left": "issue:43232", "right": "issue:43388", "accept": false, - "reason": "Generation kwargs update after sync_gpus and metric gathering on tuple labels are unrelated." + "reason": "Generation kwargs sync after sync_gpus and gather_for_metrics label trimming are distinct distributed-training bugs." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch and Qwen3.5 save_pretrained checkpoint corruption are different failure points." + "reason": "Qwen3-VL weight-shape mismatch on load and Qwen3.5 save_pretrained corruption are related only at a very broad model-family level." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "CLIPVision hidden_states output regression is not the same as attention kernel divergence under torch.compile." + "reason": "Hidden_states omission in CLIPVisionModel and attention-kernel divergence under torch.compile are different issues." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "PatchTST post_init policy and Qwen VL video-input regression are unrelated." + "reason": "patchtsmixer post_init policy and Qwen video-input regression are unrelated model families and code paths." }, { "left": "issue:43493", @@ -9808,61 +9776,61 @@ "left": "issue:44987", "right": "issue:46178", "accept": false, - "reason": "Both are loading-related, but one is a model/package load failure and the other is a backend registry error; no shared underlying fix is evident." + "reason": "Loading a specific model under transformers>=5.1.0 and missing BACKENDS_MAPPING support for tensorflow_text are unrelated failures." }, { "left": "issue:43873", "right": "issue:44871", "accept": false, - "reason": "Quantization offloading and Gemma-3 eos_token_id inconsistency are separate bugs." + "reason": "Quantization/offloading behavior and Gemma-3 eos_token_id configuration mismatch are different bugs." }, { "left": "issue:42907", "right": "issue:45698", "accept": false, - "reason": "Saving dequantized models and loading the wrong custom module after save_pretrained are both serialization-adjacent, but they are not the same bug." + "reason": "Saving dequantized Ministrals/Devstrals and loading the wrong custom module after save_pretrained are not the same defect." }, { "left": "issue:44442", "right": "issue:45803", "accept": false, - "reason": "Tokenizer loading failure is unrelated to Fuyu tensor conversion error handling." + "reason": "Tokenizer loading and Fuyu tensor conversion exception handling are unrelated." }, { "left": "issue:44464", "right": "issue:46421", "accept": false, - "reason": "Chunked generation output inconsistency and cache lazy initialization causing recompilation are different symptoms and likely different fixes." + "reason": "Chunked generation inconsistency with compiled forward and cache lazy initialization causing recompilation are related only superficially." }, { "left": "issue:43749", "right": "issue:43931", "accept": false, - "reason": "FSDP CPU RAM efficient loading breakage and Qwen3-VL shape mismatch are unrelated loading problems." + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL shape mismatch are separate loading problems." }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Generic config API discrepancy and composite Qwen text_config unwrapping are not the same issue." + "reason": "Gemma3TextConfig API discrepancy and Qwen composite-config unwrapping are both config-related but not the same underlying bug." }, { "left": "issue:44479", "right": "issue:45290", "accept": false, - "reason": "Video-input regression in Qwen VL models and chat-template tool-call crash are separate feature paths." + "reason": "Video-input regression in Qwen VL models and chat-template crashes on tool-call assistant messages are different features." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "Python 3.13 decorator parsing/import issue and DeepSpeed ZeRO3 deque underflow are unrelated." + "reason": "Python 3.13 JIT parsing/IndentationError and DeepSpeed ZeRO3 deque underflow are unrelated." }, { "left": "issue:44190", "right": "issue:44496", "accept": false, - "reason": "Local dataset loading in a training script and unrecognized model_type during AutoModel load are different failure modes." + "reason": "Local dataset loading in an example script and an unrecognized custom model type are unrelated." } ] }, @@ -9939,8 +9907,8 @@ "issue:43643|issue:44913", "issue:43295|issue:43825", "issue:43550|issue:43701", - "issue:43650|issue:46202", "issue:25251|issue:28282", + "issue:43650|issue:46202", "issue:43606|issue:43701", "issue:45632|issue:46424", "issue:42617|issue:43575", @@ -9950,217 +9918,217 @@ "issue:43295|issue:44871", "issue:43316|issue:43716", "issue:44610|issue:46369", - "issue:43316|issue:44877", "issue:43065|issue:44617", + "issue:43316|issue:44877", "issue:43540|issue:44514", - "issue:43404|issue:43441", "issue:44871|issue:45584", + "issue:43404|issue:43441", "issue:44077|issue:44534", "issue:45803|issue:46365", "issue:44162|issue:44792", "issue:45803|issue:46170", "issue:41950|issue:43475", "issue:44871|issue:46050", - "issue:43452|issue:44496", "issue:43646|issue:44913", + "issue:43452|issue:44496", "issue:43761|issue:44479" ], - "prepared_review_unit_hash": "c2c9bd015aec009c7b3983aaa054d6e8ec2a159f7cb6fcb1c1f1f7e172c965b5", + "prepared_review_unit_hash": "9f84122799a74d02b7a7e6e12f4cf5eeb3bc290dffbf3a738eb349525566281b", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 35736, - "estimated_input_tokens": 8934, - "estimated_eval_tokens": 18124 + "serialized_chars": 35731, + "estimated_input_tokens": 8933, + "estimated_eval_tokens": 18122 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a broad mix of unrelated Transformers issues. The soft pairs mostly share superficial keywords like configs, loading, or model families, but they do not describe the same underlying bug or change path. No PRs are present, and there is no single canonical issue for the whole cluster.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "This cluster is mostly a grab bag of unrelated Transformers issues spanning model loading, processor regressions, video/multimodal bugs, cache behavior, and a few v5 regressions. I don\u2019t see any soft pair that looks like the same concrete bug/change, so all candidate duplicate edges should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43646 is the broadest, most umbrella-like report in the set (`Transformers 5.0.0 breaks custom model initialization`) and is the closest thing to a cluster anchor, though this is not a true duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:43646 is the best single issue to represent the cluster because it is the most general v5-regression report and could plausibly subsume nearby initialization-related breakages, but the overall set is still too heterogeneous for a strong canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "Both are config-related, but one is about trust_remote_code returning incomplete fields while the other is GPTNeoX rotary_pct not persisting on reload." + "reason": "Both are config/load behavior bugs, but one is about trust_remote_code dropping fields and the other about rotary_pct not round-tripping; different code paths and fixes." }, { "left": "issue:43295", "right": "issue:43825", "accept": false, - "reason": "Different regressions: processor.tokenizer/images behavior in v4.57.5 vs an incorrect pipeline translation error message in v5." + "reason": "Both mention v5 regressions, but one is processor.tokenizer / image passing behavior and the other is an incorrect pipeline error message for translation tasks; not the same bug." }, { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "torch.compile with SDPA is a model execution bug; resume_from_checkpoint key mismatch is a training checkpoint-loading issue." + "reason": "Unrelated failures: torch.compile with SDPA in Bamba vs resume_from_checkpoint key mismatch." }, { - "left": "issue:43650", - "right": "issue:46202", + "left": "issue:25251", + "right": "issue:28282", "accept": false, - "reason": "No meaningful overlap; one is a vague/off-topic issue, the other is a Gemma pipeline roundtrip tokenization bug." + "reason": "Pipeline top_k output shape regression vs missing PyTorch import error; no shared underlying bug." }, { - "left": "issue:25251", - "right": "issue:28282", + "left": "issue:43650", + "right": "issue:46202", "accept": false, - "reason": "Nested-list output from pipeline top_k is unrelated to the PyTorch missing import error for AutoModel." + "reason": "Placeholder/body noise vs Gemma pipeline roundtrip issue; no meaningful technical overlap." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device mismatch and checkpoint resume key mismatch are unrelated failure modes in different code paths." + "reason": "CPU offload device mismatch vs checkpoint key mismatch; distinct subsystems and failure modes." }, { "left": "issue:45632", "right": "issue:46424", "accept": false, - "reason": "Both mention cache behavior, but one is a trust_remote_code path collision and the other is max_cache_length handling for quantized cache only." + "reason": "Both involve cache, but one is a trust_remote_code path collision and the other is max_cache_length handling for quantized cache; different defects." }, { "left": "issue:42617", "right": "issue:43575", "accept": false, - "reason": "Running 3d_parallel.py and OOM when loading Qwen2-57B with tensor parallelism are different issues." + "reason": "3d_parallel.py launch failure vs tensor-parallel OOM on Qwen2-57B; not the same problem." }, { "left": "issue:43575", "right": "issue:43742", "accept": false, - "reason": "OOM from tensor-parallel loading a large model is unrelated to a key error when loading MobileLLM-125M." + "reason": "OOM during TP loading of a large Qwen model vs KeyError loading MobileLLM-125M; unrelated." }, { "left": "issue:16998", "right": "issue:22355", "accept": false, - "reason": "DeBERTa-V3 model_max_length behavior and missing transformers.onnx module are unrelated." + "reason": "model_max_length behavior in DeBERTa-V3 vs missing transformers.onnx module; no duplicate relationship." }, { "left": "issue:27957", "right": "issue:32090", "accept": false, - "reason": "Flash Attention 2 support for XLMRoberta is unrelated to a Trainer broadcast TypeError on NoneType." + "reason": "Flash Attention 2 support for XLMRoberta vs Trainer broadcast TypeError; different code paths." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "One is a regression around processor.tokenizer and image passing; the other is an eos_token_id config mismatch in Gemma-3." + "reason": "One is a processor/tokenizer regression in v5, the other is an eos_token_id config mismatch in Gemma-3; similar theme but not the same bug." }, { "left": "issue:43316", "right": "issue:43716", "accept": false, - "reason": "Gemma3TextConfig API discrepancy and image preprocessor dtype mismatch are different bugs." + "reason": "API discrepancy in Gemma3TextConfig vs dtype mismatch between image preprocessor and model; unrelated." }, { "left": "issue:44610", "right": "issue:46369", "accept": false, - "reason": "OmDet-Turbo image size mismatch is unrelated to sam3-litetext memory overflow." + "reason": "OmDet-Turbo preprocessing size mismatch vs Sam3-LiteText memory overflow; different models and failure modes." }, { - "left": "issue:43316", - "right": "issue:44877", + "left": "issue:43065", + "right": "issue:44617", "accept": false, - "reason": "Both are configuration-related, but they concern different models and different problems: API discrepancy vs strict config loading failure." + "reason": "Dummy Conv2d implementation issue vs CUDA OOM in Sam3Video; not the same underlying defect." }, { - "left": "issue:43065", - "right": "issue:44617", + "left": "issue:43316", + "right": "issue:44877", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d and Sam3Video CUDA OOM are not the same bug, even if they share the Sam3 family name." + "reason": "Gemma3TextConfig API discrepancy vs strict config blocking granite_speech load; different config-validation problems." }, { "left": "issue:43540", "right": "issue:44514", "accept": false, - "reason": "Qwen3OmniMoe video input validation error and Qwen2_5_VLProcessor batched apply_chat_template crash are different code paths." + "reason": "Different Qwen multimodal bugs: video input ValueError in Qwen3OmniMoe vs batched chat template crash in Qwen2_5_VLProcessor." }, { - "left": "issue:43404", - "right": "issue:43441", + "left": "issue:44871", + "right": "issue:45584", "accept": false, - "reason": "lm_head weight tying in Mistral3ForConditionalGeneration and FlashAttention failure in Ministral-3 are distinct model issues." + "reason": "Gemma eos_token_id inconsistency vs Whisper empty-transcription failure after align_special_tokens; unrelated." }, { - "left": "issue:44871", - "right": "issue:45584", + "left": "issue:43404", + "right": "issue:43441", "accept": false, - "reason": "Gemma-3 eos_token_id inconsistency is unrelated to Whisper empty-transcription failure after align_special_tokens." + "reason": "Weight tying bug in Mistral3ForConditionalGeneration vs FlashAttention failure in Ministral-3; same family, different concrete issues." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "patchtsmixer post_init validation and non-persistent buffer junk values are unrelated Transformer v5 issues." + "reason": "Optional post_init policy issue vs non-persistent buffers filled with junk; unrelated v5 behavior changes." }, { "left": "issue:45803", "right": "issue:46365", "accept": false, - "reason": "Bare except swallowing errors in FuyuBatchFeature is unrelated to importing Trainer failing because of datasets." + "reason": "Bare except swallowing interrupts in Fuyu conversion vs Trainer import conflict with datasets; no shared bug." }, { "left": "issue:44162", "right": "issue:44792", "accept": false, - "reason": "ESM2 being broken and a failed janus image-generation test do not describe the same defect." + "reason": "ESM2 regression impacting workflows vs janus image-generation test failure; not the same code path." }, { "left": "issue:45803", "right": "issue:46170", "accept": false, - "reason": "FuyuBatchFeature exception handling and a Windows/ROCm import dependency issue are unrelated." + "reason": "Error swallowing in Fuyu conversion vs Windows/ROCm import failure due to torch.distributed dependency; unrelated." }, { "left": "issue:41950", "right": "issue:43475", "accept": false, - "reason": "Video-classification pipeline choosing image processors is a pipeline bug; missing fpn_position_embeddings is a SAM3 video model output issue." + "reason": "Video-classification pipeline using image processors vs missing fpn_position_embeddings on Sam3VisionEncoderOutput; different bugs." }, { "left": "issue:44871", "right": "issue:46050", "accept": false, - "reason": "Gemma-3 eos_token_id inconsistency is unrelated to sinq quantization save/load breakage." + "reason": "Gemma eos_token_id config mismatch vs sinq quantization save/load breakage; unrelated." }, { - "left": "issue:43452", - "right": "issue:44496", + "left": "issue:43646", + "right": "issue:44913", "accept": false, - "reason": "gguf_file handling for tokenizer/model loading and unrecognized model_type in config.json are different loading failures." + "reason": "Broad v5 custom initialization breakage vs GPTNeoX rotary_pct not persisting on reload; different symptoms and fixes." }, { - "left": "issue:43646", - "right": "issue:44913", + "left": "issue:43452", + "right": "issue:44496", "accept": false, - "reason": "Transformers 5 custom model initialization breakage and GPTNeoXConfig rotary_pct reset are unrelated." + "reason": "gguf_file handling for tokenizer/model loading vs unrecognized model_type in config.json; both loading-related but not the same defect." }, { "left": "issue:43761", "right": "issue:44479", "accept": false, - "reason": "CLIPVisionModel hidden_states regression and Qwen video input regression are separate model-specific bugs." + "reason": "CLIPVisionModel hidden_states regression vs Qwen video-input regression in v5.3.0; unrelated model families and code paths." } ] }, @@ -10238,8 +10206,8 @@ "issue:45127|issue:45200", "issue:43673|issue:44743", "issue:43572|issue:46178", - "issue:43504|issue:45698", "issue:44617|issue:45584", + "issue:43504|issue:45698", "issue:44734|issue:45406", "issue:45542|issue:46234", "issue:43673|issue:46421", @@ -10255,219 +10223,216 @@ "issue:35141|issue:42490", "issue:44479|issue:44734", "issue:43827|issue:43994", - "issue:44855|issue:45588", "issue:41720|issue:45245", + "issue:44855|issue:45588", "issue:43065|issue:44610", "issue:44617|issue:46369", "issue:44933|issue:45362", "issue:44464|issue:45563", "issue:44521|issue:45381" ], - "prepared_review_unit_hash": "92036fce97b29fe54cf88b8aec182eb9e2412ef80b718295966d3fb53c08ad95", + "prepared_review_unit_hash": "87b73cd22c6a783ba4a6150a7134fec1ade35ef397a47de61e81de5bc01efbeb", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36134, - "estimated_input_tokens": 9034, - "estimated_eval_tokens": 18324 + "serialized_chars": 36132, + "estimated_input_tokens": 9033, + "estimated_eval_tokens": 18322 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated issues. The only close thematic family is cache/chunked-prefill behavior, but the packet does not justify merging any of those into a confirmed duplicate set.", - "confidence": 0.84, - "canonical_issue_reason": "issue:43673 is still the best anchor because it states the most concrete cache/chunked_prefill regression. issue:46421 is related in theme, but the evidence here only supports a nearby candidate, not the same bug.", + "summary": "The cluster does not cohere around one underlying bug or change. The soft-similarity pairs are mostly unrelated regressions across different models, loaders, and runtime paths, so I reject all candidate duplicate links.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the items span unrelated bugs/regressions (generation/cache, multimodal input, config loading, backend/import errors, docs, and CI/test failures).", "canonical_pr_reason": null, - "best_issue_reason": "issue:43673 is the strongest representative issue in the cluster: specific symptom, specific subsystem, and the clearest shared theme among the few cache-related candidates.", + "best_issue_reason": "No clear global best issue for deduping; the cluster is too heterogeneous to anchor on one representative bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43388", "right": "issue:43688", "accept": false, - "reason": "Different bugs: tuple-label handling in metric gathering vs auxiliary-loss normalization." + "reason": "Different bugs: batch-label gathering vs auxiliary-loss normalization." }, { "left": "issue:45127", "right": "issue:45200", "accept": false, - "reason": "Different failure modes and code paths: LoRA merge collapse vs missing multimodal token-type defaults." + "reason": "Different areas: LoRA/vocab merge collapse vs Gemma4 multimodal token defaults." }, { "left": "issue:43673", "right": "issue:44743", "accept": false, - "reason": "Both mention cache-like behavior, but the mechanisms differ; this is related, not confidently the same bug." + "reason": "Both involve cache/state, but the concrete failures differ (missing cache in chunked prefill vs recurrent state reset with seq_len>1)." }, { "left": "issue:43572", "right": "issue:46178", "accept": false, - "reason": "Unrelated config regression vs backend-registration/import error." + "reason": "Unrelated: StableLm config field regression vs TensorFlow backend mapping error." }, { - "left": "issue:43504", - "right": "issue:45698", + "left": "issue:44617", + "right": "issue:45584", "accept": false, - "reason": "Different loading bugs: legacy-field preset loading vs wrong custom module selection after save/load." + "reason": "Different models and failures: Sam3Video OOM vs Whisper empty-transcription generation crash." }, { - "left": "issue:44617", - "right": "issue:45584", + "left": "issue:43504", + "right": "issue:45698", "accept": false, - "reason": "Completely different symptoms and models: OOM vs empty-transcription generation failure." + "reason": "Different load/save bugs: legacy preset field vs wrong custom module selection after save_pretrained." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "Both touch serving, but one is KV-cache continuation indexing and the other is a missing processor attribute." + "reason": "Serve KV-cache indexing crash is unrelated to Gemma4Processor missing _tokenizer." }, { "left": "issue:45542", "right": "issue:46234", "accept": false, - "reason": "Separate import/backend issues with no shared concrete code path." + "reason": "Different import/backend issues: undefined TF backend vs missing falcon-mamba availability helper." }, { "left": "issue:43673", "right": "issue:46421", "accept": false, - "reason": "Shared chunked-prefill/cache theme, but the packet does not establish the same underlying root cause; treat as a related duplicate candidate only." + "reason": "Both mention chunked prefill/cache, but one is missing cache creation and the other is lazy-init recompilation; not clearly the same bug." }, { "left": "issue:43525", "right": "issue:43531", "accept": false, - "reason": "Different model/config regressions: pad_token_id absence vs sliding_window behavior." + "reason": "Different Qwen/Llama config issues: missing pad_token_id vs sliding_window behavior." }, { "left": "issue:45381", "right": "issue:46369", "accept": false, - "reason": "Different problems: video position IDs vs memory overflow on another model." + "reason": "Different problems: Qwen2.5-VL video position IDs vs Sam3-LiteText memory overflow." }, { "left": "issue:44792", "right": "issue:44977", "accept": false, - "reason": "Unrelated failures: Janus test case vs Qwen3.5 flash-attention generation." + "reason": "Unrelated model/test failure vs flash-attention generation bug." }, { "left": "issue:45563", "right": "issue:45584", "accept": false, - "reason": "Different generation issues: stale warning handling vs empty-output crash after token alignment." + "reason": "Stale generate warning is unrelated to Whisper's empty transcription failure." }, { "left": "issue:43761", "right": "issue:44242", "accept": false, - "reason": "Different bugs in different subsystems: CLIP hidden-states output vs router-logit loss handling." + "reason": "Hidden-states regression is unrelated to missing load-balancing loss." }, { "left": "issue:44792", "right": "issue:45600", "accept": false, - "reason": "Different CI/model issues with no shared underlying mechanism." + "reason": "Janus image-generation test failure is unrelated to broken auto_mappings imports." }, { "left": "issue:43931", "right": "issue:44060", "accept": false, - "reason": "Different Qwen3-family issues: weight-shape mismatch vs a tied-weights warning." + "reason": "Both are Qwen3-family issues, but one is a weight-shape load error and the other is an incorrect tied-weights warning." }, { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "Different loading failures in different workflows/models." + "reason": "Local dataset loading failure is unrelated to multimodal model load mismatch." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "Serialization/buffer behavior vs config reload regression; not the same bug." + "reason": "Non-persistent buffer junk and rotary_pct reload loss are different config/runtime bugs." }, { "left": "issue:35141", "right": "issue:42490", "accept": false, - "reason": "Unrelated: embedding reinitialization on resize vs local-path import warning." + "reason": "Embedding reinit after resizing is unrelated to the trailing-slash local path import warning." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Different regressions: video-input behavior in Qwen models vs serve KV-cache indexing." + "reason": "Both are runtime regressions, but one is video-input handling and the other is KV-cache continuation indexing." }, { "left": "issue:43827", "right": "issue:43994", "accept": false, - "reason": "Docs cleanup is unrelated to the SigLIP2 inference bug." + "reason": "Docs referencing pipeline() is unrelated to siglip2 producing nonsensical outputs." }, { - "left": "issue:44855", - "right": "issue:45588", + "left": "issue:41720", + "right": "issue:45245", "accept": false, - "reason": "Different runtime/import errors in unrelated modules." + "reason": "Qwen3 device-mapping cuda assert is unrelated to the 2^24 category limit runtime error." }, { - "left": "issue:41720", - "right": "issue:45245", + "left": "issue:44855", + "right": "issue:45588", "accept": false, - "reason": "Distinct failures: CUDA assert on device mapping vs category-count limit error." + "reason": "Python 3.13 decorator parsing error is unrelated to the flash_attention s_aux=None crash." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Different model/processor mismatches with no shared concrete code path." + "reason": "Dummy Conv2d in Sam3PixelDecoder is unrelated to OmDet-Turbo's input-size mismatch." }, { "left": "issue:44617", "right": "issue:46369", "accept": false, - "reason": "Both are memory-related, but they involve different models and different failure modes." + "reason": "Same broad SAM family, but different model variants and different memory issues." }, { "left": "issue:44933", "right": "issue:45362", "accept": false, - "reason": "Unrelated import error vs Qwen3.5 chat crash." + "reason": "Missing import from image_utils is unrelated to Qwen3.5-35B chat crashes." }, { "left": "issue:44464", "right": "issue:45563", "accept": false, - "reason": "Different generation issues: compiled-forward inconsistency vs stale warning text." + "reason": "Chunked-generation inconsistency with compiled forward is unrelated to a stale warning about num_return_sequences." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "Both are multimodal, but assistant-mask generation and video position IDs are not the same bug." + "reason": "All-zero assistant masks for multimodal inputs is a different bug from Qwen2.5-VL video position IDs." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The cluster-level summary is mostly grounded, but one soft-edge verdict is too strong for the evidence in the packet. For issue:43673 vs issue:46421, the titles support a related chunked-prefill/cache theme, but not confidently the same underlying bug or root cause. That should be phrased more conservatively as a likely related duplicate candidate, not a confirmed same-bug pair. Because the accepted edge overstates what the packet shows, I\u2019m rejecting." - }, - "evaluator_used": true, - "retried": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -10548,204 +10513,210 @@ "issue:44737|issue:45803", "issue:43116|issue:44190", "issue:43295|issue:45276", - "issue:44938|issue:45542", "issue:45276|issue:45584", - "issue:45632|issue:46421", + "issue:44938|issue:45542", "issue:36246|issue:41628", + "issue:45632|issue:46421", "issue:46082|issue:46121", - "issue:43404|issue:45216", "issue:45245|issue:45584", + "issue:43404|issue:45216", "issue:42175|issue:44279", "issue:36246|issue:38617", - "issue:44246|issue:45003", "issue:41950|issue:44016", + "issue:44246|issue:45003", "issue:44186|issue:44734", "issue:44683|issue:45446" ], - "prepared_review_unit_hash": "43710a4208f402b94db8ffd8f49f0e2bfcb3898d6a18a30db9efa4b37732fc3c", + "prepared_review_unit_hash": "c62e0eeee1d1b3660d6dcd1c962e5ab4486c2454299dca83ef3e8e46a7ee2e17", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 27, - "serialized_chars": 35353, - "estimated_input_tokens": 8839, - "estimated_eval_tokens": 17934 + "serialized_chars": 35358, + "estimated_input_tokens": 8840, + "estimated_eval_tokens": 17936 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: most pairs only share broad keywords like import, config, regression, or CI, not the same underlying bug. I would not merge any of the proposed soft edges.", - "confidence": 0.88, - "canonical_issue_reason": "Issue 41628 is the broadest user-facing API/import failure in the set and is a reasonable anchor for the small import-error subset, though the overall cluster is not a true duplicate group.", + "summary": "This cluster is mostly a grab bag of distinct Transformers bugs and regressions, with only superficial similarity around imports, configs, or pipeline behavior. None of the proposed soft pairs look like the same underlying change, so I would not merge any of them as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "issue:41628 is the cleanest broad, user-facing import regression report in the set, but the overall cluster is too heterogeneous to treat as one duplicate group.", "canonical_pr_reason": null, - "best_issue_reason": "41628 is the most general and representative issue among these disparate reports; it is the best single anchor only because no tighter canonical duplicate exists.", + "best_issue_reason": "issue:41628 is the best anchor issue because it describes a clear top-level API failure with broad impact and easy reproduction compared with the more model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43644", "right": "issue:44913", "accept": false, - "reason": "Different bug classes: buffer initialization at load time vs config persistence/reload defaulting." + "reason": "Both involve config/state handling, but one is junk non-persistent buffers and the other is rotary_pct not persisting; different bugs and code paths." }, { "left": "issue:43976", "right": "issue:44855", "accept": false, - "reason": "Both are Python-version-related failures, but one is package compatibility and the other is a JIT parsing/IndentationError in a model import path." + "reason": "Both are Python compatibility failures, but one is a version-loading problem and the other is a DebertaV2 torch.jit parsing IndentationError on 3.13." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "Unrelated: torch.compile/SDPA model execution vs local dataset loading in an example script." + "reason": "One is a Bamba torch.compile/SDPA failure; the other is a local dataset loading issue in an image-classification example script." }, { "left": "issue:44462", "right": "issue:45907", "accept": false, - "reason": "Different repo/file-listing behavior: tokenizer.json selection vs forwarding a revision argument." + "reason": "Tokenizer.json precedence and list_repo_files revision forwarding are unrelated repository/file-access bugs." }, { "left": "issue:25251", "right": "issue:27957", "accept": false, - "reason": "Pipeline top_k output-shape bug is unrelated to Flash Attention 2 support on XLMRoberta." + "reason": "Top-k pipeline output shape and XLMRoberta Flash Attention 2 are unrelated pipeline/model issues." }, { "left": "issue:44077", "right": "issue:44877", "accept": false, - "reason": "Both touch config handling, but they are different model families and different failure modes." + "reason": "These are both config-related, but one is about disallowing optional post_init and the other about strict config loading for granite_speech." }, { "left": "issue:45127", "right": "issue:45357", "accept": false, - "reason": "Both are regression/save_pretrained-related, but they concern different models and different state being saved." + "reason": "LoRA merge collapse with tied embeddings and incorrect visual encoder keys in save_pretrained are different save/merge regressions." }, { "left": "issue:43540", "right": "issue:44734", "accept": false, - "reason": "Video-input processing in Qwen3OmniMoe is unrelated to KV-cache continuation indexing in transformers serve." + "reason": "Qwen3OmniMoe video processing and serve KV-cache continuation indexing are different failures in different stacks." }, { "left": "issue:44112", "right": "issue:45600", "accept": false, - "reason": "Both cause CI failures, but the affected code and root causes are unrelated." + "reason": "A stale GraniteSpeech CI test and broken auto_mappings references to removed configs are unrelated CI breakages." }, { "left": "issue:45632", "right": "issue:46234", "accept": false, - "reason": "Cache-path collision for local models is not the same as a missing availability helper in import_utils." + "reason": "Cache-path collisions for trust_remote_code and a missing import_utils symbol are not the same bug." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "Single-layer sparse-layer config bug is unrelated to a local dataset loading failure." + "reason": "SwitchTransformers sparse-layer config creation and a local dataset loading failure are unrelated." }, { "left": "issue:44737", "right": "issue:45803", "accept": false, - "reason": "Device placement bug in XLNet and bare-except error swallowing in Fuyu are different code paths and defects." + "reason": "XLNet CPU device placement and a bare-except swallowing errors in FuyuBatchFeature are different code-path defects." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Example multi-label classification output bug is unrelated to local dataset loading." + "reason": "A multi-label classification example returning empty results is unrelated to loading a local dataset in another example script." }, { "left": "issue:43295", "right": "issue:45276", "accept": false, - "reason": "Both involve multimodal/model plumbing, but one is processor/tokenizer regression and the other is embedding resize propagation." + "reason": "Processor/tokenizer access regression and Gemma4 embedding resize propagation are different bugs." }, { - "left": "issue:44938", - "right": "issue:45542", + "left": "issue:45276", + "right": "issue:45584", "accept": false, - "reason": "Python 3.14 import failure is not the same as missing TensorFlow backend when only tensorboard is installed." + "reason": "Gemma4 embedding resizing and Whisper empty-transcription generation are unrelated model behaviors." }, { - "left": "issue:45276", - "right": "issue:45584", + "left": "issue:44938", + "right": "issue:45542", "accept": false, - "reason": "Embedding-resize propagation bug is unrelated to Whisper empty-transcription generation after token alignment." + "reason": "Python 3.14 import failure and missing TensorFlow backend due to tensorboard-only install are separate environment issues." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but they involve different symbols and likely different missing-export causes." + "reason": "Both are import errors, but they concern different missing symbols and different modules; no evidence of one underlying fix." }, { - "left": "issue:46082", - "right": "issue:46121", + "left": "issue:45632", + "right": "issue:46421", "accept": false, - "reason": "Config validation for head_dim is unrelated to rope-parameter conversion failing on a JSON-loaded list." + "reason": "Trust-remote-code cache collisions and chunked-prefill lazy cache initialization are different caching problems." }, { - "left": "issue:43404", - "right": "issue:45216", + "left": "issue:46082", + "right": "issue:46121", "accept": false, - "reason": "Both concern saving or tying model weights, but they affect different models and different checkpoint logic." + "reason": "LlamaConfig head_dim validation and rope-validation list handling are distinct config-validation bugs." }, { "left": "issue:45245", "right": "issue:45584", "accept": false, - "reason": "Category-count limit error is unrelated to Whisper generation on empty transcription." + "reason": "A categories-limit runtime error and Whisper empty transcription are unrelated failures." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Mistral3 weight tying and Qwen3.5 checkpoint saving are different model serialization/tie-weights issues." }, { "left": "issue:42175", "right": "issue:44279", "accept": false, - "reason": "Missing TensorFlow in the torch extra install is not the same as a generic dependency issue." + "reason": "Missing TensorFlow in the torch extra and a generic dependency issue are not the same concrete bug." }, { "left": "issue:36246", "right": "issue:38617", "accept": false, - "reason": "Different missing imports in different modules; too broad to be the same underlying bug." + "reason": "These are different missing-import regressions in different modules, with no shared failing code path." }, { - "left": "issue:44246", - "right": "issue:45003", + "left": "issue:41950", + "right": "issue:44016", "accept": false, - "reason": "Import slowness and unsafe sys.modules access are related at most indirectly, not the same bug." + "reason": "A pipeline looking for image processors and a notebook syntax error are completely unrelated." }, { - "left": "issue:41950", - "right": "issue:44016", + "left": "issue:44246", + "right": "issue:45003", "accept": false, - "reason": "Pipeline image-processor lookup bug is unrelated to a notebook syntax error." + "reason": "Slow import and unsafe sys.modules access are different import-time issues with different root causes." }, { "left": "issue:44186", "right": "issue:44734", "accept": false, - "reason": "Tokenizer padding/NER crash is unrelated to serve KV-cache continuation indexing." + "reason": "LayoutLMv2 tokenizer crashes and serve KV-cache continuation indexing are unrelated tokenizer/server bugs." }, { "left": "issue:44683", "right": "issue:45446", "accept": false, - "reason": "Both mention flex_attention and torch versions, but one is a compiled-kernel failure and the other is a version-check/import guard bug; not the same concrete change." + "reason": "Both mention flex_attention, but one is a compiled flex_attention failure on newer torch and the other is a version-check/import bug for AuxRequest." } ] }, @@ -10771,7 +10742,6 @@ "nodes": [ "issue:27957", "issue:28282", - "issue:32090", "issue:35141", "issue:42907", "issue:43295", @@ -10812,9 +10782,11 @@ "issue:45593", "issue:45600", "issue:45759", + "issue:46082", "issue:46123", "issue:46234", "issue:46369", + "issue:46399", "issue:46424" ], "soft_pairs": [ @@ -10826,15 +10798,14 @@ "issue:44112|issue:44964", "issue:44016|issue:44220", "issue:27957|issue:35141", - "issue:46123|issue:46369", "issue:43479|issue:45375", + "issue:46123|issue:46369", "issue:42907|issue:43441", "issue:43531|issue:45600", "issue:43976|issue:44987", "issue:43295|issue:45335", "issue:43526|issue:43901", "issue:28282|issue:35141", - "issue:32090|issue:35141", "issue:44496|issue:45042", "issue:43441|issue:45588", "issue:45372|issue:46234", @@ -10849,72 +10820,73 @@ "issue:45335|issue:45584", "issue:44496|issue:45759", "issue:44075|issue:44368", - "issue:43577|issue:45216" + "issue:43577|issue:45216", + "issue:46082|issue:46399" ], - "prepared_review_unit_hash": "e33d1756db839fc9d0964ef9de9431cc94d1a22bf8e31369d75d0a43c208516e", + "prepared_review_unit_hash": "aca53eca54d40d4ac5ba212c2c634db28875966f941987c182910290fda52c24", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 36810, - "estimated_input_tokens": 9203, - "estimated_eval_tokens": 18662 + "serialized_chars": 37406, + "estimated_input_tokens": 9352, + "estimated_eval_tokens": 18960 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is a grab bag of unrelated issue reports across model loading, configs, generation, caching, and docs. None of the soft pairs looks like the same underlying bug/change closely enough to merge, so I reject all candidate duplicates.", - "confidence": 0.96, - "canonical_issue_reason": "There is no clear canonical issue for the whole set because the items span many unrelated bugs rather than one duplicate cluster.", + "summary": "The soft-pair suggestions are mostly false positives: they share broad keywords like loading, config, cache, or model families, but the underlying bugs and code paths differ. No single issue stands out as a true canonical duplicate target for this heterogeneous set.", + "confidence": 0.88, + "canonical_issue_reason": "No clear canonical issue: the cluster spans many unrelated bugs across different models, processors, configs, and utilities, so none is a reliable duplicate anchor.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44964 is the cleanest standalone report: specific model, clear failure mode, and a concrete regression description.", + "best_issue_reason": "No global best issue: there isn\u2019t a single issue here that represents the whole set or is clearly the central duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43673", "right": "issue:46424", "accept": false, - "reason": "Both mention cache behavior, but one is about GenerationMixin chunked_prefill cache availability and the other about max_cache_length only applying to Quantized Cache; different bugs." + "reason": "Both mention cache behavior, but one is about chunked_prefill cache availability and the other about max_cache_length on quantized cache only; different bugs." }, { "left": "issue:43650", "right": "issue:44220", "accept": false, - "reason": "No real overlap beyond being bug reports; ADD THE DATA is unrelated to `_torch_extract_fbank_features()`." + "reason": "Unrelated: one is a vague data request, the other is a specific _torch_extract_fbank_features() issue in audio preprocessing." }, { "left": "issue:44964", "right": "issue:45600", "accept": false, - "reason": "Both are model-loading related, but they fail for different reasons and different model families; not the same underlying issue." + "reason": "Different failures: Phi-4 multimodal loading vs auto_mappings referencing removed Sam3LiteText configs." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "One is missing fields after `trust_remote_code=True`, the other is an unrecognized model/config.json problem; distinct code paths." + "reason": "Both are config-loading problems, but one is about trust_remote_code dropping fields and the other about an unrecognized model missing model_type." }, { "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "Different models and different import/load failures; too broad to be duplicates." + "reason": "Different models and failure modes: Qwen2.5-VL import error vs Phi-4 multimodal load issue." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "CI device-override test flake in GraniteSpeech is unrelated to Phi-4 multimodal loading." + "reason": "CI stale test failure is unrelated to Phi-4 multimodal model loading." }, { "left": "issue:44016", @@ -10926,151 +10898,151 @@ "left": "issue:27957", "right": "issue:35141", "accept": false, - "reason": "Flash Attention on XLM-Roberta is unrelated to token embedding resizing reinitialization." + "reason": "XLMRoberta FlashAttention support and token-embedding reinit on resize are different code paths." }, { - "left": "issue:46123", - "right": "issue:46369", + "left": "issue:43479", + "right": "issue:45375", "accept": false, - "reason": "MaskGenerationPipeline batch-finalization bug is unrelated to sam3-litetext memory overflow." + "reason": "Both are multimodal config issues, but they affect different fields and model families; not the same underlying bug." }, { - "left": "issue:43479", - "right": "issue:45375", + "left": "issue:46123", + "right": "issue:46369", "accept": false, - "reason": "Both concern config handling, but they are different model-specific field/default issues and not the same bug." + "reason": "MaskGenerationPipeline batch-handling bug is unrelated to sam3-litetext memory overflow." }, { "left": "issue:42907", "right": "issue:43441", "accept": false, - "reason": "One is a save/dequantization problem and the other is a FlashAttention load failure for a specific model; not the same issue." + "reason": "Same model family, but one is about saving dequantized checkpoints and the other about FlashAttention loading; different bugs." }, { "left": "issue:43531", "right": "issue:45600", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to stale auto_mappings references." + "reason": "Qwen3-MoE sliding_window bug is unrelated to auto_mappings referencing removed Sam3LiteText configs." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Python version compatibility and loading a specific model are different failures." + "reason": "Python version support issue and a specific model-loading failure are not the same bug." }, { "left": "issue:43295", "right": "issue:45335", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to t5gemma embedding resize behavior." + "reason": "One is a processor/tokenizer regression with images; the other is resize_token_embeddings not updating decoder.embed_tokens." }, { "left": "issue:43526", "right": "issue:43901", "accept": false, - "reason": "Image label reduction bug and pipeline docs mismatch are unrelated." + "reason": "BEiTImageProcessorFast label reduction bug and a docs mismatch for TextClassificationPipeline are unrelated." }, { "left": "issue:28282", "right": "issue:35141", "accept": false, - "reason": "Missing PyTorch import error is unrelated to embedding reinitialization." - }, - { - "left": "issue:32090", - "right": "issue:35141", - "accept": false, - "reason": "Trainer `_gpu_broadcast_one` NoneType error is unrelated to token embedding resizing." + "reason": "PyTorch missing import error is unrelated to embedding resizing/reinitialization." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Unrecognized model config and PIL backend requiring torchvision are different bugs." + "reason": "Unrecognized model config and PIL processor torchvision dependency regression are different problems." }, { "left": "issue:43441", "right": "issue:45588", "accept": false, - "reason": "Both mention FlashAttention, but one is model-specific and the other is an `s_aux=None` crash in the integration code path; not mergeable as one fix." + "reason": "Both touch flash attention, but the concrete failures and code paths differ: Ministral loading vs s_aux=None crash in flash_attention.py." }, { "left": "issue:45372", "right": "issue:46234", "accept": false, - "reason": "Gemma 4 processor import regression and missing `is_falcon_mamba_ssm_available` are unrelated." + "reason": "Different missing-import issues affecting different features and modules." }, { "left": "issue:44016", "right": "issue:44062", "accept": false, - "reason": "Notebook syntax error and `AddedToken()` argument conflict are separate issues." + "reason": "Notebook syntax error is unrelated to AddedToken receiving duplicate special arguments." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "Qwen3VL loading failure and Qwen3.5 save_pretrained regression are different operations and different bugs." + "reason": "Both involve Qwen multimodal models, but one is a loading regression and the other is a save_pretrained checkpoint issue." }, { "left": "issue:43673", "right": "issue:44336", "accept": false, - "reason": "Cache-related generation bug is unrelated to ANSI codes in `loading_report`." + "reason": "Cache missing during chunked_prefill and stray ANSI codes in loading_report are unrelated." }, { "left": "issue:44792", "right": "issue:46424", "accept": false, - "reason": "Janus image-generation test failure is unrelated to cache-length behavior." + "reason": "Janus image-generation test failure is unrelated to max_cache_length handling." }, { "left": "issue:43976", "right": "issue:44279", "accept": false, - "reason": "Python version support issue and generic dependency issue are not the same failure." + "reason": "Both mention dependency problems, but they are not the same concrete issue." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient checkpointing eval-mode restriction and D-FINE auxiliary-loss behavior are unrelated." + "reason": "Gradient checkpointing in eval mode and missing auxiliary losses in D-FINE are unrelated." }, { "left": "issue:43824", "right": "issue:44933", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols/modules and different root causes." + "reason": "Both are import errors, but they concern different symbols and modules." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "Docs mismatch for `return_all_scores` is unrelated to an unrecognized model config error." + "reason": "Docs for return_all_scores and an unrecognized model config are unrelated." }, { "left": "issue:45335", "right": "issue:45584", "accept": false, - "reason": "Embedding resize not propagating to decoder weights and Whisper empty-transcription failure are unrelated." + "reason": "Different model-specific bugs: t5gemma embedding resize vs Whisper empty-transcription generation failure." }, { "left": "issue:44496", "right": "issue:45759", "accept": false, - "reason": "Both are config/model-loading related, but one is an unrecognized model and the other is improper `text_config` unwrapping; different fixes." + "reason": "Both are config/model instantiation issues, but one is an unrecognized model and the other is failure to unwrap text_config." }, { "left": "issue:44075", "right": "issue:44368", "accept": false, - "reason": "Optimizer argument handling and a tie_word_embeddings warning are unrelated." + "reason": "Optimizer SGD args not being used is unrelated to the Qwen3.5 tie_word_embeddings warning." }, { "left": "issue:43577", "right": "issue:45216", "accept": false, - "reason": "BLIP2 dtype propagation bug is unrelated to Qwen3.5 checkpoint saving." + "reason": "BLIP2 dtype handling and Qwen3.5 checkpoint saving are different bugs." + }, + { + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "LlamaConfig head_dim validation and missing apply_rotary_pos_emb decorator are unrelated." } ] }, @@ -11104,6 +11076,7 @@ "issue:43479", "issue:43525", "issue:43650", + "issue:43653", "issue:43854", "issue:43901", "issue:43976", @@ -11133,27 +11106,25 @@ "issue:45397", "issue:45405", "issue:45542", + "issue:45593", "issue:45812", "issue:45823", "issue:45834", "issue:45907", - "issue:46082", "issue:46132", - "issue:46209", - "issue:46399" + "issue:46209" ], "soft_pairs": [ - "issue:46082|issue:46399", - "issue:45103|issue:45372", "issue:43650|issue:46209", + "issue:45103|issue:45372", "issue:44242|issue:45200", "issue:44279|issue:45405", "issue:38175|issue:44617", "issue:44779|issue:45812", - "issue:43299|issue:45357", "issue:43295|issue:44779", - "issue:44521|issue:45325", + "issue:43299|issue:45357", "issue:44351|issue:45278", + "issue:44521|issue:45325", "issue:41950|issue:45020", "issue:39692|issue:43901", "issue:41950|issue:45823", @@ -11170,204 +11141,205 @@ "issue:44016|issue:44291", "issue:43441|issue:45834", "issue:44623|issue:45397", - "issue:44623|issue:44749" + "issue:44623|issue:44749", + "issue:43653|issue:45593" ], - "prepared_review_unit_hash": "f06dac216e089302bf4b684deaad88acd7f7eb88ffd8d137522a19f687e64d5e", + "prepared_review_unit_hash": "8599896de06695f65cebcfef329568bc7d5519bcb5ab0159c0fad2d082a82543", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35596, - "estimated_input_tokens": 8899, - "estimated_eval_tokens": 18054 + "serialized_chars": 35557, + "estimated_input_tokens": 8890, + "estimated_eval_tokens": 18036 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated issue reports across different subsystems; the soft-similarity pairs do not look like true duplicates or the same fixable change. I would reject all soft edges. If one issue must represent the set, issue 44521 is the strongest canonical candidate because it has a clear concrete repro, active discussion, and the most inbound references among the open issues.", - "confidence": 0.31, - "canonical_issue_reason": "Issue 44521 is the best representative only in a weak sense: it describes a concrete, reproducible multimodal bug, has the most external references, and is more actionable than the other candidates.", + "summary": "The items are mostly a heterogeneous grab-bag of tokenizer, processor, multimodal, config, CI, and documentation regressions. The soft pairs share broad themes, but I don\u2019t see two artifacts in any pair describing the same concrete bug/change closely enough to merge as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "No single canonical issue stands out: the cluster is too heterogeneous, and the overlapping titles reflect broad subsystem similarity rather than one underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 is the most suitable global representative because it is specific, well-scoped, and has the strongest evidence trail; the rest are mostly orthogonal regressions, docs issues, or broad dependency complaints.", + "best_issue_reason": "No global best issue for deduping; any single issue would be arbitrary because the set spans unrelated failure modes and code paths.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:46082", - "right": "issue:46399", + "left": "issue:43650", + "right": "issue:46209", "accept": false, - "reason": "Different problems: LlamaConfig validation for explicit head_dim vs a Qwen3.5 rotary-registration crash. Same broad area, not the same underlying bug." + "reason": "Completely unrelated: placeholder data report vs FP8/B200 inference failure." }, { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "Auto-docstring annotation handling is unrelated to the mistral_common import failure in Gemma 4 processor loading." - }, - { - "left": "issue:43650", - "right": "issue:46209", - "accept": false, - "reason": "Placeholder issue title versus an FP8 inference failure on B200; no shared code path or symptom." + "reason": "Different bugs: auto-docstring annotation handling vs Gemma 4 processor import failure." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "MoE load-balancing loss gating is a different bug from Gemma 4's mm_token_type_ids defaulting behavior." + "reason": "Separate training issues: router/load-balancing loss vs Gemma 4 mm_token_type_ids defaulting." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "Generic dependency trouble is not the same as bumping MIN_PEFT_VERSION to an unreleased release." + "reason": "Only vaguely dependency-related; transformers compatibility issue is not the same as an unreleased peft version bump." }, { "left": "issue:38175", "right": "issue:44617", "accept": false, - "reason": "SigLIP2 zero probabilities and Sam3Video CUDA OOM are unrelated failures." + "reason": "Unrelated model failures: SigLIP2 zero probabilities vs Sam3Video OOM." }, { "left": "issue:44779", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different code paths; too broad to merge as one bug." - }, - { - "left": "issue:43299", - "right": "issue:45357", - "accept": false, - "reason": "Model loading failure for Qwen3VL MoE is distinct from incorrect visual-encoder keys during save_pretrained." + "reason": "Both tokenizer regressions, but different model families and likely different code paths." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Processor.tokenizer API regression is not the same as Deepseek tokenization producing wrong results." + "reason": "Processor.tokenizer regression is distinct from Deepseek tokenization output changes." }, { - "left": "issue:44521", - "right": "issue:45325", + "left": "issue:43299", + "right": "issue:45357", "accept": false, - "reason": "All-zero assistant masks in multimodal chat templates is unrelated to Qwen2.5-VL rope-index scaling." + "reason": "Both Qwen multimodal issues, but loading failure and save_pretrained key serialization are different bugs." }, { "left": "issue:44351", "right": "issue:45278", "accept": false, - "reason": "A specific missing HybridCache import is not the same as a broad set of import errors after upgrade." + "reason": "Import-related, but one missing symbol and the other a broad import-error report are not the same concrete defect." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Different multimodal logic: assistant_masks bug vs rope_index temporal position scaling." }, { "left": "issue:41950", "right": "issue:45020", "accept": false, - "reason": "Video pipeline picking image processors is unrelated to remote_code breakages in recent transformers versions." + "reason": "Video-classification processor lookup bug is unrelated to remote_code breakage." }, { "left": "issue:39692", "right": "issue:43901", "accept": false, - "reason": "SigLIP2 example mistakes and TextClassificationPipeline docs drift are separate documentation problems." + "reason": "Both docs issues, but they concern different examples and behaviors." }, { "left": "issue:41950", "right": "issue:45823", "accept": false, - "reason": "Image-processor lookup in video classification and Gemma4 PLE device mismatch are different bugs." + "reason": "No shared bug: image-processor lookup in video pipeline vs Gemma4 device placement mismatch." }, { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 and tensorboard/tf backend installation are unrelated." + "reason": "Different subsystems entirely: flex_attention compilation vs TensorFlow/tensorboard backend detection." }, { "left": "issue:43976", "right": "issue:45405", "accept": false, - "reason": "Python version support regression is not the same as a PEFT minimum-version bump." + "reason": "Unrelated versioning problems: Python support mismatch vs unreleased peft dependency." }, { "left": "issue:44462", "right": "issue:45081", "accept": false, - "reason": "AutoTokenizer ignoring repository tokenizer.json and a Mistral regex-patch crash are different loader bugs." + "reason": "Both tokenizer loading bugs, but one ignores tokenizer.json and the other crashes in Mistral regex patching." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Saving the wrong tokenizer_class and ignoring tokenizer.json on load are related to tokenizers, but they are distinct defects." + "reason": "Serialization/save_pretrained mismatch vs AutoTokenizer loading preference are different directions of the pipeline." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Qwen3VL loading regression and GLM-4.7 Flash unit-test loading issue are separate model-specific failures." + "reason": "Different model-loading failures for different models and triggers." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Phi4MultimodalConfig defaulting subconfigs and Llama4Config missing pad_token_id are not the same bug." + "reason": "Distinct config classes and fields: multimodal defaults vs missing pad_token_id." }, { "left": "issue:43450", "right": "issue:44279", "accept": false, - "reason": "Video processor batching shape problems are unrelated to a generic dependency issue." + "reason": "Video processor shape bug is unrelated to a generic dependency issue." }, { "left": "issue:45081", "right": "issue:45907", "accept": false, - "reason": "Tokenizer regex patch crash and list_repo_files revision forwarding concern completely different code paths." + "reason": "Tokenizer regex crash vs API forwarding regression in list_repo_files are unrelated." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "Serialization inconsistency for lm_head.weight is separate from resize_token_embeddings not propagating across Gemma4 embedding/output layers." + "reason": "Both touch model weights, but serialization inconsistency and resize_token_embeddings behavior are different code paths." }, { "left": "issue:43650", "right": "issue:46132", "accept": false, - "reason": "No concrete shared bug: one is an empty placeholder, the other is a behavior change in AttentionInterface.register." + "reason": "Placeholder issue vs attention registration behavior change are unrelated." }, { "left": "issue:44016", "right": "issue:44291", "accept": false, - "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated." + "reason": "Notebook syntax error is unrelated to init_empty_weights TypeError." }, { "left": "issue:43441", "right": "issue:45834", "accept": false, - "reason": "FlashAttention OOM/crash for Ministral and OCR index error in Kosmos2.5 are distinct issues." + "reason": "Different model-specific failures: FlashAttention on Ministral vs long OCR index error in Kosmos2.5." }, { "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "processor.save_pretrained missing files and Gemma-4 zero3 from_pretrained failure are different load/save problems." + "reason": "Both mention save/from_pretrained, but processor file omission and Gemma-4 zero3 loading are different defects." }, { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "Missing processor files and a post-upgrade performance regression are not the same underlying defect." + "reason": "Processor save_pretrained bug is unrelated to the reported performance regression." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer decode bug is unrelated to D-FINE auxiliary-loss behavior." } ] }, @@ -11400,7 +11372,6 @@ "issue:43441", "issue:43525", "issue:43643", - "issue:43653", "issue:43673", "issue:43723", "issue:43761", @@ -11422,6 +11393,7 @@ "issue:44336", "issue:44361", "issue:44373", + "issue:44496", "issue:44521", "issue:44589", "issue:44671", @@ -11441,7 +11413,6 @@ "issue:46399" ], "soft_pairs": [ - "issue:43653|issue:45593", "issue:42994|issue:43317", "issue:45600|issue:45712", "issue:33290|issue:45656", @@ -11464,208 +11435,209 @@ "issue:41628|issue:43976", "issue:44188|issue:44336", "issue:43784|issue:45542", - "issue:44336|issue:44373", "issue:44295|issue:45397", + "issue:44336|issue:44373", "issue:45103|issue:46399", "issue:45561|issue:45563", - "issue:44242|issue:45593" + "issue:44242|issue:45593", + "issue:44246|issue:44496" ], - "prepared_review_unit_hash": "964d4f66fef47928b4568a074f48dca80c736ce221e427ec4af75c391e2a0c6a", + "prepared_review_unit_hash": "5dee2111a29cde0f93b2633c8a58a05f69a5463b797935521145995176e7999a", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 35865, - "estimated_input_tokens": 8967, - "estimated_eval_tokens": 18190 + "serialized_chars": 35889, + "estimated_input_tokens": 8973, + "estimated_eval_tokens": 18202 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items span many unrelated regressions and docs/packaging issues across tokenizers, model loading, quantization, chat templates, and attention kernels. None of the soft pairs look like the same underlying bug/change, so there is no clear canonical duplicate anchor here.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "No soft pairs look like true duplicates; they mostly share broad subsystems or regression patterns but target different code paths, models, or packaging/docs issues.", + "confidence": 0.93, + "canonical_issue_reason": "issue:44521 is the strongest representative issue in this set: it is open, has the most visible user impact and corroboration, and describes a concrete multimodal `apply_chat_template` regression.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:44521 is the best single issue to keep as the cluster representative because it is current, well-discussed, and reproducible; the rest are clearly separate bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43653", - "right": "issue:45593", - "accept": false, - "reason": "BigBirdTokenizer special-token decode bug and D-FINE auxiliary-loss behavior are unrelated subsystems and failure modes." - }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Quantized model saving failure and dequantized model offload loading failure are different code paths, not one concrete bug." + "reason": "Both involve quantization/offload, but one is about saving quantized models and the other about loading a dequantized model with `device_map=auto`; different failure modes." }, { "left": "issue:45600", "right": "issue:45712", "accept": false, - "reason": "Both touch repo hygiene/CI, but one is stale auto_mappings references and the other is leftover dummy classes; not the same issue." + "reason": "Both are repo-maintenance/CI problems, but one is stale `auto_mappings.py` references and the other is leftover dummy classes leaking into `dir(transformers)`." }, { "left": "issue:33290", "right": "issue:45656", "accept": false, - "reason": "Both mention DeepSpeed/optimizers, but one is OOM with Adafactor and the other is double optimizer steps; different bugs." + "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:44261", "right": "issue:44263", "accept": false, - "reason": "MLA q_a_layernorm config precision bug and GlmMoeDsaIndexer torch.split return-value issue are unrelated." + "reason": "Completely different code paths: missing `rms_norm_eps` in MLA q_a_layernorm versus `torch.split()` return handling in GlmMoeDsaIndexer." }, { "left": "issue:44361", "right": "issue:45593", "accept": false, - "reason": "Tokenizer AttributeError on tasks and D-FINE loss omission are unrelated model families and failures." + "reason": "Tokenizer AttributeError on tasks is unrelated to D-FINE auxiliary losses when denoising is off." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "Both are v5 regressions, but CamemBERT masked-LM output and Deepseek tokenizer output are different code paths and fixes." + "reason": "Both are v5 regressions, but one is CamemBERT masked-LM predictions and the other is Deepseek tokenizer behavior; different components." }, { "left": "issue:42175", "right": "issue:43976", "accept": false, - "reason": "Package backend dependency omission and Python-version compatibility failure are different packaging problems." + "reason": "Packaging/install backend issue versus Python-version compatibility issue; no shared underlying bug." }, { "left": "issue:44016", "right": "issue:45823", "accept": false, - "reason": "Notebook syntax error and Gemma4 device-mismatch forward bug are unrelated." + "reason": "A notebook syntax error and a Gemma4 device-mismatch forward bug are unrelated." }, { "left": "issue:43901", "right": "issue:43994", "accept": false, - "reason": "Docs mismatch for TextClassificationPipeline return_all_scores is not the same as SigLIP2 producing bad predictions." + "reason": "Docs mentioning `return_all_scores` is not the same as bad `AutoModel`/pipeline outputs for SigLIP2." }, { "left": "issue:43441", "right": "issue:44683", "accept": false, - "reason": "FlashAttention failure for Ministral-3 and compiled flex_attention failure on torch>=2.9 involve different kernels, versions, and fixes." + "reason": "Both are attention-kernel regressions, but they affect different kernels, versions, and code paths; too broad to merge." }, { "left": "issue:43723", "right": "issue:44162", "accept": false, - "reason": "AutoTokenizer loading regression and ESM2 model breakage are different components, not one underlying defect." + "reason": "Tokenizer loading failure and ESM2 runtime breakage are different issues." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Generation cache/chunked_prefill issue and CLIPVisionModel hidden_states regression are separate model-path bugs." + "reason": "Generation cache/chunked-prefill bug versus CLIPVision hidden-states regression; distinct modules." }, { "left": "issue:44297", "right": "issue:45397", "accept": false, - "reason": "Tokenizer save metadata mismatch and Gemma4 zero3 from_pretrained failure are unrelated loading/saving paths." + "reason": "Tokenizer save metadata mismatch and Gemma4 zero3 `from_pretrained` crash are unrelated." }, { "left": "issue:43262", "right": "issue:44521", "accept": false, - "reason": "Both mention apply_chat_template, but one is audio sampling-rate defaulting and the other is multimodal assistant_masks; not the same bug." + "reason": "Both touch `apply_chat_template` in multimodal contexts, but one is audio sampling-rate defaulting and the other is assistant-mask generation; different bugs." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "Missing model export/import and Python version incompatibility are unrelated." + "reason": "Missing model import and Python-version support are unrelated." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Llama4Config missing pad_token_id and glm-4v-9b loading failure are different model-specific issues." + "reason": "Llama4 missing `pad_token_id` and GLM-4V loading failure are different model-specific loading problems." }, { "left": "issue:43901", "right": "issue:44246", "accept": false, - "reason": "Pipeline docs mismatch and slow import performance are unrelated." + "reason": "A documentation mismatch and intermittent import slowness are not the same underlying issue." }, { "left": "issue:43525", "right": "issue:43643", "accept": false, - "reason": "Both concern config fields, but one is a missing model attribute and the other is trust_remote_code losing fields; not the same underlying bug." + "reason": "Both involve config fields, but one is a specific model config missing `pad_token_id` while the other is `trust_remote_code` dropping fields on `AutoConfig`; different causes." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Float8 storage deserialization error and Qwen3.5 chat crash are unrelated failures." + "reason": "Float8 storage support error versus Qwen3.5 chat crash; unrelated." }, { "left": "issue:41628", "right": "issue:43976", "accept": false, - "reason": "AutoImageProcessor import failure and Python compatibility issue are unrelated." + "reason": "`AutoImageProcessor` import failure is unrelated to the Python-version compatibility bug." }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "Attention-kernel divergence under torch.compile and ANSI-code emission in loading_report are unrelated." + "reason": "Torch.compile attention-kernel divergence and stray ANSI codes in loading_report are unrelated." }, { "left": "issue:43784", "right": "issue:45542", "accept": false, - "reason": "Sentence-transformers import NameError and tf-backend undefined error from extras installation are different dependency/import problems." + "reason": "Both are import-time dependency issues, but they come from different missing backends and different code paths; not the same bug." }, { - "left": "issue:44336", - "right": "issue:44373", + "left": "issue:44295", + "right": "issue:45397", "accept": false, - "reason": "ANSI code generation and wrong docstring for position_ids are not the same bug." + "reason": "Buffer read error for `position_ids` and Gemma4 zero3 loading failure are different problems." }, { - "left": "issue:44295", - "right": "issue:45397", + "left": "issue:44297", + "right": "issue:44373", "accept": false, - "reason": "position_ids buffer read error and Gemma4 zero3 loading failure are unrelated." + "reason": "Tokenizer save config mismatch and a wrong docstring are clearly unrelated." }, { "left": "issue:45103", "right": "issue:46399", "accept": false, - "reason": "auto_docstring crashes with future annotations and Qwen3.5 apply_rotary_pos_emb registration failure are unrelated code paths." + "reason": "Auto-docstring failure with future annotations and missing decorator registration for rotary embeddings are different issues." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "pytest-xdist file race and paged generate stale warning are unrelated testing vs generation issues." + "reason": "Pytest-xdist race in testing utils and a stale generate warning are unrelated." }, { "left": "issue:44242", "right": "issue:45593", "accept": false, - "reason": "Missing MoE load-balancing loss and D-FINE auxiliary-loss omission are both loss-accounting bugs, but in different models and conditions." + "reason": "Both are losses gated by flags, but they affect different models and different loss terms; not the same code-path bug." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import slowness and unrecognized-model config errors are unrelated." } ] }, @@ -11738,7 +11710,6 @@ "issue:46424" ], "soft_pairs": [ - "issue:44246|issue:44496", "issue:42491|issue:43299", "issue:43531|issue:45362", "issue:43526|issue:45003", @@ -11770,64 +11741,58 @@ "issue:45292|issue:45701", "issue:44561|issue:46399" ], - "prepared_review_unit_hash": "9062e9398c00e283e41c408169c21cb5b22a1a0f731b39ac8dec78923c3a0900", + "prepared_review_unit_hash": "7aa61f99bdb17fe0a6aff436ffb66c96174f113dadde1735fddf6bc0ce8d357f", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 36322, - "estimated_input_tokens": 9081, - "estimated_eval_tokens": 18418 + "soft_pair_count": 30, + "serialized_chars": 36021, + "estimated_input_tokens": 9006, + "estimated_eval_tokens": 18268 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are a grab bag of unrelated Transformers regressions and feature requests across model loading, tokenizers, pipelines, audio, caching, DeepSpeed, docs, and runtime utilities. None of the soft edges look like the same underlying bug or change, so this should not be collapsed into one duplicate cluster.", - "confidence": 0.98, - "canonical_issue_reason": null, + "summary": "This cluster is heterogeneous and mostly contains separate model-specific regressions and import/API breakages, with only weak topical overlap around v5 compatibility. I found no soft pair that is clearly the same underlying bug/change.", + "confidence": 0.62, + "canonical_issue_reason": "Issue 44561 is the broadest and most representative regression: a v5 API removal breaking trust_remote_code model loading. It best captures the cluster\u2019s repeated theme of version-upgrade breakages.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44561 is the broadest regression-style report and has inbound references, but the cluster is too heterogeneous for it to serve as a true canonical duplicate target.", + "best_issue_reason": "44561 is the most general, actionable, and cluster-like issue among the set. It describes a concrete upstream compatibility regression rather than a niche model symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44246", - "right": "issue:44496", - "accept": false, - "reason": "Different failures: import slowdown vs unrecognized model/config handling." - }, { "left": "issue:42491", "right": "issue:43299", "accept": false, - "reason": "Both involve Qwen3-MoE family, but one is LoRA loading on hf4.x and the other is a v5 model-loading regression for Qwen3VL MoE." + "reason": "Both mention Qwen3 MoE/v5 loading regressions, but one is LoRA compatibility and the other is Qwen3VL model loading; not the same concrete bug." }, { "left": "issue:43531", "right": "issue:45362", "accept": false, - "reason": "Different code paths: sliding-window behavior in Qwen3-MoE vs chat crash on Qwen3.5-35B." + "reason": "Different failures in different flows: sliding-window handling vs chat crash." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug is unrelated to unsafe sys.modules access in modeling_utils." + "reason": "Beit image processor label reduction vs unsafe sys.modules access are unrelated code paths." }, { "left": "issue:41950", "right": "issue:44220", "accept": false, - "reason": "Video-classification pipeline processor selection and audio fbank extraction are separate subsystems." + "reason": "Video-classification pipeline processor lookup and audio fbank feature extraction are different subsystems." }, { "left": "issue:42886", @@ -11836,154 +11801,154 @@ "reason": "Offline tokenizer cache loading and AttentionInterface.register behavior are unrelated." }, { - "left": "issue:44246", + "left": "issue:44242", "right": "issue:44871", "accept": false, - "reason": "Import performance issue vs Gemma-3 eos_token_id mismatch; no shared bug path." + "reason": "Router-logits loss handling and Gemma eos_token_id configuration are distinct bugs." }, { "left": "issue:44336", "right": "issue:44485", "accept": false, - "reason": "ANSI logging output and GLM-5 RoPE implementation are unrelated." + "reason": "ANSI output in loading_report and GLM-5 RoPE implementation are unrelated." }, { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but they affect different symbols and different packages/models." + "reason": "Both are import errors, but they are different missing symbols from different packages and likely different causes." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Different missing imports: Qwen2_5_VLImageProcessor versus HybridCache." + "reason": "Missing Qwen2_5_VLImageProcessor export and missing HybridCache export are separate import failures." }, { "left": "issue:44734", "right": "issue:46424", "accept": false, - "reason": "KV-cache continuation indexing bug is not the same as max_cache_length handling in quantized cache." + "reason": "Both involve cache behavior, but one is KV-cache continuation tensor indexing and the other is max_cache_length semantics." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "sys.modules access in modeling_utils is unrelated to pytest-xdist file races." + "reason": "Unsafe sys.modules access and a pytest-xdist file race are unrelated." }, { "left": "issue:44336", "right": "issue:46424", "accept": false, - "reason": "Terminal ANSI output bug and cache-length configuration bug are different concerns." + "reason": "Terminal ANSI output and cache-length configuration are unrelated." }, { "left": "issue:44779", "right": "issue:45292", "accept": false, - "reason": "DeepSeek tokenizer regression and resize_token_embeddings output embedding mismatch are unrelated." + "reason": "Tokenizer correctness in v5 and resize_token_embeddings behavior are different bugs." }, { "left": "issue:44295", "right": "issue:44749", "accept": false, - "reason": "Position_ids buffer access error and version-related tokenization slowdown do not match." + "reason": "A buffer/position_ids error and a performance regression are not the same issue." }, { "left": "issue:43475", "right": "issue:46369", "accept": false, - "reason": "SAM 3 attribute error and SAM3 memory overflow are different bugs." + "reason": "SAM3 vision encoder missing attribute vs SAM3 litetext memory overflow are different failures." }, { "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "Load-balancing loss omission in MoE output vs LwDetrImageLoss AMP crash are separate issues." + "reason": "Both are loss-related, but they affect different models and code paths." }, { "left": "issue:36246", "right": "issue:43596", "accept": false, - "reason": "Qwen2.5-VL import error and DeepSpeed ZeRO3 empty-deque IndexError are unrelated." + "reason": "ImportError for a model processor and a DeepSpeed/Bert index error are unrelated." }, { "left": "issue:46291", "right": "issue:46399", "accept": false, - "reason": "Kernel version/revision import failure and missing decorator for apply_rotary_pos_emb are different code paths." + "reason": "Both mention kernels, but one is a dependency version requirement and the other is a missing registration decorator; different bugs." }, { "left": "issue:45859", "right": "issue:46399", "accept": false, - "reason": "CPU offload incompatibility in Qwen3_5MoeTextRotaryEmbedding is not the same as kernel registration failure." + "reason": "CPU offload incompatibility and use_kernels registration failure are different Qwen3.5 problems." }, { "left": "issue:43673", "right": "issue:44561", "accept": false, - "reason": "Chunked-prefill cache missing in GenerationMixin and trust_remote_code breakage from removed API are unrelated." + "reason": "Chunked-prefill cache handling and removal of is_torch_fx_available are different regressions." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "Docs mismatch for return_all_scores does not match the sys.modules access bug." + "reason": "Docs mismatch for return_all_scores and unsafe sys.modules access are unrelated." }, { "left": "issue:44162", "right": "issue:45200", "accept": false, - "reason": "ESM2 breakage and Gemma 4 token-type-id defaults affect different model families and mechanisms." + "reason": "ESM2 breakage and Gemma 4 token-type-id handling are different model-specific issues." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "Nemotron-specific Flash Attention 2 incompatibility is not the same as the generic Flash Attention 2 issue." + "reason": "A model not supporting Flash Attention 2 yet is not the same as the broad Flash Attention 2 topic." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Unrecognized model/config parsing is unrelated to unsafe sys.modules access." + "reason": "Unrecognized model config and sys.modules access are unrelated." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "Qwen3 MOE weight conversion with accelerate+deepspeed is a different failure mode from the ZeRO3 deque IndexError." + "reason": "Qwen3 MOE weight conversion and DeepSpeed ZeRO3 deque/index errors are different loading failures." }, { "left": "issue:45292", "right": "issue:45584", "accept": false, - "reason": "resize_token_embeddings mismatch and Whisper empty-transcription generation failure are unrelated." + "reason": "resize_token_embeddings/output_embeddings and Whisper empty-transcription generation are unrelated." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio chat-template sampling-rate default bug does not match import slowdown." + "reason": "Audio processor sampling-rate defaulting and import latency are unrelated." }, { "left": "issue:45412", "right": "issue:46123", "accept": false, - "reason": "RT-DETR memory release and MaskGenerationPipeline partial-batch dropping are separate bugs." + "reason": "RT-DETR memory leaks and MaskGenerationPipeline final-batch dropping are different bugs." }, { - "left": "issue:45292", - "right": "issue:45701", + "left": "issue:44561", + "right": "issue:46399", "accept": false, - "reason": "Output embedding resizing bug and version-dependent tokenization changes are different issues." + "reason": "Both are v5 regressions, but one is trust_remote_code breakage from a removed helper and the other is a Qwen3.5 kernels registration crash." }, { - "left": "issue:44561", - "right": "issue:46399", + "left": "issue:45003", + "right": "issue:45561", "accept": false, - "reason": "Removed API breakage for trust_remote_code models is not the same as missing apply_rotary_pos_emb registration." + "reason": "Unsafe sys.modules access and captured_info.txt race conditions are unrelated." } ] }, @@ -12077,8 +12042,8 @@ "issue:44995|issue:45588", "issue:44368|issue:45563", "issue:44188|issue:46202", - "issue:45987|issue:46082", "issue:44877|issue:46170", + "issue:45987|issue:46082", "issue:43931|issue:44368", "issue:44188|issue:46421", "issue:43065|issue:46123", @@ -12090,226 +12055,226 @@ "issue:44869|issue:45356", "issue:43701|issue:44749" ], - "prepared_review_unit_hash": "0cd8bfbb1d7a1919267d51b2b2d3a58a42620d738bf5d98e1604e56d40129aed", + "prepared_review_unit_hash": "827441067ed47529d2259b67bce9f1f4c57295c6ac23ced32903ec23801f9f29", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37436, - "estimated_input_tokens": 9359, - "estimated_eval_tokens": 18974 + "serialized_chars": 37441, + "estimated_input_tokens": 9361, + "estimated_eval_tokens": 18978 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are mostly unrelated singleton bugs across tokenizer loading, strict config handling, multimodal model wiring, DeepSpeed, Whisper, and performance regressions. A few pairs share broad themes, but none look like the same concrete bug/change or something that could realistically be merged into one fix PR.", - "confidence": 0.92, - "canonical_issue_reason": "No true canonical duplicate stands out; if forced to pick a representative, issue 44877 is the broadest of the small strict-config subset, but the overall set is too heterogeneous to treat as one cluster.", + "summary": "This cluster is mostly heterogeneous: tokenizer/save_pretrained/config regressions, SAM3/Qwen video issues, Gemma/DeepSpeed loading problems, and a few cache/compile bugs. I do not see a safe duplicate set among the soft pairs.", + "confidence": 0.18, + "canonical_issue_reason": "No single canonical issue is safe here; the cluster spans several unrelated bug families, and the strongest similarities are only broad subsystem overlap.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44877 is the most representative of the closest mini-theme (strict config / multimodal loading), but it is still not a good global duplicate anchor for this mixed set.", + "best_issue_reason": "No global best issue is a good canonical representative for the whole cluster; at most, a few pairs form narrow subthemes, but they are not the same underlying defect.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45375", "right": "issue:45759", "accept": false, - "reason": "Both involve config handling in Qwen multimodal models, but one is a dropped field under @strict and the other is from_config not unwrapping text_config; different code paths and fixes." + "reason": "Same Qwen config area, but different concrete bugs: missing field vs text_config unwrapping." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Shared loose theme of strict config loading, but the reported failures are different models and different missing-field behavior." + "reason": "Both are config-loading problems, but they affect different models and code paths." }, { "left": "issue:45685", "right": "issue:46431", "accept": false, - "reason": "Completely different subsystems: MPS histogram on Int vs processor chat-template availability." + "reason": "Unrelated failures: MPS histogram support vs missing chat template." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "Both are loading/serialization-adjacent, but one is a model load failure and the other is tokenizer decode error masking; not the same bug." + "reason": "Both touch loading/tokenizer behavior, but the root causes are different." }, { "left": "issue:45397", "right": "issue:45656", "accept": false, - "reason": "Both mention DeepSpeed, but one is a Zero3 from_pretrained issue and the other is duplicate optimizer stepping; unrelated root causes." + "reason": "Both mention DeepSpeed-related behavior, but they are separate loading/optimizer bugs." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "Both touch tied/embedded weights, but one is device-dependent serialization and the other is resize_token_embeddings not updating decoder.embed_tokens." + "reason": "Both involve embedding tying/resize behavior, but in different models and mechanisms." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "Different models and failures: stale indexer cache on second forward pass vs Gemma4 mm_token_type_ids defaulting." + "reason": "Cache state corruption is unrelated to Gemma4 token-type defaults." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "One is a DeepSpeed ZeRO3 deque error, the other is a broken auto_mappings reference to removed Sam3LiteText configs." + "reason": "Checkpoint resume failure is unrelated to auto_mappings cleanup." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "Both are model-specific runtime errors, but they hit different SAM/Qwen video paths and different missing objects/value checks." + "reason": "Both are video/model issues, but they fail in different components and ways." }, { "left": "issue:43906", "right": "issue:46202", "accept": false, - "reason": "The first is just an isolated reproduction of another issue; the second is a Gemma tokenizer roundtrip bug. No concrete overlap shown." + "reason": "Different model reproductions with unrelated decoding/tokenization behavior." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Tokenizer cache offline loading and Gemma4 CUDA OOM are unrelated." + "reason": "Tokenizer offline cache bug and Gemma4 OOM are unrelated." }, { "left": "issue:44479", "right": "issue:46170", "accept": false, - "reason": "Video-input regression in Qwen models vs Windows/ROCm import dependency issue; different layers and fixes." + "reason": "Both are regressions, but one is video-input handling and the other is Windows/ROCm import dependency." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "BigBird tokenizer special-token registration and ESM2 breakage are separate bugs." + "reason": "Tokenizer special-token registration and ESM2 breakage are different bugs." }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "Tied-weights warning in Qwen3-Next and PatchTSMixer post_init policy are unrelated." + "reason": "Different features and failures: tied-weights warning vs optional post_init policy." }, { "left": "issue:44483", "right": "issue:46202", "accept": false, - "reason": "API request rejection in v5.3 vs Gemma pipeline roundtrip failure; no shared code path." + "reason": "Chat API request rejection is unrelated to the Gemma pipeline roundtrip issue." }, { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "One is a tied-embeddings warning during LoRA fine-tuning, the other is a flash-attention generation regression; not the same defect." + "reason": "These are separate warnings/issues in different code paths." }, { "left": "issue:43065", "right": "issue:43408", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d and sam3_video/sam3_tracker type mismatch are distinct SAM3 issues." + "reason": "Both SAM3-related, but one is a dummy layer and the other is a wrong-model-type instantiation warning." }, { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "GlmMoeDsa stale cache on second forward is unrelated to flash_attention.py crashing on sink-less models with s_aux=None." + "reason": "Cache/indexer state bug is unrelated to flash-attention s_aux handling." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "Tie_word_embeddings warning during training and stale paged-generate warning are unrelated warnings." + "reason": "Both are warning-related, but they concern different models and behaviors." }, { "left": "issue:44188", "right": "issue:46202", "accept": false, - "reason": "torch.compile branching divergence in attention kernels and Gemma tokenizer roundtrip are different problem areas." + "reason": "Compile-time attention branching and Gemma tokenizer roundtrip are unrelated." }, { - "left": "issue:45987", - "right": "issue:46082", + "left": "issue:44877", + "right": "issue:46170", "accept": false, - "reason": "StaticCache return type contract and LlamaConfig head_dim validation are separate API/validation issues." + "reason": "Config strictness and Windows ROCm import dependency are different problems." }, { - "left": "issue:44877", - "right": "issue:46170", + "left": "issue:45987", + "right": "issue:46082", "accept": false, - "reason": "Strict config loading for granite_speech and Windows/ROCm import failure are unrelated." + "reason": "StaticCache return-type bug and LlamaConfig head_dim validation are unrelated." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch is not the same as a tied-weights warning in Qwen3.5 fine-tuning." + "reason": "Model weight-shape mismatch is unrelated to the tie_word_embeddings warning." }, { "left": "issue:44188", "right": "issue:46421", "accept": false, - "reason": "Both mention compile/cache behavior, but one is about attention kernel divergence and the other about lazy cache init causing recompilation." + "reason": "Both mention compile/cache behavior, but the concrete failures differ." }, { "left": "issue:43065", "right": "issue:46123", "accept": false, - "reason": "SAM3 decoder dummy layer and MaskGenerationPipeline final-batch dropping are unrelated." + "reason": "SAM3 pixel-decoder dummy layer and mask batching logic are unrelated." }, { "left": "issue:44297", "right": "issue:44623", "accept": false, - "reason": "Both are save_pretrained-related, but one is tokenizer_config class mismatch while the other is missing processor files." + "reason": "Similar save_pretrained surface, but tokenizer_class mismatch and missing processor files are distinct defects." }, { "left": "issue:45584", "right": "issue:46170", "accept": false, - "reason": "Whisper empty-transcription failure and Windows ROCm import dependency issue are unrelated." + "reason": "Whisper empty-transcription failure is unrelated to the ROCm import issue." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "MLukeTokenizer task AttributeError and a broad Transformers slowdown regression are not the same bug." + "reason": "Tokenizer AttributeError and post-upgrade slowdown are unrelated." }, { "left": "issue:44623", "right": "issue:45656", "accept": false, - "reason": "Processor save_pretrained missing files and duplicate DeepSpeed optimizer steps are separate issues." + "reason": "Processor serialization and duplicate optimizer steps are different bugs." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "Performance regression after upgrading Transformers vs DeepSpeed optimizer-step duplication are different failures." + "reason": "Training slowdown after upgrade does not match the deepspeed step-count bug." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Whisper trailing replacement-character crash and Kimi-K2.5 tokenizer codec regression are different tokenizer/decoding bugs." + "reason": "Whisper timestamp decoding and Kimi tokenizer regression are unrelated." }, { "left": "issue:43701", "right": "issue:44749", "accept": false, - "reason": "resume_from_checkpoint key mismatch and a general slowdown regression do not share a concrete root cause." + "reason": "Checkpoint key mismatch is unrelated to the tokenizer/filtering slowdown issue." } ] }, @@ -12408,226 +12373,226 @@ "issue:41628|issue:44351", "issue:44945|issue:45753", "issue:44222|issue:45310", - "issue:46077|issue:46123", "issue:42994|issue:43824", + "issue:46077|issue:46123", "issue:44038|issue:45310", "issue:36010|issue:41628", "issue:42831|issue:44222" ], - "prepared_review_unit_hash": "cb09a8d59267a202b187e4efc36840ce4dd20bcb3d0ef679c316e25ed43f80ef", + "prepared_review_unit_hash": "d742f5fee4e7ad83b573b548343f587cdf9be1214229485a57abd5d18277cdc1", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36350, - "estimated_input_tokens": 9088, - "estimated_eval_tokens": 18432 + "serialized_chars": 36348, + "estimated_input_tokens": 9087, + "estimated_eval_tokens": 18430 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This packet has two clear duplicate clusters, not just one: the Qwen3ForCausalLM VRAM-leak pair (42673/43704) and the Qwen3.5 MoE from_pretrained regression cluster (45310/45375/45478). Most other soft edges remain false positives across different code paths or model families.", - "confidence": 0.91, - "canonical_issue_reason": "issue:42673 is the strongest global canonical issue because it is an exact duplicate of 43704 with the same title and the richest discussion trail; the packet also contains a second clear duplicate cluster around Qwen3.5 MoE (45310/45375/45478).", + "summary": "Most pairs are only loosely related by broad Transformer/model-family themes, not true duplicates. The strongest confirmed duplicate in the soft edges is the Qwen3ForCausalLM VRAM leak pair (42673/43704). The packet also contains another exact-title duplicate pair, issue:45310 and issue:45478, so the cluster is not dominated by a single unique duplicate case.", + "confidence": 0.89, + "canonical_issue_reason": "issue:42673 is the earlier report in the exact duplicate pair with issue:43704, and both describe the same Qwen3ForCausalLM VRAM leak symptom.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the best overall issue representative in this packet: it is an exact duplicate pair and has the clearest supporting context. The Qwen3.5 MoE cluster is also clearly duplicate-like, but 42673/43704 is the cleanest anchor.", + "best_issue_reason": "issue:42673 is the strongest issue representative because it is the earliest exact duplicate match for the Qwen3ForCausalLM VRAM leak and has the most canonical wording.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Different bugs: custom model init / Transformers v5 breakage versus ESM2 workflow breakage." + "reason": "Both are Transformers v5 regressions, but they affect different model paths and symptoms." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Both are generation-related, but they hit different code paths: sync_gpus kwargs update vs chunked_prefill cache lookup." + "reason": "Generation-related, but sync_gpus handling and chunked_prefill cache behavior are different bugs." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "Generic placeholder report versus a specific gemma-4 zero3 from_pretrained failure." + "reason": "Generic bug report vs Gemma4 ZeRO3 from_pretrained failure; not the same defect." }, { "left": "issue:44871", "right": "issue:46132", "accept": false, - "reason": "Unrelated issues: eos_token_id mismatch versus AttentionInterface.register behavior change." + "reason": "EOS token config mismatch and AttentionInterface.register behavior change are different issues." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Different areas and symptoms: Qwen3-VL-Moe incompatibility versus missing files in processor.save_pretrained." + "reason": "Different concrete failures: Qwen3-VL-Moe v5 breakage vs processor.save_pretrained missing files." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Saving a quantized model and loading Qwen2.5-GGUF are different operations and different failure modes." + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are distinct code paths." }, { "left": "issue:43475", "right": "issue:45381", "accept": false, - "reason": "Different models and errors: SAM 3 video encoder attribute missing versus qwen2.5-vl vision position ids." + "reason": "Different model families and failure modes: SAM 3 output attribute error vs qwen2.5-vl video position IDs." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "SAM 3 video attribute failure is unrelated to trust_remote_code returning missing config fields." + "reason": "SAM 3 Video attribute error and AutoConfig missing fields under trust_remote_code are unrelated." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "PyTorch-not-found ImportError versus a sentence-transformers loading hang are not the same bug." + "reason": "Missing PyTorch ImportError and Sentence Transformers getting stuck are unrelated." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 API warning and InternVL2 meta-initialization incompatibility are unrelated." + "reason": "TF32 API settings and remote-code meta-init incompatibility are different problems." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols and likely different fixes." + "reason": "Both are import errors, but for different symbols and different export paths." }, { "left": "issue:45375", "right": "issue:45478", - "accept": true, - "reason": "Same underlying Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0; the titles differ slightly, but the failure family matches." + "accept": false, + "reason": "Same model family, but missing config field vs from_pretrained error is not enough to treat as one bug." }, { "left": "issue:45406", "right": "issue:46202", "accept": false, - "reason": "Different Gemma-related failures: Gemma4 processor missing _tokenizer versus pipeline roundtrip on \u2581." + "reason": "Gemma4Processor missing _tokenizer and Gemma pipeline roundtrip failure are different code paths." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized model saving failure is unrelated to Llama4Config missing pad_token_id." + "reason": "Quantized model saving failure and Llama4Config missing pad_token_id are different bugs." }, { "left": "issue:43723", "right": "issue:46326", "accept": false, - "reason": "Tokenizer loading in v5 and continue_final_message for mistral_common are different problems." + "reason": "Tokenizer loading in v5 and continue_final_message behavior are distinct issues." }, { "left": "issue:45310", "right": "issue:45375", - "accept": true, - "reason": "Same Qwen3.5 MoE from_pretrained failure family; 45375 appears to be the more specific symptom report." + "accept": false, + "reason": "Related Qwen3.5 issues, but one is a from_pretrained error and the other is a missing config field." }, { "left": "issue:43784", "right": "issue:46399", "accept": false, - "reason": "Sentence-transformers import NameError and Qwen3.5 kernel registration failure are unrelated." + "reason": "Sentence-transformers import NameError and apply_rotary_pos_emb registration failure are unrelated." }, { "left": "issue:42673", "right": "issue:43704", "accept": true, - "reason": "Exact duplicate: same title, same Qwen3ForCausalLM VRAM leak when using multiple dataloader threads." + "reason": "Exact same title and same VRAM leak symptom; clearly the same underlying bug." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Mistral tokenizer regex patch crash and auto_docstring annotations crash are different code paths." + "reason": "Different functions and different contexts: mistral regex patch vs auto-docstring annotation handling." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization/offloading behavior and pipeline-parallel output correctness are distinct issues." + "reason": "Offloading with quantization and incorrect pipeline-parallel output are distinct failures." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation kwargs update after sync_gpus is unrelated to attention-kernel divergence under torch.compile." + "reason": "Different generation/attention paths with no concrete evidence of the same defect." }, { "left": "issue:44964", "right": "issue:45137", "accept": false, - "reason": "Phi-4 multimodal loading failure and DeepSpeed ZeRO3 deque underflow are unrelated." + "reason": "Phi-4 multimodal load failure and DeepSpeed ZeRO3 deque error are unrelated." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "Different missing imports: AutoImageProcessor versus HybridCache." + "reason": "Both are import/export regressions, but they concern different missing names and APIs." }, { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Pipeline-parallel incorrect output is not the same as a Qwen3_5 infinite loop on a specific image." + "reason": "Incorrect pipeline-parallel output and Qwen3_5 infinite loop on one image are different bugs." }, { "left": "issue:44222", "right": "issue:45310", "accept": false, - "reason": "FP8 save_pretrained for MoE versus Qwen3.5 MoE from_pretrained error are different operations and symptoms." + "reason": "FP8 save_pretrained and Qwen3.5 MoE from_pretrained are opposite directions and not the same issue." }, { - "left": "issue:46077", - "right": "issue:46123", + "left": "issue:42994", + "right": "issue:43824", "accept": false, - "reason": "Gemma4 attention-mask bug and MaskGenerationPipeline partial-batch dropping are unrelated." + "reason": "Quantized model saving and Qwen2.5-VL import failure are unrelated." }, { - "left": "issue:42994", - "right": "issue:43824", + "left": "issue:46077", + "right": "issue:46123", "accept": false, - "reason": "Quantized model saving failure and Qwen2.5-VL import failure are different bugs." + "reason": "Gemma4 attention-mask construction and final-partial-batch handling in MaskGenerationPipeline are different components." }, { "left": "issue:44038", "right": "issue:45310", "accept": false, - "reason": "Qwen3-VL-Moe breakage and Qwen3.5 MoE from_pretrained error are different model families and failures." + "reason": "Both involve Qwen and Transformers v5, but the concrete failures differ." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but they involve different symbols and different root causes." + "reason": "GenerationMixin import failure and AutoImageProcessor import failure are separate missing-symbol regressions." }, { "left": "issue:42831", "right": "issue:44222", "accept": false, - "reason": "FineGrainedFP8 accuracy issues are not the same as FP8 save_pretrained failures." + "reason": "FineGrainedFP8 accuracy issues and FP8 save_pretrained are not the same bug." } ] }, "evaluator_result": { "accept": false, - "feedback": "The duplicate summary is not fully grounded: besides issue:42673/43704, the packet also contains an exact-title Qwen3.5 MoE cluster (issue:45310/45375/45478) that the summary downgrades to only a \u201cplausible near-duplicate.\u201d The soft-edge calls are mostly conservative, but the narrative overstates heterogeneity by implying only one clear duplicate pair." + "feedback": "Reject: the summary overstates uniqueness. The packet contains another obvious duplicate pair with the exact same title, issue:45310 and issue:45478, so it is not grounded to say the Qwen3ForCausalLM VRAM leak is the only clear duplicate. The 42673/43704 canonical choice is plausible, but the overall summary should be more conservative." }, "evaluator_used": true, "retried": true, @@ -12702,8 +12667,8 @@ "issue:42175|issue:45712", "issue:44496|issue:45561", "issue:43504|issue:45561", - "issue:43526|issue:45561", "issue:42175|issue:44933", + "issue:43526|issue:45561", "issue:43976|issue:44933", "issue:42898|issue:44568", "issue:43377|issue:44242", @@ -12719,9 +12684,9 @@ "issue:42898|issue:44779", "issue:43784|issue:43824", "issue:44016|issue:46231", - "issue:44734|issue:45103", "issue:42898|issue:45356", "issue:43874|issue:45092", + "issue:44734|issue:45103", "issue:44295|issue:45656", "issue:44075|issue:46231", "issue:42898|issue:44448", @@ -12729,235 +12694,228 @@ "issue:44060|issue:45125", "issue:45405|issue:45712", "issue:44297|issue:45656", - "issue:44933|issue:45712", - "issue:44222|issue:44623" + "issue:44933|issue:45712" ], - "prepared_review_unit_hash": "5b4450f4418144bda2cd3114808ab09759691b54303d68c49dbd5eea41d8170d", + "prepared_review_unit_hash": "6708dd8dc5db9cd34b315ba62e88025c5b6e4257c0281fd1f607c2ffab23d7f8", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 33, - "serialized_chars": 37802, - "estimated_input_tokens": 9451, - "estimated_eval_tokens": 19158 + "soft_pair_count": 32, + "serialized_chars": 37503, + "estimated_input_tokens": 9376, + "estimated_eval_tokens": 19008 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a loose, heterogeneous cluster of mostly unrelated Transformers bug reports. A few are superficially similar v5 tokenizer regressions, but none of the soft-edge pairs look like the same exact underlying bug or fix.", - "confidence": 0.87, - "canonical_issue_reason": "Issue 42898 is the closest umbrella candidate because it captures the broad v5 tokenizer behavior-regression theme better than the more narrow model-specific reports, though the cluster is not a true duplicate set.", + "summary": "Mostly unrelated Transformers issues grouped by lexical similarity; I found no true duplicate pairs and no PRs in this cluster.", + "confidence": 0.93, + "canonical_issue_reason": "issue:42898 is the clearest representative of the repeated tokenizer/v5 regression theme, but the cluster is otherwise heterogeneous.", "canonical_pr_reason": null, - "best_issue_reason": "42898 is the most representative issue to anchor this cluster: it is broad, regression-oriented, and sits near several other tokenizer/v5 reports, even if they are not actual duplicates.", + "best_issue_reason": "issue:42898 has the most broadly framed, user-visible bug report among the mixed issues, so it is the best fallback representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Different failure modes: generation kwargs handling after sync_gpus vs a Janus model test failure." + "reason": "Different bugs: generation kwargs sync handling vs a Janus image-generation test failure." }, { "left": "issue:42175", "right": "issue:45712", "accept": false, - "reason": "Unrelated: packaging/backend dependency selection vs leftover dummy classes in a repo hygiene check." + "reason": "Unrelated: missing TensorFlow backend in a torch extra vs leftover dummy class cleanup." }, { "left": "issue:44496", "right": "issue:45561", "accept": false, - "reason": "Different code paths: model config loading error vs xdist race on captured_info.txt." + "reason": "Unrelated: model config loading error vs a pytest-xdist race in test utilities." }, { "left": "issue:43504", "right": "issue:45561", "accept": false, - "reason": "BEiT pretrained preset loading bug is unrelated to the testing race condition." + "reason": "Different areas and failure modes: BEiT preset loading vs captured_info.txt race." }, { - "left": "issue:43526", - "right": "issue:45561", + "left": "issue:42175", + "right": "issue:44933", "accept": false, - "reason": "BEiTImageProcessorFast label reduction bug is unrelated to the xdist file race." + "reason": "No shared bug: backend dependency packaging vs a nonexistent import from image_utils." }, { - "left": "issue:42175", - "right": "issue:44933", + "left": "issue:43526", + "right": "issue:45561", "accept": false, - "reason": "Backend extra installation issue and missing import from image_utils are different problems." + "reason": "Different defects: BEiT image processor label reduction vs testing race condition." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "Python version compatibility regression is unrelated to the missing import issue." + "reason": "Different problems: Python version compatibility vs import error in image_utils." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer-related v5 regressions, but they affect different behaviors and likely different code paths." + "reason": "Both are tokenizer regressions, but they affect different behaviors and likely different code paths." }, { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "Missing padding-mask support in MIMI encoder is unrelated to MoE load-balancing loss handling." + "reason": "Different bugs: MIMI padding-mask inconsistency vs MoE load-balancing loss gating." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "Embedding dtype drift in eval is unrelated to the DeepSpeed ZeRO3 deque IndexError." + "reason": "Different failures: embedding dtype drift vs DeepSpeed ZeRO3 deque underflow." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "Loading hang in sentence-transformers is unrelated to post_init reinitializing resized embeddings." + "reason": "Unrelated: Sentence Transformers load hang vs token-embedding resize/post_init reinit bug." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Processor save_pretrained missing files is unrelated to the Qwen3.5 MoE from_pretrained error." + "reason": "Different code paths: processor save_pretrained file omission vs Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLflow reporting on failed jobs is unrelated to DeepSpeed Adafactor OOM." + "reason": "Different operational failures: MLflow reporting on failed training jobs vs OOM with Adafactor in DeepSpeed." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Phi4Multimodal config defaults and Whisper trailing-replacement decode crash are different bugs." + "reason": "Unrelated model/config init bug vs Whisper word-timestamp decode crash." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "Vitpose model breakage is unrelated to Gemma 4 mm_token_type_ids defaults." + "reason": "Different multimodal model issues: broken VitPose models vs Gemma 4 mm_token_type_ids defaulting." }, { "left": "issue:42898", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions in v5, but they involve different model families and different failure mechanisms." + "reason": "Both are tokenizer regressions, but they concern different models and distinct symptoms." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Config initialization bug is unrelated to ModelOutput key assignment behavior." + "reason": "Different artifacts: multimodal config initialization vs ModelOutput key assignment behavior." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "AutoTokenizer loading failure and add_special_tokens BOS/EOS regression are distinct issues." + "reason": "Related only by tokenizer vocabulary of v5; one is AutoTokenizer loading, the other add_special_tokens output behavior." }, { "left": "issue:42898", "right": "issue:44779", "accept": false, - "reason": "Different tokenizer regressions affecting different models and likely different implementation paths." + "reason": "Both are v5 tokenizer regressions, but the affected models and failure mechanisms differ." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Import-time NameError in sentence-transformers is unrelated to missing Qwen2.5-VL export." + "reason": "Different import errors in different packages/models: sentence-transformers import vs Qwen2.5-VL symbol export." }, { "left": "issue:44016", "right": "issue:46231", "accept": false, - "reason": "Notebook syntax error and documentation parameter mismatch are unrelated." - }, - { - "left": "issue:44734", - "right": "issue:45103", - "accept": false, - "reason": "KV-cache continuation indexing crash is unrelated to auto-docstring annotation handling." + "reason": "Different scopes: notebook syntax error vs documentation parameter-name mismatch." }, { "left": "issue:42898", "right": "issue:45356", "accept": false, - "reason": "Both mention tokenizer regressions, but the Kimi codec/warning issue is a different concrete bug." + "reason": "Both mention tokenizer regressions, but Kimi-K2.5 codec handling is a separate bug from clean_up_tokenization_spaces." }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Missing image-processor method and remote-code meta-init incompatibility are unrelated." + "reason": "Different multimodal loading failures: missing image-processor method vs legacy remote-code/meta-init incompatibility." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Different failures: serve KV-cache continuation indexing vs auto_docstring parsing under future annotations." }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "Position_ids buffer read error is unrelated to optimizer step being called twice under deepspeed." + "reason": "Different bugs: buffer registration/reading issue vs optimizer step being called twice with DeepSpeed." }, { "left": "issue:44075", "right": "issue:46231", "accept": false, - "reason": "Optimizer argument handling bug is unrelated to a documentation name mismatch." + "reason": "Different areas: optimizer args ignored vs documentation parameter mismatch." }, { "left": "issue:42898", "right": "issue:44448", "accept": false, - "reason": "Both are v5 behavior regressions, but they concern different models and concrete tokenizer/output paths." + "reason": "Different tokenizer/output regressions: Pegasus v4-v5 output mismatch vs clean_up_tokenization_spaces change." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "Save_pretrained on FP8 MoE and tokenizer_class mismatch are both save-related, but not the same underlying bug." + "reason": "Unrelated: FP8 MoE save_pretrained vs tokenizer_class mismatch in tokenizer_config.json." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Incorrect tied-weights warning is unrelated to missing _tp_plan for tensor parallelism." + "reason": "Different Qwen3-related issues: tied-weights warning bug vs missing tensor-parallel plan." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "PEFT version pinning issue is unrelated to leftover dummy classes." + "reason": "Unrelated: PEFT version pin bump vs dummy class cleanup." }, { "left": "issue:44297", "right": "issue:45656", "accept": false, - "reason": "Tokenizer config mismatch is unrelated to DeepSpeed optimizer stepping twice." + "reason": "Different bugs: tokenizer save metadata vs DeepSpeed optimizer step duplication." }, { "left": "issue:44933", "right": "issue:45712", "accept": false, - "reason": "Missing image_utils import is unrelated to dummy class cleanup." - }, - { - "left": "issue:44222", - "right": "issue:44623", - "accept": false, - "reason": "Both involve save_pretrained, but one is FP8 MoE model saving and the other is processor file emission; different code paths." + "reason": "Unrelated import/export issue vs leftover dummy classes in the repo." } ] }, @@ -13004,7 +12962,6 @@ "issue:44279", "issue:44295", "issue:44297", - "issue:44360", "issue:44373", "issue:44448", "issue:44484", @@ -13032,12 +12989,13 @@ ], "soft_pairs": [ "issue:27957|issue:30333", + "issue:44222|issue:44623", "issue:30333|issue:34689", "issue:44448|issue:46132", "issue:42831|issue:44623", "issue:43479|issue:43644", - "issue:46209|issue:46431", "issue:30333|issue:36331", + "issue:46209|issue:46431", "issue:30333|issue:35141", "issue:22355|issue:28282", "issue:45245|issue:46132", @@ -13050,226 +13008,225 @@ "issue:45381|issue:46305", "issue:44206|issue:45103", "issue:43232|issue:46424", - "issue:44295|issue:44623", "issue:44734|issue:46421", + "issue:44295|issue:44623", "issue:44222|issue:44484", "issue:44263|issue:45092", "issue:43232|issue:45834", - "issue:44373|issue:44485", "issue:44190|issue:46082", + "issue:44373|issue:44485", "issue:43723|issue:44871", "issue:43723|issue:44991", - "issue:44279|issue:46328", - "issue:44263|issue:44360" + "issue:44279|issue:46328" ], - "prepared_review_unit_hash": "56da1e2ca1bbe875b83fc2ce9b3f73c4ba3c7140efa4ed74c7ac7b02ea4a87c4", + "prepared_review_unit_hash": "a24255677f9903317e6b86c1e803bbfc444ff7c24d090402dcf2e2e8b1b15478", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36413, - "estimated_input_tokens": 9104, - "estimated_eval_tokens": 18464 + "serialized_chars": 35873, + "estimated_input_tokens": 8969, + "estimated_eval_tokens": 18194 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous. The only mildly coherent sub-area is tokenizer-loading regressions around v5, but most pairs are clearly different bugs, so I would not merge any of the soft edges.", - "confidence": 0.9, - "canonical_issue_reason": "No single canonical issue fits the whole cluster; issue 43723 is the closest representative of the small tokenizer-loading v5 subtheme, but the rest of the set spans unrelated failures.", + "summary": "This set is not a true duplicate cluster: the soft pairs share only shallow wording (imports, save_pretrained, cache, tokenizer, etc.) but describe different bugs or features. I would reject all soft edges and avoid picking a canonical issue from this batch.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical issue fits this set; the items span unrelated regressions, feature requests, and dependency/import problems, and the soft links are only superficial lexical matches.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43723 is the best representative because it is the broadest and most central of the tokenizer-loading regression reports, though it still should not absorb the unrelated issues here.", + "best_issue_reason": "There is no strong global best issue representative because there is no shared underlying bug/change across the listed issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27957", "right": "issue:30333", "accept": false, - "reason": "Different failures: Flash Attention 2 model support vs MLflow reporting after a training-job crash." + "reason": "Different problems: Flash Attention 2/model support vs MLFlow not reporting failed jobs." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "Both mention save_pretrained, but one is FP8/MoE saving behavior and the other is processor file persistence." }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "Unrelated symptoms: MLflow state reporting vs model loading regression for Llama 3.2 vision." + "reason": "Unrelated failures: MLFlow job state reporting vs Llama 3.2 vision model loading regression." }, { "left": "issue:44448", "right": "issue:46132", "accept": false, - "reason": "Both mention behavior changes, but one is Pegasus output drift across versions and the other is AttentionInterface registration semantics." + "reason": "Different code paths: Pegasus v4/v5 output change vs AttentionInterface.register changing callback behavior." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "Different subsystems: FineGrainedFP8 accuracy vs processor.save_pretrained missing files." + "reason": "FP8 accuracy regression is unrelated to processor.save_pretrained missing files." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Config initialization bug for multimodal defaults vs non-persistent buffers being filled with junk." + "reason": "Config default initialization bug vs non-persistent buffer junk during load; different mechanisms." }, { - "left": "issue:46209", - "right": "issue:46431", + "left": "issue:30333", + "right": "issue:36331", "accept": false, - "reason": "FP8 inference failure on B200 is unrelated to chat-template availability in processors." + "reason": "Both are trainer-related, but one is MLFlow reporting after failure and the other is a signature mismatch in compute_loss." }, { - "left": "issue:30333", - "right": "issue:36331", + "left": "issue:46209", + "right": "issue:46431", "accept": false, - "reason": "Training-job reporting vs Trainer API incompatibility; not the same bug." + "reason": "FP8 inference on B200 is unrelated to chat template availability in processor.save_pretrained/apply_chat_template." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "MLflow failure handling is unrelated to embedding resize/post_init reinitialization." + "reason": "Different areas: MLFlow reporting vs embedding resize/post_init reinitialization." }, { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Two different import errors with different missing dependencies/modules, not one underlying issue." + "reason": "Both are import errors, but one is missing transformers.onnx and the other is missing PyTorch for AutoModel." }, { "left": "issue:45245", "right": "issue:46132", "accept": false, - "reason": "Category-limit runtime error is unrelated to attention registration behavior." + "reason": "Category-limit RuntimeError is unrelated to AttentionInterface.register behavior changes." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer save metadata mismatch vs Qwen3.5 MoE from_pretrained failure are different code paths." + "reason": "Tokenizer save metadata mismatch is different from a Qwen3.5 MoE from_pretrained error." }, { "left": "issue:43122", "right": "issue:46209", "accept": false, - "reason": "Tokenizer behavior change vs FP8 inference on B200 are unrelated." + "reason": "Tokenizer behavior change is unrelated to FP8 inference failure on B200." }, { "left": "issue:43475", "right": "issue:46305", "accept": false, - "reason": "SAM 3 Video attribute missing vs MoonshineStreaming mask off-by-one are different bugs." + "reason": "SAM 3 missing attribute during vision encoding is unrelated to a streaming mask off-by-one bug." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Different import failures from different packages: transformers vs huggingface_hub." + "reason": "Both are import errors, but they involve different missing symbols from different packages and likely different dependency mismatches." }, { "left": "issue:45071", "right": "issue:46082", "accept": false, - "reason": "Type-checking regression in PretrainedConfig is not the same as LlamaConfig head_dim validation." + "reason": "Type-checking regression in v5.4.0 is unrelated to LlamaConfig head_dim validation." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "Both are tokenizer-loading failures, but the reports are too specific and could stem from different checkpoint/model quirks." + "reason": "Tokenizer loading failure in v5 and Qwen3.5 flash-attention generation failure are different bugs." }, { "left": "issue:45381", "right": "issue:46305", "accept": false, - "reason": "Video vision_position_ids issue is not the same as an off-by-one sliding-window mask bug." + "reason": "Both mention positional/window logic, but they affect different models and code paths; not the same bug." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "Feature-extractor arg handling is unrelated to auto_docstring crashing on future annotations." + "reason": "Feature extractor arg handling and auto_docstring annotation parsing are unrelated." }, { "left": "issue:43232", "right": "issue:46424", "accept": false, - "reason": "Generation kwargs update after sync_gpus is a different cache-path issue than max_cache_length honoring." + "reason": "Both touch generation/cache, but one is sync_gpus model kwargs handling and the other is cache length configuration." }, { - "left": "issue:44295", - "right": "issue:44623", + "left": "issue:44734", + "right": "issue:46421", "accept": false, - "reason": "Buffer access error and processor save_pretrained missing files are different behaviors." + "reason": "Serve-time KV-cache tensor indexing crash is not the same as lazy cache init causing recompilation." }, { - "left": "issue:44734", - "right": "issue:46421", + "left": "issue:44295", + "right": "issue:44623", "accept": false, - "reason": "Both involve cache, but one is tensor indexing in serve and the other is lazy cache initialization/recompilation." + "reason": "Position_ids buffer access bug is unrelated to processor.save_pretrained missing files." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "FP8 MoE save_pretrained bug is unrelated to the default 50GB shard-size question." + "reason": "FP8 MoE save_pretrained behavior and max_shard_size default are different save_pretrained concerns." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "torch.split return handling in GlmMoeDsaIndexer is unrelated to InternVL2 meta-init incompatibility." + "reason": "torch.split return semantics in a specific indexer are unrelated to remote-code checkpoint meta-initialization compatibility." }, { "left": "issue:43232", "right": "issue:45834", "accept": false, - "reason": "Generation/cache kwargs bug does not match Kosmos2.5 OCR index overflow." + "reason": "Both are cache/generation related in a broad sense, but they are distinct failures with different roots." }, { - "left": "issue:44373", - "right": "issue:44485", + "left": "issue:44190", + "right": "issue:46082", "accept": false, - "reason": "Docstring mistake for position_ids is not the same as a GLM-5 RoPE implementation bug." + "reason": "Dataset loading in an example script is unrelated to LlamaConfig head_dim validation." }, { - "left": "issue:44190", - "right": "issue:46082", + "left": "issue:44373", + "right": "issue:44485", "accept": false, - "reason": "Dataset-loading script issue is unrelated to LlamaConfig head_dim validation." + "reason": "Wrong docstring is documentation-only and not the same as the GLM-5 RoPE implementation bug." }, { "left": "issue:43723", "right": "issue:44871", "accept": false, - "reason": "Tokenizer loading failure in v5 is not the same as inconsistent eos_token_id configuration." + "reason": "Tokenizer loading failure and inconsistent eos_token_id configuration are separate issues." }, { "left": "issue:43723", "right": "issue:44991", "accept": false, - "reason": "Related surface area, but one is generic tokenizer loading and the other is a model-specific regression; not enough to treat as the same bug." + "reason": "Both are tokenizer-related, but they concern different model-specific loading failures rather than the same defect." }, { "left": "issue:44279", "right": "issue:46328", "accept": false, - "reason": "General dependency issue is unrelated to SequenceFeatureExtractor.pad performance behavior." - }, - { - "left": "issue:44263", - "right": "issue:44360", - "accept": false, - "reason": "Both mention GlmMoeDsaIndexer, but one is torch.split return values and the other is a missing ReLU; different fixes." + "reason": "Generic dependency issue is unrelated to the SequenceFeatureExtractor.pad performance regression." } ] }, @@ -13310,17 +13267,16 @@ "issue:43408", "issue:43479", "issue:43550", - "issue:43576", "issue:43644", "issue:43646", "issue:43761", - "issue:43784", "issue:43824", "issue:44038", "issue:44261", "issue:44263", "issue:44295", "issue:44297", + "issue:44360", "issue:44368", "issue:44373", "issue:44483", @@ -13344,17 +13300,18 @@ ], "soft_pairs": [ "issue:44483|issue:45406", + "issue:44263|issue:44360", "issue:42907|issue:46050", - "issue:45092|issue:45357", "issue:30990|issue:38175", + "issue:45092|issue:45357", "issue:43644|issue:46305", - "issue:44261|issue:45092", "issue:44829|issue:46002", "issue:42831|issue:45310", + "issue:44261|issue:45092", "issue:42831|issue:44038", "issue:43408|issue:46402", - "issue:44263|issue:44373", "issue:41628|issue:43824", + "issue:44263|issue:44373", "issue:43479|issue:44913", "issue:39401|issue:44568", "issue:43761|issue:46082", @@ -13363,218 +13320,217 @@ "issue:42757|issue:45278", "issue:27957|issue:30990", "issue:42886|issue:42994", - "issue:42831|issue:45753", "issue:22355|issue:28218", "issue:44368|issue:46305", + "issue:42831|issue:45753", "issue:43408|issue:46369", "issue:43232|issue:45593", "issue:44295|issue:44297", "issue:43550|issue:46132", - "issue:43646|issue:46002", - "issue:43576|issue:43784" + "issue:43646|issue:46002" ], - "prepared_review_unit_hash": "2d91638450f4808ecdb8e02a78fbb7a422ad6add32417def116d040ba396faf2", + "prepared_review_unit_hash": "ec974d5f551e1961e2182c91e5df479081bc836bef29e278d3ed9351c2af3a5e", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36089, - "estimated_input_tokens": 9023, - "estimated_eval_tokens": 18302 + "serialized_chars": 35477, + "estimated_input_tokens": 8870, + "estimated_eval_tokens": 17996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These candidates are mostly unrelated regressions that only share broad subsystem similarity (tokenizers, save/load, imports, multimodal models). None of the soft-similarity pairs look like the same underlying bug or a single mergeable change.", - "confidence": 0.84, - "canonical_issue_reason": "No single canonical issue stands out: the set spans many distinct problems, and the soft pairs are too broad to treat as duplicates of one another.", + "summary": "This cluster is mostly lexical noise: the issues span unrelated bugs, import errors, tokenizer quirks, multimodal regressions, quantization, and config problems. None of the soft pairs look like the same underlying defect or change.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the cluster is heterogeneous and the soft pairs do not converge on one underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42886 is the strongest standalone report here because it is specific, reproducible, and version-scoped (offline cache load regression).", + "best_issue_reason": "No single issue is a safe representative for deduping; the items target different subsystems and root causes.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "Different failures and code paths: chat/completions request validation vs Gemma4Processor missing _tokenizer." + "reason": "Chat/completions request handling and Gemma4Processor _tokenizer import are unrelated bugs." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "torch.split return values in GlmMoeDsaIndexer is unrelated to the DSA ReLU discussion." }, { "left": "issue:42907", "right": "issue:46050", "accept": false, - "reason": "Both involve model saving/loading under quantization, but they describe different quantization workflows and failure modes." - }, - { - "left": "issue:45092", - "right": "issue:45357", - "accept": false, - "reason": "Both touch multimodal model persistence, but one is meta-init incompatibility for old InternVL2 checkpoints and the other is incorrect visual encoder keys in Qwen3.5 save_pretrained." + "reason": "Both involve saving/loading quantized models, but they describe different failures and code paths." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Unrelated: Sentence-Transformers loading hang vs SIGLIP2 zero-probability outputs." + "reason": "Sentence Transformers loading hang and SIGLIP2 zero probabilities are different problems." }, { - "left": "issue:43644", - "right": "issue:46305", + "left": "issue:45092", + "right": "issue:45357", "accept": false, - "reason": "Different bugs: non-persistent buffer corruption vs an off-by-one in MoonshineStreaming mask logic." + "reason": "InternVL2 meta-init incompatibility and Qwen3.5 visual-key save regression are distinct issues." }, { - "left": "issue:44261", - "right": "issue:45092", + "left": "issue:43644", + "right": "issue:46305", "accept": false, - "reason": "Different model families and different issues: MLA rms_norm_eps precision mismatch vs InternVL2 meta-init loading." + "reason": "Non-persistent buffer junk and MoonshineStreaming window off-by-one are unrelated." }, { "left": "issue:44829", "right": "issue:46002", "accept": false, - "reason": "Both mention training instability, but the concrete bugs differ: flash_attention_3 degeneration vs NaNs at checkout/init." + "reason": "Flash Attention 3 training degeneration and NaNs after checkout/init are different failure modes." }, { "left": "issue:42831", "right": "issue:45310", "accept": false, - "reason": "Different subsystems: FineGrainedFP8 accuracy vs Qwen3.5-Moe from_pretrained error." + "reason": "FineGrainedFP8 accuracy and Qwen3.5 MoE from_pretrained errors are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "MLA layernorm epsilon precision bug is unrelated to InternVL2 meta initialization." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "Different problems despite both being model-specific: FP8 accuracy regression vs Qwen3-VL-Moe bug." + "reason": "FineGrainedFP8 accuracy issues do not match the Qwen3-VL-Moe bug." }, { "left": "issue:43408", "right": "issue:46402", "accept": false, - "reason": "Same model family, but one is a warning about model type mismatch and the other is missing text encoder weight loading; not the same bug." + "reason": "sam3_tracker warning and sam3 text-encoder weights not loading are different regressions." }, { - "left": "issue:44263", - "right": "issue:44373", + "left": "issue:41628", + "right": "issue:43824", "accept": false, - "reason": "Completely different: torch.split behavior in GlmMoeDsaIndexer vs a docstring issue for position_ids." + "reason": "Both are import errors, but they concern different missing classes and likely different fixes." }, { - "left": "issue:41628", - "right": "issue:43824", + "left": "issue:44263", + "right": "issue:44373", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols and likely different root causes." + "reason": "torch.split return values and a wrong docstring for position_ids are unrelated." }, { "left": "issue:43479", "right": "issue:44913", "accept": false, - "reason": "Unrelated config bugs: Phi4Multimodal default subconfig initialization vs GPTNeoX rotary_pct reload behavior." + "reason": "Phi4MultimodalConfig default subconfigs and GPTNeoXConfig rotary_pct reload are different config bugs." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer-related, but one is offset_mapping for Qwen3 and the other is special-token insertion for mdeberta-v3." + "reason": "Wrong offset_mapping and missing BOS/EOS insertion are separate tokenizer behaviors." }, { "left": "issue:43761", "right": "issue:46082", "accept": false, - "reason": "Different code paths: CLIPVision forward hidden_states regression vs LlamaConfig head_dim validation." + "reason": "CLIPVision hidden_states regression and LlamaConfig head_dim validation are unrelated." }, { "left": "issue:22355", "right": "issue:27957", "accept": false, - "reason": "No overlap beyond being Transformers issues: missing transformers.onnx module vs Flash Attention 2 behavior on XLM-RoBERTa." + "reason": "Missing transformers.onnx module and XLMRoberta Flash Attention 2 are unrelated." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Both concern tokenizer save/load, but one is offline cache loading and the other is an incorrect tokenizer decoder type in v5." + "reason": "Offline cache loading and tokenizer decoder type mismatch are different tokenizer issues." }, { "left": "issue:42757", "right": "issue:45278", "accept": false, - "reason": "Different import problems: one missing huggingface_hub symbol, the other a broad set of post-upgrade import errors." + "reason": "General import errors after upgrade are broader, but not the same concrete bug as is_offline_mode import failure." }, { "left": "issue:27957", "right": "issue:30990", "accept": false, - "reason": "Different failures: Flash Attention 2 with XLM-RoBERTa vs Sentence-Transformers getting stuck loading." + "reason": "Flash Attention 2 support and Sentence Transformers loading hang are unrelated." }, { "left": "issue:42886", "right": "issue:42994", "accept": false, - "reason": "Different persistence bugs: offline tokenizer cache loading vs quantized model saving failure." - }, - { - "left": "issue:42831", - "right": "issue:45753", - "accept": false, - "reason": "Different model behaviors: FP8 accuracy regression vs Qwen3_5 looping on a specific image." + "reason": "Offline tokenizer cache loading and quantized model saving failure are unrelated." }, { "left": "issue:22355", "right": "issue:28218", "accept": false, - "reason": "Different tokenizer/module issues: missing transformers.onnx vs an added-token spacing bug." + "reason": "ONNX import failure and added-token spacing bug are different tokenizer topics." }, { "left": "issue:44368", "right": "issue:46305", "accept": false, - "reason": "Different concerns: a tie_word_embeddings warning during LoRA fine-tuning vs a sliding-window mask off-by-one." + "reason": "tie_word_embeddings warning and MoonshineStreaming mask off-by-one are unrelated." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FineGrainedFP8 accuracy and Qwen3_5 infinite-loop-on-image are unrelated." }, { "left": "issue:43408", "right": "issue:46369", "accept": false, - "reason": "Same family name only; warning about sam3_tracker instantiation is not the same as memory overflow in sam3-litetext." + "reason": "sam3_tracker warning and sam3-litetext memory overflow are different bugs." }, { "left": "issue:43232", "right": "issue:45593", "accept": false, - "reason": "Unrelated generation-state bug vs D-FINE auxiliary-loss behavior." + "reason": "Generation kwargs after sync_gpus and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:44295", "right": "issue:44297", "accept": false, - "reason": "Different persistence issues: reading a registered position_ids buffer vs tokenizer_class mismatch in tokenizer_config.json." + "reason": "position_ids buffer read error and tokenizer_class mismatch in config are unrelated." }, { "left": "issue:43550", "right": "issue:46132", "accept": false, - "reason": "Both involve attention code, but one is a torch.compile SDPA failure and the other is a behavior change in AttentionInterface.register." + "reason": "Bamba SDPA torch.compile failure and AttentionInterface.register behavior change are different code paths." }, { "left": "issue:43646", "right": "issue:46002", "accept": false, - "reason": "Different roots: custom model initialization regression vs NaNs in classification heads." - }, - { - "left": "issue:43576", - "right": "issue:43784", - "accept": false, - "reason": "Different import/regression contexts: transformers env CLI breakage vs NameError when importing sentence-transformers." + "reason": "Non-persistent buffer junk and NaNs in classification heads after init are not the same defect." } ] }, @@ -13605,13 +13561,13 @@ "issue:39401", "issue:41628", "issue:42371", - "issue:42757", "issue:42907", "issue:43116", "issue:43122", "issue:43408", "issue:43452", "issue:43475", + "issue:43576", "issue:43756", "issue:43784", "issue:43824", @@ -13648,11 +13604,12 @@ "issue:46326" ], "soft_pairs": [ + "issue:43576|issue:43784", "issue:42371|issue:44263", "issue:44117|issue:44704", + "issue:30990|issue:33357", "issue:44188|issue:44373", "issue:43756|issue:44060", - "issue:30990|issue:33357", "issue:42907|issue:44945", "issue:44261|issue:44373", "issue:43122|issue:44521", @@ -13677,223 +13634,222 @@ "issue:44038|issue:44222", "issue:43824|issue:44351", "issue:45216|issue:45478", - "issue:22355|issue:29159", - "issue:42757|issue:44351" + "issue:22355|issue:29159" ], - "prepared_review_unit_hash": "834a060be9644a29eacc780009ed4e13f1981503368c2ee62b06856875111590", + "prepared_review_unit_hash": "a9e3233c155ca08402033b9b2799e91f50832fc6781a20a5cf00e167523f6952", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36663, - "estimated_input_tokens": 9166, - "estimated_eval_tokens": 18588 + "serialized_chars": 36649, + "estimated_input_tokens": 9163, + "estimated_eval_tokens": 18582 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose set of unrelated Transformers issues: import failures, tokenizer/chat-template regressions, and model-specific save/load/config bugs. The soft similarities are mostly surface-level; I don\u2019t see any pair that clearly shares the same underlying bug or change.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 45278 is the broadest umbrella in the set (post-upgrade import failures) and is the closest thing to a cluster anchor, but it still does not actually cover most of the tokenizer/model-specific reports.", + "summary": "The soft pairs are mostly false positives: they span unrelated tokenizer regressions, import errors, model-loading bugs, chat-template issues, and model-specific architecture problems. I would not merge any of the proposed pairs as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": "No single strong canonical issue emerges because the set is heterogeneous. If a representative is required, issue 43122 is the cleanest tokenizer-regression example, but it does not cover the rest of the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45278 is the most general and representative issue title in the group, with the widest overlap to the import-error subset.", + "best_issue_reason": "Issue 43122 is the most general and clearly described tokenization regression among the candidates, so it is the best single representative; however, the overall set is too broad to treat as one duplicate cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Different bugs: one is the `transformers env` CLI in v5, the other is a `NameError` when importing sentence-transformers." + }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "Different problems: TF32 settings API guidance vs a torch.split return-value issue in a specific indexer." + "reason": "Unrelated topics: TF32 settings/API behavior versus a `torch.split()` indexing/return-value issue." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "Tokenizer mapping fallback bug vs AutoProcessor kwargs propagation; different code paths and symptoms." + "reason": "Both involve loader paths, but one is `TOKENIZER_MAPPING_NAMES` returning None and the other is `AutoProcessor.from_pretrained` not forwarding kwargs; not the same code-path bug." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Different failure modes: Sentence Transformers loading hangs versus a MacOS bus error with a CLIP model." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "A torch.compile attention-kernel divergence bug is unrelated to a docstring issue." + "reason": "One is a torch.compile branching issue in attention kernels; the other is a wrong docstring for `position_ids`." }, { "left": "issue:43756", "right": "issue:44060", "accept": false, - "reason": "Different model-specific logic bugs: RoPE layer dropping vs tied-weights warning in Qwen3-Next." - }, - { - "left": "issue:30990", - "right": "issue:33357", - "accept": false, - "reason": "Both are load/runtime failures, but one is Sentence-Transformers hanging and the other is a MacOS bus error for a CLIP model." + "reason": "Distinct model-specific issues: Smollm3 RoPE layer handling versus Qwen3-Next tied-weights warning miswiring." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Dequantized save_pretrained failure is not the same as incorrect output under pipeline parallelism." + "reason": "Different symptoms and subsystems: dequantized save failures versus incorrect output under pipeline parallelism." }, { "left": "issue:44261", "right": "issue:44373", "accept": false, - "reason": "Numerical precision/config bug in MLA q_a_layernorm vs a wrong docstring for position_ids." + "reason": "MLA layernorm precision/config issue versus a documentation-only `position_ids` bug." }, { "left": "issue:43122", "right": "issue:44521", "accept": false, - "reason": "Tokenizer regression across versions is separate from multimodal assistant_mask generation in apply_chat_template." + "reason": "Tokenizer regression versus multimodal assistant_mask generation in `apply_chat_template`; related area, but not the same bug." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Qwen3 offset_mapping bug vs a model loading failure for a different model; not the same defect." + "reason": "Different tokenizer problems: wrong `offset_mapping` for Qwen3 versus a specific model load failure for `cjvt/sleng-bert`." }, { "left": "issue:41628", "right": "issue:45278", "accept": false, - "reason": "A single missing AutoImageProcessor import is not clearly the same underlying bug as a broad report of many import errors." + "reason": "Both are import-related, but one is missing `AutoImageProcessor` and the other is many import errors after a version upgrade; too broad and not the same fix." }, { "left": "issue:43122", "right": "issue:46326", "accept": false, - "reason": "Different tokenizer regressions: offset/tokenization differences vs continue_final_message handling for mistral_common tokenizers." + "reason": "Both are tokenizer-related, but one is tokenization output drift across versions and the other is `continue_final_message` behavior for mistral_common tokenizers." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals and GLM-5 GenerationConfig validation are unrelated." + "reason": "Different features: dequantized model saving versus GLM-5 `GenerationConfig` validation." }, { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Python version compatibility issue vs wrong outputs with pipeline parallelism; different layers of failure." + "reason": "Python-version compatibility problem versus pipeline-parallelism output corruption." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "Passing kwargs to cached_file is unrelated to assistant messages with tool calls crashing in apply_chat_template." + "reason": "`AutoProcessor.from_pretrained` kwarg forwarding is unrelated to assistant messages with tool calls crashing in `apply_chat_template`." }, { "left": "issue:42907", "right": "issue:44964", "accept": false, - "reason": "Dequantized save failure is not the same as a Phi-4 multimodal loading failure." + "reason": "Saving dequantized models and loading a multimodal Phi-4 model are different code paths and symptoms." }, { "left": "issue:43784", "right": "issue:45600", "accept": false, - "reason": "Import-time NameError in sentence-transformers integration vs stale auto_mappings references to removed Sam3LiteText configs." + "reason": "Importing sentence-transformers with a `NameError` is unrelated to `auto_mappings.py` referencing removed Sam3LiteText configs." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Qwen3-VL-Moe loading/regression and tokenizer.save_pretrained class-name mismatch are different bugs." + "reason": "Qwen3-VL-Moe load bug versus tokenizer class metadata mismatch on `save_pretrained`; different bugs." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Wrong offset_mapping and zero assistant_masks are different tokenizer/template behaviors." + "reason": "Wrong `offset_mapping` for Qwen3 is not the same as assistant mask generation for multimodal chat templates." }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "run_classification.py multi-label output bug vs Gemma 4 mm_token_type_ids defaulting issue." + "reason": "Multi-label classification example script bug versus Gemma 4 `mm_token_type_ids` defaulting; different affected APIs." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "DSA indexer lacking a ReLU and GLM-5 RoPE implementation are separate architecture/model concerns." + "reason": "Indexer missing a ReLU is unrelated to GLM-5 RoPE implementation details." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Qwen3-VL-Moe bug vs a position_ids buffer access error; no shared underlying failure." + "reason": "Qwen3-VL-Moe loading issue versus a `position_ids` buffer access error; not the same bug." }, { "left": "issue:44625", "right": "issue:44829", "accept": false, - "reason": "Qwen3.5 num_labels propagation and flash_attention_3 training collapse are unrelated." + "reason": "Qwen3.5 `num_labels` propagation and flash_attention_3 training degradation are unrelated." }, { "left": "issue:43408", "right": "issue:43475", "accept": false, - "reason": "Both are SAM 3 Video related, but one is a model-type warning and the other is a missing attribute in encoder outputs." + "reason": "Different SAM 3 problems: model-type warning on instantiation versus missing `fpn_position_embeddings` attribute." }, { "left": "issue:43122", "right": "issue:44514", "accept": false, - "reason": "Tokenization differences across versions are separate from batched apply_chat_template crashes with padding=False." + "reason": "Both touch template/tokenization flows, but one is version-dependent tokenization drift and the other is a batched chat-template crash with `padding=False`." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both mention tokenizer output changes, but they concern different models and different regressions." + "reason": "Different tokenizer regressions affecting different models and behaviors; not the same underlying defect." }, { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Saving dequantized models is unrelated to gguf_file breaking from_pretrained for tokenizer/model loading." + "reason": "Dequantized save failure and `gguf_file` breaking `from_pretrained` are separate loader/saver issues." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Qwen3-VL-Moe loading regression and FP8 save_pretrained behavior are distinct failures." + "reason": "Qwen3-VL-Moe load bug versus FP8 `save_pretrained` for moe models; different code paths." }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Cannot import a specific Qwen2.5-VL class vs missing HybridCache import; different missing symbols and code paths." + "reason": "Both are import errors, but they concern different symbols (`Qwen2_5_VLForConditionalGeneration` vs `HybridCache`) and different fixes." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both involve Qwen3.5 and version regressions, but one is save_pretrained checkpoint correctness and the other is from_pretrained loading failure." + "reason": "Saved-checkpoint regression for Qwen3.5 is not the same as a Qwen3.5 moe `from_pretrained` error." }, { "left": "issue:22355", "right": "issue:29159", "accept": false, - "reason": "An ONNX import error and tokenizer slow/fast inconsistency are unrelated issues." - }, - { - "left": "issue:42757", - "right": "issue:44351", - "accept": false, - "reason": "Different missing imports from different packages/symbols; not the same underlying bug." + "reason": "Completely unrelated: missing `transformers.onnx` module versus inconsistent slow/fast tokenizer behavior." } ] }, @@ -13927,6 +13883,7 @@ "issue:42222", "issue:42371", "issue:42673", + "issue:42757", "issue:43064", "issue:43066", "issue:43116", @@ -13944,8 +13901,9 @@ "issue:44075", "issue:44117", "issue:44188", + "issue:44222", + "issue:44351", "issue:44355", - "issue:44380", "issue:44485", "issue:44704", "issue:44779", @@ -13955,18 +13913,17 @@ "issue:45072", "issue:45083", "issue:45278", - "issue:45357", "issue:45446", "issue:45478", "issue:45563", "issue:45588", "issue:45753", "issue:46002", - "issue:46077", "issue:46231", "issue:46328" ], "soft_pairs": [ + "issue:42757|issue:44351", "issue:36296|issue:43116", "issue:43278|issue:44945", "issue:42175|issue:44938", @@ -13987,211 +13944,210 @@ "issue:44038|issue:45478", "issue:43976|issue:45446", "issue:43116|issue:43825", - "issue:45357|issue:45478", + "issue:44222|issue:45478", "issue:45083|issue:45446", "issue:43064|issue:43673", "issue:33453|issue:39401", "issue:43937|issue:45563", "issue:39401|issue:44779", - "issue:44117|issue:45588", - "issue:44380|issue:46077" + "issue:44117|issue:45588" ], - "prepared_review_unit_hash": "f429a24c4f1ee3db94f85beff435dc13f401fa262b625b91e8287eb8398bc56c", + "prepared_review_unit_hash": "d61bdc5604ddd74c2994f246c520f232c4d26ab0f8dc4000def0e981e16596cd", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 35679, - "estimated_input_tokens": 8920, - "estimated_eval_tokens": 18096 + "serialized_chars": 35562, + "estimated_input_tokens": 8891, + "estimated_eval_tokens": 18038 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a grab-bag of unrelated Transformers issues, mostly around v5 regressions, tokenizers, loading, training, and model-specific bugs. None of the soft pairs look like the same underlying defect, so all soft edges should be rejected. If a single representative issue is needed, issue 45278 is the broadest umbrella, but it is not a true duplicate target for the rest of the cluster.", - "confidence": 0.89, - "canonical_issue_reason": "Issue 45278 is the broadest representative of the cluster because it summarizes multiple import failures after a version upgrade, but the cluster overall is too heterogeneous to have a true canonical duplicate.", + "summary": "This cluster is a heterogeneous mix of unrelated Transformers issues. None of the proposed soft edges look like the same underlying bug or change, so all should be rejected. There are no PRs in the set.", + "confidence": 0.96, + "canonical_issue_reason": "issue:45278 is the broadest, most representative problem statement in the set: a general import-regression report after upgrading versions. It is still not a duplicate of the other issues, but it is the best single umbrella-style issue here.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45278 is the best umbrella issue among this set, since it captures a broad release-regression/import-failure theme and is more general than the other very specific bug reports.", + "best_issue_reason": "issue:45278 is the most suitable representative issue because it describes a clear, high-level regression affecting multiple imports, unlike the narrower model- or task-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports from different packages/symbols: `is_offline_mode` from `huggingface_hub` vs `HybridCache` from `transformers`." + }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Tensor-parallel training state bug vs. multi-label example-script output bug; different code paths and symptoms." + "reason": "Tensor-parallel training bug versus an example-script multi-label classification output issue; different code paths and symptoms." }, { "left": "issue:43278", "right": "issue:44945", "accept": false, - "reason": "Embedding dtype drift during evaluate vs. wrong output under pipeline parallelism; both are training-time issues but not the same bug." + "reason": "BF16-to-FP32 dtype change in evaluate is unrelated to incorrect output under pipeline parallelism." }, { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Missing TensorFlow backend in extras vs. Python 3.14 import/load failure; unrelated failures." + "reason": "Packaging/backend dependency omission for `pip install '.[torch]'` is unrelated to a Python 3.14 load failure." }, { "left": "issue:43066", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but one is a decoder-type mismatch and the other is incorrect Deepseek tokenization results; not the same concrete defect." + "reason": "Both mention tokenizers, but one is a decoder-type regression and the other is incorrect Deepseek tokenizer results; not the same bug." }, { "left": "issue:43493", "right": "issue:45753", "accept": false, - "reason": "Different model families and different failures: SigLIP2 implementation discrepancy vs. Qwen3.5 image-driven infinite loop." + "reason": "Different models and failures: SigLIP2 HF/JAX discrepancy versus Qwen3_5 infinite-loop behavior on a specific image." }, { "left": "issue:22355", "right": "issue:29127", "accept": false, - "reason": "Importing transformers.onnx missing vs. improving an error message for LayoutLMv3 box handling; unrelated." + "reason": "Module import failure for `transformers.onnx` is unrelated to a LayoutLMv3 error-message clarity issue." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "UMT5 weight loading bug vs. AutoProcessor not forwarding kwargs to cached_file; different loading layers." + "reason": "Different loading paths: missing `embed_tokens.weight` in UMT5Encoder versus kwargs propagation in `AutoProcessor.from_pretrained`." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in multithreaded Qwen3ForCausalLM usage vs. dtype mismatch in bfloat16 inference; different problems." + "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM use is unrelated to bfloat16 dtype mismatch CI/inference failures." }, { "left": "issue:38617", "right": "issue:45278", "accept": false, - "reason": "A specific missing symbol import vs. a broad collection of import errors after upgrade; same general area, but not the same bug." + "reason": "A single missing symbol import is not the same as a broad upgrade-induced wave of import errors." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "GenerationConfig validation failure vs. RoPE implementation discussion; distinct GLM-5 issues." + "reason": "GenerationConfig validation failure and GLM-5 RoPE discussion are different GLM-5 issues." }, { "left": "issue:44188", "right": "issue:46328", "accept": false, - "reason": "torch.compile attention kernel branching bug vs. a padding performance inefficiency; not a duplicate." + "reason": "Torch.compile attention-kernel divergence is unrelated to SequenceFeatureExtractor padding performance." }, { "left": "issue:44355", "right": "issue:46231", "accept": false, - "reason": "Compiled Python file loading error vs. documentation parameter-name mismatch; unrelated." + "reason": "Compiled Python file errors and a documentation parameter-name mismatch are unrelated." }, { "left": "issue:43611", "right": "issue:46002", "accept": false, - "reason": "base_model_prefix loading regression vs. NaNs in classification heads after init/checkout; different failure modes." + "reason": "Base-model-prefix loading regression is unrelated to NaNs in classification heads after checkout/init." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Example-script multi-label output bug vs. SigLIP2 model discrepancy; unrelated code paths." + "reason": "Example-script multi-label classification empties and SigLIP2 implementation discrepancy are unrelated." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 configuration API guidance vs. missing image-patch method in GLM46V; unrelated." + "reason": "TF32 behavior warning/documentation issue is unrelated to a missing ImageProcessor method causing an AttributeError." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "Different model-specific breakages: vitpose models vs. Ministral-3 FlashAttention failure." + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are different model-specific bugs." }, { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "Optimizer argument handling vs. train-then-evaluate failure; different training lifecycle bugs." + "reason": "SGD optimizer arguments not being used is unrelated to `trainer.evaluate()` failing after `trainer.train()`." }, { "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3 MoE and Transformers v5, but one is a broader compatibility complaint and the other is a from_pretrained error on a later version; not the same concrete bug." + "reason": "Different Qwen3 MoE variants and different failures: one is a broad v5 bug report, the other is a from_pretrained error for Qwen3.5 Moe." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python-version incompatibility vs. a PyTorch version-check bug in flex_attention; different version gates." + "reason": "Python version support regression and a PyTorch version-check bug for AuxRequest import are not the same issue." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Empty multi-label results in an example script vs. an incorrect pipeline error message; different concerns." + "reason": "Multi-label classification example output bug is unrelated to a pipeline translation-task support/error-message issue." }, { - "left": "issue:45357", + "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "save_pretrained writes wrong visual encoder keys vs. Qwen3.5 from_pretrained error; same model family but opposite paths and not the same bug." + "reason": "FP8 save_pretrained behavior for MoE models is not the same as a Qwen3.5 Moe loading error." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Qwen3 Omni-MoE feature-length helper behavior vs. flex_attention import/version check; unrelated." + "reason": "Qwen3_omni_moe feature-length helper behavior is unrelated to the flex_attention AuxRequest version check." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Distributed training optimizer-state corruption vs. generation cache missing during chunked prefill; different subsystems." + "reason": "Wrong optimizer states under FSDP2/PEFT/cpu_ram_efficient_loading is unrelated to missing GenerationMixin cache in chunked prefill." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Tokenizer loading regression vs. wrong offset mapping in a specific tokenizer; both tokenizer-related, but not the same defect." + "reason": "Tokenizer-loading regression is too broad to be the same as a specific wrong offset_mapping bug." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "GenerationConfig invalid on GLM-5 vs. stale warning for num_return_sequences in paged generate; unrelated generation issues." + "reason": "Invalid GenerationConfig and a stale num_return_sequences warning in paged generate are distinct generation issues." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Different tokenizer bugs on different models: Qwen3 offset mapping vs. Deepseek incorrect outputs." + "reason": "Both are tokenizer correctness bugs, but they affect different models and different behaviors (`offset_mapping` vs general incorrect results)." }, { "left": "issue:44117", "right": "issue:45588", "accept": false, - "reason": "Tokenizer mapping returning None vs. flash_attention crash on s_aux=None; both involve None, but entirely different code paths." - }, - { - "left": "issue:44380", - "right": "issue:46077", - "accept": false, - "reason": "GPT2 attention scaling ignored in SDPA/FlashAttention vs. Gemma4 bidirectional-attention masking bug; different model/backend logic." + "reason": "Tokenizer mapping fallback bug and flash_attention crashing on `s_aux=None` are unrelated code paths." } ] }, @@ -14223,8 +14179,6 @@ "issue:42490", "issue:42617", "issue:42673", - "issue:42831", - "issue:43066", "issue:43122", "issue:43408", "issue:43421", @@ -14240,6 +14194,7 @@ "issue:44220", "issue:44261", "issue:44291", + "issue:44380", "issue:44410", "issue:44466", "issue:44484", @@ -14259,12 +14214,14 @@ "issue:45446", "issue:45561", "issue:46002", + "issue:46077", "issue:46178", "issue:46231", "issue:46298", "issue:46328" ], "soft_pairs": [ + "issue:44380|issue:46077", "issue:43827|issue:45446", "issue:42673|issue:43906", "issue:43450|issue:43986", @@ -14292,218 +14249,210 @@ "issue:44484|issue:45310", "issue:38617|issue:44220", "issue:43937|issue:44496", - "issue:44933|issue:45278", - "issue:43066|issue:43122", - "issue:42371|issue:42831" + "issue:44933|issue:45278" ], - "prepared_review_unit_hash": "d93eec35bab5d4e47be0f67bdfe85e0f58cb84e9e839bf015db19dca36856823", + "prepared_review_unit_hash": "b788be39585c367334634831900814a5d6a2a675e1cf91401fe051e235bce57e", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36493, - "estimated_input_tokens": 9124, - "estimated_eval_tokens": 18504 + "soft_pair_count": 29, + "serialized_chars": 36265, + "estimated_input_tokens": 9067, + "estimated_eval_tokens": 18390 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: most items are unrelated standalone issues spanning tokenization, model loading, imports, training, docs, and testing. I found no soft-edge pair that confidently describes the same underlying bug/change, so no duplicates should be merged here.", - "confidence": 0.96, - "canonical_issue_reason": "Issue 43122 is the broadest, most user-facing regression report in the only partially coherent subtheme (tokenization behavior changing across versions), so it is the best available representative if one must be chosen.", + "summary": "This cluster is mostly a mix of unrelated transformer issues: import-regression reports, tokenizer/model-specific bugs, video/model loading failures, and a few docs/testing/perf nits. I did not find any convincing duplicate pairs; the closest thing to a hub is the broad import-regression report #45278.", + "confidence": 0.63, + "canonical_issue_reason": "No single true duplicate stands out across this mixed set; #45278 is the broadest representative of the import-regression subgroup.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43122 is the most general and actionable issue here; it captures a cross-version regression with a concrete user symptom better than the narrower or tangential reports. That said, the cluster overall does not look like a true duplicate set.", + "best_issue_reason": "#45278 is the most general, high-signal issue here and the best umbrella candidate if one issue had to represent the set.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "Different concrete bugs: GPT2 attention scaling under SDPA/FlashAttention vs Gemma4 building the wrong attention mask." + }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs about removed pipeline APIs vs a flex_attention import/version-check bug; different subsystems and failures." + "reason": "Docs cleanup about pipeline() references is unrelated to a PyTorch version check in flex_attention." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "Qwen3ForCausalLM VRAM leak in multithreaded dataloaders vs an isolated reproduction of another issue; no clear same bug." + "reason": "One is a Qwen3 VRAM leak in multi-threaded loading; the other is only a reproduction of a different issue." }, { "left": "issue:43450", "right": "issue:43986", "accept": false, - "reason": "Batched video processor shape bug vs crash loading video models without torchvision; different code paths." + "reason": "Batched video-shape handling and missing-torchvision import failure are different code paths." }, { "left": "issue:43673", "right": "issue:46328", "accept": false, - "reason": "Generation cache missing during chunked_prefill vs a SequenceFeatureExtractor.pad performance issue; unrelated behavior." + "reason": "Generation cache during chunked prefill is unrelated to feature-extractor padding conversion overhead." }, { "left": "issue:44466", "right": "issue:46002", "accept": false, - "reason": "lm_head serialization inconsistency vs NaNs in classification heads; different symptoms and likely causes." + "reason": "Serialization of tied weights is not the same bug as NaNs after checkout/init." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "init_empty_weights argument TypeError vs PreTrainedConfig type-checking regression; not the same defect." + "reason": "Unexpected init_empty_weights arguments and PretrainedConfig type-checking are distinct regressions." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "Invalid GenerationConfig for GLM-5 vs Qwen3.5 flash-attention generation failure; different model-specific bugs." + "reason": "GLM-5 generation-config validation and Qwen3.5 flash-attention generation failure are different problems." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "PEFT minimum-version bump issue vs PyTorch version check in flex_attention import; unrelated." + "reason": "A PEFT minimum-version bump is unrelated to the flex_attention import-version guard." }, { "left": "issue:44117", "right": "issue:45081", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES None-handling vs Mistral regex patch crash; both tokenizer-related but distinct failures." + "reason": "TOKENIZER_MAPPING_NAMES returning None and Mistral regex patch crashes are different tokenizer-loading bugs." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "A vague testing_utils bug vs a pytest-xdist race on captured_info.txt; same area, not the same concrete bug." + "reason": "A generic testing_utils bug is not the same concrete race on captured_info.txt under xdist." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Missing Qwen3next projections vs Qwen3.5 MoE from_pretrained error; different models and code paths." + "reason": "Qwen3Next missing projections and Qwen3.5 MoE from_pretrained failures are different model/config issues." }, { "left": "issue:43122", "right": "issue:46298", "accept": false, - "reason": "Tokenizer behavior regression vs noisy logging message; clearly unrelated." + "reason": "Tokenization differences across versions are unrelated to a noisy logging message." }, { "left": "issue:45341", "right": "issue:46328", "accept": false, - "reason": "testing_utils issue vs SequenceFeatureExtractor.pad inefficiency; unrelated." + "reason": "testing_utils is unrelated to SequenceFeatureExtractor padding performance." }, { "left": "issue:38617", "right": "issue:41628", "accept": false, - "reason": "Different import errors for different symbols; same general theme, but not the same underlying bug." + "reason": "Both are import errors, but they point to different missing symbols and no shared code path is evident." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "FA3 training degeneration vs Qwen3.5 save_pretrained regression; different code paths and effects." + "reason": "FlashAttention-3 training degeneration and Qwen3.5 checkpoint serialization are different failures." }, { "left": "issue:42222", "right": "issue:46178", "accept": false, - "reason": "VitPose model breakage vs tensorflow_text backend mapping error; unrelated subsystems." + "reason": "Vitpose model loading and a BACKENDS_MAPPING backend-name error are unrelated." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel runtime failure vs a testing_utils bug; no shared underlying change." + "reason": "3d_parallel.py runtime failure is not the same as a testing_utils issue." }, { "left": "issue:45362", "right": "issue:46328", "accept": false, - "reason": "Qwen3.5 chat crash vs audio feature padding inefficiency; unrelated." + "reason": "Qwen3.5 chat crashes and a padding-performance issue are different code paths." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "sam3_video/sam3_tracker type warning vs MobileLLM key error; different model-loading problems." + "reason": "A model-type warning for sam3_* is unrelated to a key error loading MobileLLM-125M." }, { "left": "issue:39401", "right": "issue:43421", "accept": false, - "reason": "Qwen3 offset_mapping bug vs runtime special-token post-processor update issue; similar domain, but distinct concrete bugs." + "reason": "Wrong offset_mapping and runtime post-processor synchronization after special-token edits are different tokenizer behaviors." }, { "left": "issue:42673", "right": "issue:43582", "accept": false, - "reason": "Qwen3 VRAM leak vs Apple Silicon warmup TypeError; unrelated." + "reason": "A Qwen3 VRAM leak is unrelated to an AppleSilicon caching_allocator_warmup TypeError." }, { "left": "issue:42490", "right": "issue:42673", "accept": false, - "reason": "Trailing-slash local path import warning vs Qwen3 VRAM leak; different issues." + "reason": "Trailing-slash local-model import warnings and Qwen3 VRAM leaks do not match." }, { "left": "issue:44877", "right": "issue:46231", "accept": false, - "reason": "Strict config loading for granite_speech vs a docs parameter-name typo; not duplicates." + "reason": "Strict config loading and a documentation parameter-name mismatch are not the same bug." }, { "left": "issue:42371", "right": "issue:44261", "accept": false, - "reason": "TF32 API behavior vs MLA q_a_layernorm precision bug; unrelated." + "reason": "TF32 API guidance and RMS norm epsilon precision mismatch are different issues." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "Question about max_shard_size vs Qwen3.5 from_pretrained error; different topics." + "reason": "A shard-size question is unrelated to a Qwen3.5 from_pretrained error." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "ImportError for layer_type_validation vs audio fbank extraction TypeError; unrelated." + "reason": "Configuration import failure and an fbank extraction issue are unrelated." }, { "left": "issue:43937", "right": "issue:44496", "accept": false, - "reason": "GLM-5 GenerationConfig invalid vs unrecognized model/config.json issue; different model-loading failures." + "reason": "Invalid GenerationConfig and unrecognized model config are different loader/generation failures." }, { "left": "issue:44933", "right": "issue:45278", "accept": false, - "reason": "Specific missing import from image_utils vs broad import errors after an update; related theme but not the same bug." - }, - { - "left": "issue:43066", - "right": "issue:43122", - "accept": false, - "reason": "Wrong tokenizer decoder type in v5 vs changed tokenization across versions; plausible relation, but not enough evidence of one underlying defect." - }, - { - "left": "issue:42371", - "right": "issue:42831", - "accept": false, - "reason": "TF32 behavior settings vs FineGrainedFP8 accuracy issue; different numerical concerns." + "reason": "Both mention imports, but one is a specific missing image_utils import while the other is a broad post-upgrade import-regression report; not enough to treat as the same bug." } ] }, @@ -14531,9 +14480,12 @@ "issue:41628", "issue:42175", "issue:42222", + "issue:42371", + "issue:42831", "issue:43012", - "issue:43064", + "issue:43066", "issue:43097", + "issue:43122", "issue:43335", "issue:43352", "issue:43441", @@ -14549,7 +14501,6 @@ "issue:44117", "issue:44261", "issue:44263", - "issue:44315", "issue:44360", "issue:44410", "issue:44483", @@ -14567,16 +14518,16 @@ "issue:45464", "issue:45478", "issue:45542", - "issue:45588", "issue:45656", "issue:45753", - "issue:46050", "issue:46123", "issue:46172", "issue:46231", "issue:46369" ], "soft_pairs": [ + "issue:43066|issue:43122", + "issue:42371|issue:42831", "issue:44410|issue:45478", "issue:44075|issue:45656", "issue:42222|issue:45542", @@ -14585,8 +14536,8 @@ "issue:43012|issue:43577", "issue:42175|issue:42222", "issue:43673|issue:45341", - "issue:44263|issue:44485", "issue:41628|issue:45071", + "issue:44263|issue:44485", "issue:43874|issue:44261", "issue:45071|issue:45542", "issue:27957|issue:43352", @@ -14609,255 +14560,246 @@ "issue:45440|issue:46172", "issue:43531|issue:45341", "issue:44261|issue:44485", - "issue:43986|issue:45372", - "issue:43097|issue:45588", - "issue:43097|issue:46050", - "issue:43064|issue:44315" + "issue:43986|issue:45372" ], - "prepared_review_unit_hash": "deff011648ea0fedc261152f10f5fa9d2f23d3a278a602b8b57334373858644a", + "prepared_review_unit_hash": "744bbc464c8edf75794f5562988c9c8b324c10f7fefe275f81d091926eb2460e", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 36, - "serialized_chars": 38232, - "estimated_input_tokens": 9558, - "estimated_eval_tokens": 19372 + "soft_pair_count": 35, + "serialized_chars": 37770, + "estimated_input_tokens": 9443, + "estimated_eval_tokens": 19142 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "All candidate soft edges look like false positives: they share broad themes (flash attention, loading, dtype, training) but point to different models, different code paths, and different failure modes.", - "confidence": 0.94, - "canonical_issue_reason": "issue:43097 is the cleanest standalone regression report here: versioned API removal, clear impact, and inbound references.", + "summary": "The items are thematically related to Transformers regressions, but the soft pairs mostly point to different bugs, models, or code paths rather than true duplicates. I would reject every soft edge and treat this as a heterogeneous set, not a single dedup cluster.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "If one issue must anchor follow-up, issue:43097 is the best fit because it is specific, well-scoped, and clearly describes a user-facing breakage.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Both concern tokenizer behavior in v5, but one is about an incorrect decoder type and the other about changed tokenization output; likely related, not the same underlying bug." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 API guidance and FineGrainedFP8 accuracy are different precision/configuration problems with no shared concrete failure mode." + }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Different models and different FlashAttention symptoms; one is a missing projection path, the other is an unsupported/load-time error." + "reason": "One is a missing layer-0 projection bug in qwen3next; the other is a from_pretrained error for Qwen3.5 MoE. Same family, different code-path issue." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "Both concern optimizers, but one says SGD args are ignored while the other is about double optimizer steps under DeepSpeed." + "reason": "SGD arguments being ignored is about optimizer parameter handling; double optimizer.step() under DeepSpeed is a separate training loop bug." }, { "left": "issue:42222", "right": "issue:45542", "accept": false, - "reason": "Unrelated subsystems: vitpose model breakage vs TensorFlow backend installation." + "reason": "VitPose model breakage and missing TensorFlow backend when only tensorboard is installed are unrelated package/model loading issues." }, { "left": "issue:44483", "right": "issue:45464", "accept": false, - "reason": "Both mention chat/completions, but one is a generic request rejection and the other is a streaming inference issue on a specific model." + "reason": "Both involve chat/completions, but one is a generic request rejection and the other is a streaming inference failure on a specific Qwen3.5 model." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Versioned API removal vs unexpected position-id keys; these are different regressions." + "reason": "Removing tie_embeddings_and_encoder_decoder and unexpected position-id keys are different breaking changes with different affected code paths." }, { "left": "issue:43012", "right": "issue:43577", "accept": false, - "reason": "Both involve dtype/precision, but one is a compile-time PyTorch warning and the other is a load-time float32 mismatch." + "reason": "Both mention dtype/precision, but one is a PyTorch warning during bfloat16 compilation while the other is a BLIP2 loading bug leaving Qformer in float32." }, { "left": "issue:42175", "right": "issue:42222", "accept": false, - "reason": "TensorFlow packaging issue vs vitpose model breakage; no shared bug." + "reason": "A TensorFlow backend/install issue is not the same bug as VitPose model breakage." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "Generation cache bug vs testing_utils bug; unrelated code paths." + "reason": "Generation cache missing during chunked prefill and a testing_utils bug are unrelated." }, { - "left": "issue:44263", - "right": "issue:44485", + "left": "issue:41628", + "right": "issue:45071", "accept": false, - "reason": "GLM-related, but DSA indexer split behavior and RoPE implementation are different failures." + "reason": "AutoImageProcessor import failure and PretrainedConfig type checking regression affect different APIs and code paths." }, { - "left": "issue:41628", - "right": "issue:45071", + "left": "issue:44263", + "right": "issue:44485", "accept": false, - "reason": "Import/export issue for AutoImageProcessor vs PretrainedConfig type checking; unrelated." + "reason": "A torch.split return-value issue in one GLM component is not the same as the GLM-5 RoPE implementation bug." }, { "left": "issue:43874", "right": "issue:44261", "accept": false, - "reason": "Different multimodal components with different missing fields/methods." + "reason": "Missing get_number_of_image_patches in an image processor and rms_norm_eps precision mismatch are unrelated multimodal/model config bugs." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "PretrainedConfig typing regression is unrelated to tensorboard/TF backend installation." + "reason": "Type-checking regression in PretrainedConfig is unrelated to a missing TensorFlow backend dependency." }, { "left": "issue:27957", "right": "issue:43352", "accept": false, - "reason": "Both are FlashAttention-2 complaints, but they concern different models and likely different support gaps." + "reason": "Flash Attention 2 support issues for XLM-Roberta and Nemotron are separate model-specific capability gaps." }, { "left": "issue:44263", "right": "issue:46172", "accept": false, - "reason": "Different DeepSeek/GLM code paths; no shared underlying bug." + "reason": "DSA indexer output handling and deepseek-v4 comb.to(dtype).transpose(-1, -2) are different implementation bugs." }, { "left": "issue:44360", "right": "issue:46172", "accept": false, - "reason": "DSA indexer activation bug vs DeepSeek transpose bug; unrelated." + "reason": "A missing ReLU in the DSA indexer is unrelated to the deepseek-v4 tensor transpose/dtype issue." }, { "left": "issue:44484", "right": "issue:45478", "accept": false, - "reason": "Save-pretrained sharding policy vs Qwen3.5 MoE loading; no overlap." + "reason": "save_pretrained shard sizing and Qwen3.5 MoE loading failure are not the same bug." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Both are Qwen-family load issues, but one is a shape mismatch and the other is a missing projection layer." + "reason": "Model weight-shape mismatch for Qwen3-VL-30B-A3B-Instruct and missing projections in qwen3next are different model-loading/model-architecture issues." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "Different FlashAttention problems on different models; not the same bug." + "reason": "FlashAttention failure in Ministral-3 and degenerate training with flash_attention_3 are different failures, even if both mention flash attention." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Qwen3-VL shape mismatch vs a broader Qwen3-VL-Moe issue; too different to merge." + "reason": "Both are Qwen3-VL-related, but one is a weight-shape mismatch and the other is a transformers 5.0 loading issue for Qwen3-VL-Moe; not the same concrete bug." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "Encoder-decoder embedding tie removal vs gguf_file loading regression; unrelated." + "reason": "A removed config attribute and gguf_file breaking from_pretrained are unrelated regression types." }, { "left": "issue:43335", "right": "issue:45071", "accept": false, - "reason": "SwitchTransformers config bug vs a type-checking regression; different areas." + "reason": "SwitchTransformers sparse-layer construction and PretrainedConfig type checking are separate configuration bugs." }, { "left": "issue:45003", "right": "issue:45341", "accept": false, - "reason": "Modeling_utils sys.modules access vs testing_utils; unrelated." + "reason": "Unsafe sys.modules access in modeling_utils and a testing_utils bug do not share the same code path." }, { "left": "issue:44933", "right": "issue:46231", "accept": false, - "reason": "Image utils import exposure vs documentation parameter naming; unrelated." + "reason": "A missing import from image_utils and a documentation parameter-name mismatch are unrelated." }, { "left": "issue:45753", "right": "issue:46369", "accept": false, - "reason": "Specific image loop bug vs memory overflow in a different model; different symptoms and models." + "reason": "An infinite loop on a specific Qwen3_5 image and SAM3 LiteText memory overflow are distinct model/runtime failures." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "Video-model dependency crash vs testing_utils bug; unrelated." + "reason": "AutoProcessor crash without torchvision and a testing_utils bug are unrelated." }, { "left": "issue:43673", "right": "issue:46123", "accept": false, - "reason": "Chunked-prefill cache bug vs partial-batch mask pipeline bug; unrelated." + "reason": "Cache handling in GenerationMixin and MaskGenerationPipeline partial-batch dropping are different pipeline/generation bugs." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Tokenizer mapping None handling vs a specific model-loading failure; too speculative to merge." + "reason": "TOKENIZER_MAPPING_NAMES returning None and a transformers>=5.1.0 model-loading failure are separate lookup/loading problems." }, { "left": "issue:43986", "right": "issue:44484", "accept": false, - "reason": "Video-model dependency crash vs save_pretrained shard-size question; unrelated." + "reason": "Video-model loading without torchvision and save_pretrained shard size behavior are unrelated." }, { "left": "issue:43874", "right": "issue:44263", "accept": false, - "reason": "Image patch counting vs torch.split return handling; different components." + "reason": "Different GLM/VL components and different failures; no shared concrete bug." }, { "left": "issue:45440", "right": "issue:46172", "accept": false, - "reason": "Different DeepSeek versions and different code paths; not the same underlying issue." + "reason": "DeepSeekV3MoE divergence from remote implementation and deepseek-v4 tensor handling are different model-specific issues." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding_window bug vs testing_utils bug; unrelated." + "reason": "Qwen3-MoE sliding_window behavior and a testing_utils bug are unrelated." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "LayerNorm eps config vs RoPE implementation; different model internals." + "reason": "Missing rms_norm_eps in MLA q_a_layernorm and GLM-5 RoPE implementation are separate precision/architecture bugs." }, { "left": "issue:43986", "right": "issue:45372", "accept": false, - "reason": "Video processor dependency crash vs ReasoningEffort import error; unrelated." - }, - { - "left": "issue:43097", - "right": "issue:45588", - "accept": false, - "reason": "Embedding tie removal vs flash_attention sink-less crash; different subsystems." - }, - { - "left": "issue:43097", - "right": "issue:46050", - "accept": false, - "reason": "Embedding tie removal vs quantized model save/load corruption; unrelated." - }, - { - "left": "issue:43064", - "right": "issue:44315", - "accept": false, - "reason": "FSDP/PEFT optimizer state bug vs Liger kernel application bug; different training paths." + "reason": "A crash due to missing torchvision and a mistral_common import error affecting Gemma 4 processor loading are different dependency problems." } ] }, @@ -14905,7 +14847,7 @@ "issue:44077", "issue:44117", "issue:44162", - "issue:44222", + "issue:44315", "issue:44351", "issue:44368", "issue:44380", @@ -14917,11 +14859,12 @@ "issue:44938", "issue:45103", "issue:45375", - "issue:45478", "issue:45479", "issue:45584", + "issue:45588", "issue:45712", "issue:45753", + "issue:46050", "issue:46132", "issue:46172", "issue:46178", @@ -14930,11 +14873,14 @@ "issue:46421" ], "soft_pairs": [ + "issue:43097|issue:45588", + "issue:43097|issue:46050", "issue:38617|issue:42757", + "issue:43064|issue:44315", "issue:43502|issue:43992", - "issue:42890|issue:46298", - "issue:44410|issue:45375", "issue:44062|issue:44117", + "issue:44410|issue:45375", + "issue:42890|issue:46298", "issue:44038|issue:44410", "issue:38617|issue:44351", "issue:43097|issue:43873", @@ -14961,229 +14907,246 @@ "issue:44380|issue:46132", "issue:44380|issue:45479", "issue:44077|issue:46365", - "issue:43064|issue:44380", - "issue:44222|issue:45478" + "issue:43064|issue:44380" ], - "prepared_review_unit_hash": "1c88ab08286e9e828ab23022810778d7f3d0183cd491a534bde506cd26b1640f", + "prepared_review_unit_hash": "de9d4bfc1473316aacd6b3568ab8ae3c80be3aefdb954aa1259ed9022b13b361", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 36838, - "estimated_input_tokens": 9210, - "estimated_eval_tokens": 18676 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38093, + "estimated_input_tokens": 9524, + "estimated_eval_tokens": 19304 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous: import/export regressions, model-family-specific bugs, save/load issues, warning cleanup, and trainer/runtime problems. None of the soft pairs are close enough to be treated as the same underlying bug or a mergeable single PR.", - "confidence": 0.97, - "canonical_issue_reason": "Issue 43873 is the strongest representative by activity and breadth, but the cluster is not truly coherent; it only fits the quantization/offloading subtopic, not the other reports.", + "summary": "This cluster is mostly a false merge: the items span unrelated import errors, model-specific bugs, training regressions, and warnings. I found no soft-edge pair that looks like the same underlying fix/change, so all candidate links should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 43097 is the strongest representative only by prominence and external references; it is a broad Transformers v5 regression report, while the rest of the items are mostly unrelated bugs. There is no true duplicate canonical across the whole set.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43873 has the richest discussion and inbound references, making it the best available anchor, though it is not a good duplicate target for most of the cluster.", + "best_issue_reason": "43097 is the best single issue to stand in for the cluster because it has the broadest user impact and the most linkage, but it should not be treated as a duplicate target for the other issues.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Different bugs: removed tie_embeddings API vs flash_attention crash on s_aux=None." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Different areas: API removal regression vs broken quantized save/load after sinq." + }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Different missing imports from different modules; unrelated regressions." + "reason": "Both are import errors, but for different symbols and modules; not the same failure." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Different training problems: FSDP2/PEFT optimizer-state mismatch vs Liger Kernel not applied with model_init." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "One is a local_files_only network-leak bug, the other is a missing weight on UMT5 load." + "reason": "Different code paths: local_files_only still making requests vs missing embed_tokens.weight on load." }, { - "left": "issue:42890", - "right": "issue:46298", + "left": "issue:44062", + "right": "issue:44117", "accept": false, - "reason": "Test flakiness from missing seed vs a noisy logging message; not the same bug." + "reason": "Tokenizer construction TypeError vs tokenizer mapping returning None; not the same bug." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Both are Qwen-family issues, but they describe different config/model failures in different code paths." + "reason": "Different Qwen issues: missing projections in a model layer vs config field dropped by @strict." }, { - "left": "issue:44062", - "right": "issue:44117", + "left": "issue:42890", + "right": "issue:46298", "accept": false, - "reason": "Tokenizer AddedToken argument error vs tokenizer mapping None handling; distinct failures." + "reason": "Unrelated: flaky test seeding issue vs noisy logging message." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "Different Qwen model bugs; one is a v5 VL-Moe failure, the other missing projections in qwen3next." + "reason": "Different Qwen model bugs with different symptoms and code paths." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols and likely different fixes." + "reason": "Separate missing imports for different names; not a single underlying export bug." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "Tie-embeddings API removal vs quantization/offloading behavior; unrelated." + "reason": "API removal regression vs quantization/offloading behavior; too different to merge." }, { "left": "issue:44936", "right": "issue:46365", "accept": false, - "reason": "Trainer.evaluate failure after train is unrelated to importing Trainer with datasets installed." + "reason": "Trainer evaluation failure after train vs import failure for Trainer due to datasets; unrelated." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "AutoProcessor kwargs forwarding bug vs auto-docstring annotation crash; different subsystems." + "reason": "AutoProcessor kwargs forwarding vs docstring annotation crash; different subsystems." }, { "left": "issue:44380", "right": "issue:46421", "accept": false, - "reason": "GPT2 attention scaling with SDPA/FlashAttention vs cache lazy init recompilation; not the same issue." + "reason": "Attention backend config ignored vs cache lazy-init recompilation; not the same fix." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "Video-model AutoProcessor crash without torchvision vs tokenizer mapping None handling." + "reason": "AutoProcessor/torchvision crash vs tokenizer mapping None handling; unrelated." }, { "left": "issue:43824", "right": "issue:43986", "accept": false, - "reason": "Missing Qwen2.5-VL export vs torchvision-dependent AutoProcessor crash; distinct root causes." + "reason": "Import error for a specific VL class vs video-model crash without torchvision; different causes." }, { "left": "issue:44938", "right": "issue:46178", "accept": false, - "reason": "Python 3.14 load failure vs unsupported tensorflow_text backend mapping; different environment/backend problems." + "reason": "Python 3.14 load failure vs invalid BACKENDS_MAPPING backend error; unrelated." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type and an isolated reproduction of another issue are not the same concrete bug." + "reason": "Wrong tokenizer decoder type vs isolated reproduction of an older issue; not the same concrete bug." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing embed_tokens.weight on load vs max_shard_size behavior in save_pretrained." + "reason": "Missing weight during load vs max_shard_size default question; different concerns." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training blockers vs Bamba torch.compile/SDPA failure; unrelated model bugs." + "reason": "xLSTM training bugs vs Bamba compile/SDPA failure; different models and failures." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "bfloat16 precision warning vs compile-time SDPA failure; different code paths." + "reason": "PyTorch warning on bfloat16 compile vs model compile failure under SDPA; not duplicate." }, { "left": "issue:45712", "right": "issue:46298", "accept": false, - "reason": "Leftover dummy classes vs a noisy log message; no shared bug." + "reason": "Dummy class cleanup vs noisy logging; different maintenance issues." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Pipeline task error message vs saving Pipeline objects; same area but not the same defect." + "reason": "Pipeline error-message wording vs inability to save Pipeline objects; distinct problems." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Dequantized model save failure vs pipeline save_pretrained failure; different object types and failures." + "reason": "Dequantized save failure vs Pipeline save_pretrained issue; unrelated save paths." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "Unwanted PyTorch warning on compile vs tie_word_embeddings warning in a fine-tuning workflow; not the same bug." + "reason": "Different warnings with different triggers: bfloat16 precision warning vs tie_word_embeddings warning." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "Save failure for dequantized models vs infinite loop on a specific image; unrelated." + "reason": "Different model-specific bugs: save/dequantization failure vs infinite loop on a specific image." }, { "left": "issue:43097", "right": "issue:46298", "accept": false, - "reason": "API removal complaint vs noisy logging; unrelated." + "reason": "API removal regression vs noisy logging; no shared bug." }, { "left": "issue:43352", "right": "issue:44380", "accept": false, - "reason": "FlashAttention 2 unsupported for Nemotron vs GPT2 attention scaling ignored under SDPA/FlashAttention; different model/backend problems." + "reason": "Unsupported Flash Attention 2.0 for a model vs GPT2 attention scaling ignored under backends; different issues." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "Model/backend support issue vs Python-version compatibility failure; unrelated." + "reason": "Model-level attention backend support vs Python-version compatibility failure; unrelated." }, { "left": "issue:44162", "right": "issue:46172", "accept": false, - "reason": "ESM2 breakage vs deepseek-v4 tensor transpose bug; different models and code paths." + "reason": "ESM2 broken vs deepseek-v4 tensor transpose issue; different models and code paths." }, { "left": "issue:45584", "right": "issue:46365", "accept": false, - "reason": "Whisper empty-transcription generation bug vs Trainer import conflict with datasets; unrelated." + "reason": "Whisper empty transcription bug vs Trainer import blocked by datasets; unrelated." }, { - "left": "issue:44380", - "right": "issue:46132", + "left": "issue:44077", + "right": "issue:46365", "accept": false, - "reason": "GPT2 backend-specific config bug vs AttentionInterface.register behavior change; different functionality." + "reason": "patchtsmixer post_init policy vs Trainer import conflict; different problems." }, { "left": "issue:44380", - "right": "issue:45479", + "right": "issue:46132", "accept": false, - "reason": "Attention scaling ignored and zero-loss classification are unrelated bugs." + "reason": "Ignored attention scaling config vs AttentionInterface.register behavior change; not the same code-path bug." }, { - "left": "issue:44077", - "right": "issue:46365", + "left": "issue:44380", + "right": "issue:45479", "accept": false, - "reason": "patchtsmixer post_init validation vs Trainer import conflict; unrelated." + "reason": "Attention backend config bug vs zero-loss classification problem; unrelated." }, { - "left": "issue:44222", - "right": "issue:45478", + "left": "issue:43064", + "right": "issue:44380", "accept": false, - "reason": "FP8 save_pretrained for MoE vs Qwen3.5 MoE from_pretrained error; same family, but not enough evidence of the same concrete bug." + "reason": "Distributed optimizer-state corruption vs attention scaling backend bug; different subsystems and fixes." } ] }, @@ -15272,9 +15235,9 @@ "issue:43618|issue:43761", "issue:44483|issue:45290", "issue:43299|issue:43867", - "issue:44625|issue:46365", "issue:43066|issue:46209", "issue:36683|issue:43475", + "issue:44625|issue:46365", "issue:42994|issue:44655", "issue:43756|issue:44485", "issue:45276|issue:46077", @@ -15289,227 +15252,230 @@ "issue:43352|issue:45593", "issue:46077|issue:46082" ], - "prepared_review_unit_hash": "a5a1a576b4a1e5dd1ae08e73d7acecbf2bdd99e7a4bca4e3a38c34b7c6452552", + "prepared_review_unit_hash": "cb2a89f2931fcc35ec35a91483d27572086d280ebc09fe2e9ea9c6e5dbe71ae3", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36535, - "estimated_input_tokens": 9134, - "estimated_eval_tokens": 18524 + "serialized_chars": 36539, + "estimated_input_tokens": 9135, + "estimated_eval_tokens": 18526 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the soft pairs mostly connect unrelated regressions, import errors, model-specific bugs, and doc/backend issues. I found no pair that clearly shares the same underlying fix, so I would not consolidate them.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "Most items are unrelated standalone bugs, mostly in Transformers models, config, or packaging. The only plausible duplicate pair is the CLIP v5 regression around missing optional outputs in `CLIPVisionModel.forward`.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43761 is the clearest umbrella report for the CLIP v5 output regression: it explicitly ties the failure to `CLIPVisionModel.forward`, notes it is a v5 regression, and states the expected behavior when `output_hidden_states=True`.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a good canonical representative because the cluster spans many unrelated topics; the safest choice is to keep them separate.", + "best_issue_reason": "issue:43761 is the best representative issue among the plausible duplicates because it is explicit, regression-focused, and broad enough to cover the paired CLIP output symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36683", "right": "issue:43582", "accept": false, - "reason": "Gemma3 config vocab_size missing vs Apple Silicon allocator warmup TypeError are unrelated bugs in different code paths." + "reason": "Different bugs in unrelated areas: Gemma3 config missing `vocab_size` vs Apple Silicon allocator warmup TypeError." }, { "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "One is a chat completions API request rejection, the other is a chat template crash on batched input; different failures and fixes." + "reason": "Both relate to chat/VL workflows, but one is API request validation and the other is a chat-template crash on batched input; not the same code-path bug." }, { "left": "issue:42175", "right": "issue:46365", "accept": false, - "reason": "Packaging backend selection and Trainer import conflicts with datasets are separate dependency issues." + "reason": "Separate packaging/import problems with different symptoms and triggers." }, { "left": "issue:43208", "right": "issue:43526", "accept": false, - "reason": "xLSTM training bugs and BeitImageProcessorFast reduce_labels behavior are unrelated model/component bugs." + "reason": "Unrelated model/training bugs; no shared underlying change." }, { "left": "issue:42490", "right": "issue:43012", "accept": false, - "reason": "A local-path import warning and a bfloat16 compilation warning are different warnings with different causes." + "reason": "Both mention warnings, but one is a local-path import warning and the other is a PyTorch precision warning; different fixes." }, { "left": "issue:43012", "right": "issue:44683", "accept": false, - "reason": "PyTorch precision warning and compiled flex_attention failure are not the same bug or code path." + "reason": "Precision warning vs compiled flex-attention failure; distinct issues." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "An import error for layer_type_validation and unwanted network requests under local_files_only are unrelated." + "reason": "ImportError for a missing utility vs incorrect network access despite `local_files_only=True`." }, { "left": "issue:42757", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but for different symbols from different packages and different fixes." + "reason": "Both are import errors, but for different symbols, modules, and breakages; not one bug." }, { "left": "issue:45600", "right": "issue:46298", "accept": false, - "reason": "Broken auto_mappings references and a noisy logging message are not the same underlying issue." + "reason": "Breaking CI from removed config references vs a noisy logging message." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Both mention Qwen3-VL-Moe, but one is a loading regression and the other is fp8 inference support; not the same concrete bug." + "reason": "Same model family, but loading regression and FP8 inference support are different problems." }, { "left": "issue:45698", "right": "issue:46172", "accept": false, - "reason": "save_pretrained/from_pretrained custom module mismatch and a DeepSeek-v4 tensor transpose issue are unrelated." + "reason": "save/load custom-module corruption vs a DeepSeek tensor operation bug; unrelated." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "Missing CLIPOutput attentions and GPT2 attention scaling ignored under SDPA/FlashAttention are separate attention-path bugs." + "reason": "CLIP attentions regression vs GPT-2 attention scaling with SDPA/FlashAttention; different code paths." }, { "left": "issue:43618", "right": "issue:43761", - "accept": false, - "reason": "Both are CLIP regressions, but one drops attentions and the other drops hidden_states; distinct outputs and likely distinct fixes." + "accept": true, + "reason": "Both point to the same CLIP v5 forward-output regression where optional outputs are no longer populated; one PR could plausibly restore both attentions and hidden states." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "Chat completions request handling and apply_chat_template tokenization crash are different layers of the stack." + "reason": "Both involve chat-related APIs, but the failures are different: endpoint request rejection vs template crash with tool-call assistant messages." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Qwen3.5 save_pretrained corruption and a sorted-state_dict load error are both serialization-related but not clearly the same bug." - }, - { - "left": "issue:44625", - "right": "issue:46365", - "accept": false, - "reason": "Qwen3.5 num_labels propagation and Trainer import resolution are unrelated." + "reason": "Different model-loading failures with no clear shared root cause." }, { "left": "issue:43066", "right": "issue:46209", "accept": false, - "reason": "Tokenizer decoder type mismatch and FP8 inference failure on B200 are completely different problems." + "reason": "Tokenizer decoder regression vs FP8 inference on B200; unrelated." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Gemma3 vocab_size missing and Sam3 video output missing fpn_position_embeddings are unrelated attribute errors." + "reason": "Different missing-attribute bugs in different models/configs." + }, + { + "left": "issue:44625", + "right": "issue:46365", + "accept": false, + "reason": "Qwen3.5 config propagation bug vs Trainer import issue; unrelated." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving failure and Pipeline.save_pretrained failure are different save paths and likely different bugs." + "reason": "Both involve saving, but quantized model saving and Pipeline save_pretrained are distinct behaviors and fixes." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "RoPE layer dropping in Smollm3 and GLM-5 RoPE implementation concerns are model-specific and not the same concrete change." + "reason": "Both mention RoPE, but they concern different models and different defects." }, { "left": "issue:45276", "right": "issue:46077", "accept": false, - "reason": "Gemma4 embedding resize propagation and Gemma4 attention-mask construction are separate bugs in the same model family." + "reason": "Same Gemma4 family, but token embedding resizing and bidirectional attention masks are separate bugs." }, { "left": "issue:43901", "right": "issue:44655", "accept": false, - "reason": "Pipeline docs mentioning return_all_scores and pipeline persistence failure are unrelated." + "reason": "Docs mismatch for `return_all_scores` vs pipeline serialization; unrelated." }, { "left": "issue:36010", "right": "issue:42757", "accept": false, - "reason": "Importing GenerationMixin and importing is_offline_mode are unrelated import breakages from different modules." + "reason": "Both are import errors, but for different packages and symbols; not the same underlying breakage." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "FlashAttention 2 support missing for NemotronHForCausalLM and invalid GLM-5 GenerationConfig are different model/runtime issues." + "reason": "Flash Attention support gap vs invalid GenerationConfig error; different code paths." }, { "left": "issue:43611", "right": "issue:44655", "accept": false, - "reason": "base_model_prefix loading regression and Pipeline.save_pretrained failure are different code paths." + "reason": "Model loading with `base_model_prefix` vs saving Pipeline objects; unrelated." }, { "left": "issue:36683", "right": "issue:44077", "accept": false, - "reason": "Gemma3 config attribute error and patchtsmixer post_init validation are unrelated." + "reason": "Gemma3 config bug vs a patchtsmixer post_init validation rule; different issues." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "A load failure with sorted state_dict and a Qwen3.5 save_pretrained regression are both serialization-related but not clearly the same bug." + "reason": "State-dict sorting/load failure and a Qwen3.5 save_pretrained regression are both serialization-adjacent but not the same bug." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Python version compatibility failures on different versions are not the same underlying issue." + "reason": "Both are Python-version compatibility reports, but for different versions and likely different breakpoints." }, { "left": "issue:44625", "right": "issue:46077", "accept": false, - "reason": "Qwen3.5 label propagation and Gemma4 attention-mask behavior are unrelated." + "reason": "Qwen config propagation bug vs Gemma4 attention mask bug; unrelated." }, { "left": "issue:43352", "right": "issue:45593", "accept": false, - "reason": "FlashAttention support in NemotronHForCausalLM and D-FINE auxiliary-loss behavior are unrelated." + "reason": "Flash Attention not supported for Nemotron vs D-FINE auxiliary-loss behavior; unrelated." }, { "left": "issue:46077", "right": "issue:46082", "accept": false, - "reason": "Gemma4 attention-mask bug and LlamaConfig head_dim validation are different model/config issues." + "reason": "Different Gemma/Llama config behaviors and different failure modes." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet. The only accepted pair, issue:43618 and issue:43761, is a plausible single CLIPVisionModel.forward v5 regression affecting optional outputs, and the other soft-edge rejections stay conservative rather than overclaiming duplicate status. The overall characterization of the remaining items as unrelated standalone bugs is supported by their titles and descriptions." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -15602,239 +15568,239 @@ "issue:44360|issue:45440", "issue:43502|issue:44062", "issue:42890|issue:44964", - "issue:42890|issue:44112", "issue:42890|issue:45600", + "issue:42890|issue:44112", "issue:43723|issue:44556", "issue:44387|issue:44556", "issue:44222|issue:45125", "issue:43010|issue:43065" ], - "prepared_review_unit_hash": "b97d84826f76645dddbc4757c6e60766e20ccc64073e74d982cd592d3984518f", + "prepared_review_unit_hash": "0f58f8d9bcd99b4302e16e3a6d46af2a95d9320745371ef344f490edd90f00d7", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 36819, - "estimated_input_tokens": 9205, - "estimated_eval_tokens": 18666 + "serialized_chars": 36822, + "estimated_input_tokens": 9206, + "estimated_eval_tokens": 18668 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most items are distinct regression reports across different models, configs, and runtime paths. Only the import-performance pair looks like the same underlying issue; the rest should stay separate.", - "confidence": 0.32, - "canonical_issue_reason": "Issue 44246 is the best representative bug report: it is broad, user-facing, and matches the import-time regression theme better than the more implementation-specific 41669.", + "summary": "Mostly a grab bag of unrelated Transformers issues. The only plausible duplicate pair is the import-time slowdown report (41669/44246); the rest are distinct model-, config-, loading-, training-, or test-specific bugs.", + "confidence": 0.82, + "canonical_issue_reason": "issue:41669 is the most concrete root-cause report in the set: it explains a specific import-time performance regression from `import *` usage, and it best matches the only plausible duplicate report here (issue:44246).", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44246 is the strongest standalone issue in the set because it describes the general user symptom ('import transformers takes long sometimes') with the broadest impact.", + "best_issue_reason": "issue:41669 is the best canonical issue for the small import-performance subcluster because it is specific, actionable, and closer to a root cause than the more generic symptom report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36683", "right": "issue:43525", "accept": false, - "reason": "Both are missing-config-attribute reports, but they affect different models and different fields, so they are not the same bug." + "reason": "Different model configs and different missing attributes (`vocab_size` vs `pad_token_id`)." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "One is about CLIPOutput attentions, the other about Qwen3.5 save_pretrained checkpoint keys; different code paths." + "reason": "Different bugs: missing `attentions` assignment in CLIP outputs vs Qwen3.5 save-pretrained checkpoint corruption." }, { "left": "issue:36683", "right": "issue:44625", "accept": false, - "reason": "Both involve config attributes, but one is missing vocab_size and the other is num_labels propagation; not a duplicate." + "reason": "Both are config-related, but they concern different fields and model families (`vocab_size` vs `num_labels`)." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "Different models and different load failures; both are from_pretrained-related, but not the same concrete bug." + "reason": "Different models and different loading failures (`UMT5Encoder` missing weights vs Qwen3.5 MoE from_pretrained error)." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Optimizer/FSDP state issues and a cache-strategy typo are unrelated." + "reason": "Unrelated: distributed trainer optimizer-state bug vs a typo in cache strategies." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "One is a no_grad decoration request, the other is a broad import regression after upgrade; no shared concrete bug." + "reason": "Different topics: `update()` no-grad decoration vs broad import errors after upgrade." }, { "left": "issue:44075", "right": "issue:46365", "accept": false, - "reason": "SGD argument handling and importing Trainer with datasets installed are unrelated." + "reason": "Optimizer argument handling is unrelated to an import failure when `datasets` is installed." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Both are import-time failures, but one is about compiled Python files and the other is a torch.jit parsing issue on a specific model." + "reason": "Both involve import/parse failures, but one is a generic compiled-file error and the other is a Python 3.13 `torch.jit.script` parsing issue." }, { "left": "issue:43867", "right": "issue:46202", "accept": false, - "reason": "State-dict ordering/load failure and Gemma pipeline roundtrip token issues are different problems." + "reason": "Different loading/round-trip failures for different pipelines; no shared root cause is evident." }, { "left": "issue:43867", "right": "issue:45375", "accept": false, - "reason": "Different Qwen/Gemma-related serialization issues with different missing fields and failure modes." + "reason": "Different Qwen/Gemma loading paths and different missing pieces (`state_dict` ordering vs a dropped vision config field)." }, { "left": "issue:44938", "right": "issue:45071", "accept": false, - "reason": "Python 3.14 import failure and PretrainedConfig type-checking breakage are not the same underlying bug." + "reason": "Python 3.14 import/load breakage is not the same as `PretrainedConfig` type-checking regression." }, { "left": "issue:36683", "right": "issue:43064", "accept": false, - "reason": "Missing config attributes in a model config and FSDP2 optimizer-state corruption are unrelated." + "reason": "Different subsystems: config attribute error vs trainer/FSDP/PEFT optimizer-state corruption." }, { "left": "issue:41669", "right": "issue:44246", "accept": true, - "reason": "Both describe the same import-time performance regression; 41669 points to the likely root cause of the slow 'import transformers' symptom." + "reason": "Both describe the same import-time slowdown; 41669 identifies a concrete likely cause, while 44246 is the symptom report." }, { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "A flaky SamHQ integration test and an xdist race in patched testing utils are different test failures." + "reason": "Separate test failures: missing `set_seed()` in SamHQ integration tests vs xdist race on `captured_info.txt`." }, { "left": "issue:43010", "right": "issue:45850", "accept": false, - "reason": "A specific no_grad/decorator request is not the same as a broad 'breaks too many models' complaint." + "reason": "One is a specific cache/layer update bug; the other is an overly broad complaint about many models breaking in v5." }, { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Same general area (Qwen3.5 load), but different missing-key failures on different models/paths." + "reason": "Same general area of model loading, but different model families and no shared concrete failure mode." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "Both involve save/load regressions, but the affected keys and model structures are different." + "reason": "Different bugs in different Qwen3.5 save/load flows; not the same underlying issue." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "No_grad/cache update behavior and Python-version compatibility are unrelated." + "reason": "Unrelated regression in `update()` handling vs a Python-version compatibility problem." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Both mention tensor parallelism, but one is a training bug and the other is a load-time OOM for a specific model." + "reason": "Tensor-parallel training bug vs tensor-parallel load OOM are related only at a broad subsystem level, not the same concrete defect." }, { "left": "issue:43618", "right": "issue:44077", "accept": false, - "reason": "CLIPOutput attention fields and patchtsmixer post_init are unrelated." + "reason": "Different models and different issues: CLIP attentions output vs PatchTSMixer `post_init` API policy." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "Torch dependency import error and tokenizer-loading regression are different failures." + "reason": "One is a missing-PyTorch import error for AutoModel; the other is a tokenizer-loading regression." }, { "left": "issue:43618", "right": "issue:46285", "accept": false, - "reason": "Different CLIP structure changes affecting different classes and fields." + "reason": "Different CLIP-related regressions with different symptoms; one is missing attentions, the other is a structural mismatch between CLIP text model classes." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "CLIPOutput attentions and LwDetrImageLoss AMP/CUDA crash are unrelated." + "reason": "No shared root cause: CLIP output regression vs float16 AMP CUDA crash in another loss path." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "A missing ReLU in a discussion issue and a DeepSeekV3 implementation divergence are not the same bug." + "reason": "Different model/code-path problems: DSA indexer ReLU discussion vs DeepSeekV3MoE parity divergence." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "Local-files-only API requests and AddedToken keyword conflicts are unrelated tokenizer/loading issues." + "reason": "Unrelated: offline-loading API requests vs tokenizer `AddedToken` keyword collision." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "SamHQ test seeding flakiness and Phi-4 multimodal loading are unrelated." + "reason": "Different failures in different tests/models; no evidence they are the same bug." }, { "left": "issue:42890", - "right": "issue:44112", + "right": "issue:45600", "accept": false, - "reason": "Test seed flakiness and GraniteSpeech device-override CI failure are different test bugs." + "reason": "CI test flakiness from missing seeds is unrelated to stale `auto_mappings.py` config references." }, { "left": "issue:42890", - "right": "issue:45600", + "right": "issue:44112", "accept": false, - "reason": "SamHQ test instability and stale auto_mappings references are unrelated." + "reason": "Both are CI/test issues, but they involve different tests and different failure causes." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Both are loading regressions after v5, but one is tokenizer.from_pretrained and the other is checkpoint reload compatibility; too broad to merge." + "reason": "Related only at a very high level (`from_pretrained`/reload regressions); the concrete failures differ." }, { "left": "issue:44387", "right": "issue:44556", "accept": false, - "reason": "Int4 quantization OOM and checkpoint reload failure are different problems." + "reason": "Different problems: int4 quantization memory growth vs checkpoint reload compatibility." }, { "left": "issue:44222", "right": "issue:45125", "accept": false, - "reason": "FP8 save_pretrained behavior and missing tensor-parallel plan are different code paths." + "reason": "Same model family, but different concrete defects: FP8 save-pretrained behavior vs missing tensor-parallel plan." }, { "left": "issue:43010", "right": "issue:43065", "accept": false, - "reason": "Cache/layer no_grad decoration and a dummy Conv2d in Sam3PixelDecoder are unrelated." + "reason": "Completely different bugs: cache/layer update semantics vs a dummy Conv2d in Sam3PixelDecoder." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative on the soft-edge pairs. The only accepted pair, issue:41669 with issue:44246, is a plausible same regression family (import-time slowdown). The rest are reasonably kept separate given their distinct model paths, failure modes, or runtime contexts." + "feedback": "The summary is grounded in the packet and stays conservative. The one accepted pair, issue:41669/issue:44246, is the only clearly supported duplicate-like match, and the remaining soft-edge rejections do not overclaim shared root causes." }, "evaluator_used": true, "retried": false, @@ -15932,8 +15898,8 @@ "issue:44961|issue:45468", "issue:43576|issue:44938", "issue:44222|issue:44230", - "issue:44485|issue:44961", "issue:44908|issue:44961", + "issue:44485|issue:44961", "issue:46143|issue:46144", "issue:41628|issue:42548", "issue:36296|issue:43901", @@ -15942,20 +15908,20 @@ "issue:43673|issue:44492", "issue:44623|issue:45230" ], - "prepared_review_unit_hash": "fb37283afa56ef5728214050fcb562fc2c2ca68e3f7ea37216945a878c4ab1a0", + "prepared_review_unit_hash": "b62d4a07638a52c2d8b554da8524cf1736dba8988c3b0734cfec74a40df71634", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 36, - "serialized_chars": 37695, + "serialized_chars": 37696, "estimated_input_tokens": 9424, "estimated_eval_tokens": 19104 }, @@ -15964,228 +15930,228 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: it mixes v5 regressions, import failures, model-loading bugs, docs issues, and performance complaints. I don\u2019t see any soft pair that clearly describes the same underlying bug or change, so I\u2019d keep them separate.", - "confidence": 0.76, - "canonical_issue_reason": "issue:42548 is the most central high-impact regression here: a core API import failure involving `PreTrainedModel`, which is broader and more representative than the specialized model/config or docs items.", + "summary": "This cluster is mostly a set of unrelated closed issues that happen to share broad Transformers/v5 or import/loading themes. I found no soft pair that looks like the same underlying bug or change, so all soft edges are rejected.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43576 is the best anchor for the cluster because it is a concrete v5 regression, has the most similarity links, and is representative of the broad \u201cTransformers v5 broke something\u201d subgroup.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42548 is the best single representative for this mixed cluster because it is a concrete, foundational import break with broad downstream impact and clear triage value.", + "best_issue_reason": "issue:43576 is the strongest representative issue in this set: it is specific enough to be actionable, yet broad enough to sit near several other v5-related loading/import regressions.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43576", "right": "issue:46132", "accept": false, - "reason": "Different subsystems and symptoms: a broken `transformers env` command vs altered `AttentionInterface.register` behavior." + "reason": "Both touch attention-related behavior in Transformers v5, but one is a broken CLI/env command and the other is a change in AttentionInterface.register semantics; not the same bug." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "Unrelated bugs: offline/local file fetching vs GLM-5 RoPE implementation." + "reason": "One is about unwanted network requests despite local_files_only=True; the other is a GLM-5 RoPE implementation issue. Different code paths and symptoms." }, { "left": "issue:42548", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but they concern different symbols and likely different code paths (`PreTrainedModel` vs `HybridCache`)." + "reason": "Both are import errors, but for different symbols/modules (PreTrainedModel vs HybridCache). These are distinct API breakages." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "Same model family, but one is specifically about missing `_tp_plan` for tensor parallelism while the other is a generic `from_pretrained` failure; not enough to treat as the same bug." + "reason": "Both concern Qwen3.5 MoE, but one is missing tensor-parallel planning and the other is a from_pretrained error in transformers>=5.4.0. Too different to merge." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Completely different problems: compiled Python file failures vs performance regression after upgrade." + "reason": "Both mention loading/performance after upgrade, but one is compiled Python file errors and the other is data filtering slowdown. Not the same underlying defect." }, { "left": "issue:30333", "right": "issue:33453", "accept": false, - "reason": "Different areas: MLFlow reporting on failed jobs vs tokenizer loading regression." + "reason": "MLFlow reporting after failed training and tokenizer loading regression are unrelated subsystems and failure modes." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Both involve Qwen3.5 Moe, but one is tensor-parallel plan handling and the other is a config field being dropped; distinct changes." + "reason": "Both are Qwen3.5 MoE-related, but one is tensor-parallel support and the other is a config field being dropped by @strict. Different bugs." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Both are generation/cache-related, but one is about `@torch.no_grad` decoration and the other is about `_update_model_kwargs_for_generation` after `sync_gpus`; not the same concrete bug." + "reason": "One is about no_grad decoration on cache/layer update methods; the other is generation kwargs handling after sync_gpus. Similar area, but not the same code-path problem." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Offline download behavior and FP8 save/pretrained issues are unrelated." + "reason": "One is an unexpected network request issue; the other is an FP8 save_pretrained moe bug. Different operations and failure points." }, { "left": "issue:30990", "right": "issue:31795", "accept": false, - "reason": "Stuck model loading vs documentation confusion; different issue types and scopes." + "reason": "Sentence Transformers loading hangs versus documentation confusion for input_ids/past_key_values. One is a runtime bug, the other is docs wording." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Both are loading regressions, but they affect different model/version combinations and do not clearly share a code-path." + "reason": "Both involve v5 model loading regressions, but they affect different model families and likely different serialization assumptions." }, { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "No meaningful overlap; local-files-only networking bug vs an unrelated placeholder issue." + "reason": "The local_files_only network-request bug is unrelated to an issue titled only 'racoon'; there is no evidence of the same defect." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "Different failures: FineGrainedFP8 accuracy issue vs missing image-processor method." + "reason": "Both are accuracy/processor-related, but one is FineGrainedFP8 accuracy and the other is a missing image-processor method causing AttributeError." }, { "left": "issue:30333", "right": "issue:36296", "accept": false, - "reason": "MLFlow reporting bug and tensor-parallel training bug are unrelated." + "reason": "MLFlow reporting on failure and tensor parallel training are separate concerns with no shared code path indicated." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Compiled-file loading errors and `PretrainedConfig` type-checking regression are different issues." + "reason": "Compiled file loading errors and PretrainedConfig type-checking breakage are unrelated kinds of regressions." }, { "left": "issue:43827", "right": "issue:44961", "accept": false, - "reason": "Docs referencing removed `pipeline()` API vs unrelated placeholder issue." + "reason": "Docs still referencing pipeline() and a one-word issue title do not indicate the same underlying bug." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Different failure modes: compiled Python file errors vs pipeline `save_pretrained` support." + "reason": "Compiled Python file errors and inability to save Pipeline objects are different problems in different parts of the API." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Tokenizer-loading regression and Llama 3.2 Vision model-loading failure are related only at a very broad level, not the same bug." + "reason": "Tokenizer loading regression and Llama 3.2 Vision model loading breakage are distinct loading failures affecting different components." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Liger Kernel application issue and DSA indexer discussion are unrelated." + "reason": "Liger Kernel not applied in model_init and a DSA indexer lacking ReLU are unrelated implementation issues." }, { "left": "issue:27957", "right": "issue:33453", "accept": false, - "reason": "Flash Attention 2 support for XLM-Roberta vs tokenizer-loading regression are different problems." + "reason": "Flash Attention 2 support for XLMRoberta and a tokenizer loading regression are different feature/regression areas." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "Import error from `huggingface_hub` and broken `transformers env` are distinct issues." + "reason": "Importing is_offline_mode from huggingface_hub and a broken transformers env command are both v5-adjacent, but they are different missing-API vs CLI issues." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "CLI/environment command breakage is unrelated to `PretrainedConfig` type checking." + "reason": "A broken env command and a PretrainedConfig typing regression are unrelated user-facing regressions." }, { "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "Different failures: broken env command vs missing Qwen2.5-VL model class import." + "reason": "The env command failure and a missing Qwen2.5-VL class import are different import/CLI breakages." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "Documentation confusion about `forward` args vs missing `layer_type_validation` import." + "reason": "Documentation confusion for model.forward and a missing layer_type_validation import are not the same bug." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "No duplicate signal: placeholder issue vs Gemma-4 audio positional-encoding bug." + "reason": "The 'racoon' issue provides no evidence of overlap with Gemma-4 audio positional encoding." }, { "left": "issue:43576", "right": "issue:44938", "accept": false, - "reason": "Broken CLI command and Python 3.14 load failure are different regression reports." + "reason": "A broken transformers env command and Python 3.14 load failures are both environment-related, but the concrete failures differ." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Same broad FP8/Moe area, but one is about saving pretrained weights and the other is about inference support; not the same concrete change." + "reason": "Both mention FP8, but one is saving moe checkpoints and the other is inference support for specific Qwen3-VL/Qwen3.5 models." }, { - "left": "issue:44485", + "left": "issue:44908", "right": "issue:44961", "accept": false, - "reason": "GLM-5 RoPE implementation issue and placeholder issue are unrelated." + "reason": "Inverse_sqrt scheduler kwargs handling and the unrelated 'racoon' issue are not the same defect." }, { - "left": "issue:44908", + "left": "issue:44485", "right": "issue:44961", "accept": false, - "reason": "Scheduler kwargs bug vs unrelated placeholder issue." + "reason": "GLM-5 RoPE implementation and an unrelated placeholder title do not match." }, { "left": "issue:46143", "right": "issue:46144", "accept": false, - "reason": "Related RoFormer area, but one is about missing `**kwargs` propagation and the other about attention-interface usage; not clearly the same bug." + "reason": "One is about **kwargs not being propagated in RoFormer methods; the other is that RoFormer attention does not use the attention interface. Related subsystem, but distinct bugs." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "Both are import errors, but they target different public symbols and likely different fixes." + "reason": "AutoImageProcessor import failure and PreTrainedModel import failure are separate missing exports." }, { "left": "issue:36296", "right": "issue:43901", "accept": false, - "reason": "Tensor-parallel training bug and docs mentioning `return_all_scores` are unrelated." + "reason": "Tensor parallel training bug and text-classification docs mentioning return_all_scores are unrelated." }, { "left": "issue:44315", "right": "issue:46209", "accept": false, - "reason": "Liger Kernel/model-init issue vs FP8 inference on B200; different code paths." + "reason": "Liger Kernel application in model_init and FP8 inference on B200 are different performance/runtime issues." }, { "left": "issue:36296", "right": "issue:43526", "accept": false, - "reason": "Tensor-parallel training bug and BeitImageProcessorFast label reduction bug are unrelated." + "reason": "Tensor parallel training and BeitImageProcessorFast reduce_labels behavior are different areas with no shared failing path." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Runtime cache-missing regression vs a typo in cache strategies; not the same underlying bug." + "reason": "Generation cache missing during chunked_prefill and a typo in cache strategies are not the same bug." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "Specific processor `save_pretrained` file-missing bug vs a generic bug report placeholder." + "reason": "Missing files on processor.save_pretrained and a generic 'Bug report' have no evidence of being the same issue." } ] }, @@ -16302,232 +16268,232 @@ "issue:44273|issue:44908", "issue:43576|issue:46218" ], - "prepared_review_unit_hash": "e9f972fdb0ca00d78a0582deadb097709c98af9e6f3a6365986560231ade560c", + "prepared_review_unit_hash": "61872d20b5b7ecafbe5b708a2081f3d2c850dcc65314b2a3e1989c14d7dd7445", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 42, - "serialized_chars": 39678, - "estimated_input_tokens": 9920, - "estimated_eval_tokens": 20096 + "serialized_chars": 39687, + "estimated_input_tokens": 9922, + "estimated_eval_tokens": 20100 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is heterogeneous: most items are unrelated v5 regressions, model-specific bugs, import/load failures, or docs/test issues. No pull requests are present, and none of the soft pairs look like true duplicates of the same concrete bug/change.", - "confidence": 0.74, - "canonical_issue_reason": "Issue 45278 is the closest thing to a cluster anchor because it captures a broad post-upgrade regression pattern, but the overall set is too mixed for a strong canonical duplicate.", + "summary": "These are mostly unrelated issue reports. The only soft match I\u2019d accept is the import-time slowdown pair, which appears to describe the same performance regression from different angles.", + "confidence": 0.81, + "canonical_issue_reason": "issue:41669 is the most concrete, actionable report for the import slowdown, with a specific suspected cause and clear performance impact; issue:44246 is the broader symptom report.", "canonical_pr_reason": null, - "best_issue_reason": "45278 is the most usable issue anchor among the group: broad, concrete, and aligned with several upgrade-regression complaints, though still not a precise duplicate target.", + "best_issue_reason": "issue:41669 best represents the underlying bug because it identifies the likely root cause and the exact regression surface.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45362", "right": "issue:45850", "accept": false, - "reason": "Different scope: a specific Qwen3.5 chat crash vs a vague broad complaint about v5 breaking many models." + "reason": "Specific Qwen3.5 chat crash vs a broad complaint that many models break; no shared concrete code path." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading problems and a TensorFlow backend/import issue are different failures." + "reason": "Lazy loading malfunction and TensorFlow backend installation error are different failures." }, { "left": "issue:44230", "right": "issue:46095", "accept": false, - "reason": "FP8 support for Qwen3-VL/Qwen3.5 MoE is unrelated to a DeepSeek weight-conversion question." + "reason": "FP8 inference support request and DeepSeek weight-conversion question are unrelated." }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "Qwen3-VL-Moe regression and missing _tp_plan for Qwen3.5 MoE are different code-path issues." + "reason": "Same model family, but one is a Qwen3-VL-Moe compatibility bug and the other is a tensor-parallel planning gap." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "A Qwen3-VL hallucination/error and a PretrainedConfig type-checking regression are unrelated." + "reason": "Qwen3-VL bbox output problem vs a PretrainedConfig type-checking regression; different bugs." }, { "left": "issue:44273", "right": "issue:45003", "accept": false, - "reason": "Lazy loading is not the same as unsafe sys.modules access / import-time behavior in modeling_utils." + "reason": "Both touch import/lazy-loading behavior, but one is a lazy-loading correctness issue and the other is unsafe sys.modules access." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "Broken `transformers env` command and checkpoint reload incompatibility are distinct issues." + "reason": "Broken env command and checkpoint reload incompatibility are separate v5 regressions." }, { "left": "issue:44393", "right": "issue:45753", "accept": false, - "reason": "Different model-specific bugs: Qwen3-VL bbox output vs Qwen3_5 image loop." + "reason": "BBox hallucination/error and infinite-loop-on-one-image are not the same symptom or code path." }, { "left": "issue:43525", "right": "issue:45850", "accept": false, - "reason": "Llama4Config pad_token_id missing is a concrete config bug, not the same as a broad v5-breakage complaint." + "reason": "A specific Llama4Config attribute error is not the same as a generic claim that many models break." }, { "left": "issue:41669", "right": "issue:43673", "accept": false, - "reason": "Import-star slowdown and GenerationMixin cache missing during chunked prefill are different performance/behavior problems." + "reason": "Import-time slowdown from import * usage is different from a missing GenerationMixin cache during chunked_prefill." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both are upgrade-related, but one is checkpoint reload compatibility and the other is meta-initialization with remote-code checkpoints." + "reason": "Both are v5 reload incompatibilities, but they involve different loading scenarios and model-specific failures." }, { - "left": "issue:43576", + "left": "issue:43572", "right": "issue:45850", "accept": false, - "reason": "A broken CLI command and a vague generic v5 complaint are not the same bug." + "reason": "StableLmConfig pad_token_idx regression is too specific to match the broad breakage complaint." }, { "left": "issue:44492", "right": "issue:45850", "accept": false, - "reason": "A cache-strategy typo and a broad breakage complaint are unrelated." + "reason": "A typo in cache-strategy docs/metadata is unrelated to a general model-breakage report." }, { "left": "issue:45278", "right": "issue:45850", "accept": false, - "reason": "Both are broad upgrade regressions, but one is import errors and the other is general model breakage; not the same underlying bug." + "reason": "General import errors after upgrade and a broad 'many models break' complaint are not the same concrete bug." }, { "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "TF32 settings guidance and cache-strategy typo are unrelated documentation/config issues." + "reason": "TF32 API guidance and a cache-strategy typo are unrelated." }, { "left": "issue:41669", "right": "issue:46218", "accept": false, - "reason": "Import slowdown from lazy loading and a Qwen3_5 import failure are different import-path problems." + "reason": "Import slowdown and a specific missing Qwen3_5ForCausalLM import are different failures." }, { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "sam3 tracker warning and Liger Kernel not applied in model_init are different model-init issues." + "reason": "Wrong model-type warning and Liger Kernel not being applied with model_init are unrelated." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "Missing set_seed in tests and inverse_sqrt scheduler kwargs being ignored are unrelated." + "reason": "Missing set_seed in tests and inverse_sqrt scheduler kwargs handling are unrelated." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "Tokenizer class mismatch and a generic bug report are not enough to indicate the same bug." + "reason": "Tokenizer save metadata mismatch is not the same as a generic bug report." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "Position_ids buffer handling and a generic bug report are unrelated." + "reason": "position_ids buffer access error is unrelated to the generic bug report title." }, { "left": "issue:44246", "right": "issue:44273", "accept": false, - "reason": "Import slowdown and lazy loading malfunction are related broadly, but not the same concrete bug." + "reason": "Both concern import/lazy-loading behavior, but one is import slowness and the other is a lazy-loading correctness issue; not the same concrete bug." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Nemotron Flash Attention 2 support and Qwen3-VL 2D bbox errors are different model-specific issues." + "reason": "Flash Attention 2 support warning and Qwen3-VL bbox output failure are different issues." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "CLI environment command failure and scheduler kwargs handling are unrelated." + "reason": "Broken env command and scheduler kwargs handling are unrelated." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python 3.9 compatibility and Qwen3-VL bbox hallucination are unrelated." + "reason": "Python version support breakage and Qwen3-VL output errors are unrelated." }, { - "left": "issue:44908", - "right": "issue:45071", + "left": "issue:44704", + "right": "issue:44908", "accept": false, - "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are different subsystems." + "reason": "AutoProcessor kwargs passthrough and inverse_sqrt scheduler kwargs are different subsystems." }, { - "left": "issue:44704", - "right": "issue:44908", + "left": "issue:44908", + "right": "issue:45071", "accept": false, - "reason": "AutoProcessor kwargs passthrough and inverse_sqrt scheduler kwargs are different bugs despite both involving kwargs." + "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are unrelated." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star performance regression and missing `HybridCache` import are different import-related issues." + "reason": "Import-time performance regression is not the same as failing to import HybridCache." }, { "left": "issue:44492", "right": "issue:46328", "accept": false, - "reason": "Cache-strategy typo and SequenceFeatureExtractor.pad conversion inefficiency are unrelated." + "reason": "Cache-strategy typo and SequenceFeatureExtractor.pad performance issue are unrelated." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Import slowdown and a testing_utils bug are different code paths." + "reason": "Import slowdown regression and a testing_utils bug are different code paths." }, { "left": "issue:43976", "right": "issue:45850", "accept": false, - "reason": "Python version compatibility and vague v5 breakage are not the same bug." + "reason": "Python 3.9/3.10 support mismatch is not the same as a broad post-v5 breakage complaint." }, { "left": "issue:44393", "right": "issue:45478", "accept": false, - "reason": "Qwen3-VL bbox output issue and Qwen3.5 MoE from_pretrained error are different model failures." + "reason": "Qwen3-VL bbox hallucination/error and Qwen3.5 Moe from_pretrained failure are different model bugs." }, { "left": "issue:43976", "right": "issue:46218", "accept": false, - "reason": "Python 3.9 compatibility and Qwen3_5 import failure are distinct issues." + "reason": "Python-version compatibility and a Qwen3_5ForCausalLM import failure are unrelated." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Optional post_init policy and `_get_tied_weight_keys` AttributeError are unrelated." + "reason": "Optional post_init policy and _get_tied_weight_keys crashing on list input are unrelated." }, { "left": "issue:43827", @@ -16539,49 +16505,31 @@ "left": "issue:31795", "right": "issue:33453", "accept": false, - "reason": "Forward-doc confusion and tokenizer loading regression do not describe the same bug." + "reason": "Documentation confusion about forward args and a tokenizer-loading regression are unrelated." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Model_init applying Liger Kernel and flash_attention_3 training collapse are different problems." + "reason": "Liger Kernel not applied with model_init and degenerate training under flash_attention_3 are different bugs." }, { "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Broken `transformers env` command and lazy loading malfunction are not the same issue." + "reason": "Broken env command and lazy loading not working are different issues, even if both are v5 regressions." }, { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "A generic bug report and a Qwen3.5 MoE from_pretrained error are not the same concrete bug." + "reason": "A placeholder 'Bug report' title does not establish the same concrete failure as the Qwen3.5 Moe from_pretrained error." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer/ReLU discussion and Gemma4Audio positional encoding bug are unrelated." - }, - { - "left": "issue:44655", - "right": "issue:44861", - "accept": false, - "reason": "Pipeline save_pretrained issue and tied-weight-key AttributeError are different behaviors." - }, - { - "left": "issue:44273", - "right": "issue:44908", - "accept": false, - "reason": "Lazy loading malfunction and inverse_sqrt scheduler kwargs are unrelated." - }, - { - "left": "issue:43576", - "right": "issue:46218", - "accept": false, - "reason": "The `transformers env` command bug and Qwen3_5 import failure are different failures." + "reason": "DSA indexer ReLU discussion and Gemma4AudioRelPositionalEncoding are unrelated." } ] }, @@ -16702,20 +16650,20 @@ "issue:37934|issue:43668", "issue:42936|issue:43668" ], - "prepared_review_unit_hash": "cf88283b83197fc2b9c5389e4fafcacb72c445ec9a4dfce505afd1e7c6f0733b", + "prepared_review_unit_hash": "f54ee0083abfc14deb9c31d9c977c5c440d58c72ba0348d30bb8254a352b7382", "component_budget": { "node_count": 453, "item_count": 453, - "soft_pair_count": 1679, - "serialized_chars": 868496, - "estimated_input_tokens": 217124, - "estimated_eval_tokens": 434504 + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 47, - "serialized_chars": 40312, + "serialized_chars": 40309, "estimated_input_tokens": 10078, "estimated_eval_tokens": 20412 }, @@ -16724,48 +16672,48 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: most items are distinct model, tokenizer, scheduler, import, or docs issues. I would not merge any of the soft edges; the only plausible umbrella is the broad v5-breakage report, but it is not a true duplicate target for the rest.", - "confidence": 0.81, - "canonical_issue_reason": "issue:45850 is the broadest representative because it explicitly describes many models breaking after v5; it is the closest umbrella among these mixed regression reports.", + "summary": "This cluster is mostly a set of unrelated Transformer issues that happen to share broad loading/model/tokenizer language. I don\u2019t see a true duplicate set; the soft links are mostly spurious similarity matches.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 45850 is the closest broad umbrella for the cluster because it describes widespread 5.0 breakage, but the underlying reports are still too diverse to be real duplicates.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45850 is the best issue representative for the cluster as a whole, but only as a loose umbrella, not as a duplicate of the other items.", + "best_issue_reason": "45850 is the best single representative of the overall theme because it is the broadest regression report in the set, though it is not a tight canonical duplicate anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46018", "right": "issue:46172", "accept": false, - "reason": "Both are DeepSeek-V4-related, but one is about shared-expert gating and the other is a tensor/casting bug; different concrete code paths." + "reason": "Different DeepSeek-V4 bugs: gated expert logic vs a tensor transpose/dtype issue." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Different models and bugs: Qwen3VL timestamp calculation vs GLM-5 RoPE implementation." + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are unrelated bugs." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Processor timestamp math and scheduler kwarg handling are unrelated issues." + "reason": "Vision processor timestamp math is unrelated to scheduler kwargs handling." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import-star performance cleanup and a VRAM leak in dataloader threads are unrelated." + "reason": "Import star cleanup and a multithreaded VRAM leak are different problems." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Confusing forward docs and a model-loading breakage for Llama 3.2 are different problems." + "reason": "Documentation confusion in forward args is not the same as a model-loading regression." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Slow checkpoint loading is not the same as forward-argument documentation confusion." + "reason": "Slow checkpoint loading and forward-API documentation are unrelated." }, { "left": "issue:43704", @@ -16777,241 +16725,241 @@ "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "GLM-5 RoPE behavior and Gemma-4 audio positional encoding are different code paths." + "reason": "Both touch positional encoding, but they are different models and different concrete failures." }, { "left": "issue:43298", "right": "issue:46095", "accept": false, - "reason": "Audio continuous batching support and a DeepSeek-V4 weight-conversion question are not the same bug." + "reason": "Audio continuous batching support and a DeepSeek-V4 weight-conversion question are unrelated." }, { "left": "issue:45850", "right": "issue:46218", "accept": false, - "reason": "A broad v5 regression umbrella is not the same as a specific Qwen3_5 import failure." + "reason": "A broad 5.0 breakage complaint is not the same as a specific missing import for Qwen3_5ForCausalLM." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 API guidance and a VRAM leak are unrelated." + "reason": "TF32 settings guidance is unrelated to a VRAM leak." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "Local-files-only API requests and Qwen3VL timestamp calculation are different issues." + "reason": "Local-files-only network leakage and timestamp calculation are different code paths." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Timestamp calculation and missing save_pretrained files do not share the same bug." + "reason": "Timestamp math and missing save_pretrained files are unrelated processor bugs." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Qwen3VL processor math and Qwen3ForCausalLM VRAM leak are unrelated." + "reason": "Processor timestamp logic is unrelated to multithreaded memory leakage." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split return-value handling and Gemma-4 audio positional encoding are different problems." + "reason": "torch.split return handling is unrelated to Gemma4 audio positional encoding." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch is unrelated to Gemma-4 audio positional encoding." + "reason": "CPU-offload device mismatch and audio positional encoding are different model-specific issues." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "MLA q_a_layernorm eps handling is not the same as Gemma-4 audio rel-pos encoding." + "reason": "RMS norm epsilon precision mismatch is not the same as Gemma4 audio positional encoding." }, { "left": "issue:44351", "right": "issue:46218", "accept": false, - "reason": "Missing HybridCache import and missing Qwen3_5ForCausalLM module are distinct import failures." + "reason": "Different import failures for different symbols/models; not the same missing export." }, { "left": "issue:43352", "right": "issue:46218", "accept": false, - "reason": "Flash Attention support limitation and module import failure are not the same bug." + "reason": "Flash Attention unsupported and missing Qwen3_5 import are unrelated." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Both are tokenizer-related, but one is a general behavior change and the other is a specific decoder-type mistake." + "reason": "Both are tokenizer-v5 reports, but one is broad behavior drift and the other is a specific decoder-type bug." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs handling and a nonsense issue titled 'Unemployment' are unrelated." + "reason": "Scheduler kwargs bug and spam/off-topic issue are unrelated." }, { "left": "issue:44492", "right": "issue:44998", "accept": false, - "reason": "Cache-strategy typo and an unrelated placeholder issue are not duplicates." + "reason": "Cache-strategy typo and unrelated spam issue are not duplicates." }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions regression has nothing to do with the unrelated 'Unemployment' issue." + "reason": "CLIPOutput attentions regression is unrelated to the off-topic item." }, { "left": "issue:44998", "right": "issue:45850", "accept": false, - "reason": "A broad v5 regression report is not the same as an unrelated placeholder issue." + "reason": "Spam/off-topic text is not a duplicate of a broad 5.0 regression report." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "Tied-weight-key attribute crash and chat-template tool-call crash are different code paths." + "reason": "Tied-weight key crash and apply_chat_template tool-call crash are different failure modes." }, { "left": "issue:44938", "right": "issue:44998", "accept": false, - "reason": "Python 3.14 import failure is unrelated to the placeholder issue." + "reason": "Python 3.14 load failure is unrelated to the off-topic item." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Missing pad_token_id in Llama4Config is unrelated to tied-weight-key handling." + "reason": "Missing pad_token_id and tied-weight key handling are unrelated." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "Slow checkpoint loading and a missing configuration_utils import symbol are unrelated." + "reason": "Slow checkpoint loading and an ImportError for layer_type_validation are unrelated." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "StableLm pad_token_idx regression and tied-weight-key crash are different bugs." + "reason": "StableLmConfig pad token regression is unrelated to tied-weight key crashes." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "FP8 support request and missing processor save_pretrained files are unrelated." + "reason": "FP8 inference support and missing processor save files are unrelated." }, { "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 jobs is unrelated to checkpoint loading speed." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "Slow from_pretrained checkpoint loading is unrelated to tensor-parallel training bugs." + "reason": "Checkpoint loading performance and tensor-parallel training bugs are different issues." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "FP8 support request and Qwen3.5 MoE from_pretrained error are different issues." + "reason": "FP8 inference support and Qwen3.5 MoE from_pretrained failure are not the same bug." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "Slow checkpoint loading and tokenizer loading regression are not the same concrete bug." + "reason": "Checkpoint loading slowness and tokenizer-loading regression are different problems." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Audio continuous batching support and FP8 inference support are different feature gaps." + "reason": "Audio continuous batching support and FP8 support are unrelated feature gaps." }, { "left": "issue:43717", "right": "issue:46144", "accept": false, - "reason": "Mamba init_weights usage and RoFormer attention interface are unrelated model implementations." + "reason": "Mamba init_weights usage and RoFormer attention-interface usage are unrelated architecture issues." }, { "left": "issue:42936", "right": "issue:46018", "accept": false, - "reason": "Mask2Former ignore_value handling is unrelated to DeepSeek-V4 shared-expert gating." + "reason": "Mask2former ignore_value handling is unrelated to DeepSeek-V4 gating." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "A ModernBERT type hint issue and tokenizer regex warnings are unrelated." + "reason": "A type-hint typo in config is unrelated to tokenizer regex warnings." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Mask2Former ignore_value and kwargs not passed to loss calculation are different bugs." + "reason": "Ignoring a configured loss value and missing kwargs in loss calculation are different bugs." }, { "left": "issue:43240", "right": "issue:46144", "accept": false, - "reason": "Loss-function kwargs plumbing and RoFormer attention interface are unrelated." + "reason": "Loss kwargs handling and RoFormer attention interface are unrelated." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2Former ignore_value and tokenizer regex warnings do not describe the same defect." + "reason": "Mask2former ignore_value and tokenizer regex warnings do not share a code path." }, { "left": "issue:42936", "right": "issue:46144", "accept": false, - "reason": "Mask2Former ignore_value handling is unrelated to RoFormer attention plumbing." + "reason": "Different model subsystems: segmentation loss handling vs attention interface." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "Mask2Former ignore_value and a typo in quantization_operation are unrelated." + "reason": "Ignore_value usage and a typo in quantization operation naming are unrelated." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Loss kwargs routing and ModernBERT norm_eps typing are unrelated." + "reason": "Loss kwargs forwarding and a config type-hint mistake are unrelated." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "ModernBERT type hints and a quantization typo are different issues." + "reason": "Type-hint correction and a typo in a conversion API name are different fixes." }, { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "Llama4TextL2Norm/RMS norm question and ModernBERT type hint mismatch are unrelated." + "reason": "Llama4 norm semantics and a ModernBERT type-hint issue are unrelated." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Mask2Former ignore_value and ModernBERT norm_eps typing are different bugs." + "reason": "Segmentation ignore_value logic is unrelated to a config annotation typo." } ] }, @@ -17044,7 +16992,7 @@ "issue:10519|issue:11005", "issue:11005|issue:11307" ], - "prepared_review_unit_hash": "ea237469e10375cb350bcd88b70798ea4bb908cffff542a5d6529ba34b576e42", + "prepared_review_unit_hash": "d6b022fe048fb172de4c6c8f48e899062b498cdfe75f0986fc87a0f5d0d7d59a", "component_budget": { "node_count": 3, "item_count": 3, @@ -17066,30 +17014,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These three issues are not duplicates: they ask for unrelated features in different parts of the library (token truncation behavior, a ReduceLROnPlateau-like scheduler feature, and Wav2Vec2 word time offsets).", + "summary": "All three items are separate feature requests in different areas: token truncation behavior, a learning-rate scheduler feature, and Wav2Vec2 word timestamp offsets. No pair looks like the same underlying bug or change.", "confidence": 0.98, - "canonical_issue_reason": "No single canonical issue for this cluster; if one representative is needed, #10519 is the most concrete feature request and best specified.", + "canonical_issue_reason": "issue:10519 is the most concrete and self-contained request, with a specific API behavior change described in the title.", "canonical_pr_reason": null, - "best_issue_reason": "#10519 is the best representative issue because it is narrowly scoped and clearly describes a specific desired behavior.", + "best_issue_reason": "issue:10519 is the clearest representative of the cluster because it is specific and narrowly scoped; the other issues are unrelated in topic and subsystem.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:10519", "right": "issue:11307", "accept": false, - "reason": "Different feature requests in different subsystems: truncation strategy vs Wav2Vec2 word timestamp offsets." + "reason": "Different domains and requested changes: truncation direction for tokenization vs. Wav2Vec2 word boundary timestamps." }, { "left": "issue:10519", "right": "issue:11005", "accept": false, - "reason": "Unrelated topics: text truncation behavior vs a scheduler/learning-rate-reduction feature request." + "reason": "Unrelated features: tokenizer truncation behavior vs. ReduceLROnPlateau-like scheduler functionality." }, { "left": "issue:11005", "right": "issue:11307", "accept": false, - "reason": "No shared underlying bug or change; one is about LR scheduling, the other about extracting Wav2Vec2 word boundaries." + "reason": "No shared underlying bug/change; one asks for a scheduler feature, the other for speech model timing offsets." } ] }, @@ -17121,12 +17069,12 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "81c75c222b0cb105c0939c2146fac1c51c7745a39bedced1339cc10ab6cfe6d5", + "prepared_review_unit_hash": "213ce43d2eb4b8ba8cbecd04767d18b56af5b19550e2ee914386937bf2eb1290", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2478, + "serialized_chars": 2479, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -17134,7 +17082,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2478, + "serialized_chars": 2479, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -17143,24 +17091,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These three issues are only loosely related by NLP/model-loading context, but they describe different problems: tokenization mismatch, private-model auth/error messaging, and failure to load a public model. I would not merge any of the soft pairs as duplicates.", - "confidence": 0.91, - "canonical_issue_reason": "Issue #16618 is the broadest and most central of the three, describing an actual model loading failure rather than a narrow tokenizer or messaging variant.", + "summary": "The three issues are only loosely related by model/tokenizer/HF-loading wording, but they describe different problems: Tapas tokenization mismatch, a private-model login error message, and a failure loading bert-base-uncased. I would not merge any of the soft pairs as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "Issue #16618 is the broadest and most general of the set, with the most discussion and a user-facing model-loading failure, so it is the best single representative if one is needed.", "canonical_pr_reason": null, - "best_issue_reason": "#16618 is the best representative issue for this cluster because it is the most general user-facing failure and the closest to the soft-similarity links.", + "best_issue_reason": "#16618 is the best issue candidate because it is the clearest, most general symptom report in the cluster and most likely to attract related discussion.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Different underlying problem: #13244 is about Tapas tokenization not matching TensorFlow behavior, while #16618 is a model loading failure for bert-base-uncased." + "reason": "These appear to be different bugs: tokenizer behavior for Tapas versus a model-loading failure for bert-base-uncased. They do not share a concrete code path or fix." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "Different root cause and symptom: #15980 concerns a bad auth-related error when downloading a private model, not inability to load a public model." + "reason": "Both concern model downloads/loading, but one is specifically a private-model authentication/error-message issue while the other is a generic load failure. Too broad to treat as the same underlying bug." } ] }, @@ -17190,40 +17138,40 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "163145e1acc8c664cdc4a7e8e458b111cb5e4641ff5223e33cf6b637eb4368a6", + "prepared_review_unit_hash": "2a16aa37975dffdadd38eea6ab1a77c6f6b8a57c84f1e350b674261500d37f9c", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1588, - "estimated_input_tokens": 397, - "estimated_eval_tokens": 1050 + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1588, - "estimated_input_tokens": 397, - "estimated_eval_tokens": 1050 + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate community feature requests in the same broad area of model/runtime support, but they target different capabilities: Flash Attention 2 vs GGUF. They are not duplicates.", + "summary": "These are two separate community-contribution tracking issues for different features: Flash Attention 2 support vs GGUF support. They share a broad \u201cmore architectures\u201d theme but do not describe the same underlying bug or change.", "confidence": 0.98, - "canonical_issue_reason": "No single canonical issue: #26350 is about adding Flash Attention 2 support for more architectures, while #33260 is about adding GGUF support for more architectures. Different underlying changes.", + "canonical_issue_reason": "Not applicable as a duplicate canonical: the issues track different feature requests (Flash Attention 2 support vs GGUF support), so neither can serve as the canonical for the other.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a better global representative of the other; they track distinct feature requests with different implementation paths.", + "best_issue_reason": "Neither issue is a global representative of the other; they are unrelated feature-tracking threads in the same general area but for different deliverables.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different features and code paths: Flash Attention 2 support is unrelated to GGUF support, so they should not be merged as duplicates." + "reason": "Different requested features and likely different implementation work; same broad subsystem/theme is not enough to treat them as duplicates." } ] }, @@ -17253,7 +17201,7 @@ "soft_pairs": [ "issue:26413|issue:31501" ], - "prepared_review_unit_hash": "21c2bb534754eae5db402d13b6035191ac580027b2e47ca9a3657db972c1616a", + "prepared_review_unit_hash": "bf6f367a3d4a3b3d1f97cb00c97be0775c99036f58d79989e176e1846bf66add", "component_budget": { "node_count": 2, "item_count": 2, @@ -17275,18 +17223,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues both concern training not starting, but they do not clearly describe the same bug. One is specifically about `resume_from_checkpoint` triggering an empty `epoch_iterator` error, while the other is a broader FineWeb SLM training startup failure with no explicit indication it is the same checkpoint/iterator path.", - "confidence": 0.71, - "canonical_issue_reason": "`issue:26413` is the better canonical issue because it names the exact failing code path and error message, making it the more precise representative of the bug class.", + "summary": "These two issues both concern training failing to start, but the evidence is too thin to treat them as the same bug. The first is a specific `resume_from_checkpoint`/`epoch_iterator` failure; the second is a broader FineWeb SLM training startup problem with no explicit linkage to that code path.", + "confidence": 0.88, + "canonical_issue_reason": "Issue #26413 is the more specific, clearly framed bug report with an explicit failure message and likely reproducible code path.", "canonical_pr_reason": null, - "best_issue_reason": "`issue:26413` is the stronger issue to keep as the representative record since it is concrete and actionable; `issue:31501` is more generic and does not establish the same root cause.", + "best_issue_reason": "Issue #26413 is the best representative because it names the concrete error and underlying training control-flow problem; #31501 is too broad from the title alone.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26413", "right": "issue:31501", "accept": false, - "reason": "Reject: the overlap is only broad training-start failure. `issue:26413` is about `resume_from_checkpoint` failing with an empty `epoch_iterator`, while `issue:31501` does not provide enough evidence that it is the same underlying code-path problem." + "reason": "Both mention training not starting, but one is specifically about `resume_from_checkpoint` hitting an empty `epoch_iterator` while the other is a general FineWeb SLM startup failure. No evidence they are the same underlying bug or fixable in one change." } ] }, @@ -17345,187 +17293,187 @@ "issue:44370|issue:46226", "issue:46222|issue:46226" ], - "prepared_review_unit_hash": "d29ad8d643e0f55270113999df7dc8560e51c1bd0bc3441818585ae41f9079c2", + "prepared_review_unit_hash": "57e5ee589e06df80630543a675117e169abf39e6aa102a568e80e2930631eba3", "component_budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 24, - "serialized_chars": 13531, - "estimated_input_tokens": 3383, - "estimated_eval_tokens": 7022 + "serialized_chars": 13526, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 24, - "serialized_chars": 13531, - "estimated_input_tokens": 3383, - "estimated_eval_tokens": 7022 + "serialized_chars": 13526, + "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 mostly a false-positive similarity group of i18n doc-translation issues; the only clear duplicate pair is the two identical Thai-titled issues. The language-specific translation requests should stay separate.", - "confidence": 0.66, - "canonical_issue_reason": "Issue 46222 is the best canonical representative of the only clear duplicate pair: it matches 46226\u2019s Thai title and near-identical content, and it has more inbound references/activity, so it looks like the parent copy.", + "summary": "Using only titles, body lengths, and soft-similarity scores, this looks like two tight duplicate-like subgroups rather than one big duplicate set: three identical generic placeholder translation issues, and one identical Thai-titled pair. The language-specific translation requests (Turkish, Spanish, Romanian) appear separate and should not be merged based on this evidence alone.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 43834 is the best canonical anchor for the largest high-similarity subgroup: it has the generic placeholder title shared by the other two placeholder issues and sits in the densest cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 46222 is the strongest overall issue candidate because it is the most connected exact-duplicate-looking item in the cluster; the rest are better treated as separate translation requests.", + "best_issue_reason": "43834 is the strongest overall issue representative because it belongs to the largest near-duplicate cluster and connects to the other placeholder-title copies with very high similarity.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Same i18n translation theme, but this is a different language request, not the same underlying issue." + "reason": "Turkish translation request vs. generic placeholder issue; related topic, but not the same underlying issue." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "Same i18n translation theme, but a different language/task." + "reason": "Different language-specific request vs. generic placeholder issue; similarity is only boilerplate-level." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Template similarity only; no evidence it is the same translation request." + "reason": "Turkish request and placeholder-template issue are not the same concrete task." }, { "left": "issue:27088", "right": "issue:38435", "accept": false, - "reason": "Different language translation request; similar topic is not enough." + "reason": "Different language requests (Turkish vs. Romanian); separate issues." }, { "left": "issue:27088", "right": "issue:28936", "accept": false, - "reason": "Turkish vs Spanish docs translation are distinct issues, not duplicates." + "reason": "Different language requests (Turkish vs. Spanish); separate issues." }, { "left": "issue:28936", "right": "issue:43834", "accept": false, - "reason": "Different translation requests; likely just shared issue-template wording." + "reason": "Spanish translation request vs. generic placeholder issue; too little evidence of the same underlying task." }, { "left": "issue:28936", "right": "issue:44365", "accept": false, - "reason": "Same broad i18n topic, but not the same underlying change." + "reason": "Spanish request and placeholder-template issue look related only at the template level, not duplicate." }, { "left": "issue:28936", "right": "issue:38435", "accept": false, - "reason": "Spanish and Romanian translation requests are separate tasks." + "reason": "Spanish vs. Romanian translation requests are distinct issues." }, { "left": "issue:28936", "right": "issue:44370", "accept": false, - "reason": "Different language/request despite high template similarity." + "reason": "Spanish request vs. placeholder-template issue; not the same change or request." }, { "left": "issue:38435", "right": "issue:44370", "accept": false, - "reason": "Highly similar placeholder text, but no proof these are the same language-specific request." + "reason": "Romanian request is language-specific; placeholder issue is a different artifact." }, { "left": "issue:38435", "right": "issue:44365", "accept": false, - "reason": "Different translation requests; similarity appears to come from the template." + "reason": "Romanian request vs. generic placeholder issue; related, but not a clear duplicate." }, { "left": "issue:38435", "right": "issue:43834", "accept": false, - "reason": "Same template style, but not the same underlying language task." + "reason": "Romanian translation request is distinct from the placeholder-template issue." }, { "left": "issue:38435", "right": "issue:46222", "accept": false, - "reason": "Romanian translation issue and Thai-titled issue are unrelated." + "reason": "Romanian issue and Thai-titled issue do not look like the same translation request." }, { "left": "issue:38435", "right": "issue:46226", "accept": false, - "reason": "Romanian translation issue and Thai-titled issue are unrelated." + "reason": "Romanian issue and Thai-titled issue do not look like the same underlying task." }, { "left": "issue:43834", "right": "issue:44370", - "accept": false, - "reason": "Near-identical templated wording, but likely separate language requests." + "accept": true, + "reason": "Same generic placeholder title and extremely high similarity; highly likely duplicate." }, { "left": "issue:43834", "right": "issue:44365", - "accept": false, - "reason": "Placeholder similarity alone is not enough to treat them as duplicates." + "accept": true, + "reason": "Same generic placeholder title and extremely high similarity; highly likely duplicate." }, { "left": "issue:43834", "right": "issue:46222", "accept": false, - "reason": "Different issue content; one is a generic i18n template, the other is a specific Thai-titled report." + "reason": "Very high similarity, but different titles and no body text to show the same underlying issue." }, { "left": "issue:43834", "right": "issue:46226", "accept": false, - "reason": "Different issue content; no evidence of the same underlying bug/change." + "reason": "Very high similarity, but different titles and insufficient evidence to merge." }, { "left": "issue:44365", "right": "issue:44370", - "accept": false, - "reason": "Same generic translation template, but not enough evidence they are the same request." + "accept": true, + "reason": "Identical placeholder-template title with near-perfect similarity; highly likely duplicate." }, { "left": "issue:44365", "right": "issue:46222", "accept": false, - "reason": "Different issue content and likely different translation task." + "reason": "High textual similarity alone is not enough here; the titles are different and the evidence is too thin." }, { "left": "issue:44365", "right": "issue:46226", "accept": false, - "reason": "Different issue content and likely different translation task." + "reason": "High similarity, but different titles and no direct evidence of the same issue." }, { "left": "issue:44370", "right": "issue:46222", "accept": false, - "reason": "Different issue content; shared i18n theme is too broad." + "reason": "Placeholder-template issue vs. Thai-titled issue; not the same concrete request." }, { "left": "issue:44370", "right": "issue:46226", "accept": false, - "reason": "Different issue content; shared i18n theme is too broad." + "reason": "Placeholder-template issue vs. Thai-titled issue; not the same concrete request." }, { "left": "issue:46222", "right": "issue:46226", "accept": true, - "reason": "These have the same Thai title and near-identical metadata, so they look like the same underlying issue duplicated." + "reason": "Identical Thai title and perfect similarity; highly likely duplicate pair." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative: the only clearly duplicate-looking pair is 46222/46226, which have identical Thai titles and exact body-length match, while the other high-similarity i18n items remain unproven duplicates. Canonical choice of 46222 is also supported by its higher inbound references/activity." + "accept": false, + "feedback": "Reject: the summary overstates what the packet supports. The packet only shows titles, body lengths, and soft-similarity scores; it does not expose the bodies, and every pair is marked deterministic_accept=false. Saying the placeholder issues or Thai-titled issues are \"exact\" or even clearly duplicate is stronger than the evidence here. The cross-language separations are conservative, but the accepted soft-edge pairs need weaker wording (e.g., \"highly similar\" / \"possible duplicates\")." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -17598,142 +17546,142 @@ "issue:43837|issue:45177", "issue:45322|issue:46133" ], - "prepared_review_unit_hash": "0af0bfd2ecefe7fb724839090f249e03f4da2f973226d7c9907d837934b74d0d", + "prepared_review_unit_hash": "f1ad9f7ce3672a3f9491d76b92071c71f657be7778fad29eed149934a3a5ca6b", "component_budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21655, - "estimated_input_tokens": 5414, - "estimated_eval_tokens": 11084 + "serialized_chars": 21657, + "estimated_input_tokens": 5415, + "estimated_eval_tokens": 11086 }, "budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21655, - "estimated_input_tokens": 5414, - "estimated_eval_tokens": 11084 + "serialized_chars": 21657, + "estimated_input_tokens": 5415, + "estimated_eval_tokens": 11086 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are broad but distinct model-add/support requests across different model families and modalities; the soft similarities are topical only, so none should be deduplicated as the same underlying change.", - "confidence": 0.96, - "canonical_issue_reason": "No canonical issue: the items are separate model requests rather than duplicates of one bug or one implementation change.", + "summary": "This cluster is a loose collection of model-support requests across different architectures and modalities. Despite textual similarity, the items are not the same underlying bug/change, so the soft edges should not be merged.", + "confidence": 0.97, + "canonical_issue_reason": "If a single anchor is needed, issue 42738 is the broadest conceptual request in the set, but it is still only a rough umbrella and not a true duplicate of the model-specific requests.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 42738 is the closest umbrella-style representative, but it is still too broad to stand in for the other specific model requests.", + "best_issue_reason": "42738 is the closest representative issue because it is the most general architecture-style request; however, this cluster is heterogeneous and lacks a real canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "Different model requests (EuroBert vs DEIMv2); no shared code-path or concrete change." + "reason": "Different model families and different implementation targets; not the same change." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "Different modalities and models; a BERT config request is not the same as Qwen3-TTS support." + "reason": "EuroBert config support and Qwen3-TTS support are unrelated model additions." }, { "left": "issue:36817", "right": "issue:36879", "accept": false, - "reason": "Different models and tasks (EuroBert vs RF-DETR); not the same underlying request." + "reason": "EuroBert config vs RF-DETR: different architectures and code paths." }, { "left": "issue:36879", "right": "issue:41211", "accept": false, - "reason": "Different architectures and use cases; RF-DETR and DEIMv2 are separate model integrations." + "reason": "RF-DETR and DEIMv2 are separate vision models, not one bug or patch." }, { "left": "issue:36879", "right": "issue:43909", "accept": false, - "reason": "RF-DETR vision support and LFM2.5 Audio are unrelated model additions." + "reason": "RF-DETR is a detection model; LFM2.5 Audio is an audio model request." }, { "left": "issue:36879", "right": "issue:44121", "accept": false, - "reason": "Different model families; RF-DETR and a weight-sparse transformer are not the same fix." + "reason": "Different model classes; no shared concrete fix path." }, { "left": "issue:36879", "right": "issue:43837", "accept": false, - "reason": "Different tasks and code paths; object detection support is not the same as Qwen3-ASR support." + "reason": "RF-DETR and Qwen3-ASR are unrelated model support requests." }, { "left": "issue:36879", "right": "issue:43671", "accept": false, - "reason": "Different tasks and models; RF-DETR is unrelated to Qwen3-TTS integration." + "reason": "Different modalities and model families; not mergeable as one PR." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are distinct model-support requests." + "reason": "NeoBERT and Qwen3-ASR are unrelated requests." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "Different model families; NeoBERT support is not the same change as adding a weight-sparse transformer." + "reason": "NeoBERT support and weight-sparse transformer support are different changes." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT and LFM2.5 Audio target different models and modalities." + "reason": "BERT-like encoder work vs audio model support; distinct code paths." }, { "left": "issue:36879", "right": "issue:37015", "accept": false, - "reason": "RF-DETR and NeoBERT are separate model additions with no shared implementation." + "reason": "RF-DETR and NeoBERT are different model additions with no shared fix." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT and ModernVBERT are different model requests, not one duplicate." + "reason": "NeoBERT and ModernVBERT are separate model requests, not duplicates." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 architecture support and Qwen3-ASR support are different concrete changes." + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are unrelated." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "Different model families and implementation work; not the same bug or feature." + "reason": "MiniCPM3 and sparse-transformer support are different concrete changes." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 and ModernVBERT are separate model integrations." + "reason": "MiniCPM3 vs ModernVBERT are distinct model-support requests." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 support and Voxtral TTS support are unrelated requests." + "reason": "MiniCPM3 and Voxtral TTS address different models and modalities." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 and Qwen3-TTS are different models and likely different code paths." + "reason": "DEIMv2 and Qwen3-TTS are unrelated model support changes." }, { "left": "issue:42503", @@ -17745,103 +17693,103 @@ "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT and Qwen3-TTS are separate model additions." + "reason": "ModernVBERT and Qwen3-TTS are not the same underlying issue." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "A specific model request is not the same as the broad BERT-like-with-RoPE umbrella issue." + "reason": "ModernVBERT is a specific model request; 42738 is a broad BERT-like/RoPE family request." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "Broad BERT-like RoPE discussion and Voxtral TTS support are unrelated changes." + "reason": "BERT-like models with RoPE and Voxtral TTS are different model domains." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Same product family, but TTS and ASR are distinct tasks and support paths; not a duplicate." + "reason": "TTS and ASR are related modality families but still distinct model support requests." }, { "left": "issue:43671", - "right": "issue:43909", + "right": "issue:45027", "accept": false, - "reason": "Both are audio-related, but Qwen3-TTS and LFM2.5 Audio are different model integrations." + "reason": "Both are TTS-related, but they concern different concrete models and fixes." }, { "left": "issue:43671", - "right": "issue:44121", + "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS support is unrelated to a weight-sparse transformer request." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and implementation paths." }, { "left": "issue:43671", - "right": "issue:45027", + "right": "issue:44121", "accept": false, - "reason": "Both are TTS-related, but they target different models and likely separate implementations." + "reason": "Audio generation support and sparse-transformer support are unrelated." }, { "left": "issue:43671", - "right": "issue:45177", + "right": "issue:43909", "accept": false, - "reason": "TTS support and OCR support are different modalities and code paths." + "reason": "Qwen3-TTS and LFM2.5 Audio are different audio-model requests." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "ASR support and LFM2.5 Audio are different concrete model requests." + "reason": "Qwen3-ASR and LFM2.5 Audio 1.5B are different speech/audio model requests." }, { "left": "issue:43837", - "right": "issue:44121", + "right": "issue:45322", "accept": false, - "reason": "Different model families; ASR support is not the same change as a weight-sparse transformer." + "reason": "ASR support and EUPE encoder support are unrelated." }, { "left": "issue:43837", - "right": "issue:45322", + "right": "issue:44121", "accept": false, - "reason": "ASR support and EUPE are unrelated model requests." + "reason": "Speech recognition support and sparse-transformer support are not the same code change." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "Audio model support and SAM3-LiteText are different modalities and integrations." + "reason": "LFM2.5 Audio and SAM3-LiteText are different multimodal model requests." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Weight-sparse transformer and EUPE are unrelated model requests." + "reason": "Sparse-transformer support and EUPE model support are distinct changes." }, { "left": "issue:43837", "right": "issue:45027", "accept": false, - "reason": "ASR and TTS are related only at a high level; these are different models and changes." + "reason": "Qwen3-ASR and Voxtral TTS are separate speech-model requests." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT and Voxtral TTS are separate model-support requests." + "reason": "NeoBERT and Voxtral TTS are unrelated model support requests." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "ASR support and OCR support are different tasks and code paths." + "reason": "ASR support and OCR support are different modalities and code paths." }, { "left": "issue:45322", "right": "issue:46133", "accept": false, - "reason": "EUPE and TIPSv2 are distinct model requests with no shared concrete implementation." + "reason": "EUPE and TIPSv2 are different vision-language model requests with no shared concrete fix." } ] }, @@ -17871,40 +17819,40 @@ "soft_pairs": [ "issue:38740|issue:46443" ], - "prepared_review_unit_hash": "722064d5ee11aa90fe34c1f38f6283e781a6befcd5602459a007541d583342f5", + "prepared_review_unit_hash": "f0b3266543caebdedb1a6f0b0b7a1214fe1a62ad63a81a62c990734b56707193", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1505, - "estimated_input_tokens": 377, - "estimated_eval_tokens": 1010 + "serialized_chars": 1503, + "estimated_input_tokens": 376, + "estimated_eval_tokens": 1008 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1505, - "estimated_input_tokens": 377, - "estimated_eval_tokens": 1010 + "serialized_chars": 1503, + "estimated_input_tokens": 376, + "estimated_eval_tokens": 1008 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are not duplicates: one is a closed docs request about adding Pruna, the other is an open refactor for Vision RoPE. They do not describe the same underlying bug or change.", - "confidence": 0.99, - "canonical_issue_reason": "No canonical issue for deduping here; the issues are unrelated in both topic and scope.", + "summary": "The two issues are unrelated: one is a docs request for adding Pruna as an optimization framework, while the other is a Vision RoPE refactor. They should not be treated as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue: these are about different topics and do not describe the same underlying problem or change.", "canonical_pr_reason": null, - "best_issue_reason": "If a single issue must be preferred as the cluster anchor, issue 46443 is the better candidate because it is the active, substantive code-refactor issue; issue 38740 is a closed documentation request.", + "best_issue_reason": "Issue 46443 is the more substantive standalone code-related item, but it is not a duplicate of issue 38740.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:38740", "right": "issue:46443", "accept": false, - "reason": "Different subjects: docs/optimization-framework addition vs Vision RoPE refactor. No shared code path or underlying change." + "reason": "Different subjects and scopes: documentation/framework listing vs Vision RoPE refactor. No shared underlying bug or change." } ] }, @@ -17934,12 +17882,12 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "e11b208165da97e22fdbb7086e4d71c9b9dc55b461df3a365d7627ae424d0dd0", + "prepared_review_unit_hash": "28133fcd8bcfac0d513f43b8f08eeb2065c2961cf47db7db52d45294c7543d83", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1479, + "serialized_chars": 1480, "estimated_input_tokens": 370, "estimated_eval_tokens": 996 }, @@ -17947,7 +17895,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1479, + "serialized_chars": 1480, "estimated_input_tokens": 370, "estimated_eval_tokens": 996 }, @@ -17956,18 +17904,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are not duplicates: one is about block decoding behavior, the other is a model-support request for DeepSeek v3.2. They reference different problems with no shared targets or code-path evidence.", - "confidence": 0.98, - "canonical_issue_reason": null, + "summary": "The two issues are not duplicates: one asks for Set Block Decoding, while the other requests DeepSeek v3.2 support. They target different features and should not be merged.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 43037 is the more specific and actionable request; 41084 is a separate, broader feature request.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good canonical representative for the other because they concern unrelated topics and likely different work streams.", + "best_issue_reason": "43037 is the best issue to keep as the representative item because it names a concrete model support request, making it easier to route and implement.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different underlying requests: block decoding vs. DeepSeek v3.2 support. No shared filenames, targets, or concrete bug/change overlap." + "reason": "Different underlying requests: block decoding vs. model support for DeepSeek v3.2. They do not describe the same bug or change." } ] }, @@ -17997,12 +17945,12 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "ad785f52ebb09236b27bc69d7a55e45a2bf4c8b81e793f835159150ff23fe50e", + "prepared_review_unit_hash": "5e5ad7dd4c48c7a8bafe05e002358f766a42843259ed2d87aaff7d056bfca893", "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 }, @@ -18010,7 +17958,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1584, + "serialized_chars": 1583, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -18019,18 +17967,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are related only at a very high level: both mention custom trainers and Trainer behavior, but they describe different problems and different code paths. They should not be treated as duplicates.", - "confidence": 0.94, - "canonical_issue_reason": "No single canonical issue is appropriate here because the issues are about distinct concerns rather than the same underlying bug.", + "summary": "These issues are only loosely related through Trainer customization, but they target different behaviors: one is about `remove_unused_column` handling, the other about FLOPS estimation for custom trainers. They do not look like the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "If one issue must represent the cluster, #43598 is the cleaner representative because it states a concrete Trainer customizability problem more directly.", + "best_issue_reason": "If one must be surfaced as the representative issue, `#43598` is the more concrete Trainer API/customization request; `#43600` is a separate FLOPS-estimation problem.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "They share the broad Trainer/custom-trainer theme, but one is about remove_unused_column customizability and the other is about FLOPS estimation; these are separate issues." + "reason": "Reject: both mention custom trainers, but they affect different code paths and outcomes (`remove_unused_column` behavior vs FLOPS estimation). This is broad thematic overlap, not a duplicate underlying problem." } ] }, @@ -18060,7 +18008,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "85c571b4709ae53dcf161b8db6186fa5ba8c8ae1e300415cf27f023f03590ae0", + "prepared_review_unit_hash": "da18fee54d1845ec5e5d6d30049a40e382a3e47b542d55b27214a4ae7543ead9", "component_budget": { "node_count": 2, "item_count": 2, @@ -18082,18 +18030,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one requests Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration, while the other asks for AutoModelForSequenceClassification support for Qwen3.5/Qwen3_5Config. They target different model families and different auto-class pathways.", - "confidence": 0.94, - "canonical_issue_reason": null, + "summary": "These are not duplicates: one asks for Qwen3-Omni auto-registration for generation, while the other requests sequence-classification support for Qwen3.5. They target different model families and different AutoModel heads.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the requests are about different Qwen variants and different auto-class targets.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43684 is the more general API-registration request, but it still does not subsume the Qwen3.5 sequence-classification support request, so it is only the closest representative, not a true duplicate canonical.", + "best_issue_reason": "Issue 43684 is the broader, more central model-registration request, so it is the better representative if one must be chosen.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different underlying change: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration vs Qwen3.5 AutoModelForSequenceClassification support. They involve different configs and different auto-class mappings." + "reason": "Different model families and different tasks (conditional generation vs sequence classification); not the same underlying bug or change." } ] }, @@ -18130,70 +18078,70 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "6e46efc713737abab951e690686cbfd03f8870544048d677076ef103c7fb90de", + "prepared_review_unit_hash": "9bacf7033ca970dac859ace97a1bd90280f26f9b3f7568b2148b96e760e61c81", "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": 4267, + "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": 4267, + "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 very low-information, mostly title-only reports. The three 'huggingface' items are similar to each other, but there is not enough concrete bug/change detail to safely treat any pair as the same underlying issue; 'Fett' looks even less connected.", - "confidence": 0.42, - "canonical_issue_reason": "issue #43845 is the cleanest cluster representative: it matches the repeated 'huggingface' title pattern and is the earliest of the near-identical subgroup.", + "summary": "These are low-signal issues with generic or empty titles and no links, filenames, or targets. There is not enough evidence to treat any pair as the same underlying bug, though the three \"huggingface\" issues look superficially similar.", + "confidence": 0.35, + "canonical_issue_reason": "Issue 43859 is the most recent and the only one with any discussion activity, so it is the least empty representative among otherwise sparse reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue #43859 is the strongest standalone issue in the set because it has the only discussion activity, so it is slightly more substantiated than the other stubs.", + "best_issue_reason": "43859 is the best issue candidate only by marginal metadata signal; none of the issues contains enough substantive description to establish a clearly better canonical report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Only soft title similarity; the titles are different and there is no concrete description showing the same bug or request." + "reason": "Titles are unrelated/generic and there is no body detail or linkage showing the same bug." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Only soft title similarity; 'Fett' and 'huggingface' do not establish the same underlying issue." + "reason": "No substantive evidence connects \"Fett\" to the \"huggingface\" report beyond weak similarity." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "Only soft title similarity with no substantive issue details to confirm a shared problem." + "reason": "Insufficient evidence that these describe the same underlying issue." }, { "left": "issue:43845", "right": "issue:43846", "accept": false, - "reason": "The titles match, but both issues are too vague to confirm the same concrete bug or change; reject on insufficient evidence." + "reason": "Both titles are identical, but the reports are empty enough that identical wording alone is not enough to confirm a duplicate bug." }, { "left": "issue:43845", "right": "issue:43859", "accept": false, - "reason": "Both are generic 'huggingface' stubs, but there is no description-level evidence that they refer to the same underlying issue." + "reason": "Same generic title, but no descriptive content, links, or targets to prove the same underlying problem." }, { "left": "issue:43846", "right": "issue:43859", "accept": false, - "reason": "Both are generic 'huggingface' stubs, but the available metadata is too sparse to safely merge them as duplicates." + "reason": "Same generic title, but no concrete issue details to establish a true duplicate relationship." } ] }, @@ -18234,88 +18182,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "3f5797f8d4a1c7ba87f8b37d8ad9e1d9382c684404771c594def9066d6a85879", + "prepared_review_unit_hash": "debbb935d687b4dc0314ed21df69244119f189dc9b7dea445bc9d0a96689fc63", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5785, - "estimated_input_tokens": 1447, - "estimated_eval_tokens": 3150 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5785, - "estimated_input_tokens": 1447, - "estimated_eval_tokens": 3150 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All five items look like unrelated low-signal issues (spam or generic placeholders), not duplicates of a single underlying bug or change.", - "confidence": 0.98, - "canonical_issue_reason": "None of the issues describe the same concrete problem; they are spam or generic placeholders, so there is no defensible canonical issue.", + "summary": "These are low-signal issue reports (mostly spam-like titles) with no bodies or issue targets. They do not show evidence of a shared underlying bug, so none of the soft pairs should be merged as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 44418 is the only somewhat descriptive title, but it is still a vague documentation suggestion and not a strong duplicate anchor.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both are spam placeholders with no concrete shared bug or change." + "reason": "Both are generic spam-like issues with no concrete description; same title text alone is not enough to establish the same underlying problem." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Both are spam placeholders with no concrete shared bug or change." + "reason": "Same as above: no actionable body or shared target, so this looks like separate spam rather than a duplicate bug report." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "One is spam, the other is a vague documentation suggestion; not the same underlying issue." + "reason": "One is spam-like and the other is a vague documentation suggestion; different problem types and no evidence of a shared issue." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "Spam placeholder and unrelated 'Google LLC' title; no shared problem." + "reason": "Titles are unrelated and there is no body detail or target to indicate the same underlying issue." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "Generic documentation suggestion vs spam placeholder; no evidence of the same issue." + "reason": "A documentation enhancement request is not the same underlying issue as an empty spam report." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "Generic documentation suggestion vs spam placeholder; no evidence of the same issue." + "reason": "Different artifact types and no shared concrete bug/change to merge." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "Both are spam placeholders, but there is no concrete bug/change to treat as a duplicate." + "reason": "Even though both titles are identical spam placeholders, there is no substantive content tying them to one bug or change." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "Spam placeholder and unrelated title; not the same underlying issue." + "reason": "No shared concrete subject matter; one is spam-like, the other has a distinct title." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "Spam placeholder and unrelated title; not the same underlying issue." + "reason": "No evidence of the same underlying problem; the reports are too vague and unrelated." } ] }, @@ -18345,12 +18293,12 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "452a709afc47290b078df51d489c86c8e5a9efac6d0b72ed804b018062f9a3f9", + "prepared_review_unit_hash": "532ca6ccf3a0a227d8b4776db4a7de0875ea7beb06af7378d40e61ee854f991b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1646, + "serialized_chars": 1645, "estimated_input_tokens": 412, "estimated_eval_tokens": 1080 }, @@ -18358,7 +18306,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1646, + "serialized_chars": 1645, "estimated_input_tokens": 412, "estimated_eval_tokens": 1080 }, @@ -18367,27 +18315,27 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two closed issues are very likely duplicates: they have identical titles and highly similar content. #45145 is the better canonical thread because it has the same title and one inbound reference, making it slightly more established.", - "confidence": 0.86, - "canonical_issue_reason": "#45145 is the stronger canonical issue because it shares the identical title with #45151 and has one inbound reference, while #45151 has none.", + "summary": "The two closed issues have the same title and highly similar body length, so they appear to be duplicate reports of the same energy-reduction proposal. Issue #45145 is the better canonical because it was created first and already has an inbound reference.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 is the earliest report and already has inbound reference activity, while #45151 appears to be the later duplicate of the same issue content.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is the best single issue to keep because it appears to be the earlier/more linked instance of the same report and serves as the most central discussion thread.", + "best_issue_reason": "#45145 is the best representative because it is the first-created instance of the identical issue and is likely the original thread for discussion.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", "accept": true, - "reason": "Likely duplicate: the titles are identical, both issues are closed, and the content appears substantially similar. #45145 is slightly better connected due to its inbound reference." + "reason": "Same title and very similar body structure/length; these look like duplicate reports of the same underlying issue." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The grounding is a bit too strong for the packet. What is supported: the two issues have identical titles, both are closed, and #45145 has one inbound reference versus none for #45151, so preferring #45145 as the canonical thread is reasonable. What is not fully supported: calling them the \"exact same proposal,\" saying there are \"no meaningful distinguishing details,\" or asserting they are \"clearly the same underlying issue\" goes beyond the evidence available here. The soft-edge verdict should be phrased more conservatively as a likely duplicate based on identical title and soft similarity, not as a definitive match." + "accept": true, + "feedback": "The conclusion is grounded in the packet: the two closed issues have identical titles, very similar body lengths, and a soft-similarity match. Choosing #45145 as canonical is also conservative because it was created first and has one inbound reference. The soft-edge accept looks reasonable, though the confidence is a bit high given the evidence is still mostly similarity-based." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -18465,7 +18413,7 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "dba55bba673bb7044b65e349f15a5b6efc7e1d9544c54ff7f2ea134d42bda2ff", + "prepared_review_unit_hash": "7b76c4d324e633500176b38c88aa3974c16636b2454b3e95158ff4267cdaf8b6", "component_budget": { "node_count": 2, "item_count": 2, @@ -18487,18 +18435,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are related to 4-bit inference in the quantization stack, but they describe different failures: one is a uint8/torch.finfo TypeError in Gemma4 audio modules, the other is a BitsAndBytesConfig regression where llm_int8_skip_modules drops the default lm_head exclusion and triggers an AssertionError. They should not be deduplicated.", - "confidence": 0.95, - "canonical_issue_reason": "Issue 45674 is the better representative title for a concrete, configuration-specific regression in 4-bit inference; issue 45672 is a separate module/dtype-specific bug in Gemma4 audio paths.", + "summary": "These are not duplicates: one reports a Gemma4 audio-module failure from `torch.finfo()` on `uint8` weights during 4-bit NF4 inference, while the other reports a `BitsAndBytesConfig.llm_int8_skip_modules` interaction that drops the default `lm_head` exclusion and triggers an `AssertionError` in 4-bit inference.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 45674 is the most broadly phrased and actionable of the two, but it still should only represent its own skip-modules regression, not the Gemma4 uint8 finfo failure.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Different underlying bugs and code paths: torch.finfo on uint8 weights in Gemma4 audio modules versus llm_int8_skip_modules clearing the default lm_head exclusion and causing an AssertionError. Same subsystem, but not the same fix." + "reason": "Different root causes and code paths: dtype handling in Gemma4 audio modules versus BitsAndBytesConfig skip-module logic affecting `lm_head` exclusion. They share the same broad 4-bit inference area but not the same underlying bug." } ] }, @@ -18528,40 +18476,40 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "ed3698ef4c4f00ed1dcec75ff2b39843aa272712bdf19e3161d5c2b22ff7ddb1", + "prepared_review_unit_hash": "6a707e5e3448a544ad2b41e5935385a589effb8a31fa7f934e27c53bf68d01f0", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are unrelated empty entries with different personal names/dates; there is no evidence they describe the same bug or request.", - "confidence": 0.99, - "canonical_issue_reason": null, + "summary": "Two closed issues with personal-name/date titles and no body content or technical context; they do not appear to describe the same bug or request.", + "confidence": 0.98, + "canonical_issue_reason": "No strong canonical issue exists; if one must be chosen, issue #45762 is marginally earlier, but it has no meaningful technical content to serve as a true duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": "No issue is a good canonical choice because neither issue contains actionable bug-report content and they do not match each other.", + "best_issue_reason": "Neither issue is a good cluster anchor because both are effectively empty and unrelated, but #45762 is the least arbitrary choice due to earlier creation time.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "Different names and dates, no body, no shared technical context, and no evidence of the same underlying issue." + "reason": "Reject: the titles reference different people and different dates, and there is no shared technical description, issue target, or evidence of the same underlying bug." } ] }, @@ -18591,40 +18539,40 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "46a2a19a2ea32003c87a27c172d543e76b5694bb5ed3653937c1937c7b6b5b0c", + "prepared_review_unit_hash": "3a2a237c243cb1d88df9a47d67bf10f49cbb026ff974555c919ceb6b60b1e6dc", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1481, - "estimated_input_tokens": 371, - "estimated_eval_tokens": 998 + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1481, - "estimated_input_tokens": 371, - "estimated_eval_tokens": 998 + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are not duplicates: they have unrelated personal-name/date titles, no body content, and no shared links or targets indicating a common bug or request.", - "confidence": 0.98, - "canonical_issue_reason": "No clear canonical issue exists here; both issues lack descriptive content and do not appear to report the same underlying problem.", + "summary": "These two closed issues do not look like duplicates of the same bug or change; they are just sparse title-only issues for different named people, with no shared context beyond a surname/date pattern.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the titles refer to different people and contain no shared problem statement or actionable issue context.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a strong representative for a duplicate cluster, because neither contains enough detail to serve as a meaningful canonical bug report.", + "best_issue_reason": "Neither issue is a meaningful representative for a duplicate cluster; both are unrelated, body-less entries.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "These issues do not describe the same bug or change; the titles are unrelated personal entries, and there is no supporting body text or linkage." + "reason": "Reject: only superficial similarity (same surname and date-like text). The titles identify different people and there is no evidence of the same underlying bug or request." } ] }, @@ -18654,12 +18602,12 @@ "soft_pairs": [ "issue:45810|issue:46295" ], - "prepared_review_unit_hash": "9f28e86dfb1f7f3e25815a4e430c734e2eaa48f9872527aa1965b78d1422b7f1", + "prepared_review_unit_hash": "7543f6da070c70847fe812244cb6ba089674f6ce0df8e5960366590f8b1c0970", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1541, + "serialized_chars": 1542, "estimated_input_tokens": 386, "estimated_eval_tokens": 1028 }, @@ -18667,7 +18615,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1541, + "serialized_chars": 1542, "estimated_input_tokens": 386, "estimated_eval_tokens": 1028 }, @@ -18676,18 +18624,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are not duplicates: one requests a new Qwen3_5 token-classification model class, while the other asks for FP8 training support. They target different features and code paths.", - "confidence": 0.97, - "canonical_issue_reason": "issue:45810 is the more concrete and self-contained request, making it the better representative of this cluster if one must be chosen.", + "summary": "The two issues are unrelated: one requests a new Qwen3_5 token-classification/value-model class, while the other asks for FP8 training support. They do not describe the same underlying bug or change.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45810 is the strongest issue to keep as the cluster representative because it specifies a clear model addition with a narrower scope.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45810", "right": "issue:46295", "accept": false, - "reason": "Reject: these are unrelated requests. One is for adding a token-classification model class for Qwen3_5, the other is for FP8 training support; they do not describe the same underlying bug or change." + "reason": "Different feature requests in different areas (model class addition vs FP8 training support); no shared code-path or duplicated underlying change." } ] }, @@ -18727,12 +18675,12 @@ "issue:45871|issue:45873", "issue:45872|issue:45873" ], - "prepared_review_unit_hash": "045e5d4a2df5277b2d526a4dd445a9b157e3dc97acb487dcd2f642451dab8f57", + "prepared_review_unit_hash": "42b0bfa2c742e9248f855f7bcf0e2981a6633f062df11639c2920f1abd55205c", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5727, + "serialized_chars": 5726, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -18740,7 +18688,7 @@ "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5727, + "serialized_chars": 5726, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -18749,66 +18697,66 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The packet contains two separate exact-duplicate issue pairs: 45870/45871 and 45872/45873. Issue 45869 is unrelated to both pairs.", - "confidence": 0.93, - "canonical_issue_reason": "45870 is a good canonical representative for the first exact-duplicate pair (45870/45871). The packet also contains a separate exact-duplicate pair (45872/45873), so this is only a per-pair canonical, not a single representative for the whole packet.", + "summary": "This cluster most plausibly splits into two likely duplicate issue pairs (45870/45871 and 45872/45873) plus one outlier (45869). The evidence supports duplication at the pair level, but not a single canonical issue for the whole cluster.", + "confidence": 0.88, + "canonical_issue_reason": "No single canonical issue cleanly covers the whole cluster. The strongest reading is two likely duplicate pairs\u201445870/45871 and 45872/45873\u2014with 45869 as an unrelated outlier. Without body-text evidence, I would not collapse the two topical pairs into one issue.", "canonical_pr_reason": null, - "best_issue_reason": "45870 is the best overall issue representative because it is the earliest issue in the earliest exact-duplicate pair present; 45869 is unrelated, and 45872/45873 is a separate duplicate pair.", + "best_issue_reason": "As a global representative, 45870 is the most defensible anchor only because it is the earlier item in one of the two likely duplicate pairs. That said, 45872 is equally plausible for the second pair, so the cluster is better handled by splitting than by picking one overall canonical issue.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45869", "right": "issue:45870", "accept": false, - "reason": "Different topics: a financial dataset listing versus a DataForge example request." + "reason": "Financial dataset bundle vs. a prediction example request. The shared financial wording is too broad to treat them as the same issue." }, { "left": "issue:45869", "right": "issue:45871", "accept": false, - "reason": "Different topics: a financial dataset listing versus a DataForge example request." + "reason": "Financial dataset bundle vs. a prediction example request. Similar theme, but clearly different underlying requests." }, { "left": "issue:45870", "right": "issue:45871", "accept": true, - "reason": "Exact duplicate pair: same title and same body length." + "reason": "Strong duplicate candidate: same title, same body length, and the closest similarity signal in the set. I cannot prove exact textual duplication, but they look like the same issue report." }, { "left": "issue:45870", "right": "issue:45872", "accept": false, - "reason": "Related broad area, but different request: one is a DataForge example, the other adds a scraping service." + "reason": "Both mention DataForge and financial time series, but one is an example request and the other adds a scraping service. Different scope, so not the same underlying issue." }, { "left": "issue:45870", "right": "issue:45873", "accept": false, - "reason": "Related broad area, but different request: one is a DataForge example, the other adds a scraping service." + "reason": "Different request scope: the prediction example and the scraping-service variant are not the same issue." }, { "left": "issue:45871", "right": "issue:45872", "accept": false, - "reason": "Related broad area, but different request: one is a DataForge example, the other adds a scraping service." + "reason": "Different request scope: example/prediction versus prediction plus scraping service. Too different to merge as duplicates." }, { "left": "issue:45871", "right": "issue:45873", "accept": false, - "reason": "Related broad area, but different request: one is a DataForge example, the other adds a scraping service." + "reason": "Different request scope: example/prediction versus prediction plus scraping service. Not the same underlying issue." }, { "left": "issue:45872", "right": "issue:45873", "accept": true, - "reason": "Exact duplicate pair: same title and same body length." + "reason": "Strong duplicate candidate: same title, same body length, and the strongest similarity signal for this pair. There is no evidence of a separate underlying request." } ] }, "evaluator_result": { "accept": false, - "feedback": "Rejecting because the representative/canonical rationale overstates the evidence. The packet shows two exact-duplicate pairs: 45870/45871 and 45872/45873, and 45870 is earlier than 45872, so 45872 is not the \"earliest\" duplicate overall. Also, \"matching metadata\" is stronger than what the packet supports; we only see matching titles and body lengths, not identical metadata in general. The summary that the set splits into two duplicate pairs and that 45869 is unrelated is grounded, but the best/canonical issue reasoning is not sufficiently conservative." + "feedback": "The split into two topical pairs plus one outlier is plausible, but the writeup is too certain for the evidence shown. The packet only gives soft-similarity, identical titles, and body lengths; it does not prove the issues are exact textual duplicates. Also, 45870/45871 and 45872/45873 have identical titles, so the claim that one has the \"cleaner, descriptive title\" is not grounded. Please downgrade the duplicate language to a more conservative phrasing unless you have body-text evidence." }, "evaluator_used": true, "retried": true, @@ -18835,7 +18783,7 @@ "soft_pairs": [ "issue:46439|issue:46441" ], - "prepared_review_unit_hash": "9591d506e9c3006aa750cc24244ad9bd73837e0eba530c7e73d92d57d1608f45", + "prepared_review_unit_hash": "5505c40d92acf2d9e764bb68ac0c13d8590b835a908af386f9c1ce781e5752a6", "component_budget": { "node_count": 2, "item_count": 2, @@ -18857,18 +18805,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both issues involve static-cache failures around linear-attention, but they describe different observed failures: one is a general early-initialization shape corruption, the other is a Qwen3.5-specific KeyError for `linear_attention`. They are related, but not clearly the same concrete bug.", - "confidence": 0.76, - "canonical_issue_reason": "Issue 46439 is the better canonical issue because it describes the broader underlying static-cache/linear-attention failure mode in a model-agnostic way, whereas 46441 looks like a specific crash manifestation on Qwen3.5.", + "summary": "Two related static-cache crash reports, but they appear to describe different failure modes: one is linear-attention shape corruption during early initialization, the other is a Qwen3.5 `KeyError` for `linear_attention`.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 46439 is the better canonical issue because it describes the underlying static-cache/linear-attention corruption more generally, rather than one model-specific symptom.", "canonical_pr_reason": null, - "best_issue_reason": "46439 is the best issue to anchor the cluster since it is more general and technically descriptive of the shared subsystem and failure class.", + "best_issue_reason": "46439 is the strongest representative for the cluster because its title points to the broader mechanism and likely root cause affecting linear-attention layers under static cache.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46439", "right": "issue:46441", "accept": false, - "reason": "Related area, but not the same concrete bug: 46439 reports shape corruption from static-cache early initialization, while 46441 reports a `KeyError: 'linear_attention'` on Qwen3.5. The symptoms and likely fix scope are not clearly identical enough to merge as duplicates." + "reason": "Both involve static cache and linear attention, but one reports shape corruption in early initialization while the other is a `KeyError` in Qwen3.5; these do not clearly look like the same underlying bug." } ] }, @@ -18928,7 +18876,7 @@ "pull_request:43823|pull_request:45213", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "acba087464fa8b7ad595de933f7f6c531066dd76b73e8056eee3c508d73c8e2f", + "prepared_review_unit_hash": "0774dc55518e9ce607d8a143712671358f4ecabf8d2e3e6c5e6dddf7e9369317", "component_budget": { "node_count": 14, "item_count": 14, @@ -18950,138 +18898,132 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated PRs that only share broad topical similarity. The only clear duplicate pair is the two loss_utils PRs (43251 and 43254), which both patch the same cross-entropy helper for the same issue. Everything else is too different in code path and intent to merge as duplicates.", - "confidence": 0.96, + "summary": "This cluster is mostly a false-positive similarity group of unrelated pull requests. The only true duplicate-like pair is the two `fixed_cross_entropy` PRs (#43251 and #43254), which implement the same loss-function kwargs support in the same file and target the same issue. All other pairs are broad-subsystem or surface-level lookalikes and should stay separate.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the best canonical PR for the duplicate pair: it targets the same loss_utils code path, has the explicit issue link, and presents the more direct implementation of the cross-entropy kwargs fix.", + "canonical_pr_reason": "PR #43251 is the best canonical representative because it has the clearest, most explicit scope (`Fix(43240): pass kwargs to nn.functional.cross_entropy`) and fully covers the same concrete change as #43254.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the strongest representative artifact in this cluster because it is the only one that cleanly matches another PR as a true duplicate and has the clearest, most specific fix description.", + "best_pr_reason": "PR #43251 is the best fit for the cluster only because it is the sole legitimate duplicate anchor; it is more precise than #43254 and directly describes the code-path fix in `fixed_cross_entropy`.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "VideoPrism model/docs addition vs. executorch dynamic-shape export test; different features and code paths." + "reason": "Unrelated changes: VideoPrism model/docs/Auto wiring vs an ExecuTorch dynamic-shape test." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Mistral tokenizer helper methods vs. adding MobileLLM; unrelated model and tokenizer work." + "reason": "Tokenizer API additions for MistralCommon vs adding a new MobileLLM model." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer method addition vs. export test infrastructure; not the same bug or change." + "reason": "Missing tokenizer methods vs an ExecuTorch export test; no shared bug/change." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor-from-pretrained robustness changes vs. MobileLLM model addition; only broad subsystem overlap." + "reason": "Processor/pretrained robustness fixes vs a new model implementation; different code paths." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Cross-entropy helper fix vs. MobileLLM addition; completely different code paths." + "reason": "Loss-function kwargs support vs adding MobileLLM; unrelated." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Loss helper kwargs fix vs. Mistral4 test fixes; not the same concrete change." + "reason": "Loss utility kwargs fix vs Mistral4 test/model adjustments; not the same change." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Executorch export test vs. Mistral4 tests; both are tests, but they address different behavior." + "reason": "ExecuTorch export test vs Mistral4 fixes; only superficial similarity around testing." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Export dynamic-shape test vs. model addition; unrelated." + "reason": "Dynamic-shape export test vs new MobileLLM model; different artifacts and intent." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "MobileLLM addition vs. config resolution order fix; different areas and objectives." + "reason": "Model addition vs `cls.config` resolution fix; unrelated." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM model addition vs. Falcon refactoring; only broad model-code overlap." + "reason": "New MobileLLM model vs Falcon output-capture refactor; different concerns." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon output-collection refactor vs. model-creation skill docs; not the same change." + "reason": "Falcon refactor vs model-creation skill docs; not the same underlying bug/change." }, { "left": "pull_request:43995", "right": "pull_request:44827", "accept": false, - "reason": "Falcon model refactor vs. Mistral4 test fixes; different model families and fixes." + "reason": "Falcon output-interface refactor vs Mistral4 test fixes; different code paths." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test fixes vs. layrnorm-to-layernorm rename; separate bug fixes in different code paths." + "reason": "Mistral4 test/model changes vs a conversion-mapping typo fix; only broad model-maintenance similarity." }, { "left": "pull_request:44827", "right": "pull_request:46267", "accept": false, - "reason": "Mistral4 test fixes vs. serve chat parsing updates; unrelated functionality." + "reason": "Mistral4 changes vs chat response parsing/serve updates; unrelated." }, { "left": "pull_request:44827", "right": "pull_request:46048", "accept": false, - "reason": "Mistral4 test fixes vs. TDT loss kernel integration; different subsystems and problems." + "reason": "Mistral4 changes vs TDT loss kernel integration; different subsystem and goal." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 test fixes vs. config resolution order fix; no shared concrete bug." + "reason": "Mistral4 fixes vs `cls.config` resolution order fix; unrelated implementation areas." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 test fixes vs. model-creation skill scaffolding; not a duplicate fix." + "reason": "Mistral4 fixes vs model-creation skill draft; not the same change." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "MobileLLM model addition vs. model-creation skill; one is code, the other is process/docs." - }, - { - "left": "pull_request:43995", - "right": "pull_request:44827", - "accept": false, - "reason": "Falcon refactor vs. Mistral4 tests; different models and different fixes." + "reason": "MobileLLM model addition vs model-creation skill docs and scaffolding; not duplicates." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both PRs modify src/transformers/loss/loss_utils.py to extend fixed_cross_entropy with the same cross-entropy kwargs support for the same issue (43240); these are the same underlying fix." + "reason": "Both PRs change `src/transformers/loss/loss_utils.py` to add the same `cross_entropy` kwargs support (`weight`, `label_smoothing`) for the same issue target #43240; these could plausibly be merged into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The output is well grounded in the packet. The only accepted pair, 43251/43254, shares the same file, the same issue target (43240), and the same cross-entropy kwargs fix, so treating them as mergeable duplicates is conservative. The other soft-edge rejections are also cautious and consistent with the diff previews and titles." + "feedback": "The summary is well-grounded in the packet and the only accepted pair (#43251/#43254) is supported by the shared file and shared issue target. The rejections stay conservative: the other pairs are indeed only soft-similarity matches across unrelated subsystems. No overstatement stands out." }, "evaluator_used": true, "retried": false, @@ -19110,12 +19052,12 @@ "pull_request:40546|pull_request:42781", "pull_request:42781|pull_request:46331" ], - "prepared_review_unit_hash": "3799d265c7e9acb00b6854d40408e0b971a1c7e270c6f88b0a8b10a913128dba", + "prepared_review_unit_hash": "048c241623a37ccab685e6cf0820bf4958a309a4a7ca2a7ce5045442cadbe6ac", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10865, + "serialized_chars": 10868, "estimated_input_tokens": 2717, "estimated_eval_tokens": 5690 }, @@ -19123,7 +19065,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10865, + "serialized_chars": 10868, "estimated_input_tokens": 2717, "estimated_eval_tokens": 5690 }, @@ -19132,24 +19074,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are not duplicates: one is the initial VibeVoice implementation, one adds VibeVoice Realtime on top of it, and one is an unrelated Parakeet-RNNT addition. The shared auto-mapping/doc filenames are broad integration files, not evidence of the same underlying change.", - "confidence": 0.96, + "summary": "This cluster contains one base VibeVoice implementation PR, one separate VibeVoice Realtime extension, and one unrelated Parakeet-RNNT PR. The VibeVoice PRs are not duplicates; the Parakeet PR is clearly separate.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 40546 is the best representative because it is the foundational VibeVoice implementation and the broadest, most complete base change in the cluster.", + "canonical_pr_reason": "PR 40546 is the clearest representative because it is the foundational VibeVoice implementation and the earliest, broadest change set in the cluster.", "best_issue_reason": null, - "best_pr_reason": "PR 40546 is the strongest standalone PR to anchor the cluster: it introduces VibeVoice itself, whereas 42781 is a later incremental realtime extension and 46331 is unrelated.", + "best_pr_reason": "PR 40546 is the best cluster representative because it introduces the core VibeVoice model support; the other PRs either extend a different VibeVoice variant or belong to a different model family.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Different scope: 40546 implements core VibeVoice, while 42781 adds VibeVoice Realtime and an acoustic tokenizer. They overlap in shared integration files, but they are not the same concrete change and would not plausibly be merged into one PR without losing the incremental boundary." + "reason": "Both target VibeVoice, but 42781 is a separate VibeVoice Realtime addition rather than the same concrete bug/change as the base implementation. The scope is overlapping, not duplicate." }, { "left": "pull_request:42781", "right": "pull_request:46331", "accept": false, - "reason": "Unrelated models and code paths. 42781 is VibeVoice Realtime; 46331 is Parakeet-RNNT with RNNT loss/modeling changes. The overlap is only in generic ASR auto-mapping/docs infrastructure, not the same bug or feature." + "reason": "These are different model families and code paths: VibeVoice Realtime versus Parakeet-RNNT. Shared auto-registration files are too broad to indicate the same change." } ] }, @@ -19179,7 +19121,7 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "ee3181073c7820f12a1c330f43354a0f7177344bdc7449fc75ca3293bcd55b64", + "prepared_review_unit_hash": "b2e223d7701d36f399031f0681aa6ab79df050b91868752fc9207a963e16b5f7", "component_budget": { "node_count": 2, "item_count": 2, @@ -19201,18 +19143,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs are unrelated feature additions in different model families: TimesFM covariate forecasting vs. DINOv3 image classification support. They do not look like the same underlying change.", + "summary": "The two pull requests are unrelated feature additions: one adds TimesFM covariate forecasting support, and the other adds DINOv3 image classification support. They should not be clustered as duplicates or merged as the same change.", "confidence": 0.99, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No single canonical PR: the PRs address different models, different code paths, and different user-facing features with no meaningful overlap.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "Neither PR is a good global best representative for this cluster because they are independent, unrelated feature PRs rather than duplicates of one another.", "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different subsystems, different model architectures, and different user-facing capabilities; not the same fix or change." + "reason": "Different subsystems and different changes: TimesFM covariate forecasting versus DINOv3 image classification. They do not appear to fix the same underlying bug or implement the same change." } ] }, @@ -19307,7 +19249,7 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "c82ab77d61683735b95d29e5de820fbaeffb7a26f673093d706a79dba1922344", + "prepared_review_unit_hash": "815535a6a04af0fa15091711233a3517a145b1b46cd256c6fdbb067740de30ce", "component_budget": { "node_count": 3, "item_count": 3, @@ -19329,24 +19271,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are model-addition scaffolds in the same broad Transformers area, but they target different models and code paths, so they are not duplicates.", + "summary": "All three are distinct model-addition pull requests (DeepSeek OCR, Isaac, Kimi2-6) with separate model directories, docs, and auto-registration changes. They are superficially similar template-style PRs, but not duplicates of the same underlying bug or feature.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45630 is the strongest representative of the cluster: it is the most extensive and mature model-addition PR, with the broadest set of model-specific files and the highest review activity.", + "canonical_pr_reason": "No canonical PR: these are separate feature PRs for different models, not the same change.", "best_issue_reason": null, - "best_pr_reason": "PR #45630 is the best overall representative because it has the most complete integration surface (config, modeling, processors, auto mappings, tests) and the most discussion, making it the most informative anchor for this cluster.", + "best_pr_reason": "No PR is a true duplicate target; 45630 is the most extensive/reviewed, but it still implements a different model than the others.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different model additions: DeepSeek OCR vs Isaac. They share boilerplate doc/auto-registration files, but the model-specific code, architecture, and tests are distinct, so this is not the same underlying change." + "reason": "DeepSeek OCR vs Isaac: different model families, different new files under src/transformers/models/deepseek_ocr/ vs src/transformers/models/isaac/, so not the same code-path change." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "Different model additions: Isaac vs Kimi2-6. Despite overlapping Transformers scaffolding, they introduce separate model families with different implementation files and behaviors, so they should not be merged as duplicates." + "reason": "Isaac vs Kimi2-6: both add new model scaffolding, but for unrelated models with different implementations, configs, and tests; not mergeable as one duplicate PR." } ] }, @@ -19376,7 +19318,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "edc0265a1c07ca0049fcdb564c284df72b898e0c26b78de639f497c2b6c16aab", + "prepared_review_unit_hash": "d16b2045a30c238cb12f6443652aa2a2473f90e529de9548343cc26332e402e6", "component_budget": { "node_count": 2, "item_count": 2, @@ -19398,18 +19340,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related workflow-maintenance PRs, but they address different upgrade targets and different concrete problems. PR 42975 is a Node 24 compatibility refresh for GitHub Actions; PR 42976 is a separate latest-version bump focused on Docker/build actions. I would not merge them as duplicates.", + "summary": "Two unrelated workflow maintenance PRs: one upgrades many GitHub Actions for Node 24 compatibility, while the other bumps docker/build-push-action in several Docker-related workflows. They overlap in some files but address different actions and different change scopes.", "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 42975 is the more canonical PR for this cluster because it has a clear, specific goal (GitHub Actions Node 24 compatibility) and updates multiple workflow actions in a coherent pass.", + "canonical_pr_reason": "PR 42975 is the broader and more representative workflow-actions upgrade, spanning many files and explicitly targeting Node 24 compatibility; PR 42976 is a narrower Docker action version bump.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the best representative because its scope and title align around one concrete maintenance change, whereas 42976 is a narrower version bump across a different set of actions.", + "best_pr_reason": "PR 42975 is the better cluster representative because its title and diff align with the overarching maintenance theme and it touches the widest set of workflows.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "They overlap in workflow files, but the diffs target different actions and different upgrade goals. 42975 updates core GitHub Actions for Node 24 compatibility; 42976 updates Docker/build-related actions to newer versions. Not the same underlying change." + "reason": "They are both CI/workflow dependency upgrades, but they change different actions (`actions/*` vs `docker/build-push-action`) and do not fix the same concrete problem, so they should not be merged as duplicates." } ] }, @@ -19446,7 +19388,7 @@ "pull_request:43448|pull_request:45490", "pull_request:43451|pull_request:45490" ], - "prepared_review_unit_hash": "e6d4d9dc8eea269047a3a553c8ee5db3623ef876016d17ac8d32b6172997eb65", + "prepared_review_unit_hash": "a438447a3081ee8804976067bb52c18364402eb4edea75d8c70d164c8de5adab", "component_budget": { "node_count": 4, "item_count": 4, @@ -19468,48 +19410,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All four PRs are separate model-addition changes (ViT NEPA, Molmo, Molmo2, CTSM). They only overlap on generic registration/docs/auto-mapping files, not on the same underlying bug or feature, so none should be clustered as duplicates.", + "summary": "These are four separate model-addition PRs (ViT NEPA, Molmo, Molmo2, and CTSM), not duplicate fixes. The overlap is mostly standard docs/auto-registration boilerplate, not the same underlying change.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No true duplicate PR exists in this set; if a single representative is needed, 43451 (Molmo2) is the broadest and most reviewed model-addition PR.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "43451 is the strongest representative by scope and review activity, but it is still a distinct model addition rather than a duplicate of the others.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo). Shared files are generic docs/auto-registry boilerplate, not the same concrete code-path change." + "reason": "Different model additions: ViT NEPA vs Molmo. Shared auto/docs files are boilerplate, not the same code-path or fix." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "ViT NEPA and Molmo2 are distinct model integrations with different implementations and tests; only broad registration files overlap." + "reason": "ViT NEPA and Molmo2 introduce different models with different configs/tests; no same underlying bug or mergeable single change." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "CTSM and ViT NEPA are unrelated model adds. The overlap is limited to shared infrastructure updates, not a single fix or feature." + "reason": "Completely different models and subsystems: vision model addition vs timeseries CTSM. Shared registration files are incidental." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo vs Molmo2 are related names but different model families/codebases; each adds separate configs, processors, and model files, so they are not mergeable as one PR." + "reason": "Molmo and Molmo2 are distinct model families; overlapping docs/auto mappings do not indicate one underlying change." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo and CTSM address different model integrations. Any shared filenames are incidental repo plumbing, not the same underlying change." + "reason": "Molmo vs CTSM are unrelated model additions across different domains; no shared concrete bug or code-path." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are separate feature PRs for different models; they do not fix the same concrete code-path problem." + "reason": "Molmo2 and CTSM touch different model implementations and tests; similarity is limited to generic model-registration boilerplate." } ] }, @@ -19546,7 +19488,7 @@ "pull_request:43665|pull_request:46380", "pull_request:44815|pull_request:46380" ], - "prepared_review_unit_hash": "d32013749f0f5b36a30beb79da01bd193e6a3f8acbe94f9553f5543b82a9555b", + "prepared_review_unit_hash": "7aa242bc75b2c851c7564819643451a95426f3435f12dba0c3f44639b725689d", "component_budget": { "node_count": 4, "item_count": 4, @@ -19568,48 +19510,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These four PRs are not duplicates: they target different code paths and problems (Whisper decoding, test-only CLIP/ViT changes, FP8 dequant/loading, and generation watermarking validation). The shared \"fix\" style titles are misleading; there is no strong evidence of a single underlying bug or change.", - "confidence": 0.96, + "summary": "These four PRs are not duplicates of one another. PR 43291 is a Whisper decoding fix, PR 43665 is mostly test-file noise with arbitrary failing asserts, PR 44815 is an FP8 dequantization fix, and PR 46380 is a generation/watermarking validation fix. They touch different code paths and solve different problems.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "If a representative PR is needed, 46380 is the clearest concrete bug fix: it changes generation validation behavior, has a specific failure mode, and includes broad test coverage.", + "canonical_pr_reason": "PR 44815 is the best canonical representative because it is a substantive, self-contained bug fix with real implementation changes and targeted tests; the others are either unrelated or appear to be test-only noise.", "best_issue_reason": null, - "best_pr_reason": "46380 is the best representative because it has the most explicit production code fix and the strongest signal of a real bug resolution; the others are either unrelated or appear to be test-focused/vague.", + "best_pr_reason": "PR 44815 is the strongest example of a genuine fix: it targets a concrete runtime path in FP8 dequantization, spans the relevant implementation files, and is more coherent than the other PRs.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Different artifacts: Whisper tokenizer/timestamp decoding vs CLIP/ViT test edits. No shared bug or code path." + "reason": "Whisper tokenizer decoding changes are unrelated to the CLIP/ViT test edits; 43665 looks like test-noise rather than the same bug fix." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper decoding changes are unrelated to FP8 dequant/loading logic in core_model_loading and finegrained_fp8." + "reason": "One is a Whisper decode/tokenization fix, the other is an FP8 dequantization fix; different subsystems and different concrete bugs." }, { "left": "pull_request:43291", "right": "pull_request:46380", "accept": false, - "reason": "Whisper tokenizer behavior and generation watermarking validation are separate subsystems with different fixes." + "reason": "Whisper decoding and generation watermarking validation are separate code paths with unrelated failure modes." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "Test-only CLIP/ViT changes do not match the FP8 dequant bugfix; not the same underlying change." + "reason": "43665 only injects arbitrary failing asserts into tests, while 44815 fixes FP8 dequantization logic; not the same change or bug." }, { "left": "pull_request:43665", "right": "pull_request:46380", "accept": false, - "reason": "CLIP/ViT test edits are unrelated to generation config watermarking validation." + "reason": "The bogus test-only edits in 43665 do not match the generation/watermarking validation fix in 46380." }, { "left": "pull_request:44815", "right": "pull_request:46380", "accept": false, - "reason": "FP8 quantization/loading fixes and generation watermarking validation are distinct code paths and failure modes." + "reason": "FP8 dequantization and generation configuration watermarking validation are unrelated concrete issues and would not be merged as one PR." } ] }, @@ -19639,7 +19581,7 @@ "soft_pairs": [ "pull_request:43488|pull_request:43532" ], - "prepared_review_unit_hash": "e88ff2644ffca14120b3956c80f871bd621fe816869a9c06620dedf87095d349", + "prepared_review_unit_hash": "08248353b78bca26a3bdcce85fbd0c06dff274849370872add1e09bd1e0d2f7f", "component_budget": { "node_count": 2, "item_count": 2, @@ -19661,18 +19603,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two pull requests are unrelated: one is a repo-bot/setup version-format test PR, the other rewires CI/workflow behavior and test utilities. The only overlap is a shared Vit test file, which is not enough to indicate the same underlying change.", - "confidence": 0.96, + "summary": "These two pull requests do not look like duplicates: one is a narrow repo-bot formatting/check change, while the other is a broader workflow/testing diff. I would not merge them as the same underlying fix.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "PR 43488 is the cleaner representative: it is a small, focused change around dependency/version formatting and repo-bot checks, with a limited diff and clear intent.", "best_issue_reason": null, - "best_pr_reason": "Neither PR is a strong canonical representative because they target different code paths and purposes; they should not be clustered as duplicates.", + "best_pr_reason": "PR 43488 is the best PR to keep as the canonical example because it is narrowly scoped and self-contained; PR 43532 touches multiple workflows and test-related files, making it less representative of a single concrete fix.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "Different underlying changes: 43488 tweaks setup/versioning and a Vit test, while 43532 changes CI workflows, testing utilities, and reporting. Shared test-file overlap is incidental, not the same bug or fix." + "reason": "Different scopes and likely different purposes: 43488 is a small repo-bot/version-formatting change, while 43532 changes CI/workflow and test infrastructure across several files. They do not appear to fix the same concrete code-path problem." } ] }, @@ -19702,12 +19644,12 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "7f7186956b7273bc67ae354680d3c2de05acc03c1f1a1263cd4d861d4c67ea0d", + "prepared_review_unit_hash": "7aab7591b25c775b52a198ecc4171b3a69fe9ceade3aa67ba67646fa996c4c9e", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6082, + "serialized_chars": 6083, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -19715,7 +19657,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6082, + "serialized_chars": 6083, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -19724,18 +19666,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related documentation/API additions for new segmentation pipelines, but they target different concrete features: promptable concept segmentation vs promptable visual segmentation. They should not be treated as duplicates.", - "confidence": 0.96, + "summary": "Two distinct feature PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share infrastructure but target different tasks, models, tests, and docs, so they are not duplicates.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: the two pull requests implement different pipeline additions with different model/docs/tests and are not the same change.", + "canonical_pr_reason": "No single canonical PR for a duplicate cluster: the two PRs implement different pipelines and should remain separate changes.", "best_issue_reason": null, - "best_pr_reason": "Neither PR is a duplicate of the other; if one must be highlighted, 43613 is slightly broader in model/test coverage, but it does not subsume 43612.", + "best_pr_reason": "PR #43613 is the broader representative of the shared multimodal-segmentation theme, but it is still not a duplicate of #43612.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different underlying pipelines and code paths: concept segmentation vs visual segmentation. Shared scaffolding files are not enough to merge them as the same PR." + "reason": "Different pipeline features with different concrete targets: concept segmentation vs visual segmentation. Shared docs/registry scaffolding does not make them the same change." } ] }, @@ -19765,7 +19707,7 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "ea7dbb2813a17d550e472787361097e3adfa2d4941ddf955b2b965131cb903d4", + "prepared_review_unit_hash": "2bec84ad19f2ef6eee7c7cfcc7c4037fa487063c3531914cb75dac36fa30523c", "component_budget": { "node_count": 2, "item_count": 2, @@ -19787,18 +19729,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both items are open PRs in `trainer.py`, but they address different trainer extension points: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering for custom/PEFT trainers. They are not the same underlying change and should remain separate.", - "confidence": 0.98, + "summary": "These are two separate trainer.py enhancement PRs: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering for custom trainers/PEFT. They share a file and broad subsystem but not the same underlying change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43651 is the more concrete, core-behavior change in `training_step` and is the better representative if one PR must be chosen for the cluster.", + "canonical_pr_reason": "PR 43651 is the better canonical representative because it has more discussion and a clearly scoped functional change in the training loss path.", "best_issue_reason": null, - "best_pr_reason": "PR 43651 is the strongest standalone representative because it directly changes loss normalization behavior in the main training path; PR 43663 is a separate refactor for column filtering.", + "best_pr_reason": "PR 43651 is the strongest standalone PR in this cluster due to higher discussion activity and a concrete, user-visible behavior change; 43663 is a different trainer extensibility refactor.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Different problems and code paths: 43651 adds `_loss_is_scaled_for_ga` to control gradient-accumulation loss scaling, while 43663 adds `_get_signature_columns` to customize input column filtering. They are not plausibly one merged fix." + "reason": "Reject: they touch the same file but fix different code paths\u2014loss scaling during training vs. signature-column filtering for input preprocessing. They would not plausibly be merged as one PR." } ] }, @@ -19828,7 +19770,7 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "72ae7da7f2e8cd96f79590398fa4a3a322b5387a4d8b598d177b3e67738f8bdb", + "prepared_review_unit_hash": "64d3c9b0be0ac24b2f646f8a90d308a3c6804d2dbbb899f5cd5883773d6d9d6b", "component_budget": { "node_count": 2, "item_count": 2, @@ -19850,25 +19792,28 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: both PRs edit `src/transformers/cli/serve.py`, but they appear to fix different NameError cases in the CLI type-hint/runtime path. PR 43656 is a broader annotation/runtime typing cleanup, while PR 43842 specifically targets `TypeAdapter` import resolution when `pydantic` is absent.", - "confidence": 0.89, + "summary": "Both PRs target the same import-time `TypeAdapter` NameError in `src/transformers/cli/serve.py`; they differ mainly in implementation style, not in the underlying bug. PR 43842 is the more focused fix.", + "confidence": 0.77, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR for a duplicate cluster: the two PRs do not look like the same underlying fix and should not be merged as duplicates.", + "canonical_pr_reason": "PR 43842 is the clearest representative: it directly guards the `TypeAdapter` reference with `TYPE_CHECKING`/string annotations and is narrowly scoped to the import-time failure.", "best_issue_reason": null, - "best_pr_reason": "43842 is the more focused and clearly scoped fix for the `TypeAdapter`/`pydantic`-optional NameError in the CLI, with a minimal change set and an explicit issue target.", + "best_pr_reason": "PR 43842 is the best fit for the cluster because it targets the concrete `TypeAdapter` NameError with minimal collateral changes and an explicit issue link.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", - "accept": false, - "reason": "Reject: same file and similar title, but the diffs indicate different underlying problems. 43656 is a broader typing/annotation cleanup, while 43842 specifically handles `TypeAdapter` import availability when `pydantic` is not installed." + "accept": true, + "reason": "Both PRs fix the same `TypeAdapter` NameError in the CLI serve module. 43656 uses a broader `from __future__ import annotations` approach, while 43842 uses a direct `TYPE_CHECKING`/quoted-annotation guard; these are alternate implementations of the same underlying fix." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs touch `src/transformers/cli/serve.py`, both are titled around the `TypeAdapter` NameError, and the soft-edge reasoning stays conservative by describing them as alternate implementations of the same fix. I don\u2019t see an overstatement or a mergeability issue here." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -20053,12 +19998,12 @@ "pull_request:44044|pull_request:44129", "pull_request:44044|pull_request:44154" ], - "prepared_review_unit_hash": "f67088dd7737804559862bf5f0059202ee3f32f8f215ea1553b42b6f28b4a363", + "prepared_review_unit_hash": "06c18c038c8dea7dca482fa7227add62ee3956a53f8ba470b4e790bb9006d782", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84892, + "serialized_chars": 84890, "estimated_input_tokens": 21223, "estimated_eval_tokens": 42702 }, @@ -20066,7 +20011,7 @@ "node_count": 27, "item_count": 27, "soft_pair_count": 72, - "serialized_chars": 75088, + "serialized_chars": 75085, "estimated_input_tokens": 18772, "estimated_eval_tokens": 37800 }, @@ -20075,132 +20020,132 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a broad output-tracing refactor cluster across unrelated models. The only true duplicate/mergeable pairs are within the same concrete GPT-J/CodeGen and GPT-Neo code paths; the RegNet/FNet-CVT pair must be rejected because they are different architectures and not mergeable as one bugfix. I kept the triage conservative and only accepted same-file, same-change pairs.", - "confidence": 0.9, + "summary": "This cluster is mostly a set of independent model-by-model output-tracing refactors under the same umbrella issue (#43979). The only clear duplicate subgroup is the GPT-J / CodeGen set, with a smaller duplicate pair in GPT-Neo.", + "confidence": 0.81, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44722 is the best canonical representative because it is the most consolidated GPT-J/CodeGen output-tracing refactor, spans the copied code paths in both files, and matches the shared change pattern most completely.", + "canonical_pr_reason": "No issue artifact here; for the PR side, 44722 is the best canonical representative because it is the broadest/latest GPT-J/CodeGen output-tracing refactor and overlaps the clear duplicate subgroup.", "best_issue_reason": null, - "best_pr_reason": "PR 44722 is the strongest single representative of the cluster\u2019s real duplicate core: the GPT-J/CodeGen output-tracing migration. It is the most complete version of that concrete change and is mergeable with the other GPT-J variants.", + "best_pr_reason": "44722 is the strongest single PR to represent the duplicate subgroup: it covers the same GPT-J output-tracing change as the other GPT-J PRs and also includes the CodeGen copy, making it the broadest/revised version.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families/code paths (FNet/CVT vs GPT-J); shared issue target alone is not enough." + "reason": "Different models/files (CVT/FNet vs GPT-J); same tracking issue is not enough." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Different model families and concrete refactors; only the output-tracing theme overlaps." + "reason": "Different model families and code paths; not the same change." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Same GPT-J output-tracing refactor in the same code path; these are mergeable variants." + "reason": "Same GPT-J output-tracing refactor in `modeling_gptj.py`; these look like overlapping versions of the same change and could be merged." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "RegNet vs EfficientNet are different models and changes; same refactor theme only." + "reason": "RegNet vs EfficientNet; separate model-specific migrations." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "Different model families and code paths; not the same concrete bugfix." + "reason": "EfficientNet vs GPT-J/CodeGen; unrelated code paths." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen vs EfficientNet; only the standardized tracing pattern is shared." + "reason": "GPT-J/CodeGen vs EfficientNet; different models." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 vs DeBERTa are unrelated implementations." + "reason": "MobileNetV2 vs DeBERTa-v2; unrelated refactors." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen vs MGP-STR; same framework idea, different concrete code paths." + "reason": "GPT-J/CodeGen vs MGP-STR; different model implementations." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen vs MPT; not the same underlying change." + "reason": "GPT-J/CodeGen vs MPT; different model families." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Same GPT-Neo output-tracing migration in the same file/copy path; mergeable variants." + "reason": "Same GPT-Neo file and same output-tracing migration; one looks like a later variant of the same refactor." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo vs GPT-J/CodeGen; different model implementations." + "reason": "GPT-Neo vs GPT-J/CodeGen; different underlying model code." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen vs GPT-Neo; related refactor style, but different code paths." + "reason": "GPT-J/CodeGen vs GPT-Neo; not the same model or code path." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet vs GPT-J/CodeGen; different model families and code paths." + "reason": "ResNet vs GPT-J/CodeGen; unrelated model changes." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR vs GPT-J/CodeGen; only the decorator migration is common." + "reason": "MGP-STR vs GPT-J/CodeGen; different code paths." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet vs GPT-J; different concrete model code paths." + "reason": "ResNet vs GPT-J; separate model-specific refactors." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo vs ResNet; same refactor theme only." + "reason": "GPT-Neo vs ResNet; different model families." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet vs MPT; unrelated implementations." + "reason": "ResNet vs MPT; unrelated changes." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo vs Flaubert; different model families/code paths." + "reason": "GPT-Neo vs Flaubert; different model implementations." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo vs GPT-J/CodeGen; related theme, not the same change." + "reason": "GPT-Neo vs GPT-J/CodeGen; not the same refactor." }, { "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo vs Flaubert; only the output-tracing framework overlaps." + "reason": "GPT-Neo vs Flaubert; unrelated model code." }, { "left": "pull_request:44071", @@ -20212,25 +20157,25 @@ "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo vs GPT-J; different model families." + "reason": "GPT-Neo vs GPT-J; different underlying model." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo vs MGP-STR; different concrete changes." + "reason": "GPT-Neo vs MGP-STR; unrelated." }, { "left": "pull_request:44068", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo vs MGP-STR; same refactor style only." + "reason": "GPT-Neo vs MGP-STR; different models." }, { "left": "pull_request:44071", "right": "pull_request:44085", "accept": false, - "reason": "MPT vs GPT-J; unrelated model code." + "reason": "MPT vs GPT-J; unrelated." }, { "left": "pull_request:44059", @@ -20242,103 +20187,79 @@ "left": "pull_request:44085", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J vs MGP-STR; only the output-tracing migration is shared." + "reason": "GPT-J vs MGP-STR; not the same change." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT2 vs GPT-J/CodeGen; different model families." + "reason": "GPT2 vs GPT-J/CodeGen; different transformer blocks." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT vs Flaubert/XLM; different architectures and code paths." + "reason": "ViLT vs Flaubert/XLM; separate model migrations." }, { "left": "pull_request:44018", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo vs MPT; unrelated model implementations." + "reason": "GPT-Neo vs MPT; different model families." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer vs GPT-J/CodeGen; same refactor theme only." + "reason": "SegFormer vs GPT-J/CodeGen; unrelated code paths." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer vs LongT5; different concrete code paths." + "reason": "SegFormer vs LongT5; separate refactors." }, { "left": "pull_request:44017", "right": "pull_request:44073", "accept": false, - "reason": "SegFormer vs VisualBert; different models." + "reason": "SegFormer vs VisualBERT; different model code." }, { "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo vs MPT; not the same bug/change." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44086", - "accept": false, - "reason": "MPT vs MGP-STR; unrelated implementations." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44073", - "accept": false, - "reason": "RegNet vs VisualBert; different model families." - }, - { - "left": "pull_request:44017", - "right": "pull_request:44018", - "accept": false, - "reason": "SegFormer vs GPT-Neo; only the output-tracing theme overlaps." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44056", - "accept": false, - "reason": "RegNet vs MPNet; different concrete code paths." + "reason": "GPT-Neo vs MPT; unrelated model changes." }, { "left": "pull_request:44071", - "right": "pull_request:44129", + "right": "pull_request:44076", "accept": false, - "reason": "MPT vs SpeechT5; unrelated model implementations." + "reason": "MPT vs ImageGPT; different implementations." }, { "left": "pull_request:44074", "right": "pull_request:44101", "accept": false, - "reason": "TextNet vs Flaubert/XLM; different code paths." + "reason": "TextNet vs Flaubert/XLM; not the same change." }, { "left": "pull_request:43997", "right": "pull_request:44018", "accept": false, - "reason": "RegNet vs GPT-Neo; different model families." + "reason": "RegNet vs GPT-Neo; different model-specific refactors." }, { "left": "pull_request:44056", "right": "pull_request:44161", "accept": false, - "reason": "MPNet vs LongT5; different model code paths." + "reason": "MPNet vs LongT5; unrelated code paths." }, { "left": "pull_request:43997", "right": "pull_request:44161", "accept": false, - "reason": "RegNet vs LongT5; same refactor theme only." + "reason": "RegNet vs LongT5; separate model migrations." }, { "left": "pull_request:44074", @@ -20346,128 +20267,134 @@ "accept": false, "reason": "TextNet vs LongT5; different models." }, - { - "left": "pull_request:44018", - "right": "pull_request:44161", - "accept": false, - "reason": "GPT-Neo vs LongT5; unrelated implementations." - }, { "left": "pull_request:43997", "right": "pull_request:44071", "accept": false, - "reason": "RegNet vs MPT; different concrete model paths." + "reason": "RegNet vs MPT; unrelated." }, { "left": "pull_request:44010", "right": "pull_request:44071", "accept": false, - "reason": "SqueezeBERT vs MPT; different model families." + "reason": "SqueezeBert vs MPT; different model families." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo vs SpeechT5; not the same code path." }, { "left": "pull_request:44071", "right": "pull_request:44129", "accept": false, - "reason": "MPT vs SpeechT5; different code paths." + "reason": "MPT vs SpeechT5; unrelated model changes." }, { "left": "pull_request:44010", "right": "pull_request:44073", "accept": false, - "reason": "SqueezeBERT vs VisualBert; same theme only." + "reason": "SqueezeBert vs VisualBERT; separate refactors." }, { "left": "pull_request:43997", "right": "pull_request:44074", "accept": false, - "reason": "RegNet vs TextNet; different concrete implementations." + "reason": "RegNet vs TextNet; different model code." }, { - "left": "pull_request:44010", - "right": "pull_request:44018", + "left": "pull_request:44056", + "right": "pull_request:44101", "accept": false, - "reason": "SqueezeBERT vs GPT-Neo; unrelated model code." + "reason": "MPNet vs Flaubert/XLM; unrelated." }, { - "left": "pull_request:44056", + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT2 vs MPT; different model families." + }, + { + "left": "pull_request:44101", "right": "pull_request:44161", "accept": false, - "reason": "MPNet vs LongT5; not the same change." + "reason": "Flaubert/XLM vs LongT5; not a duplicate." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44018", + "accept": false, + "reason": "SegFormer vs GPT-Neo; different underlying model." }, { "left": "pull_request:43997", - "right": "pull_request:44071", + "right": "pull_request:44056", "accept": false, - "reason": "RegNet vs MPT; different code paths." + "reason": "RegNet vs MPNet; unrelated changes." }, { - "left": "pull_request:43996", - "right": "pull_request:44017", + "left": "pull_request:44071", + "right": "pull_request:44076", "accept": false, - "reason": "FNet/CVT vs SegFormer; only the output-tracing pattern overlaps." + "reason": "MPT vs ImageGPT; different code paths." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBERT vs DeBERTa; different model families." + "reason": "SqueezeBert vs DeBERTa-v2; unrelated refactors." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa vs TextNet; unrelated implementations." + "reason": "DeBERTa-v2 vs TextNet; different model-specific changes." }, { "left": "pull_request:44066", "right": "pull_request:44722", "accept": true, - "reason": "Same GPT-J/CodeGen output-tracing migration; 44722 is the consolidated follow-up." + "reason": "Both are GPT-J/CodeGen output-tracing refactors over the same copied block; 44722 looks like the broader/revised version." }, { "left": "pull_request:44085", "right": "pull_request:44722", "accept": true, - "reason": "Same GPT-J output-tracing refactor in the same code path; these are mergeable." + "reason": "Same GPT-J output-tracing change in `modeling_gptj.py`; 44722 subsumes it and also updates the CodeGen copy." }, { "left": "pull_request:44085", "right": "pull_request:44129", "accept": false, - "reason": "GPT-J vs SpeechT5; same issue tag does not make them duplicates." + "reason": "GPT-J vs SpeechT5; different model families." }, { "left": "pull_request:43996", "right": "pull_request:44007", "accept": false, - "reason": "FNet/CVT vs RegNet are different architectures and code paths; not mergeable as one fix." + "reason": "CVT/FNet vs RegNet/ResNet; unrelated model refactors." }, { "left": "pull_request:43996", "right": "pull_request:44129", "accept": false, - "reason": "FNet/CVT vs SpeechT5; different models and code paths." + "reason": "CVT/FNet vs SpeechT5; different code paths." }, { "left": "pull_request:44044", "right": "pull_request:44129", "accept": false, - "reason": "DeBERTa vs SpeechT5; different concrete changes." - }, - { - "left": "pull_request:44044", - "right": "pull_request:44154", - "accept": false, - "reason": "DeBERTa vs ViTS; unrelated model implementations." + "reason": "DeBERTa-v2 vs SpeechT5; not the same underlying change." } ] }, "evaluator_result": { - "accept": false, - "feedback": "Reject: one soft-edge acceptance is not grounded in the packet. The verdict for pull_request:43996 vs pull_request:44007 claims they are the same RegNet refactor in the same file, but 43996 is actually FNet/CVT (src/transformers/models/cvt/modeling_cvt.py and fnet), while 44007 is RegNet/ResNet/RT-DETR ResNet. Those changes do not look mergeable into a single PR for the same bugfix, so this should have stayed rejected. The rest of the summary is broadly conservative, but this over-acceptance is enough to fail precision." + "accept": true, + "feedback": "The summary is grounded in the packet: most PRs are separate model-specific output-tracing refactors under the same umbrella issue, and the only clearly mergeable overlap clusters are the GPT-J/CodeGen subgroup and the GPT-Neo pair. The soft-edge accept/reject calls are conservative and match the diff scope shown." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -20502,12 +20429,12 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "8a73d03bd947d99ddd94c9898e645f395b4253e949bd54821359196dad8d3464", + "prepared_review_unit_hash": "2ad8556b96deb4e5a57acbba9808c6af32da4e120d15bb7267621818ff112291", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84892, + "serialized_chars": 84890, "estimated_input_tokens": 21223, "estimated_eval_tokens": 42702 }, @@ -20515,7 +20442,7 @@ "node_count": 8, "item_count": 8, "soft_pair_count": 6, - "serialized_chars": 17125, + "serialized_chars": 17126, "estimated_input_tokens": 4282, "estimated_eval_tokens": 8820 }, @@ -20524,48 +20451,48 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are all model-specific PRs for the same broad refactor theme (standardized output tracing / capture_outputs), but they target different code paths and models, so they should not be deduplicated into one change.", - "confidence": 0.95, + "summary": "These PRs are all model-specific refactors that migrate different architectures to standardized output tracing/capture_outputs patterns. They are related in theme, but they do not describe the same bug or the same code-path change, so they should not be deduplicated together.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "44129 is the strongest standalone representative because it has the clearest intent, an explicit issue link, and a concrete output-tracing refactor in a single model.", "best_issue_reason": null, - "best_pr_reason": "PR 44073 is the clearest representative of the shared refactor pattern, but it is still only a VisualBERT-specific implementation and not a duplicate of the other model PRs.", + "best_pr_reason": "44129 is the best single PR to keep as the representative example of this refactor pattern: it is self-contained, specific, and tied to an explicit tracking issue.", "soft_edge_verdicts": [ { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "Both are standardized-output refactors, but they touch different models (FocalNet vs ImageGPT) and separate code paths." + "reason": "Different models and different implementation paths; both are similar output-tracing refactors, but they are not the same underlying fix." }, { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Same refactor style, but VisualBERT and TextNet are distinct implementations with model-specific output handling." + "reason": "VisualBert and TextNet are unrelated model codepaths, so these are parallel refactors rather than duplicate changes." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Different models and different encoder/attention internals; similar pattern only, not the same bug or change." + "reason": "Same broad standardized-output theme, but MPNet and VisualBert have distinct model-specific tracing logic and are not mergeable as one duplicate PR." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "MPNet and TextNet are separate model refactors, not one mergeable fix." + "reason": "These touch different models and different output-capture setups; similarity is thematic only." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model implementations; they share only the generic output-tracing migration theme." + "reason": "RegNet and SqueezeBert are separate architectures with separate output-handling code, so this is not the same change." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SpeechT5 has its own explicit tracing/output change and issue link, but it is not the same concrete change as SqueezeBert." + "reason": "Both are output-tracing refactors, but they affect different models and code paths; the shared concept is too broad to treat as a duplicate." } ] }, @@ -20650,298 +20577,310 @@ "pull_request:44028|pull_request:46438", "pull_request:43998|pull_request:46438" ], - "prepared_review_unit_hash": "441dc81ff174ce062fff68486e2b318c0e764787885a129d2ac7c3c19d552f9a", + "prepared_review_unit_hash": "55990ded056047f06528309ba3f83e760ed52f3c5b6e32f14c9b66587765df17", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 44, - "serialized_chars": 43568, - "estimated_input_tokens": 10892, - "estimated_eval_tokens": 22040 + "serialized_chars": 43563, + "estimated_input_tokens": 10891, + "estimated_eval_tokens": 22038 }, "budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 44, - "serialized_chars": 43568, - "estimated_input_tokens": 10892, - "estimated_eval_tokens": 22040 + "serialized_chars": 43563, + "estimated_input_tokens": 10891, + "estimated_eval_tokens": 22038 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are all related to the same broad refactor theme (output tracing / can_return_tuple), but they touch different model implementations and code paths, so they are not duplicates of one another.", + "summary": "All items are separate model-specific output-tracing refactors (or one utility speedup), not duplicates of the same bug/change. They share a common theme, but each PR targets different model implementations and code paths, so none should be merged as duplicate PRs.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: each change is scoped to a different model or utility path, so there is no one PR that subsumes the others as a duplicate.", + "canonical_pr_reason": "PR 44003 is the best representative of the cluster because it is a broader output-tracing refactor touching two related model files (FalconMamba and Mamba) and clearly reflects the common pattern in the set.", "best_issue_reason": null, - "best_pr_reason": "PR 44025 is the most coherent multi-file model refactor in the set, but it is still only the Depth Anything family and not a duplicate target for the rest.", + "best_pr_reason": "PR 44003 is the strongest single representative of the shared refactor pattern, with the broadest model coverage among the model PRs shown.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Different models (`timm_backbone` vs `upernet`); same refactor pattern, not the same bug/change." + "reason": "Same refactor theme, but different model families and files (timm_backbone vs upernet); not the same underlying change." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Different model families and forward paths; only the output-tracing pattern matches." + "reason": "Different model implementations and code paths; both are output-tracing refactors only at a high level." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Separate model implementation (`timm_backbone` vs `univnet`), so not one duplicate change." + "reason": "Separate model families (timm_backbone vs univnet); no shared concrete bug or mergeable fix." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "`timm_backbone` and `falcon_mamba/mamba` are unrelated code paths; refactor theme alone is insufficient." + "reason": "Different architectures and files; similar refactor pattern does not make them duplicates." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "Different architectures and modules; both are tracing refactors but not the same fix." + "reason": "Timm backbone refactor vs CodeGen attention/cache refactor are distinct code paths." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "`timm_backbone` and `rwkv` touch different code paths; not mergeable as one duplicate PR." + "reason": "Unrelated model families and implementations; only the generic tracing refactor theme matches." }, { "left": "pull_request:43998", "right": "pull_request:46438", "accept": false, - "reason": "Model refactor vs modular-conversion tooling; completely different change types." + "reason": "Model refactor vs modular-conversion utility speedup are unrelated changes." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "Different models (`mobilenet_v1` vs `univnet`); same tracing cleanup pattern only." + "reason": "MobileNetV1 and UnivNet are unrelated model code paths; not the same bug/change." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different subsystems and forward logic; not the same underlying bug or change." + "reason": "Different models and outputs; same broad tracing refactor only." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "Different model families; both refactor outputs, but they do not fix the same code path." + "reason": "MobileNetV1 vs Depth Anything are distinct model implementations; not duplicates." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "`mobilenet_v1` and `vision_encoder_decoder` are unrelated implementations, so not duplicates." + "reason": "Different model families and files; only similar refactoring style." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Different encoder-decoder stack vs MobileNet; too broad a similarity to merge." + "reason": "MobileNetV1 and Speech Encoder-Decoder fix different code paths." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "`vision_text_dual_encoder` and `univnet` are different model paths; same refactor intent only." + "reason": "Vision/Text dual encoder and UnivNet are unrelated models; no common concrete defect." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Different model code paths (`vision_text_dual_encoder` vs `upernet`); not a duplicate fix." + "reason": "Different model families and implementations; same tracing refactor theme only." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Different architectures and outputs; similar API cleanup, not the same change." + "reason": "Different model stacks and files; not plausibly one PR." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Vision/Text dual encoder vs CodeGen are separate code paths and fixes." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Distinct model families and output contracts; not the same concrete bug/change." + "reason": "Different model families and output semantics; not the same underlying change." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "`univnet` and `upernet` are unrelated models; only the refactor style matches." + "reason": "UnivNet and UperNet are unrelated implementations; no duplicate bug/change." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and model files; same refactor motif only." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "Different model families and code paths; not plausibly one merged PR." + "reason": "UnivNet vs Depth Anything are distinct model-specific refactors." }, { "left": "pull_request:44002", - "right": "pull_request:44029", + "right": "pull_request:44003", "accept": false, - "reason": "`upernet` and `rwkv` are unrelated implementations; same tracing cleanup is too generic." + "reason": "UperNet and Mamba/FalconMamba changes are unrelated code paths." }, { "left": "pull_request:44002", - "right": "pull_request:44003", + "right": "pull_request:44004", "accept": false, - "reason": "Different model stacks (`upernet` vs `falcon_mamba/mamba`); not the same underlying fix." + "reason": "Different models and mechanics; one is segmentation, the other CodeGen cache/output handling." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "`mamba`/`falcon_mamba` versus `codegen` are different code paths; not duplicate changes." + "reason": "Mamba/FalconMamba output tracing and CodeGen cache/output refactor address different model internals." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "Different model architectures and files; same broad refactor theme is not enough." + "reason": "Different model families and concrete changes; same high-level theme is insufficient." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "`univnet` and `codegen` are unrelated; these are not the same change." + "reason": "Separate model implementations with different forward paths and return behavior." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Different model code paths and APIs; refactor similarity only." + "reason": "Different concrete model code paths; not mergeable as a single duplicate fix." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Different wrapper models (`depth_anything` vs `vision_encoder_decoder`); not one duplicate bug." + "reason": "Depth Anything and Vision/Encoder-Decoder are unrelated models; similar refactor pattern only." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Separate model families; output-tracing refactor alone does not make them duplicates." + "reason": "Different encoder/decoder model families and files; not the same underlying issue." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "Different models and return types; not the same concrete change." + "reason": "Depth Anything vs SuperPoint target unrelated model code paths." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "`depth_anything` and `rwkv` are unrelated code paths; reject as duplicate." + "reason": "Different model families and refactor scopes; only generic tracing work overlaps." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Different models (`depth_anything` vs `dpr`); same API cleanup only." + "reason": "Depth Anything and DPR are distinct model implementations; not duplicates." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "Encoder-decoder wrappers for different modalities; not the same underlying fix." + "reason": "Vision Encoder-Decoder and Speech Encoder-Decoder are separate models with different code paths." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "Different models (`vision_encoder_decoder` vs `superpoint`); no shared code path." + "reason": "Different model families and concrete refactors; not one underlying bug." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "`vision_encoder_decoder` and `rwkv` are unrelated; only the refactor theme overlaps." + "reason": "Unrelated model internals despite similar output-tracing theme." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "Different model implementations and outputs; not a duplicate PR." + "reason": "Different architectures and files; cannot plausibly be merged into one PR." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "`speech_encoder_decoder` and `superpoint` are separate model paths; not mergeable as one change." + "reason": "Speech Encoder-Decoder and SuperPoint are unrelated changes." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "Different architectures and return paths; same tracing refactor pattern only." + "reason": "Different model families and separate output handling code paths." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "`speech_encoder_decoder` and `dpr` are unrelated implementations; reject as duplicate." + "reason": "Different model implementations; only the general refactor style matches." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "Different models (`superpoint` vs `rwkv`); too little overlap beyond cleanup style." + "reason": "SuperPoint and Mamba/FalconMamba are unrelated model code paths." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "`superpoint` and `dpr` do not share the same code path or bug." + "reason": "Different models and different forward/output semantics; not duplicates." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "Different model families and forward contracts; not the same change." + "reason": "Mamba/FalconMamba and DPR are unrelated model implementations and fixes." }, { "left": "pull_request:44000", "right": "pull_request:46438", "accept": false, - "reason": "Model code refactor vs modular conversion tool change; entirely different area." + "reason": "Vision/Text dual encoder refactor is unrelated to the modular-conversion pool reuse utility change." }, { "left": "pull_request:44027", "right": "pull_request:46438", "accept": false, - "reason": "Unrelated model wrapper refactor and tooling optimization; not duplicates." + "reason": "Speech Encoder-Decoder model refactor and modular conversion utility speedup are unrelated." }, { "left": "pull_request:44029", "right": "pull_request:46438", "accept": false, - "reason": "`rwkv` model code and modular-conversion utility are unrelated changes." + "reason": "Model output-tracing refactor vs utility process-pool optimization are different changes." }, { "left": "pull_request:44028", "right": "pull_request:46438", "accept": false, - "reason": "Different subsystem and intent; not the same underlying fix." + "reason": "SuperPoint model refactor and modular-conversion utility speedup are unrelated." }, { "left": "pull_request:43998", "right": "pull_request:46438", "accept": false, - "reason": "`timm_backbone` model refactor and conversion-tool pool reuse are unrelated." + "reason": "Model refactor vs utility performance change; no shared concrete bug or code path." } ] }, @@ -20971,12 +20910,12 @@ "soft_pairs": [ "pull_request:44662|pull_request:45401" ], - "prepared_review_unit_hash": "0f63a06fc8dbd3976104c21a78ed74a1b2d3f8ae59a39dc69c361623771c71d6", + "prepared_review_unit_hash": "32ae3f77219dce503b473f82903f8b05ebbdeb9fda6aa8867cd00bfbd6211061", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6550, + "serialized_chars": 6551, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -20984,7 +20923,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6550, + "serialized_chars": 6551, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -20993,18 +20932,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated model-addition PRs for different architectures (PenguinVL vs Voxtral TTS). They only overlap on standard Transformers integration files and documentation plumbing, not on the same concrete bug or change.", - "confidence": 0.98, + "summary": "Both pull requests are new, unrelated model integrations: one adds PenguinVL and the other adds Voxtral TTS. Their overlap is limited to shared Transformers registration/docs scaffolding, which is not enough to treat them as the same underlying change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44662 is the stronger representative of the cluster: it is a non-draft, broader model implementation with substantial tests and review activity. PR 45401 is draft and tied to a specific issue for a different model.", + "canonical_pr_reason": "No canonical PR: these are separate model additions with different code paths, tests, and targets.", "best_issue_reason": null, - "best_pr_reason": "PR 44662 is the best overall fit because it is the more mature and complete implementation PR; PR 45401 is still draft and addresses a different model entirely.", + "best_pr_reason": "Neither PR is a global best representative for this cluster because the cluster is not a true duplicate set.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different model additions with different targets and code paths. Shared auto/docs scaffolding is generic Transformers integration work, not evidence of the same underlying change." + "reason": "Reject: PenguinVL implementation and Voxtral TTS support are distinct model ports with different architecture, files, and goals; shared auto-registration/docs edits are boilerplate and do not indicate the same bug or change." } ] }, @@ -21038,7 +20977,7 @@ "pull_request:45363|pull_request:46339", "pull_request:45666|pull_request:46339" ], - "prepared_review_unit_hash": "016a1a9f0a1bf5064a8a01b24b8c6f384adcce21f9f1ddd5b589c49bb646b70f", + "prepared_review_unit_hash": "2d86ac1456ca38e68f3406fbbdb96c0fc66e5156e407b7a55026deccc8c9d2ce", "component_budget": { "node_count": 4, "item_count": 4, @@ -21060,36 +20999,36 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a single feature line around n-to-1 kernel fusion / `KernelConfig`, with one foundational API PR and two later extensions/refinements. The newest PR appears to subsume the earlier work best.", - "confidence": 0.83, + "summary": "The cluster splits into two groups: a foundational Module Fusion API PR, and a sequence of n-to-1 KernelConfig fusion PRs that look like iterative revisions of the same feature work. I would keep the Module Fusion API separate and treat the later KernelConfig PRs as the closest duplicates.", + "confidence": 0.7, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46339 is the broadest and latest representative of the same kernel-fusion feature line, explicitly extending and simplifying `KernelConfig` across the shared loading/integration paths.", + "canonical_pr_reason": "PR 46339 is the broadest and most mature KernelConfig fusion change: it refactors core loading paths, updates hub kernel integration, touches modeling utilities, and adds kernel tests. It looks like the best representative of the kernel-fusion line.", "best_issue_reason": null, - "best_pr_reason": "PR 46339 is the best single representative because it covers the shared code paths, reflects the mature form of the feature, and likely supersedes the earlier kernel-fusion iterations.", + "best_pr_reason": "PR 46339 is the best standalone candidate because it covers the most complete end-to-end n-to-1 kernel fusion flow and appears to supersede earlier incremental versions.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Related area, but not the same concrete change: 44979 introduces a module-fusion API foundation, while 45363 implements n-to-1 kernel fusion via `KernelConfig`. They could coexist as separate PRs." + "reason": "44979 introduces the new Module Fusion API foundation, while 45363 implements KernelConfig-based n-to-1 fusion on top of it. Related, but not the same concrete change." }, { "left": "pull_request:45363", "right": "pull_request:46339", "accept": true, - "reason": "These appear to target the same underlying kernel-fusion feature and code paths, with 46339 reading like an extension/simplification of 45363 rather than a separate change." + "reason": "Both target the same n-to-1 KernelConfig fusion feature and overlap on the same integration paths. 46339 looks like a later, broader revision of the same change stream." }, { "left": "pull_request:45666", "right": "pull_request:46339", "accept": true, - "reason": "Both are about the same `KernelConfig` n-to-1 kernel fusion mechanism and touch the same loading/integration files; 46339 looks like a refinement of the earlier implementation." + "reason": "These are very close revisions of the same KernelConfig n-to-1 fusion work, with strong filename overlap and the later PR reading like an extension/simplification of the earlier one." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: 44979 is a foundational Module Fusion API PR, while 45363, 45666, and 46339 are all clearly in the n-to-1 KernelConfig kernel-fusion line. The soft-edge verdicts are conservative overall: 44979 vs 45363 is correctly treated as related but not the same concrete change, and the later pairs have enough title/diff overlap and shared files to plausibly be mergeable as a single feature-line PR. No overstatement stands out." + "feedback": "The summary is grounded in the packet: 44979 is clearly a separate Module Fusion API PR, while 45363/45666/46339 all sit on the same KernelConfig n-to-1 fusion line and share overlapping files/titles. The soft-edge verdicts are conservative enough overall, and the accepted pairs look plausibly mergeable as iterative revisions of the same feature work." }, "evaluator_used": true, "retried": false, @@ -21116,12 +21055,12 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "55ba3182326445c6c6ec1db56cdf5337009b28cce93f4be57ad3d5335de4fa75", + "prepared_review_unit_hash": "46432e2da72908acd0f4ef90bfb5ba17a7639d010d79e765da202fa23756f777", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4414, + "serialized_chars": 4413, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -21129,7 +21068,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4414, + "serialized_chars": 4413, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -21138,18 +21077,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests are effectively the same fix for the same Mistral regex AttributeError in `tokenization_utils_tokenizers.py` and target the same issue (#45081). PR 45317 is the stronger representative because it includes the code fix plus a test update.", + "summary": "Both PRs target the same Mistral regex AttributeError in the same file; 45317 is the more complete fix because it includes the code change plus test coverage. They are duplicates at the PR level rather than separate changes.", "confidence": 0.98, - "canonical_issue_reason": "No issue artifact is present in this cluster; both PRs point to issue #45081.", - "canonical_pr_reason": "PR 45317 is the best canonical PR because it contains the same code-path fix as 45086 and also adds a test file, making it the more complete change.", + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45317 is the better canonical PR: it fixes the same AttributeError as 45086, but also adds tests and has a clearer, more specific description of the bug and trigger condition.", "best_issue_reason": null, - "best_pr_reason": "45317 is the best representative PR: same underlying fix as 45086, same target issue, and it includes additional test coverage.", + "best_pr_reason": "PR 45317 is the best representative because it covers the same underlying code-path fix and validates it with tests, making it the strongest standalone record of the change.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same underlying bug fix in `_patch_mistral_regex`/`fix_mistral_regex=True`, same changed code path and issue target; 45317 is just the more complete version with tests." + "reason": "Same concrete bug in _patch_mistral_regex, same touched code path and file, and the second PR is essentially a superset of the first with tests added." } ] }, @@ -21179,7 +21118,7 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "a73e4e0d0aa7b0d550beb7e5f227d4df55c5e9d944e1b86fc4a897ac1cf61dc6", + "prepared_review_unit_hash": "3f1025c6886fca19dbf229b46ae530cb503a67acdcdba1aaa763bd25d5c0be86", "component_budget": { "node_count": 2, "item_count": 2, @@ -21201,18 +21140,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related NemotronH work items, but they are not the same change: one is a narrow conversion-mapping/refactor PR, the other is a broader architecture split that introduces NemotronH Dense/Sparse variants and documentation updates. The soft duplicate edge should be rejected.", - "confidence": 0.92, + "summary": "Both PRs are about the NemotronH model family, but they target different changes: #45115 is a smaller refactor/inheritance and checkpoint-conversion mapping update, while #45569 is a broader feature split introducing NemotronHDense/NemotronHSparse plus docs/tests. They overlap in shared model files, but not in the same concrete fix.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45569 is the better canonical representative because it is the broader, more complete NemotronH redesign: it adds the dense/sparse model variants, updates auto-mapping and docs, and appears to subsume the architectural direction of the smaller refactor.", + "canonical_pr_reason": "PR #45569 is the better canonical representative because it is the larger, more user-visible NemotronH family change and includes the new dense/sparse model split, docs, and tests.", "best_issue_reason": null, - "best_pr_reason": "PR 45569 is the best representative of the cluster because it touches the main model evolution and user-facing packaging changes, while PR 45115 is a narrower internal conversion-mapping refactor.", + "best_pr_reason": "PR #45569 is the strongest overall artifact in this cluster: it covers the full architecture decomposition and documentation updates, whereas #45115 is a narrower refactor focused on inheritance and conversion mapping.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "They share NemotronH files, but the changes are not the same concrete fix. 45115 is a focused conversion-mapping/refactor PR, while 45569 is a larger split into nemotron_h_dense and nemotron_h_sparse with docs and auto-mapping updates. They could not plausibly be merged as one PR without conflating distinct work." + "reason": "Related NemotronH work, but not the same underlying change: #45115 is a refactor/conversion-mapping update; #45569 introduces new dense/sparse model variants and supporting docs/tests. Shared files alone are not enough to merge them as duplicates." } ] }, @@ -21242,12 +21181,12 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "aa0c570a6c77693f650b4fe8f57041a0a6fbb3618be39ca645b702c959d1ef70", + "prepared_review_unit_hash": "5b5c8fa5594d7567097e1b083d27787f31796e99e9e1f65d3081017ddc4a6104", "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 }, @@ -21255,7 +21194,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5466, + "serialized_chars": 5465, "estimated_input_tokens": 1367, "estimated_eval_tokens": 2990 }, @@ -21264,18 +21203,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related but not duplicates: PR #45333 adds heterogeneous per-layer config support, while PR #45332 is a broader follow-on that also changes cache, masking, modeling, integrations, and tests for heterogeneous model execution. They share some configuration plumbing, but they are not the same underlying code change.", - "confidence": 0.92, + "summary": "These two PRs are related but not duplicates: one adds heterogeneous config plumbing, while the other adds broader heterogeneous modeling/runtime support that builds on that config work. They share files because of the shared infrastructure, but they do not appear to be the same concrete change.", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the better umbrella candidate because it subsumes config handling and extends it into the actual heterogeneous modeling/runtime paths (cache, masking, modeling, integrations).", + "canonical_pr_reason": "PR #45332 is the broader umbrella change: it covers heterogeneous model support across cache/masking/modeling paths, not just config plumbing. It is the better representative of the overall feature set in this cluster.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the most comprehensive representative of the heterogeneous-model work; PR #45333 is a narrower prerequisite/config layer and not interchangeable with the broader runtime change.", + "best_pr_reason": "PR #45332 is the strongest standalone representative because it implements the end-to-end heterogeneous model behavior, whereas PR #45333 is a narrower per-layer configuration piece that could plausibly be a prerequisite or companion change.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Shared heterogeneity/config files do not make these duplicates. #45333 focuses on per-layer configuration API support, while #45332 adds broader runtime/modeling behavior that could not simply be merged into the config-only PR." + "reason": "Related feature work, but not the same underlying bug/change. #45333 is focused on per-layer configuration support; #45332 extends heterogeneous support into modeling/cache/runtime code paths and is materially broader." } ] }, @@ -21305,7 +21244,7 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "ba605432e20257d30e9752f866b3a3cd94774bbd43badd8e4ca258be5bdc4d52", + "prepared_review_unit_hash": "dc5a9bf184ffb31d52947b80ff803d0856af62b385c7e08112ed42aed74dfdc6", "component_budget": { "node_count": 2, "item_count": 2, @@ -21327,18 +21266,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are both typing-focused PRs, but they target different levels of change: one is a broad repository-wide type-checking cleanup, while the other adds specific typing protocols and helpers for a narrower modeling_utils/PEFT-related area. The overlap is too superficial to treat them as duplicates.", - "confidence": 0.9, + "summary": "Both PRs are typing-focused, but they are different changes: one is a broad type-checking sweep across many transformer modules, while the other adds specific typing protocols/helpers and related fixes. The overlap is incidental rather than the same underlying fix.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45415 is the broader umbrella change: it updates pyproject typing ignores and touches many transformers modules in a repo-wide type-checking pass. It is the best representative of the cluster if one PR must be chosen.", + "canonical_pr_reason": "No issue link exists; if a representative PR is needed, PR 45425 is the more focused typing-refinement change with substantial review activity.", "best_issue_reason": null, - "best_pr_reason": "PR #45415 is the strongest single candidate because it spans the widest scope and best captures the overall typing-cleanup theme of the cluster.", + "best_pr_reason": "PR 45425 is the better cluster representative because it targets a narrower, more concrete typing API/stub adjustment, whereas PR 45415 is a broad type-checking pass across many files.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Both are typing-related, but #45415 is a broad type-checking sweep across many files, while #45425 adds specific typing protocol/type-alias work for modeling_utils and related integrations. They do not appear to fix the same concrete bug or change the same code path, and would not naturally merge into one PR." + "reason": "Reject: both are typing-related, but 45415 is a wide sweep to add type checking across many files, while 45425 introduces specific typing protocols and utility annotations. They do not appear to fix the same concrete code-path problem and would not naturally collapse into one PR." } ] }, @@ -21368,7 +21307,7 @@ "soft_pairs": [ "pull_request:45513|pull_request:46084" ], - "prepared_review_unit_hash": "50edc7f94b6f0987fd8f5c63470a53aff710585089ed308d2d86ec08319e9dcc", + "prepared_review_unit_hash": "9d926012c44f28379a6b5a05efee576e5597676d63cbaf8f756440c366260d57", "component_budget": { "node_count": 2, "item_count": 2, @@ -21390,18 +21329,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs both address cached multi-token forward behavior, but they target different model families and different implementation paths. They are related conceptually, not duplicates.", - "confidence": 0.91, + "summary": "Two open PRs were clustered only by superficial cache/multi-token similarity, but they address different model families and different bugs: Qwen3.5/OLMo hybrid cached forward behavior vs Mamba2 use_cache handling for seq_len > 1.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46084 is the more canonical representative: it has an explicit issue target, is tightly scoped to one concrete cache bug in Mamba2, and the diff is focused on a single code path.", + "canonical_pr_reason": "PR #46084 is the more self-contained and explicit fix: it targets a single Mamba2 cached-forward bug, includes an issue reference, and has a narrow code-path change. PR #45513 spans multiple model files and appears broader and less directly tied to one concrete bug.", "best_issue_reason": null, - "best_pr_reason": "PR #46084 is the best overall PR candidate because it most clearly fixes one specific, reproducible bug (`use_cache` with `seq_len > 1`) and is easier to map to a single underlying defect than the broader multi-model cache-forward PR.", + "best_pr_reason": "#46084 is the strongest representative because it has a clear, singular failure mode (use_cache with seq_len > 1), an explicit issue target, and a focused patch. #45513 is broader and less specific.", "soft_edge_verdicts": [ { "left": "pull_request:45513", "right": "pull_request:46084", "accept": false, - "reason": "Both involve cached forward behavior, but they are for different architectures (Qwen3.5/OLMo hybrid GDN linear attention vs Mamba2) and different code paths. They are not the same concrete bug and would not plausibly be merged into one PR." + "reason": "Different underlying bugs and code paths: #45513 fixes multi-token cached forward in Qwen3.5/OLMo hybrid linear attention, while #46084 fixes Mamba2 use_cache behavior when seq_len > 1. Shared caching language is not enough to treat them as the same change." } ] }, @@ -21431,12 +21370,12 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "6820625e613c7629bd77d22595edb044aa34c63ab7b5bb4f461fbd4ba8ed65cd", + "prepared_review_unit_hash": "418ba20040f77fc6b1de481edb94ff86349e9be09f52f2d286785470541fad0a", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4508, + "serialized_chars": 4507, "estimated_input_tokens": 1127, "estimated_eval_tokens": 2510 }, @@ -21444,7 +21383,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4508, + "serialized_chars": 4507, "estimated_input_tokens": 1127, "estimated_eval_tokens": 2510 }, @@ -21453,18 +21392,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs both touch expert-parallel loading, but they fix different bugs: one is about serializing/deriving EP config for accelerate/ZeRO-3 loading, the other is about preventing FSDP2 rank-0 broadcast from overwriting expert parameters. They should not be merged as duplicates.", - "confidence": 0.95, + "summary": "The two PRs are both EP-related, but they fix different distributed-loading bugs on different code paths: one for DeepSpeed ZeRO-3/accelerate config handling, the other for FSDP2 expert parameter overwrites during sharded loading. They should not be deduplicated.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45662 is the more concrete, self-contained bug fix with a specific failure mode and loading code path; PR #45548 is a separate configuration/initialization issue.", + "canonical_pr_reason": "PR #45662 is the more concrete representative of a distinct bugfix: it targets expert parameters being overwritten during distributed loading, with code changes in model loading and MoE/tensor-parallel integration.", "best_issue_reason": null, - "best_pr_reason": "PR #45662 best represents the cluster if one must be chosen: it has the clearest concrete symptom and targeted fix, while #45548 addresses a different EP-related loading/configuration path.", + "best_pr_reason": "PR #45662 is the strongest standalone PR here because it addresses a specific runtime failure mode in the loading path and contains the broader, more direct fix.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Both are EP-related, but #45548 fixes distributed_config serialization and a has_ep helper for accelerate/ZeRO-3 loading, while #45662 fixes expert parameters being overwritten during FSDP2/post-shard loading. Different code paths and failure modes." + "reason": "Related only by broad EP/distributed-loading theme. #45548 fixes EP + DeepSpeed ZeRO-3 loading via accelerate launch and config serialization; #45662 fixes EP + FSDP2 experts being overwritten by rank-0 broadcast. Different bug, different code path, not mergeable into one PR." } ] }, @@ -21494,12 +21433,12 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "a80383461754144d57006fc15ad1b984de9c65ca49f3b38167dd51e5591006e3", + "prepared_review_unit_hash": "eaf6d84a0003c8a6499fbbb20d754a7932901973d53fe71688207eb734b9e9f4", "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 }, @@ -21507,7 +21446,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4590, + "serialized_chars": 4589, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -21516,18 +21455,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same xdist collision around captured_info debug artifacts in testing_utils; #45645 appears to be the more complete version because it includes the core file-writing fix plus CI/workflow handling for multiple captured files.", - "confidence": 0.83, + "summary": "Both PRs address xdist-related capture of debug artifacts in `testing_utils.py`, but they are not the same change: #45639 makes the log write path worker-safe, while #45645 also updates CI artifact collection and notification handling. The overlap is thematic, not a single mergeable fix.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45645 is the better canonical representative: it includes the core xdist-safe output path change and the downstream workflow update needed to consume multiple captured_info files, so it covers the broader finalized fix.", + "canonical_pr_reason": "#45645 is the better canonical PR because it includes the core xdist-safe file handling plus the downstream CI changes needed to consume multiple captured-info files; it is the more complete end-to-end fix.", "best_issue_reason": null, - "best_pr_reason": "PR #45645 is the strongest overall fit because it solves the same captured_info collision bug as #45639 but in a more complete way, with both the producer-side fix and consumer-side CI adjustments.", + "best_pr_reason": "#45645 is the strongest representative of the underlying problem because it addresses the same xdist collision issue in `testing_utils.py` and extends it through the CI path that surfaces those artifacts.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", - "accept": true, - "reason": "Same underlying bug: xdist collisions for captured_info debug output. #45639 is the producer-side fix, and #45645 extends the same change with artifact collection updates, so they could plausibly be merged into one PR." + "accept": false, + "reason": "They share the same affected helper files and issue target, but #45645 is broader and changes CI artifact aggregation plus notification logic, so this is not a single same-scope code-path fix that could plausibly be merged as one PR." } ] }, @@ -21559,7 +21498,7 @@ "pull_request:45670|pull_request:46320", "pull_request:46320|pull_request:46401" ], - "prepared_review_unit_hash": "c2aa82e5d6ce5e64404f0ab532c7167dc40fa44da98814fde56134b8e0d9c8dd", + "prepared_review_unit_hash": "3b66a3da56305d968031eadec299630de43903458318f5d81f14b5012cf77e62", "component_budget": { "node_count": 3, "item_count": 3, @@ -21581,24 +21520,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are thematically similar only at a high level, but they target different model families and code paths. The RecurrentGemma attn-dispatch work is separate from the BLIP-2/InstructBLIP SDPA dispatch work, and the AutoModel registration tweak is unrelated.", - "confidence": 0.95, + "summary": "These are three separate PRs: one adds GLM-ASR to auto mappings, one adds attn_implementation support for RecurrentGemma, and one adds SDPA dispatch support for BLIP-2/InstructBLIP variants. The two attn_implementation PRs are thematically similar but fix different model code paths and should not be merged as duplicates.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "No true canonical PR here: 46320 and 46401 both add attn_implementation support, but for different model families and non-overlapping code paths; 45670 is a separate AutoModel mapping fix.", + "canonical_pr_reason": "PR #46320 is the strongest representative of the cluster because it is a focused, substantive model-support change with clear code, conversion, docs, and tests; however, it is still not a duplicate of the BLIP-2 dispatch PR.", "best_issue_reason": null, - "best_pr_reason": "PR 46320 is the closest representative of the shared theme because it is the broader attn_dispatch implementation work, but it should not be treated as a duplicate of the other PRs.", + "best_pr_reason": "PR #46320 is the best single representative of the attn_implementation theme, but the cluster does not contain a true duplicate PR pair.", "soft_edge_verdicts": [ { "left": "pull_request:45670", "right": "pull_request:46320", "accept": false, - "reason": "Different changes: AutoModel registry mapping for glmasr vs RecurrentGemma attn_implementation support. Different files, models, and code paths." + "reason": "Different changes: auto-model registration for GLM-ASR vs RecurrentGemma attn_implementation dispatch. They touch unrelated code paths and would not be merged into one PR." }, { "left": "pull_request:46320", "right": "pull_request:46401", "accept": false, - "reason": "Both add attn_implementation dispatch, but for different concrete architectures (RecurrentGemma vs BLIP-2/InstructBLIP). Not the same bug/change and not plausibly one merged PR." + "reason": "Both are attn_implementation support PRs, but for different architectures and model stacks (RecurrentGemma vs BLIP-2/InstructBLIP). The underlying code paths are distinct, so this is not a duplicate." } ] }, @@ -21628,7 +21567,7 @@ "soft_pairs": [ "pull_request:45720|pull_request:46259" ], - "prepared_review_unit_hash": "09d17c4d61a55028d62c3c78847528cb9fe09ee63624c9a628b83cedd92505b6", + "prepared_review_unit_hash": "3336b5bfe11d55a0da6feb563fb00b4f3e9818af3c45e663f81828da27eab9c0", "component_budget": { "node_count": 2, "item_count": 2, @@ -21650,18 +21589,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two pull requests are both linter/rule work, but they address different underlying problems in different code paths: TRF003 for model forward return behavior, and TRF018 for image/video processor boolean flag propagation. They are not duplicates and should not be merged into one PR.", - "confidence": 0.98, + "summary": "These are two unrelated open pull requests: one implements TRF003 for model forward/return_tuple handling across text/audio models, and the other adds TRF018 to detect dead do_* flags in image/video processor overrides. They share some model filenames, but not the same underlying bug or change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46259 is the better representative of the cluster because it is a self-contained rule introduction with an accompanying docs update and broad implementation across many image/video processor classes.", + "canonical_pr_reason": "No canonical PR: the two PRs address different lint/modeling-rule problems in different code paths, so they should not be merged as duplicates.", "best_issue_reason": null, - "best_pr_reason": "PR 46259 is the strongest standalone change for this cluster: it clearly defines the rule, updates the documentation, and includes the relevant code changes in one place.", + "best_pr_reason": "Neither PR is a good global representative of the other; each is the primary change for its own distinct rule and subsystem.", "soft_edge_verdicts": [ { "left": "pull_request:45720", "right": "pull_request:46259", "accept": false, - "reason": "Different rule targets and code paths: 45720 fixes TRF003 around `forward(...)/return_dict` handling in model classes, while 46259 adds TRF018 for `do_*` flags consumed by image/video processor preprocess overrides. They do not appear to fix the same concrete bug or change." + "reason": "Different underlying changes: TRF003 updates model forward signatures/tuple handling, while TRF018 adds a rule about unused do_* flags in image/video processor overrides. Shared files are incidental and not a single concrete fix." } ] }, @@ -21757,7 +21696,7 @@ "soft_pairs": [ "pull_request:45932|pull_request:46100" ], - "prepared_review_unit_hash": "6a1cbccc9d5e70c554c7a935fafe4204b836095736aa67c83bc3f861d28237e7", + "prepared_review_unit_hash": "3f72d0e198383a34e1b5c03855bc4bd1f9469d103079291b7f6218a3de180009", "component_budget": { "node_count": 2, "item_count": 2, @@ -21779,18 +21718,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates of the same fix/change. PR 45932 is a broad model/runtime change set with workflow and model code edits, while PR 46100 is a tiny test-only PR that injects failing assertions. They target different purposes and would not plausibly be merged as one change.", - "confidence": 0.96, + "summary": "Two open pull requests, but they are not duplicates: one is a broad model/workflow change PR, the other is an isolated test-migration placeholder with intentional failing asserts. They do not fix the same code-path problem and should not be merged together.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45932 is the better cluster representative because it contains the substantive code/workflow changes; PR 46100 is a trivial test-only migration/sabotage PR and does not represent the underlying code change theme of the cluster.", + "canonical_pr_reason": "PR 45932 is the more substantive representative of the cluster: it changes actual implementation/workflow files across multiple model paths, while PR 46100 is just a tiny test-only migration stub with obvious placeholder failures.", "best_issue_reason": null, - "best_pr_reason": "PR 45932 is the strongest standalone PR in the set: it has actual implementation edits across model files and workflow config, whereas PR 46100 is only a small test file edit with intentional failures.", + "best_pr_reason": "PR 45932 is the best single PR representative because it contains the real code changes; PR 46100 is not a meaningful duplicate candidate and looks like a throwaway test edit.", "soft_edge_verdicts": [ { "left": "pull_request:45932", "right": "pull_request:46100", "accept": false, - "reason": "Different underlying changes: 45932 modifies model implementation and CI secrets plumbing, while 46100 only alters a ViT test file with dummy failing assertions. They do not fix the same code-path problem and would not be merged together." + "reason": "Reject: these are unrelated changes. PR 45932 modifies CI/model code, while PR 46100 only edits a ViT test file with deliberate failing asserts. They do not appear to fix the same bug or belong in one merged PR." } ] }, @@ -21823,12 +21762,12 @@ "pull_request:46073|pull_request:46196", "pull_request:46154|pull_request:46196" ], - "prepared_review_unit_hash": "e6483c50f327cd5ac317c5978c4d4f8166e9aff65b943b0e7f76895a59f16cdd", + "prepared_review_unit_hash": "8c38e1de8a414f0d204b84543ea43e798edf151f4ccb08750b560c4e163104c7", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 8081, + "serialized_chars": 8083, "estimated_input_tokens": 2021, "estimated_eval_tokens": 4298 }, @@ -21836,7 +21775,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 8081, + "serialized_chars": 8083, "estimated_input_tokens": 2021, "estimated_eval_tokens": 4298 }, @@ -21845,30 +21784,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related feature PRs for adding `attn_implementation=\"sdpa\"` support, but they target different model implementations and code paths (CTRL, T5, SeamlessM4T/v2, wav2vec2 variants), so they should not be treated as duplicates.", - "confidence": 0.93, + "summary": "These are three separate model-specific SDPA dispatch PRs, not duplicates of one another. They share a broad theme and similar wording, but they target different code paths and model families.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46196 is the most comprehensive representative of the cluster: it spans multiple model families, includes corresponding model code changes, and adds/updates tests across several affected areas.", + "canonical_pr_reason": "No true canonical PR: each change is a distinct model-family implementation of SDPA dispatch, so they should not be deduplicated into one artifact.", "best_issue_reason": null, - "best_pr_reason": "PR #46196 is the strongest standalone candidate because it covers the broadest set of SDPA dispatch adaptations and has the largest supporting test surface, making it the most complete example of the work in this cluster.", + "best_pr_reason": "PR #46196 is the broadest representative, covering multiple related model families and the largest implementation scope, but it is still not a duplicate of the others.", "soft_edge_verdicts": [ { "left": "pull_request:46073", "right": "pull_request:46154", "accept": false, - "reason": "Both add SDPA dispatch support, but for different model code paths (CTRL vs T5). They are similar feature work, not the same concrete bug/change." + "reason": "CTRL and T5 are separate architectures with different attention code paths; these are similar feature additions, not the same fix/change." }, { "left": "pull_request:46073", "right": "pull_request:46196", "accept": false, - "reason": "They target different architectures and implementations (CTRL vs wav2vec2/SeamlessM4T families). Same broad theme, but not mergeable as one duplicate PR." + "reason": "This is CTRL-only vs wav2vec2_conformer/bert and SeamlessM4T/v2; the implementations and affected code paths are different and would not be merged into one PR." }, { "left": "pull_request:46154", "right": "pull_request:46196", "accept": false, - "reason": "These are separate model-specific SDPA enablement changes across unrelated code paths, so they should remain distinct PRs." + "reason": "T5 SDPA dispatch and the audio/seq2seq model updates in #46196 are distinct model-specific changes, not one underlying bug or patch." } ] }, @@ -21898,12 +21837,12 @@ "soft_pairs": [ "pull_request:46126|pull_request:46269" ], - "prepared_review_unit_hash": "d32ec203cbffe569c355a1e770baa0762cc9e5eff62782890a4f25432c02c900", + "prepared_review_unit_hash": "26176033073ffced1b5f25762be9e71da8d998b70527da563f12e6d620cf4d89", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5572, + "serialized_chars": 5571, "estimated_input_tokens": 1393, "estimated_eval_tokens": 3042 }, @@ -21911,7 +21850,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5572, + "serialized_chars": 5571, "estimated_input_tokens": 1393, "estimated_eval_tokens": 3042 }, @@ -21920,18 +21859,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs overlap in a few distributed-training files, but they are not the same change: #46126 is a focused EP + Trainer integration feature, while #46269 is a broad distributed-branch base touching many unrelated subsystems and CI. The soft duplicate edge should be rejected.", - "confidence": 0.94, + "summary": "These two PRs are not duplicates. PR #46126 is a focused feature/integration change around expert parallelism and Trainer wiring, while PR #46269 is a much broader distributed-branch base touching many unrelated distributed and model files plus CI. The shared filenames indicate subsystem overlap, not the same concrete fix or mergeable patch.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "#46126 is the more coherent, narrowly scoped PR with a single concrete goal (EP + Trainer integration). #46269 is an umbrella branch-base PR spanning many distributed and CI changes, so it is not a good canonical representative for this cluster.", + "canonical_pr_reason": "#46126 is the more coherent, narrowly scoped PR with a clear user-facing purpose and concrete code path changes; #46269 is too broad and looks like an integration/base branch rather than a single fix.", "best_issue_reason": null, - "best_pr_reason": "#46126 is the best PR to keep as the cluster representative because it targets one specific code path and feature. #46269 is too broad and includes many unrelated distributed-infra changes, which makes it unsuitable as a duplicate target.", + "best_pr_reason": "#46126 is the best representative PR for this cluster because it has a specific, well-defined change set and a clear title describing the feature being integrated.", "soft_edge_verdicts": [ { "left": "pull_request:46126", "right": "pull_request:46269", "accept": false, - "reason": "Shared filenames only indicate both touch distributed-training code, but the scopes differ substantially. #46126 focuses on EP/router/trainer integration, whereas #46269 is a large branch-base PR spanning FSDP, MoE, tensor parallel, CI, and loader changes. They do not appear to be the same underlying fix or mergeable into one PR." + "reason": "They overlap in distributed code, but the diffs are not the same underlying change: #46126 adds EP/Trainer integration, while #46269 is a broad distributed branch base with many extra files and CI changes. They would not plausibly be merged into one PR." } ] }, @@ -21961,7 +21900,7 @@ "soft_pairs": [ "pull_request:46138|pull_request:46432" ], - "prepared_review_unit_hash": "c46193c7cfde6645b4b96ac868cbc01322678bb36cc1f685688d883fa4ddd220", + "prepared_review_unit_hash": "3da5e3ebfb97fcbf4839d8a7f1a4e62834ea63156ef215012c63c11127b7bad9", "component_budget": { "node_count": 2, "item_count": 2, @@ -21983,18 +21922,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs are unrelated workflow maintenance changes: one adjusts how self-comment CI passes PR comments to a script, while the other hardens the SSH runner workflow by pinning a downloaded script and unsetting HF_TOKEN. They do not fix the same bug or implement the same change.", + "summary": "These are unrelated workflow PRs: one changes how self-comment CI passes message input, the other hardens SSH runner setup by pinning a downloaded script and unsetting HF_TOKEN. They do not fix the same underlying problem.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46432 is the better representative because it makes a more concrete security/operational hardening change in a single workflow, with a clearer and more substantive diff than the small CLI-argument refactor in #46138.", + "canonical_pr_reason": "PR 46432 is the stronger representative because it makes the more substantive workflow hardening change, including supply-chain pinning and secret cleanup in ssh-runner.", "best_issue_reason": null, - "best_pr_reason": "PR #46432 is the strongest standalone PR in the cluster: it addresses a specific SSH runner hardening concern by pinning an external download and clearing a sensitive environment variable.", + "best_pr_reason": "PR 46432 is the best overall PR in the cluster as a representative change: it is more concrete and security-relevant than the small CLI invocation adjustment in 46138.", "soft_edge_verdicts": [ { "left": "pull_request:46138", "right": "pull_request:46432", "accept": false, - "reason": "Different workflows, different code paths, and different purposes; one changes comment handling for self-comment CI, the other hardens SSH runner setup. They are not the same underlying change and would not be merged into one PR." + "reason": "Different workflows and different concerns: stdin handling for self-comment CI versus pinning a remote script and clearing HF_TOKEN in ssh-runner. They are not the same underlying bug/change and would not be merged into one PR." } ] }, @@ -22024,7 +21963,7 @@ "soft_pairs": [ "pull_request:46152|pull_request:46207" ], - "prepared_review_unit_hash": "9ff4a80eb05930a14ddfe31b202cac879a799149985a20c770fc48470d86f2c6", + "prepared_review_unit_hash": "efc6146224f0a56f2e416449991d5e1f0574958a8b81dd251570a041d1b47a08", "component_budget": { "node_count": 2, "item_count": 2, @@ -22046,18 +21985,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs share a broad theme of replacing source-file opening with source inspection, but they target different code paths and are not duplicate fixes.", - "confidence": 0.92, + "summary": "Two open PRs with a superficial similarity around avoiding file-open/source-reading patterns, but they target different code paths and different bugs: one in import/resource traversal, the other in attention-implementation capability detection.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46207 is the better representative because it addresses a concrete, localized bug in `_can_set_attn_implementation` and includes test coverage, whereas the other PR is a separate import-utils refactor with a different purpose.", + "canonical_pr_reason": "PR 46207 is the better representative of this cluster because it is a focused, self-contained fix in a single utility path with accompanying tests, and its title/body clearly describe the concrete behavior change.", "best_issue_reason": null, - "best_pr_reason": "PR 46207 is the strongest standalone change in this cluster: it has a clear behavioral fix, a well-defined target area, and tests. PR 46152 appears to be an independent utility-level cleanup/fix rather than the same change.", + "best_pr_reason": "PR 46207 is the strongest standalone candidate overall: it addresses a specific source-reading heuristic in modeling_utils.py and appears easier to review/merge than the more infrastructure-oriented resource-resolution change in 46152.", "soft_edge_verdicts": [ { "left": "pull_request:46152", "right": "pull_request:46207", "accept": false, - "reason": "Both replace `open()`-based source access, but they fix different modules and different behaviors: one is for `import_utils` path resolution, the other for `modeling_utils` attention-implementation detection. They are related in technique, not the same underlying bug or mergeable PR." + "reason": "They both touch source-reading/open() patterns, but the underlying problems differ: 46152 is about resolving package resources without reading self source via file paths, while 46207 changes how model classes detect support for dynamic attention implementation. These are not the same bug or mergeable into one PR." } ] }, @@ -22090,7 +22029,7 @@ "pull_request:46337|pull_request:46362", "pull_request:46362|pull_request:46368" ], - "prepared_review_unit_hash": "14221657e03160e79363cf52084d60ee4e22c838a8b2f41cbfbb639861edf112", + "prepared_review_unit_hash": "d6c4c5df68c32bf631d55c5a20f09278b880179e2fed0c4c4b0517c369b21413", "component_budget": { "node_count": 3, "item_count": 3, @@ -22112,30 +22051,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are model-specific test fixes, but they address different failure modes: Cohere2Moe is about device-map/offload setup, while Deepseek V4 and Nemotron H both skip quantized-cache generation tests for model-specific cache incompatibilities. They are related only at a very high level, not the same underlying change.", - "confidence": 0.95, + "summary": "These PRs are not true duplicates. PRs 46362 and 46368 both skip a quantized-cache generation test, but they target different model families and likely different model-specific cache implementations. PR 46337 is a separate test-environment fix for Cohere2Moe device mapping/offload behavior.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: each PR fixes a different model/test incompatibility, so none should absorb the others as duplicates.", + "canonical_pr_reason": "PR 46362 is the clearest representative of the quantized-cache test-skip pattern: it adds a focused skip for DeepSeek V4 due to an explicit cache incompatibility. The others are either for a different model family or a different failure mode entirely.", "best_issue_reason": null, - "best_pr_reason": "PR 46337 is the strongest standalone fix because it adjusts the test configuration to make an existing test pass, rather than simply skipping coverage.", + "best_pr_reason": "PR 46362 is the best single representative because it is a minimal, explicit fix for a concrete quantized-cache incompatibility in one model test. PR 46368 is similar in shape but model-specific, and PR 46337 addresses a different device-map/offload test setup issue.", "soft_edge_verdicts": [ { "left": "pull_request:46337", "right": "pull_request:46368", "accept": false, - "reason": "Different models and different failures: Cohere2Moe device-map/offload setup vs NemotronH quantized-cache incompatibility. Same broad test infrastructure, but not the same bug." + "reason": "Different concrete problems: 46337 tweaks Cohere2Moe test configuration to satisfy device-map/offload behavior, while 46368 simply skips a quantized-cache test for NemotronH. Same broad testing area, but not the same underlying bug or mergeable change." }, { "left": "pull_request:46337", "right": "pull_request:46362", "accept": false, - "reason": "Cohere2Moe is fixing accelerate/device placement for hf_device_map; Deepseek V4 is only skipping a quant-cache test. Different code paths and failure causes." + "reason": "Not the same bug. 46337 is about Cohere2Moe failing hf_device_map/offload tests; 46362 skips a DeepSeek V4 quantized-cache test because its compressor/cache design is incompatible. Different models and different failure mechanisms." }, { "left": "pull_request:46362", "right": "pull_request:46368", "accept": false, - "reason": "Both skip the same test name, but for different model internals and different cache implementations. Too model-specific to be the same underlying change." + "reason": "Both skip quant_cache tests, but for different model families with distinct cache implementations (DeepSeek V4 vs NemotronH). They are similar test workarounds, not the same concrete code-path problem, so they should not be merged as duplicates." } ] }, @@ -22165,7 +22104,7 @@ "soft_pairs": [ "pull_request:46376|pull_request:46417" ], - "prepared_review_unit_hash": "b50deb68399a31908c5922242d998ed237efc0d4e7490b83529b05eefd5dfa7c", + "prepared_review_unit_hash": "2be7108f54a818583de6cb58a02df6fdff2c275dc1f1307d11e76c15f622a02e", "component_budget": { "node_count": 2, "item_count": 2, @@ -22187,18 +22126,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs are both feature work but address unrelated areas: one adds multimodal support to continuous batching, the other adds the HunYuan VL model stack and docs. They should not be treated as duplicates or merged together.", + "summary": "The pair is not a duplicate cluster: PR #46376 adds multimodal support to continuous batching infrastructure, while PR #46417 adds the HunYuan VL model with docs, processors, modeling, and tests. They touch different code paths and represent separate changes.", "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "If one PR must represent the cluster, #46376 is the more central infrastructure change; #46417 is a separate model addition and does not overlap on the underlying bug/change.", "soft_edge_verdicts": [ { "left": "pull_request:46376", "right": "pull_request:46417", "accept": false, - "reason": "Different underlying changes and code paths: continuous-batching multimodal support vs a new model implementation/docs/test suite. They share only broad Transformers context, not the same fix or feature." + "reason": "Different scopes and code paths: continuous-batching multimodal support vs adding a new HunYuan VL model. They are not plausibly the same fix or mergeable into one PR." } ] }, @@ -22217,8 +22156,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "llm", - "cache_hit": false, + "source": "cache", + "cache_hit": true, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-46465-2", "nodes": [ diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index a6d61b53f2a409fc3c3e47ffa9c26fb89ac4b922..8b322e8cb81541f74b69b0dea75b878d2cebe339 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260607t060034z", + "analysis_id": "hybrid-model-20260607t120016z", "archived_artifacts": { - "hybrid": "snapshots/20260607T060034Z/analysis-runs/hybrid-model-20260607t060034z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260607T060034Z/analysis-runs/hybrid-model-20260607t060034z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-06-07T06:08:55Z", + "published_at": "2026-06-07T12:08:27Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260607T060034Z", + "snapshot_id": "20260607T120016Z", "variant": "hybrid" } diff --git a/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.json b/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..49604cf218af49174fb519977b0c58e081247e56 --- /dev/null +++ b/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.json @@ -0,0 +1,2600 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260607T120016Z", + "generated_at": "2026-06-07T12:08:25Z", + "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-45561-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 45561, + "canonical_pr_number": 45645, + "issue_numbers": [ + 45561 + ], + "pr_numbers": [ + 45639, + 45645 + ], + "evidence_types": [ + "closing_reference" + ], + "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-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-46463-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46463.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 46463, + "canonical_pr_number": 46465, + "issue_numbers": [ + 46463 + ], + "pr_numbers": [ + 46465, + 46470 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46465, + "right_pr_number": 46470, + "code_similarity": 0.63, + "size_similarity": 0.737, + "file_overlap": 0.333, + "area_overlap": 1.0, + "patch_similarity": 0.214, + "shared_filenames": [ + "src/transformers/models/mt5/modeling_mt5.py", + "src/transformers/models/t5/modeling_t5.py", + "src/transformers/models/umt5/modeling_umt5.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/mt5/modeling_mt5.py", + "left_ranges": [ + [ + 654, + 660 + ], + [ + 664, + 670 + ] + ], + "right_ranges": [ + [ + 654, + 660 + ], + [ + 664, + 670 + ] + ] + }, + { + "filename": "src/transformers/models/t5/modeling_t5.py", + "left_ranges": [ + [ + 643, + 649 + ], + [ + 653, + 659 + ] + ], + "right_ranges": [ + [ + 643, + 649 + ], + [ + 653, + 659 + ] + ] + }, + { + "filename": "src/transformers/models/umt5/modeling_umt5.py", + "left_ranges": [ + [ + 612, + 618 + ], + [ + 622, + 628 + ] + ], + "right_ranges": [ + [ + 612, + 618 + ], + [ + 622, + 628 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43656-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 43656, + "issue_numbers": [], + "pr_numbers": [ + 43656, + 43842 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 43656, + "right_pr_number": 43842, + "code_similarity": 0.405, + "size_similarity": 0.5, + "file_overlap": 1.0, + "area_overlap": 0.0, + "patch_similarity": 0.036, + "shared_filenames": [ + "src/transformers/cli/serve.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cli/serve.py", + "left_ranges": [ + [ + 11, + 18 + ], + [ + 30, + 36 + ], + [ + 315, + 323 + ], + [ + 665, + 671 + ], + [ + 931, + 937 + ], + [ + 1843, + 1849 + ], + [ + 1868, + 1874 + ] + ], + "right_ranges": [ + [ + 54, + 61 + ], + [ + 587, + 593 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44002-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44002, + "issue_numbers": [], + "pr_numbers": [ + 44002, + 44029 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44002, + "right_pr_number": 44029, + "code_similarity": 0.135, + "size_similarity": 0.43, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.329, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, + { + "cluster_id": "cluster-44018-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44068, + "issue_numbers": [], + "pr_numbers": [ + 44018, + 44068 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44018, + "right_pr_number": 44068, + "code_similarity": 0.766, + "size_similarity": 0.939, + "file_overlap": 1.0, + "area_overlap": 0.425, + "patch_similarity": 0.866, + "shared_filenames": [ + "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "tests/models/gpt_neo/test_modeling_gpt_neo.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "left_ranges": [ + [ + 26, + 46 + ], + [ + 138, + 143 + ], + [ + 182, + 187 + ], + [ + 283, + 295 + ], + [ + 327, + 341 + ], + [ + 348, + 354 + ], + [ + 360, + 369 + ], + [ + 399, + 406 + ], + [ + 411, + 419 + ], + [ + 428, + 433 + ], + [ + 472, + 492 + ], + [ + 507, + 513 + ], + [ + 519, + 528 + ], + [ + 541, + 559 + ], + [ + 562, + 567 + ], + [ + 595, + 601 + ], + [ + 607, + 614 + ], + [ + 627, + 643 + ], + [ + 685, + 690 + ], + [ + 708, + 714 + ], + [ + 720, + 727 + ], + [ + 740, + 757 + ], + [ + 761, + 766 + ], + [ + 780, + 786 + ], + [ + 791, + 798 + ], + [ + 807, + 822 + ], + [ + 840, + 845 + ] + ], + "right_ranges": [ + [ + 26, + 31 + ], + [ + 34, + 44 + ], + [ + 136, + 141 + ], + [ + 180, + 185 + ], + [ + 281, + 293 + ], + [ + 325, + 339 + ], + [ + 346, + 352 + ], + [ + 358, + 367 + ], + [ + 397, + 404 + ], + [ + 409, + 417 + ], + [ + 426, + 434 + ], + [ + 463, + 483 + ], + [ + 498, + 504 + ], + [ + 510, + 519 + ], + [ + 532, + 550 + ], + [ + 553, + 564 + ], + [ + 586, + 592 + ], + [ + 598, + 605 + ], + [ + 618, + 634 + ], + [ + 676, + 688 + ], + [ + 699, + 705 + ], + [ + 711, + 718 + ], + [ + 731, + 748 + ], + [ + 752, + 762 + ], + [ + 771, + 777 + ], + [ + 782, + 789 + ], + [ + 798, + 813 + ], + [ + 831, + 836 + ] + ] + }, + { + "filename": "tests/models/gpt_neo/test_modeling_gpt_neo.py", + "left_ranges": [ + [ + 458, + 464 + ] + ], + "right_ranges": [ + [ + 458, + 464 + ] + ] + } + ] + } + ] + } + ], + "duplicate_issues": [], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 45613, + "duplicate_pr_numbers": [ + 41116 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43656-2", + "canonical_pr_number": 43656, + "duplicate_pr_numbers": [ + 43842 + ], + "target_issue_number": null, + "reason": "PRs in cluster-43656-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "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-44002-2", + "canonical_pr_number": 44002, + "duplicate_pr_numbers": [ + 44029 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44002-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-44018-2", + "canonical_pr_number": 44068, + "duplicate_pr_numbers": [ + 44018 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-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 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-46463-3", + "canonical_pr_number": 46465, + "duplicate_pr_numbers": [ + 46470 + ], + "target_issue_number": 46463, + "reason": "PRs in cluster-46463-3 are treated as duplicates because they converge on issue #46463 with closing_reference, shared_issue_target 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/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..212343d31ea5961508145c12dc8beb6f2096692a --- /dev/null +++ b/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,22218 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260607T120016Z", + "generated_at": "2026-06-07T12:08:25Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "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:43296", + "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" + ], + "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: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:45237|issue:46015", + "issue:43039|issue:44805", + "issue:43866|issue:46402", + "issue:45800|issue:46304", + "issue:43296|issue:46304", + "issue:41762|issue:43596", + "issue:39290|issue:41720" + ], + "prepared_review_unit_hash": "8cb746d2b977866aed72a267fc6eca5ad93cf15a412264ceb38fcb9b8467a521", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 65, + "serialized_chars": 47165, + "estimated_input_tokens": 11792, + "estimated_eval_tokens": 23840 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly loose similarity across unrelated model, docs, dtype, and distributed-training reports. The best-supported duplicate candidates are 43596/43638 (same ZeRO-3/Bert zero-size index error family) and 45859/45902 (same Qwen3.5 MoE rotary-embedding/offload bug).", + "confidence": 0.81, + "canonical_issue_reason": "issue:43596 is the likely canonical for the ZeRO-3/Bert zero-size index-error cluster: it is broader and better-scoped than 43638, which reads like a narrower Bert/non-pretrained variant of the same failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43596 is the best issue representative overall because it is the broadest clear fit for the ZeRO-3/Bert index-error family; this is a likely duplicate candidate, not a confirmed duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different bugs: Qwen3.5 config attribute error vs streaming chat/completions failure." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different hardware/runtime failures: AMD GPU issue vs MPS histogram on Int." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Gemma3 dtype mismatch in preprocessing is unrelated to GPT-OSS AMD GPU failure." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Autocast dtype mismatch and AMD GPU failure are different code paths." + }, + { + "left": "issue:45237", + "right": "issue:46468", + "accept": false, + "reason": "Unrelated: GPT-OSS AMD GPU bug vs Nemotron example breakage." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different problems: AMD GPU support vs invalid method call in GraniteMoEHybrid." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch and ModernBERT ONNX export are unrelated." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ModernBERT export issue vs Qwen3Moe FP8 failure; same broad area only." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ONNX export bug vs Qwen3 device-mapping CUDA assert." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "Table QA crash and outdated RAG examples are different issues." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions, but they target different removed v5 references and are not the same concrete issue." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "Export error vs GenerationMixin import error are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:46165", + "accept": false, + "reason": "TRL input-embeddings bug vs resume-training/trust_remote_code issue." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask shape mismatch and Gemma4 PLE device mismatch are different failures." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Bert ZeRO-3 index error is unrelated to tensor mask shape mismatch." + }, + { + "left": "issue:43638", + "right": "issue:46165", + "accept": false, + "reason": "Different problems: Bert ZeRO-3 loading vs model resume/trust_remote_code." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "ZeRO-3 Bert index error vs mask/index shape mismatch are not the same bug." + }, + { + "left": "issue:44805", + "right": "issue:46294", + "accept": false, + "reason": "Mask/tensor shape mismatch is unrelated to video feature/token mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Different failures: tensor shape mismatch vs GPT-OSS tensor-parallel issue." + }, + { + "left": "issue:46165", + "right": "issue:46468", + "accept": false, + "reason": "Resume-training issue vs Nemotron example breakage are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Video StopIteration in Qwen3-vl-embedding is unrelated to resume training." + }, + { + "left": "issue:43596", + "right": "issue:46165", + "accept": false, + "reason": "Bert ZeRO-3 loading bug vs training-resume issue; not the same code path." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice preset bug is unrelated to ModernBERT ONNX export." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Collective mismatch in Jamba/ZeRO-3 is different from Bert zero-size index error." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same underlying ZeRO-3/Bert zero-size index error; 43638 is the narrower non-pretrained Bert variant." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross-entropy dispatch bug vs GPT-OSS TP issue." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Detokenization bug and LoRA merge/save mismatch are different issues." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime issue is unrelated to LoRA save/reload inconsistency." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "FSDP checkpointing size mismatch is not the same as mask/tensor shape mismatch." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Both mention DeepSpeed, but they are different model/error scenarios." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": true, + "reason": "Same Qwen3.5 MoE rotary-embedding/offload bug around inv_freq materialization and device handling." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "Bert ZeRO-3 index error vs GPT-OSS MoE TP issue." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "Activation-checkpointing/FSDP bug vs CustomTrainer kwarg incompatibility." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "Different bugs: Bert ZeRO-3 loading vs GPT-OSS tensor-parallel support." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Gemma3 dtype mismatch is unrelated to GraniteMoEHybrid invalid method." + }, + { + "left": "issue:45874", + "right": "issue:46294", + "accept": false, + "reason": "Expensive reverse-embedding lookup is unrelated to video token/template mismatch." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "NVFP4 runtime error and AMD GPU failure are different issues." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3 config attribute regression vs Mistral tokenizer regex crash." + }, + { + "left": "issue:45823", + "right": "issue:45874", + "accept": false, + "reason": "Gemma4 PLE device mismatch and reverse-embedding slowdown are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "Torch/transformers version incompatibility vs Qwen-3-8B-NVFP4 runtime error." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP checkpointing mismatch vs Qwen3 auto device-map CUDA assert." + }, + { + "left": "issue:43828", + "right": "issue:44512", + "accept": false, + "reason": "Autocast dtype mismatch vs docs mention of removed run command." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch and MPS histogram issue are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "TRL input embeddings bug vs MPS histogram issue." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger Kernel cross-entropy bug vs autocast dtype mismatch." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch vs CustomTrainer compute_loss signature change." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP checkpointing issue vs GenerationMixin import error." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "AMD GPU runtime issue vs outdated RAG docs/examples." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger kernel cross-entropy dispatch bug is unrelated to mask shape mismatch." + }, + { + "left": "issue:43866", + "right": "issue:46402", + "accept": false, + "reason": "Corrupted Ovis checkpoint is unrelated to Sam3Model weights not loading." + }, + { + "left": "issue:45800", + "right": "issue:46304", + "accept": false, + "reason": "Version incompatibility and USE_DISTRIBUTED=0 breakage are related only at a very broad level." + }, + { + "left": "issue:43296", + "right": "issue:46304", + "accept": false, + "reason": "PaddleOCR-VL/vLLM load failure is unrelated to the distributed-support regression." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Both are ZeRO-3 index errors, but on different models and likely different code paths." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Gemma3 config attribute regression vs Qwen3 auto device-mapping CUDA assert." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The overall clustering is conservative, but the rationale overstates evidence in a few places. In particular, calling issue:43596 the \u2018strongest confirmed duplicate pair\u2019 is too strong for a packet with only soft similarity evidence and no explicit duplicate link. Please downgrade that language to \u2018likely\u2019/\u2018best candidate\u2019 and avoid implying confirmation not present in the packet." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 2, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:36010", + "issue:39290", + "issue:41093", + "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:44291", + "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:44589|issue:46304", + "issue:45823|issue:46294", + "issue:45237|issue:45901", + "issue:43901|issue:44512", + "issue:44841|issue:45823", + "issue:43039|issue:44560", + "issue:43975|issue:46015", + "issue:45812|issue:45920", + "issue:44863|issue:46015", + "issue:45020|issue:46304", + "issue:43866|issue:44863", + "issue:44292|issue:46015", + "issue:43872|issue:46304", + "issue:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43792|issue:45901", + "issue:43854|issue:43866", + "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:45800|issue:45823", + "issue:41093|issue:41762", + "issue:43296|issue:43366", + "issue:43572|issue:45070", + "issue:44805|issue:44918", + "issue:44387|issue:45005", + "issue:43541|issue:45161", + "issue:43828|issue:45464", + "issue:44291|issue:45070", + "issue:43366|issue:45464", + "issue:43531|issue:45070", + "issue:45020|issue:45800" + ], + "prepared_review_unit_hash": "bdac03df1027d7083405d55fb0fe716ca5d59f8fd2b8227776beb56c7aaa4a4a", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 42, + "serialized_chars": 40096, + "estimated_input_tokens": 10024, + "estimated_eval_tokens": 20304 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous regression reports, not one cohesive duplicate set. The only strong duplicate-like pair is the AutoTokenizer wrong-token-IDs regression across models.", + "confidence": 0.69, + "canonical_issue_reason": "Issue #45920 is the broadest and most complete report of the tokenizer ID regression, covering multiple affected model families and the same v5-era failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "#45920 is the best umbrella issue: it is open, explicit about the shared tokenizer regression, and aggregates the widest model impact.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44589", + "right": "issue:46304", + "accept": false, + "reason": "Different failures: float8 storage lookup vs distributed support breakage. Same release family, but not the same bug." + }, + { + "left": "issue:45823", + "right": "issue:46294", + "accept": false, + "reason": "Different models and code paths: Gemma4 device placement vs Qwen3.5 video chat-template token/feature mismatch." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "AMD GPU execution issue vs table-question-answering crash; no concrete shared bug." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions, but they concern unrelated documentation updates and not one underlying code change." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Voxtral processor failure and Gemma4 device mismatch are distinct model-specific problems." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross-entropy behavior and Qwen3-vl StopIteration are unrelated code paths." + }, + { + "left": "issue:43975", + "right": "issue:46015", + "accept": false, + "reason": "Tokenizer detokenization bug and outdated RAG examples are unrelated." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Both report the same underlying AutoTokenizer v5 regression producing wrong token IDs across multiple model families." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint-loading issue is unrelated to RAG documentation examples." + }, + { + "left": "issue:45020", + "right": "issue:46304", + "accept": false, + "reason": "Broad remote_code compatibility breakage and USE_DISTRIBUTED=0 support are different regressions." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "A corrupted Ovis2 checkpoint is not the same as NemotronH implementation loading failures." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error is unrelated to documentation examples." + }, + { + "left": "issue:43872", + "right": "issue:46304", + "accept": false, + "reason": "bitsandbytes init argument incompatibility is unrelated to distributed build support." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention sliding_window, but one is a missing config attribute for vLLM/Gemma3 and the other is a Qwen3-MoE behavior issue." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both are IndexErrors, but the reported shapes/contexts differ enough that a shared root cause is not established." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Whisper model loading and table QA task crashes are unrelated." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "A unit-test model loading failure is not the same as a corrupted checkpoint report." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch version incompatibility and PretrainedConfig/pydantic regression are separate compatibility problems." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "DeepSpeed ZeRO-3 BERT loading error and torch 2.4.1/v5.8.0 incompatibility are not the same bug." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Different subsystems: detokenization vs table QA runtime crash." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper model loading and pydantic config breakage are unrelated." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Both involve Qwen3.5, but one is a chat crash and the other is a streaming API failure; different code paths." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "NVFP4 model execution error and table QA crash are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Tied-weight bug in Mistral3 is unrelated to an Ovis2 checkpoint corruption report." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 streaming API failure and MPS histogram/int error are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Both are dtype/tracing-related, but they affect different models and different execution mechanisms; not one concrete bug." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Both are training-time tensor/mask issues, but the error shapes and implicated paths differ; too little evidence for one root cause." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Remote_code breakage and torch/transformers version incompatibility are separate compatibility issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analysis is well grounded in the packet and stays conservative on the soft-edge pairs. The choice of #45920 as the umbrella/canonical issue is supported by the broader, open report, and the only accepted duplicate pair (#45812/#45920) is plausibly the same AutoTokenizer v5 token-ID regression. No 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": 21, + "review_unit_index": 3, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:39186", + "issue:39692", + "issue:40990", + "issue:41093", + "issue:41762", + "issue:42915", + "issue:43054", + "issue:43404", + "issue:43541", + "issue:43596", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43828", + "issue:43866", + "issue:43872", + "issue:43873", + "issue:43950", + "issue:43975", + "issue:43994", + "issue:44292", + "issue:44360", + "issue:44403", + "issue:44479", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45005", + "issue:45183", + "issue:45399", + "issue:45507", + "issue:45542", + "issue:45632", + "issue:45698", + "issue:45701", + "issue:45759", + "issue:45800", + "issue:45902", + "issue:46294", + "issue:46463" + ], + "soft_pairs": [ + "issue:43716|issue:46294", + "issue:43541|issue:44589", + "issue:44479|issue:46294", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45759|issue:45902", + "issue:45542|issue:45800", + "issue:39186|issue:44918", + "issue:43716|issue:44918", + "issue:40990|issue:44960", + "issue:45632|issue:45698", + "issue:43645|issue:45399", + "issue:43404|issue:45005", + "issue:42915|issue:45399", + "issue:43873|issue:45005", + "issue:44530|issue:45005", + "issue:44863|issue:45005", + "issue:45183|issue:46463", + "issue:44360|issue:44512", + "issue:44918|issue:45507", + "issue:44492|issue:44512", + "issue:43975|issue:44292", + "issue:44292|issue:45507", + "issue:43950|issue:44534", + "issue:39692|issue:43054", + "issue:43644|issue:43950", + "issue:36331|issue:41762", + "issue:43828|issue:44928", + "issue:43872|issue:44589", + "issue:43950|issue:45902", + "issue:43054|issue:45701", + "issue:43866|issue:44403", + "issue:38175|issue:43054" + ], + "prepared_review_unit_hash": "535b3f02d8e1476c6d996fbf5958398a63e018205eed12d94986283cc6f78487", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37535, + "estimated_input_tokens": 9384, + "estimated_eval_tokens": 19024 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous, but a few pairs are strong duplicates: the v5 non-persistent-buffer corruption reports, a Qwen3.5 video-token mismatch regression, and a trust_remote_code/custom-module resolution collision. Most other pairs only share broad subsystem or model-family similarity and should stay separate.", + "confidence": 0.78, + "canonical_issue_reason": "issue:43950 is the clearest root-cause report for the non-persistent-buffer regression, and it lines up directly with the shorter duplicate descriptions in 43644 and 44534.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43950 is the best canonical anchor because it is concrete, regression-focused, and names the exact failure mode and mechanism; it best represents the duplicate buffer-corruption subgroup in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43716", + "right": "issue:46294", + "accept": false, + "reason": "Different models and different multimodal failures; both are video/image input issues, but not the same concrete bug." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Unrelated runtime errors: torch dynamo grouped_mm tracing versus missing Float8 storage object." + }, + { + "left": "issue:44479", + "right": "issue:46294", + "accept": true, + "reason": "Both describe the same Qwen3.5 video-input regression, with 46294 appearing to be a specific symptom of the broader token/feature mismatch in 44479." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are IndexErrors, but the failing code paths are different: mask-length mismatch versus ZeRO-3 zero-sized parameter access." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but the reported problems differ: degraded text embeddings versus nonsensical AutoModel/pipeline output." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Different bugs: multimodal config unwrapping versus uninitialized rotary embedding memory after meta materialization." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "No shared code-path bug; one is a missing TensorFlow backend check, the other is a torch/transformers version incompatibility." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "Both mention training/loading, but they are distinct failures on different models and different paths." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Different model families and different multimodal/training failure modes." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "No meaningful overlap beyond being model-specific inference complaints." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": true, + "reason": "Both point to the same trust_remote_code/custom-module resolution problem: cache/path collision causing the wrong custom module to be loaded." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Unrelated issues: Jupyter/custom model initialization versus flash-attn2 fallback gating." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Both involve tied weights, but they affect different model classes and appear to be separate bugs rather than one mergeable fix." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Different failures: FineGrainedFP8Config on Qwen3Moe versus fallback logic blocked when fa2 is absent." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization offloading behavior is unrelated to tied-weight issues." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is a different code path than tied-weight handling." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "Checkpoint-loading failure for NemotronH is not the same bug as tied weights in translation models." + }, + { + "left": "issue:45183", + "right": "issue:46463", + "accept": false, + "reason": "Both are validation/error-message bugs, but they target different validation logic and are not clearly the same underlying defect." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "Completely unrelated: DSA indexer discussion versus outdated docs about a removed command." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Different model-specific failures with no shared concrete code-path bug." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "Both are documentation/text issues, but about different docs and different fixes." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Different models and different runtime symptoms; no evidence of the same bug." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Unrelated model execution failure versus invalid-method call in a different model implementation." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Same regression: Transformers v5 corrupts/overwrites non-persistent buffers during load/materialization." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "SigLIP2 is the shared subject, but one is a documentation example/model-processor mismatch and the other is a model-quality complaint." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Direct duplicate descriptions of the same non-persistent-buffer junk regression in Transformers v5." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Different failure modes: custom trainer signature mismatch versus DeepSpeed ZeRO-3 empty-tensor indexing." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Different models and different numeric-instability/training failures; not the same concrete bug." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Both are quantization/storage-related TypeErrors, but they involve different missing incompatibilities and different code paths." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "Both involve buffers/initialization, but the concrete bug is different: buffer corruption on load versus uninitialized rotary embedding memory." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "Too broad and not the same bug; one is SigLIP2 embedding quality, the other is tokenization changing across versions." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Unrelated issues: checkpoint corruption versus loading noise/doc-style behavior." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Same model family, but the symptoms are different enough that they do not clearly look like one underlying bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The accepted duplicate pairs are plausible from the titles alone: the non-persistent-buffer regression set (43644/43950/44534), the Qwen3.5 video-input mismatch set (44479/46294), and the trust_remote_code/custom-module loading collision set (45632/45698). The non-duplicate rejections are also appropriately cautious. I did not see a clear overstatement that would require rejection." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 4, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:36331", + "issue:41093", + "issue:41553", + "issue:41762", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43404", + "issue:43493", + "issue:43643", + "issue:43742", + "issue:43782", + "issue:43792", + "issue:43824", + "issue:43827", + "issue:43828", + "issue:43856", + "issue:43872", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44509", + "issue:44661", + "issue:44910", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45356", + "issue:45399", + "issue:45520", + "issue:45663", + "issue:45685", + "issue:45715", + "issue:45823", + "issue:45859", + "issue:45902", + "issue:46315" + ], + "soft_pairs": [ + "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:43975|issue:45520", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:43643|issue:45020", + "issue:41762|issue:42915", + "issue:43782|issue:44960", + "issue:45685|issue:45823", + "issue:43824|issue:43872", + "issue:43881|issue:43883", + "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:45020|issue:45356", + "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:36331|issue:41093" + ], + "prepared_review_unit_hash": "48b2be9ed773463c93ab5bdf3dfa7139b77a3d58c01f2187153f2a6a8bf9a887", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36831, + "estimated_input_tokens": 9208, + "estimated_eval_tokens": 18672 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated Transformer bugs and docs regressions. I only accepted a few clear duplicate pairs; most soft links are same-subsystem but not the same concrete defect.", + "confidence": 0.89, + "canonical_issue_reason": "No single cluster-wide canonical issue: the items split into separate subgroups. If forced to anchor one bug, issue:44291 is the clearest runtime regression report, and issue:44509 is the clearest docs regression report.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44291 is the strongest standalone issue here: it states the root cause clearly, is reproducible, and cleanly covers the same unexpected `_is_hf_initialized` regression seen in issue:43872.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Different failures: tokenizer loading error messaging vs missing `all_tied_weights_keys` on Molmo." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Both are Qwen3.5-MoE related, but one is a flash-attn packed-sequence crash and the other is a CPU-offload rotary embedding bug." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Distinct model/code paths: Gemma4 FSDP2 shared-kv rebuilding vs Qwen3_5Moe rotary-memory initialization." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Different problems: LoRA gradient-checkpointing ineffectiveness vs int4 quantization memory growth/OOM." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": true, + "reason": "Both describe the same SigLIP2 implementation discrepancy; one reports worse embeddings, the other names the HF-vs-JAX mismatch." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Unrelated loading issues for different models; one is a bad error message, the other is a model load failure." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading-noise complaint vs a specific ScandiBERT load failure." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Different areas: tokenizer detokenization regression vs Python 3.13 flash_attn import-key handling." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Same broad loading area, but not the same bug; one is log noise, the other is a specific model load failure." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Both involve MoE training, but checkpointing behavior and memory usage are separate defects." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Both mention remote_code, but one is missing fields from AutoConfig and the other is a broader loading regression." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Different model-specific failures: DeepSpeed ZeRO-3 load crash vs FineGrainedFP8Config failure." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Different models and symptoms; Qwen3VL weight-only loading is not the same as GLM5 loading." + }, + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "Different device/runtime issues: MPS histogram support vs Gemma4 device-map forward mismatch." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Completely different bugs: missing Qwen2.5-VL import vs `_is_hf_initialized` constructor incompatibility." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Different model-loading failures for different code paths; no shared concrete defect." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading regression vs flash-attn2 fallback being blocked by other checks." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "Remote-code loading regression and Gemma4 PLE device-mapping mismatch are unrelated." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "`add-new-model-like` tokenizer mapping bug and remote_code regression are different issues." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3 fine-tuning warning and hf4.x\u2192hf5.x LoRA incompatibility are not the same underlying bug." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Autocast dtype mismatch and device-map mismatch are separate runtime errors." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer regex patch crash vs flash-attn2 fallback gating are different code paths." + }, + { + "left": "issue:45715", + "right": "issue:46315", + "accept": false, + "reason": "Both are tokenizer performance regressions, but they hit opposite conversion APIs and are not the same bug." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Both involve model loading/remote code, but the concrete failures are different." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "MobileLLM key error and flash-attn2 fallback gating are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Tied-weight handling in Mistral3 and non-persistent buffer corruption are different regressions." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Same docs regression: both complain that v5 docs still reference removed pipeline-based summarization/translation tasks." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Same root bug: both report the unexpected `_is_hf_initialized` argument during model initialization/loading." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Qwen3MoE FP8 failure and bitsandbytes `_is_hf_initialized` incompatibility are separate regressions." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Whisper run failure and `_torch_extract_fbank_features()` issue are related at most by subsystem, not the same concrete bug." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Different training/runtime errors: unexpected `num_items_in_batch` kwarg vs mask/tensor shape mismatch." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analysis is well grounded in the packet and stays conservative on the soft-edge calls. The accepted pairs are plausibly the same concrete defect or docs regression, while the rejected pairs are mostly clearly distinct by title and failure mode. The cluster summary also accurately reflects that this is a mixed bag with only a few true duplicates." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 5, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36010", + "issue:38175", + "issue:41093", + "issue:41553", + "issue:42617", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43824", + "issue:43828", + "issue:43854", + "issue:43940", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44805", + "issue:44863", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:44991", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45584", + "issue:45588", + "issue:45663", + "issue:45685", + "issue:45823", + "issue:45834", + "issue:45902", + "issue:45923", + "issue:46402" + ], + "soft_pairs": [ + "issue:43257|issue:45305", + "issue:42947|issue:45923", + "issue:41553|issue:44991", + "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:43366|issue:43828", + "issue:45325|issue:45902", + "issue:45663|issue:45823", + "issue:43366|issue:43531", + "issue:44155|issue:45584", + "issue:43716|issue:44560", + "issue:43950|issue:45834", + "issue:43824|issue:43957", + "issue:44560|issue:45685", + "issue:43366|issue:44589", + "issue:42915|issue:43957", + "issue:43645|issue:43646", + "issue:43550|issue:44910", + "issue:44560|issue:44918", + "issue:42617|issue:43366", + "issue:36010|issue:42915", + "issue:44661|issue:45823", + "issue:44560|issue:44805", + "issue:38175|issue:43994" + ], + "prepared_review_unit_hash": "51234b80f8339ba1dc47ba9b04e9538542602add32579869b6ce12d39fbec1f9", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37155, + "estimated_input_tokens": 9289, + "estimated_eval_tokens": 18834 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issues with superficial title overlap. The only clear duplicate-style pair is 43645/43646 about the Transformers 5.0 custom-model initialization regression. The SigLIP2 pair 38175/43994 is only a plausible same-model relation from the titles, not a confirmed duplicate on the packet alone.", + "confidence": 0.79, + "canonical_issue_reason": "43645 is the broader, more self-contained report of the Transformers 5.0 custom-model initialization regression; 43646 is the narrower follow-up on the same failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "43645 is the strongest canonical issue candidate because it cleanly describes the bug and already subsumes the narrower duplicate report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Both involve DeepSpeed, but one is about weight conversion and the other about gradient averaging; different bugs." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "LoRA gradient-checkpointing behavior and a missing gradient-checkpointing flag are only loosely related." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Different models and different failures: bad Voxtral tokenizer messaging vs tokenizer loading failure for est-roberta." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are Qwen/ZeRO-3 weight-loading issues, but the titles point to different model generations and different failure descriptions." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Both are model-loading problems, but the packet does not show the same model or the same root cause." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "Qwen3.5 NaNs from SDPA/position_ids and a flash-attention AttributeError are distinct code-path failures." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": false, + "reason": "Both are mask-shape IndexErrors, but there is no evidence they are the same model or same code path." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "A broad video-input regression and a specific StopIteration in Qwen3-vl-embedding are not the same bug." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Both concern loading, but the symptoms and affected models are different." + }, + { + "left": "issue:44863", + "right": "issue:46402", + "accept": false, + "reason": "Different model families and different loading symptoms; not enough to treat as one bug." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "Unrelated training/runtime error versus image-processor segmentation-map failure." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is unrelated to NemotronH checkpoint loading." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "GGUF gpt-oss support and a Phi-tiny-MoE dtype mismatch are unrelated changes." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Different bugs: rope-index scaling for Qwen2.5-VL versus uninitialized-memory after meta-to_empty materialization." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states and Gemma4 PLE device mismatch are separate issues." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "gpt-oss GGUF support and Qwen3-MoE sliding_window behavior do not match." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "AudioFlamingo3 batched leakage and Whisper empty-transcription handling are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Mistral-3 dtype mismatch and Qwen3-vl video StopIteration are different code paths." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Non-persistent buffer corruption is unrelated to Kosmos2.5 OCR index errors." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Missing import of Qwen2_5_VLForConditionalGeneration is not the same as meta-device loading regressions." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Video-input regression and MPS histogram support are unrelated." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "gpt-oss support and missing Float8 storage are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "GenerationMixin import failure and meta-device loading breakage are different issues." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Same Transformers 5.0 custom-model initialization regression; 43646 is the narrower duplicate report." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different failures: torch.compile/SDPA on Bamba versus flash-attention illegal-memory-access on Qwen3.5." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Both mention Qwen3.5/video/embedding themes, but the concrete failures are different." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py execution failure is unrelated to gpt-oss GGUF support." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "GenerationMixin import error and FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES/add-new-model-like bug is unrelated to Gemma4 PLE device mismatch." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but the packet only shows related symptoms (zero probabilities vs nonsensical results), not enough to confirm the same underlying bug." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The review is mostly cautious, but it overstates evidence in a couple of places. In particular, treating issue:38175 / issue:43994 as a true duplicate-style pair is not well grounded from the packet alone: the titles indicate the same model family, but one is about zero probabilities and the other about nonsensical results, so this is at best a plausible relation, not a confirmed duplicate. The summary also says it found \"two true duplicate-style pairs,\" which is stronger than the available evidence supports. The 43645 / 43646 acceptance is plausible, but the overall conclusion should stay more conservative unless the issue bodies clearly establish duplication." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 6, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43708", + "issue:43720", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44075", + "issue:44164", + "issue:44291", + "issue:44387", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44960", + "issue:44977", + "issue:44991", + "issue:45030", + "issue:45276", + "issue:45292", + "issue:45313", + "issue:45335", + "issue:45357", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45507", + "issue:45812", + "issue:45834", + "issue:46050", + "issue:46468" + ], + "soft_pairs": [ + "issue:43632|issue:45834", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:44877|issue:45030", + "issue:44661|issue:44805", + "issue:43317|issue:44756", + "issue:38175|issue:39692", + "issue:43644|issue:44534", + "issue:42491|issue:43720", + "issue:41720|issue:42915", + "issue:44960|issue:44977", + "issue:42915|issue:44756", + "issue:44451|issue:44488", + "issue:45313|issue:45357", + "issue:42491|issue:45276", + "issue:45507|issue:46468", + "issue:44488|issue:44991", + "issue:43708|issue:44075", + "issue:43644|issue:43645", + "issue:39692|issue:43994", + "issue:45488|issue:45812", + "issue:43856|issue:43873", + "issue:43854|issue:43931", + "issue:43856|issue:44387", + "issue:43257|issue:43854", + "issue:45362|issue:45406", + "issue:44960|issue:46050", + "issue:42491|issue:45292", + "issue:42491|issue:43931", + "issue:44164|issue:44291", + "issue:45276|issue:45335" + ], + "prepared_review_unit_hash": "611b492fb3c971fef32030b0cb10f6e9fcb0275314ffc1a9ce89054eae79a71e", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36225, + "estimated_input_tokens": 9057, + "estimated_eval_tokens": 18370 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs look distinct from the metadata alone. The strongest duplicate candidate is 43644/44534: they share the same core wording about Transformers v5 filling non-persistent buffers with junk, so they likely describe the same bug, but the packet does not prove it and should not be treated as confirmed. The rest appear to be separate issues across different models or code paths.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 43644 is the best canonical candidate because it is the clearest, most general title for the non-persistent-buffer junk regression, and 44534 closely mirrors that wording. Based on titles alone, this looks like a probable duplicate pair rather than a confirmed one.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43644 is the best representative issue overall because it most directly names the shared regression in a model-agnostic way and is the strongest metadata match for 44534.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "Different bugs: HF init-flag regression vs. a Kosmos2.5 OCR mask/index error." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Both involve loading/training machinery, but one is a ZeRO-3 BertModel indexing failure and the other is tokenizer-mapping/model-registration logic." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Different concrete failures: ZeRO-3/BertModel index error vs. a tokenizer-mapping constraint in model registration." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Both are config-related, but they affect different models and the failures are not the same code path." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping registration bug vs. a tensor/mask shape mismatch at inference time." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Both mention loading/offload or memory, but the specific bugs are different: quantized model offload loading vs. mmap-driven OOM on Strix Halo." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "SigLIP2 output anomaly vs. documentation example errors; not the same underlying bug." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Very strong metadata match: the titles are near-paraphrases of the same Transformers v5 non-persistent-buffer junk regression, so this is the only likely duplicate pair in the packet." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Different underlying problems: Qwen3 MoE LoRA compatibility vs. BitNet packed-weight loading with accelerate." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Qwen3 auto device-map CUDA assert is unrelated to the Qwen3Moe FineGrainedFP8Config failure." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "Different models and different symptoms: GLM5 issue vs. Qwen3.5 flash-attention generation problem." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "FP8 config failure and mmap/OOM behavior are distinct bugs." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading failures, but the titles do not indicate the same concrete bug or shared code path." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Trainer resume/max_steps math bug vs. optimizer SGD arguments not being used." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Buffer initialization regression vs. Jupyter custom-model initialization issue; different failure modes." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Docs example errors vs. wrong outputs from AutoModel/pipeline; similar model family, different bug." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Different tokenizer regressions: ByteLevel/metaspace override in DeepSeek vs. Granite token-ID mismatch." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Training-memory inefficiency vs. quantization/offloading behavior; broad memory theme only." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different load failures and different model families; no clear shared fix." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "MoE weight conversion with accelerate+deepspeed is unrelated to GLM-4.7-Flash test loading." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Qwen3.5 chat crash and Gemma4 processor attribute error are distinct failures." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "GLM5 issue vs. quantized save/load regression after sinq; different affected paths." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading failures, but for different models and without evidence of a shared concrete bug." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "save/from_pretrained extra_state handling and init_empty_weights _is_hf_initialized are separate breakages." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Training-memory issue vs. int4 quantization OOM; related theme only, not the same bug." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Both involve resize_token_embeddings, but on different architectures and different embedding/output wiring." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen3.5 regressions: ZeRO-3 language_model weight loading vs. incorrect visual encoder keys on save_pretrained." + }, + { + "left": "issue:45507", + "right": "issue:46468", + "accept": false, + "reason": "GraniteMoEHybrid invalid-method bug and Nemotron example breakage are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is a bit too strong for the packet. The 43644/44534 pair is the strongest candidate because they share the same title and high similarity, but the packet does not confirm it as a duplicate, so phrases like \u201cconfirmed duplicate,\u201d \u201cone true duplicate pair,\u201d and \u201calmost verbatim\u201d overstate the evidence. The rest of the soft-edge rejections are generally conservative enough, but the canonical/best-issue rationale should be downgraded to probabilistic language grounded only in the metadata." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 7, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42947", + "issue:43335", + "issue:43381", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43828", + "issue:43854", + "issue:43873", + "issue:43957", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44479", + "issue:44488", + "issue:44756", + "issue:44863", + "issue:44877", + "issue:44910", + "issue:44912", + "issue:44928", + "issue:44960", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45685", + "issue:45698", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:46315", + "issue:46468" + ], + "soft_pairs": [ + "issue:44910|issue:45588", + "issue:43404|issue:43454", + "issue:41720|issue:44155", + "issue:43425|issue:43957", + "issue:39692|issue:44960", + "issue:44402|issue:44488", + "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:44387|issue:44756", + "issue:44292|issue:44912", + "issue:43957|issue:44756", + "issue:43335|issue:45663", + "issue:42947|issue:43381", + "issue:45685|issue:46468", + "issue:45276|issue:45292", + "issue:44863|issue:46468", + "issue:44164|issue:44756", + "issue:43854|issue:45216", + "issue:44479|issue:44877", + "issue:43381|issue:44928", + "issue:45684|issue:45698", + "issue:45588|issue:45663", + "issue:43646|issue:43708", + "issue:43645|issue:44164", + "issue:45292|issue:45335", + "issue:45812|issue:46315", + "issue:43828|issue:43957" + ], + "prepared_review_unit_hash": "acfff889b2fab1228730fd3224488a66852ca145a66055d79a2df27beb3363f2", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37467, + "estimated_input_tokens": 9367, + "estimated_eval_tokens": 18990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly heterogeneous Transformers bug reports, with a couple of tight duplicate families: resize_token_embeddings not propagating to model-specific embeddings, and untied lm_head weights in multimodal generation models. I\u2019d use the generic resize_token_embeddings report as the main canonical issue.", + "confidence": 0.62, + "canonical_issue_reason": "issue:45292 is the broadest, most model-agnostic report in the resize_token_embeddings family and cleanly subsumes the Gemma4/T5Gemma variants.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45292 is the best umbrella issue because its title captures the shared core bug without being model-specific, making it the clearest duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Both involve flash attention, but the failures are different: packed-sequence/position_ids misinterpretation vs s_aux=None handling on sink-less models." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": true, + "reason": "Both report untied lm_head weights causing bad generation in multimodal model classes; this looks like the same missing weight-tying bug." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Different symptoms and code paths: device-mapping CUDA assert on Qwen3 vs batched-inference token/embedding leakage in AudioFlamingo3." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "One is a Torch 2.10 compatibility report, the other is a meta-device model-loading regression; not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Doc-example errors for SigLIP2 and a separate GLM5 issue are unrelated." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Different problems: tokenizer vocab-size mismatch vs a model-loading failure for cjvt/sleng-bert." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the T5/apex RMSNorm memory leak." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Both mention quantization and memory, but one is offloading behavior and the other is reserved-memory/OOM under int4; not the same concrete bug." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "ImportError for GenerationMixin and Whisper pipeline return_language breakage are separate issues." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states KeyError and Qwen multimodal text_config unwrapping are different model-init problems." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "Unexpected _is_hf_initialized in init_empty_weights is unrelated to the Strix Halo mmap OOM report." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Different bugs: Gemma4 FSDP2 attention state rebuild vs Qwen3_5Moe rotary embedding CPU-offload incompatibility." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "AudioFlamingo3 token leakage is not the same as Gemma4 PLE device mismatch with device_map=auto." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "One is int4 quantization memory regression; the other is mmap/OOM on Strix Halo." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Both are loading/quantization issues, but they affect different models and different failure modes (load error vs fallback to bf16)." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading regression and Strix Halo mmap/OOM are not the same underlying bug." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction and Gemma4 FSDP2 attention state are unrelated." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is ineffective LoRA checkpointing and the other is an eval-mode restriction." + }, + { + "left": "issue:45685", + "right": "issue:46468", + "accept": false, + "reason": "MPS int histogram limitation in MoE and the Nemotron example failure are different issues." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same resize_token_embeddings bug: the Gemma4-specific report is a narrower manifestation of the general output_embeddings propagation failure." + }, + { + "left": "issue:44863", + "right": "issue:46468", + "accept": false, + "reason": "NemotronH checkpoint-loading failure and a broken official Nemotron example are too broad to treat as the same concrete bug." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "save/from_pretrained extra_state handling is unrelated to disabling mmap to avoid OOM." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "GLM-4.7-Flash load failure in tests and Qwen3.5 save_pretrained regression are separate model issues." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video-input regression for Qwen VL models and strict granite_speech config loading are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Gradient checkpointing eval-mode prohibition is not the same as the Qwen3.5 RLHF NaN/SDPA fallback regression." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Read-only permission propagation from save_pretrained and wrong custom-module loading after save_pretrained are different bugs." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Flash-attention s_aux=None handling and Gemma4 FSDP2 shared_kv_states rebuilding are unrelated." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization in Transformers 5.0 and Trainer max_steps calculation on resume are different regression classes." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Notebook-specific custom-model init breakage is not the same as save/from_pretrained extra_state handling." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same resize_token_embeddings propagation bug; t5gemma.decoder.embed_tokens is another model-specific target not being updated." + }, + { + "left": "issue:45812", + "right": "issue:46315", + "accept": false, + "reason": "Wrong token IDs for Granite models and slow convert_tokens_to_ids with many added tokens are separate tokenizer problems." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Autocast dtype mismatch on Phi-tiny-MoE and meta-device loading regressions are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays grounded in the packet and the accepted soft pairs are conservative. The resize_token_embeddings family is the clearest canonical cluster, and the untied lm_head pair is also reasonably supported by the titles. I don\u2019t see an overstatement that would require rejection." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 8, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:32090", + "issue:36032", + "issue:39404", + "issue:43299", + "issue:43425", + "issue:43452", + "issue:43582", + "issue:43596", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43828", + "issue:43856", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44265", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44661", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44805", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45701", + "issue:45759", + "issue:45823", + "issue:45859", + "issue:46050", + "issue:46326" + ], + "soft_pairs": [ + "issue:44402|issue:44451", + "issue:43425|issue:43828", + "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:44265|issue:44898", + "issue:39404|issue:44912", + "issue:44479|issue:44805", + "issue:44589|issue:45084", + "issue:43906|issue:46326", + "issue:46050|issue:46326", + "issue:43425|issue:44918", + "issue:43901|issue:44509", + "issue:43994|issue:44079", + "issue:36032|issue:43452", + "issue:43299|issue:43931", + "issue:44671|issue:44991", + "issue:43596|issue:44683", + "issue:44164|issue:45759", + "issue:44991|issue:45081", + "issue:43856|issue:44756", + "issue:39404|issue:44671", + "issue:44164|issue:44479", + "issue:44488|issue:45081", + "issue:44661|issue:44683", + "issue:44737|issue:45859", + "issue:44488|issue:44987", + "issue:32090|issue:39404", + "issue:43611|issue:43644", + "issue:43582|issue:43632" + ], + "prepared_review_unit_hash": "e19ee05a3bcef4c609d105c04a88556d70a96a5d0512302213d7efd3da323f55", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38204, + "estimated_input_tokens": 9551, + "estimated_eval_tokens": 19358 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose set of unrelated Transformers regression reports, many around v5/model-loading/tokenizer behavior but not true duplicates. No PRs are present; among the issues, the broadest and clearest representative is the base_model_prefix loading regression.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43611 is the most general, user-facing loading regression in the set and best matches the cluster\u2019s broad v5 breakage theme.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43611 is the strongest cluster representative: concrete, widely applicable, and closer to the shared model-loading regression pattern than the more model-specific tickets.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Both are tokenizer/model-loading complaints, but they involve different models and different failure modes; not the same bug." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "One is a general torch version incompatibility report, the other a Phi MoE autocast dtype mismatch; distinct problems." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are unrelated runtime issues." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Torch 2.9 flex_attention compilation failure is not the same as Gemma4 PLE device-map mismatch." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper return_language regression and a general tokenization change report are different behaviors and code paths." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Model loading via base_model_prefix and custom model initialization in Jupyter are separate regressions." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "Version incompatibility versus a Gemma4 forward device mismatch; not one underlying bug." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "These are both compatibility-style reports, but the affected paths and symptoms are different." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache/Qwen3.5 crash is unrelated to the 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 PagedAttentionCache linear_attention crash." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export failures for torch_compilable_check and Perceiver non-default resolution failure are different bugs." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper pipeline regression and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "A video-input regression across Qwen VL variants does not match the generic mask/index shape mismatch." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Missing Float8 storage and non-template-node compilation are unrelated." + }, + { + "left": "issue:43906", + "right": "issue:46326", + "accept": false, + "reason": "Reproduction of issue 38071 and mistral_common continue_final_message behavior are different tokenizer concerns." + }, + { + "left": "issue:46050", + "right": "issue:46326", + "accept": false, + "reason": "Quantized save/load breakage and a mistral_common message-format bug are not the same issue." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch version incompatibility and Qwen3.5 SFT input-unpacking failure are distinct problems." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs-related, but they document different deprecated/changed pipeline behaviors." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "SigLIP2 incorrect outputs and ModelOutput key assignment are separate functional bugs." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5 tokenizer special-token conflict and gguf_file loading failures affect different code paths." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both mention Qwen3-VL loading, but one is MoE/dev-version breakage and the other is a weight-shape mismatch; not the same root cause." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "CamemBERT masked-LM prediction regression and EMBEDDIA tokenizer loading failure are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "DeepSpeed zero3/BertModel index error is not the compiled flex_attention torch-version bug." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "extra_state save/load handling and composite Qwen config unwrapping are different serialization/configuration issues." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Two tokenizer-loading problems, but with different models and different failure mechanisms; do not dedupe." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3 MoE training memory inefficiency and Strix Halo mmap OOM are not the same underlying problem." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline behavior and CamemBERT masked-LM predictions are unrelated model-specific regressions." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "Serialization extra_state handling and Qwen VL video-input regression do not match." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Loading a specific BERT model and the Mistral regex-patch tokenizer crash are different bugs." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "A model-listing/add-new-model-like failure is unrelated to the flex_attention compilation regression." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "Both are device-placement related, but one is an XLNet CPU-arange inefficiency and the other is CPU-offload incompatibility in Qwen3_5Moe; not the same fix." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Two separate model-loading failures with different models and symptoms." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error and Whisper return_language regression are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "base_model_prefix loading breakage and non-persistent buffer junk-fill are distinct v5 regressions." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "AppleSilicon allocator warmup TypeError and the _is_hf_initialized flag breakage 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": 21, + "review_unit_index": 9, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:34567", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:42175", + "issue:42915", + "issue:43262", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43421", + "issue:43452", + "issue:43482", + "issue:43540", + "issue:43541", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43746", + "issue:43883", + "issue:43994", + "issue:44206", + "issue:44220", + "issue:44403", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44530", + "issue:44534", + "issue:44610", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45507", + "issue:45584", + "issue:45987", + "issue:46178", + "issue:46285", + "issue:46326", + "issue:46402" + ], + "soft_pairs": [ + "issue:46285|issue:46402", + "issue:43716|issue:45084", + "issue:40990|issue:43421", + "issue:36032|issue:42915", + "issue:43643|issue:44530", + "issue:34567|issue:40444", + "issue:44451|issue:44987", + "issue:44811|issue:45584", + "issue:43541|issue:45084", + "issue:42175|issue:46178", + "issue:40444|issue:46326", + "issue:42915|issue:44530", + "issue:43644|issue:43646", + "issue:43452|issue:43482", + "issue:44206|issue:44220", + "issue:43883|issue:44488", + "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:43404|issue:44466", + "issue:43883|issue:44991", + "issue:44403|issue:45081", + "issue:42915|issue:45084", + "issue:43994|issue:44610", + "issue:43322|issue:43540" + ], + "prepared_review_unit_hash": "7dc17ff6832b9bab97b108a477eae5c48b9cf08685a240a829f3941cea283639", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37416, + "estimated_input_tokens": 9354, + "estimated_eval_tokens": 18964 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous: the soft pairs span unrelated tokenizer, model-loading, vision, audio, and backend issues. I only see one likely duplicate pair in the audio feature-extraction path; the rest look like false positives rather than the same underlying bug.", + "confidence": 0.43, + "canonical_issue_reason": "issue:44206 is the clearest concrete bug report in the only likely duplicate family: it names the exact failing audio feature-extraction path, version, and crash mode. The companion issue is more generic, so 44206 is the better anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44206 is the strongest canonical candidate because it identifies a specific regression (`LasrFeatureExtractor` / `_torch_extract_fbank_features` / unsupported `center` arg) rather than a broad symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46285", + "right": "issue:46402", + "accept": false, + "reason": "Different models and failures: CLIP text-model structure inconsistency vs Sam3 text-encoder weights not loading." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Unrelated bugs: image preprocessor dtype mismatch vs Mistral regex patch/tokenizer attribute crash." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "Perplexity regression on a specific model is not the same as runtime special-token/post-processor update behavior." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "Tokenizer class conflict for T5 is unrelated to Qwen3Moe FP8 loading failure." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "AutoConfig missing fields under `trust_remote_code` is a different path from Qwen3.5 PagedAttentionCache group-type crashes." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "Trainer token-count tracking and Qwen2.5-VL multi-image iterable dataset failures are separate problems." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Both are model-loading failures, but they affect different repos and likely different code paths; not enough to merge as one bug." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both involve Whisper, but one is decode skip-special-tokens behavior and the other is empty-transcription generation after align_special_tokens." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Torch dynamo grouped_mm tracing failure is unrelated to the non-template-node compile error." + }, + { + "left": "issue:42175", + "right": "issue:46178", + "accept": false, + "reason": "Package extra/backend installation coverage is different from an invalid BACKENDS_MAPPING entry for tensorflow_text." + }, + { + "left": "issue:40444", + "right": "issue:46326", + "accept": false, + "reason": "Qwen2.5-VL iterable dataset multi-image handling is unrelated to mistral_common continue_final_message behavior." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Different Qwen-family failures: FineGrainedFP8Config loading vs linear_attention group-type crash in PagedAttentionCache." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both mention Transformers v5 breakage, but one is about non-persistent buffers and the other about custom model initialization; not the same concrete bug." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": false, + "reason": "GGUF loading is the broad theme, but one is an API/path break around `gguf_file`, while the other is a specific Qwen2.5-GGUF v5 load failure." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": true, + "reason": "These point to the same audio feature-extraction failure path; the generic `_torch_extract_fbank_features()` report matches the unsupported-arg crash described in 44206." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Missing `all_tied_weights_keys` in Molmo is unrelated to a model-loading failure for cjvt/sleng-bert." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer special-token method conflict is unrelated to custom model initialization breaking in notebooks." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "Device-map/offload loading failure is a different bug from GGUF loading regression in Transformers v5." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Compile-time template-node error is unrelated to GraniteMoEHybrid calling an invalid method." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3.5 hidden-state output bug and Qwen2.5-VL rope-index scaling bug affect different models and code paths." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are version-related loading failures, but for different models and different tokenizer/model code paths." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked-generation inconsistency under compiled forward is not the same as `StaticCache.get_seq_length()` returning the wrong type." + }, + { + "left": "issue:43746", + "right": "issue:46326", + "accept": false, + "reason": "PEFT local-checkpoint loading for GraniteSpeech is unrelated to mistral_common `continue_final_message` handling." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio processor sampling-rate defaults and MIMI encoder padding-mask behavior are different issues." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Generic custom-model initialization breakage is not the same as non-persistent buffers being filled with junk." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Custom model initialization regression and junk non-persistent buffers are related only at a high level, not the same underlying bug." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Weight-tying failure in Mistral3 is different from device-dependent serialization inconsistency for `lm_head.weight`." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Different components and failure modes: Molmo tied-weights attribute error vs tokenizer loading for est-roberta." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "General loading noise is not the same as a Mistral tokenizer regex patch crash." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "Qwen3Moe FP8 loading and template-node compilation failures are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Nonsensical model outputs and a processor/model input-size mismatch are distinct problems." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Llava Next loading segfault and Qwen3OmniMoe video-processing ValueError are different code paths and symptoms." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is appropriately conservative and grounded in the packet. The only accepted soft pair, issue:44206 vs issue:44220, is supported by closely related audio feature-extraction titles, and the remaining rejects are framed as different models, code paths, or failure modes without overstating equivalence. The canonical choice and low confidence also fit the evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 10, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:33666", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42915", + "issue:43089", + "issue:43296", + "issue:43377", + "issue:43404", + "issue:43421", + "issue:43425", + "issue:43531", + "issue:43611", + "issue:43632", + "issue:43742", + "issue:43792", + "issue:43819", + "issue:44062", + "issue:44079", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44479", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45663", + "issue:45987", + "issue:46209", + "issue:46315", + "issue:46326" + ], + "soft_pairs": [ + "issue:43377|issue:43819", + "issue:43089|issue:45987", + "issue:44451|issue:45081", + "issue:44987|issue:45081", + "issue:44991|issue:46315", + "issue:44265|issue:44610", + "issue:44220|issue:44683", + "issue:42915|issue:44291", + "issue:45216|issue:45357", + "issue:44062|issue:44987", + "issue:36032|issue:43742", + "issue:42915|issue:44589", + "issue:44479|issue:45325", + "issue:46209|issue:46326", + "issue:36032|issue:41720", + "issue:43632|issue:44756", + "issue:43404|issue:45127", + "issue:43531|issue:44877", + "issue:40444|issue:45325", + "issue:43296|issue:43531", + "issue:43611|issue:44534", + "issue:43425|issue:44403", + "issue:40990|issue:44811", + "issue:43792|issue:44589", + "issue:44403|issue:44991", + "issue:33666|issue:40444", + "issue:43632|issue:44534", + "issue:43421|issue:44568", + "issue:44683|issue:45020", + "issue:44079|issue:45663" + ], + "prepared_review_unit_hash": "65371c118f57de265cdb65d75bab5f514022acc734533b0f9c61c7889b381b5e", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36153, + "estimated_input_tokens": 9039, + "estimated_eval_tokens": 18334 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose similarity set of unrelated Transformers issues across different models/subsystems. I found one plausible duplicate pair around the Qwen3.5 save_pretrained checkpoint regression; the rest should stay separate.", + "confidence": 0.87, + "canonical_issue_reason": "If forced to pick a representative issue, issue 45216 is the best canonical: it states the broader Qwen3.5 save_pretrained checkpoint regression, while 45357 reads like a narrower follow-up about the same broken serialization path.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45216 is the best global representative because it is broader, earlier, and more likely to capture the root save_pretrained regression for Qwen3.5; no other issue cleanly covers the rest of the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Different bugs: MIMI batching/padding-mask mismatch vs DAC from_latents forward/STE mismatch." + }, + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Unrelated symptoms: generation GPU-sync/dispatch overhead vs StaticCache.get_seq_length returning the wrong type." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-loading failures, but the model and root cause differ: ScandiBERT load failure vs Mistral regex patch crash." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Different failures: loading physical-intelligence/fast vs Mistral tokenizer regex/backend-tokenizer crash." + }, + { + "left": "issue:44991", + "right": "issue:46315", + "accept": false, + "reason": "One is a tokenizer load failure, the other is a performance regression in convert_tokens_to_ids." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "Different code paths: torch.export exportability with torch_compilable_check vs a processor/model input-size mismatch." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio fbank feature extraction bug vs compiled flex_attention regression on newer torch; not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Distinct initialization/runtime problems: FineGrainedFP8Config failure vs unexpected _is_hf_initialized argument in init_empty_weights." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Same underlying save_pretrained serialization regression for Qwen3.5; 45357 looks like a narrower continuation about incorrect visual encoder keys." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken special-argument conflict is unrelated to loading physical-intelligence/fast." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Different tokenizer/model issues: T5 tokenizer special-token method conflict vs MobileLLM-125M key error on load." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "FP8 config failure and missing Float8 storage are separate runtime compatibility issues." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Related Qwen VL area, but not the same concrete bug: broad video-input regression across several models vs still-image temporal position scaling in Qwen2.5-VL." + }, + { + "left": "issue:46209", + "right": "issue:46326", + "accept": false, + "reason": "Unrelated: FP8 inference on B200 vs continue_final_message behavior for mistral_common tokenizers." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "T5 tokenizer loading conflict is unrelated to Qwen3 auto device mapping cudaErrorAssert." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "Different regressions: _is_hf_initialized handling in Transformers v5 vs mmap/OOM behavior on Strix Halo." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both involve tied embeddings, but the bugs differ: missing lm_head tying in Mistral3 vs LoRA merge collapse with extended vocabulary." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3-MoE sliding_window bug and granite_speech strict-config loading are unrelated." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Multi-image iterable finetuning failure vs Qwen2.5-VL still-image temporal position scaling; different symptoms and code paths." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "PaddleOCR-VL loading failure is unrelated to Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "base_model_prefix loading breakage and non-persistent buffer junk are separate v5 regressions." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch 2.10 incompatibility is not the same as noisy transformer loading logs." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity regression on gpt-oss-20b is unrelated to Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper runtime failure and missing Float8 storage type are different compatibility bugs." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Loading noise vs tokenizer load failure; not the same issue." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both are Qwen2-VL-family training issues, but one is multi-GPU training and the other is iterable-dataset multi-image finetuning; not the same bug." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "_is_hf_initialized regression and non-persistent buffer junk are separate Transformers v5 problems." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime post-processor update support is not the same as add_special_tokens failing to add BOS/EOS for a specific tokenizer." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and remote_code model loading failures are unrelated." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment bug and Gemma4 shared_kv_states FSDP2 error are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays appropriately conservative. The only accepted duplicate pair, issue:45216 and issue:45357, is clearly aligned by title and scope, and the remaining soft-edge rejections are mostly cautious distinctions across different models/subsystems. No obvious overstatement or mergeability problem." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 11, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:35707", + "issue:40444", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43381", + "issue:43454", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43701", + "issue:43749", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44462", + "issue:44466", + "issue:44534", + "issue:44589", + "issue:44829", + "issue:44898", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45542", + "issue:45812", + "issue:45834", + "issue:45923", + "issue:46402" + ], + "soft_pairs": [ + "issue:43381|issue:45923", + "issue:44291|issue:44589", + "issue:44829|issue:45479", + "issue:44462|issue:45812", + "issue:43482|issue:45538", + "issue:43482|issue:43525", + "issue:35707|issue:40444", + "issue:43317|issue:43856", + "issue:41553|issue:43582", + "issue:42175|issue:45542", + "issue:43818|issue:46402", + "issue:43818|issue:45357", + "issue:44079|issue:44534", + "issue:41720|issue:45081", + "issue:43531|issue:44291", + "issue:43377|issue:45491", + "issue:40444|issue:45381", + "issue:44265|issue:45072", + "issue:43644|issue:44079", + "issue:45325|issue:45381", + "issue:43749|issue:44079", + "issue:43611|issue:43646", + "issue:42175|issue:43296", + "issue:43792|issue:43957", + "issue:43994|issue:44898", + "issue:45325|issue:45834", + "issue:43454|issue:44466", + "issue:43632|issue:45072", + "issue:42915|issue:43482", + "issue:43701|issue:45325" + ], + "prepared_review_unit_hash": "1ac6c6af845b96a00a91ece34571672f10f7a39f268dc731cff80b0b12764b92", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36863, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated transformer regressions. The only soft edge that looks like the same underlying bug is the Qwen2.5-VL position-id regression pair (45325/45381); the rest are too symptom- or model-specific to merge safely.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 45325 is the clearest root-cause report: it pinpoints the Qwen2.5-VL `get_rope_index` position-id scaling bug, which can reasonably subsume the closely related video-input report.", + "canonical_pr_reason": null, + "best_issue_reason": "45325 is the best representative issue because it names the concrete failing code path and is the most general description of the accepted duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "Different bugs: eval-mode gradient checkpointing use versus a missing `supports_gradient_checkpointing` flag." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Unrelated TypeErrors in different loading paths (`init_empty_weights` vs float8 storage lookup)." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Different training failures: FlashAttention-3 instability versus zero loss for `num_labels=1` classification." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but one ignores `tokenizer.json` while the other yields wrong Granite token IDs." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Different failures: GGUF loading under v5 versus an absurd `model_max_length` on CLIPTokenizer." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "No shared code path; GGUF loading and `Llama4Config.pad_token_id` are separate issues." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Different multimodal problems: progressive generation with `past_key_values` versus multi-image iterable finetuning." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Offload/device-map loading failure and Qwen3 MoE training memory use are not the same bug." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad tokenizer error messaging is unrelated to the Apple Silicon allocator warmup TypeError." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Both mention TF/backend packaging, but they fail at different points and are not the same fix." + }, + { + "left": "issue:43818", + "right": "issue:46402", + "accept": false, + "reason": "Different model families and symptoms: Video-LLaVA weight/temporal-attention issues versus Sam3Model text-encoder loading." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Video-LLaVA architecture issues are unrelated to Qwen3.5 save_pretrained key serialization." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer junk are distinct v5 regressions." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Qwen3 auto device-map CUDA assert is unrelated to Mistral tokenizer regex patching." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Sliding-window handling in Qwen3-MoE is unrelated to the `_is_hf_initialized` init_empty_weights issue." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Both involve batching/padding, but one is missing padding-mask support in MIMI and the other is NaNs from Gemma3 sliding-window all-padding cases." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL multimodal issues, but one is multi-image finetuning and the other is video position-id handling." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "Export with `torch_compilable_check` and bf16 dtype mismatches are different problems." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "One is a v5 non-persistent buffer regression; the other is `ModelOutput` key assignment behavior." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Both point at the same Qwen2.5-VL vision/temporal position-id computation in 5.3.0; a single fix to `get_rope_index`/position-id handling could plausibly cover both still-image and video inputs." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading and ModelOutput key assignment are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are broad v5 regressions, but one is `base_model_prefix` loading and the other custom model initialization; not the same concrete bug." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Packaging/backend installation issue versus a PaddleOCR-VL loading regression." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Whisper runtime failure and meta-device model loading are separate issues." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model failures with different code paths and reproduction conditions." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Qwen2.5-VL position-id scaling and Kosmos2.5 long-OCR index errors are different bugs." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Both mention `lm_head`, but one is missing weight tying and the other is inconsistent serialization depending on device." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "`_is_hf_initialized` breakage is unrelated to bf16 dtype mismatches." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different model/load failures: Qwen3MoE FP8 config versus Qwen2.5-GGUF loading under v5." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL position-id scaling." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The result is grounded in the packet and the only accepted soft edge (45325/45381) is plausibly the same Qwen2.5-VL position-id bug. The remaining pair judgments are conservative and do not overclaim sameness from the available titles alone." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 12, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42947", + "issue:43257", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43454", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44877", + "issue:44898", + "issue:44991", + "issue:45072", + "issue:45325", + "issue:45584", + "issue:45701", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45920", + "issue:45923", + "issue:46050", + "issue:46291", + "issue:46305", + "issue:46315" + ], + "soft_pairs": [ + "issue:44291|issue:44877", + "issue:44479|issue:45823", + "issue:43606|issue:44265", + "issue:43873|issue:45701", + "issue:43296|issue:45800", + "issue:43262|issue:43746", + "issue:40990|issue:45584", + "issue:34634|issue:35707", + "issue:44568|issue:45920", + "issue:35707|issue:41720", + "issue:44186|issue:44265", + "issue:43881|issue:44488", + "issue:43606|issue:43819", + "issue:39692|issue:40444", + "issue:41720|issue:45823", + "issue:43454|issue:43550", + "issue:46305|issue:46315", + "issue:43377|issue:44610", + "issue:43454|issue:45072", + "issue:43317|issue:43873", + "issue:43257|issue:45923", + "issue:40990|issue:42947", + "issue:43377|issue:44898", + "issue:43322|issue:46050", + "issue:43572|issue:45759", + "issue:43454|issue:43720", + "issue:45325|issue:46315", + "issue:41720|issue:43322", + "issue:43749|issue:44534", + "issue:44991|issue:46291" + ], + "prepared_review_unit_hash": "413c4918b048acf428dccef48ade10eec2c347dbceb87b5371269ce9ea5362b2", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36338, + "estimated_input_tokens": 9085, + "estimated_eval_tokens": 18426 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pair looks like the same concrete bug/change; the list is a mix of unrelated model-loading, tokenizer, offload, and multimodal regression reports. If a single representative issue is needed, the broad offloading/quantization report is the best anchor.", + "confidence": 0.83, + "canonical_issue_reason": "issue:43873 is the broadest and most central offloading/quantization bug report in the set, so it is the best umbrella issue if one representative must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 has the clearest generic scope around offloading plus quantization, making it the most suitable cluster representative among these unrelated reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Different failure modes: init_empty_weights/_is_hf_initialized vs strict config loading for granite_speech." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Both are multimodal regressions, but one is a video-input regression and the other is a device-mismatch forward failure." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "Bark CPU offload device mismatch is unrelated to torch.export.export failing on torch_compilable_check." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Offloading/quantization bug vs tokenizer behavior change; not the same code-path problem." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different incompatibilities: PaddleOCR-VL loading in vLLM vs torch/transformers version mismatch." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sampling-rate default is unrelated to PEFT adapter checkpoint loading." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "Perplexity regression on a language model is unrelated to Whisper empty-transcription generation failure." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "Voice preset handling in Bark is unrelated to progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "Specific BOS/EOS insertion bug for one tokenizer is not the same as broad wrong-token-IDs regression across multiple models." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Generation cache/input-embed issue vs auto device-mapping cudaErrorAssert are different problems." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer crash on NER/padding is unrelated to export failures involving torch_compilable_check." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Model loading failure for glm-4v-9b vs tokenizer loading failure for cjvt/sleng-bert are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Both mention loading/latents, but one is CPU offload device mismatch and the other is DAC from_latents vs forward mismatch." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "SigLIP2 documentation errors are unrelated to Qwen2.5-VL finetuning with multi-image iterable datasets." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both are device-related, but one is auto device-map crash on A800 and the other is a forward-time device mismatch in Gemma4 PLE." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "lm_head tying/garbage generation is unrelated to torch.compile SDPA failure." + }, + { + "left": "issue:46305", + "right": "issue:46315", + "accept": false, + "reason": "Streaming mask off-by-one is unrelated to tokenizer performance regression." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "Padding-mask batch inconsistency in MIMI is unrelated to OmDet-Turbo processor image-size mismatch." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Weight tying bug in AyaVision is unrelated to dtype mismatch in SwitchTransformers/TimmWrapperModel inference." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Similar offload/quantization theme, but one is dequantized-model loading on gpu+cpu offload and the other is a broader offloading regression; not enough evidence of one concrete bug." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to a missing supports_gradient_checkpointing flag." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "High perplexity on a model is unrelated to gradient checkpointing not being effective with LoRA." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "MIMI encoder batching bug is unrelated to Perceiver image classification at non-default resolution." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Llava Next segfault on load is unrelated to sinq quantized model save/load breakage." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to AutoModelForCausalLM.from_config not unwrapping text_config." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "AyaVision lm_head tying bug is unrelated to BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:45325", + "right": "issue:46315", + "accept": false, + "reason": "Qwen2.5-VL rope-index regression is unrelated to slow convert_tokens_to_ids behavior." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "cudaErrorAssert under auto device mapping is not the same as a Llava Next loading segfault." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP CPU RAM efficient loading is unrelated to non-persistent buffer initialization junk in v5." + }, + { + "left": "issue:44991", + "right": "issue:46291", + "accept": false, + "reason": "Tokenizer loading regression for EMBEDDIA/est-roberta is unrelated to kernels v0.15 requiring version/revision on import." + } + ] + }, + "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": 21, + "review_unit_index": 13, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:36010", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42617", + "issue:42915", + "issue:43065", + "issue:43295", + "issue:43317", + "issue:43404", + "issue:43425", + "issue:43450", + "issue:43525", + "issue:43526", + "issue:43550", + "issue:43577", + "issue:43606", + "issue:43708", + "issue:43720", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44008", + "issue:44186", + "issue:44387", + "issue:44423", + "issue:44466", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45704", + "issue:45823", + "issue:46291", + "issue:46402" + ], + "soft_pairs": [ + "issue:42617|issue:42915", + "issue:41720|issue:44977", + "issue:44466|issue:45357", + "issue:43526|issue:43883", + "issue:43881|issue:44991", + "issue:44387|issue:45704", + "issue:40990|issue:45701", + "issue:36331|issue:42617", + "issue:43065|issue:46402", + "issue:44186|issue:45072", + "issue:36331|issue:42915", + "issue:43450|issue:45701", + "issue:44877|issue:44991", + "issue:43873|issue:44756", + "issue:43550|issue:43720", + "issue:43577|issue:43708", + "issue:43404|issue:45357", + "issue:44008|issue:45823", + "issue:42175|issue:46291", + "issue:41720|issue:42617", + "issue:44423|issue:45406", + "issue:43550|issue:43606", + "issue:44610|issue:44898", + "issue:44186|issue:44610", + "issue:43295|issue:45701", + "issue:42915|issue:43317", + "issue:45632|issue:45684", + "issue:36010|issue:42617", + "issue:43425|issue:46291", + "issue:41720|issue:45161", + "issue:43525|issue:46291", + "issue:44671|issue:44743" + ], + "prepared_review_unit_hash": "6ea41a1f64abaf757a91a9b736e5c31da52186c8811aa7cb5cdeeeb7af02f907", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36821, + "estimated_input_tokens": 9206, + "estimated_eval_tokens": 18668 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These soft links are mostly keyword-similar but not true duplicates: they span different models, code paths, and failure modes (trainer, serving, quantization, tokenization, processors, loading). I would reject every soft edge.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue fits the cluster: the linked issues describe distinct bugs rather than one underlying defect.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the most discussion-heavy and has inbound references, so it is the closest thing to a hub, but it is still too broad to serve as a true duplicate representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Different failures in different contexts: 3d_parallel.py execution vs FineGrainedFP8Config loading." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Both involve Qwen3/Qwen3.5, but one is device-map CUDA assert and the other is flash-attention generation behavior." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Different bugs: device-dependent lm_head serialization vs incorrect visual encoder keys on save_pretrained." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "Unrelated components: BeitImageProcessorFast label reduction vs Molmo tied-weights attribute error." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Different models and subsystems: glm-4v loading vs tokenizer loading for est-roberta." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Different root causes: int4 quantization memory growth vs apex RMSNorm memory leak." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity regression on one model/dataset is not the same bug as a tokenization behavior change." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Trainer compute_loss signature breakage is unrelated to running 3d_parallel.py." + }, + { + "left": "issue:43065", + "right": "issue:46402", + "accept": false, + "reason": "Sam3 pixel decoder dummy conv2d is a different issue from Sam3 text encoder weights not loading." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer padding/truncation crash and bfloat16 dtype mismatch are different code paths." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "CustomTrainer API breakage is unrelated to Qwen3Moe FP8 loading failure." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video processor batch-shape bug is unrelated to tokenization changes." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Strict config loading for granite_speech is not the same as tokenizer loading failure." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Quantization/offloading behavior and mmap OOM on Strix Halo are different problems." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba is unrelated to packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "BLIP2 dtype propagation and Trainer resume max_steps calculation are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Tied lm_head weights in Mistral3 are a different defect from Qwen3.5 visual encoder key serialization." + }, + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3n forward variable collision and Gemma4 PLE device mismatch are distinct bugs." + }, + { + "left": "issue:42175", + "right": "issue:46291", + "accept": false, + "reason": "Missing TensorFlow backend in torch extras is unrelated to kernels v0.15 version/revision requirements." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Different models and symptoms: auto device-map CUDA assert vs 3d_parallel.py execution failure." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Both involve serve, but the crashes come from different causes: string `.to` misuse vs missing `_tokenizer`." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Different device/runtime issues: torch.compile SDPA failure vs CPU-offload device mismatch." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both are processor/image-size issues, but they affect different models and preprocessing behavior." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer crash in LayoutLMv2 and OmDet-Turbo image-size mismatch are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor.tokenizer regression is not the same as version-dependent tokenization changes." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Qwen3Moe FP8 loading and dequantized device_map offload failures are different bugs." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collision and read-only permission propagation are unrelated save/load issues." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "Importing GenerationMixin and running 3d_parallel.py are unrelated failures." + }, + { + "left": "issue:43425", + "right": "issue:46291", + "accept": false, + "reason": "Torch 2.10 incompatibility is unrelated to kernels version/revision handling." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Auto device-mapping CUDA assert is not the same as tensor-parallel support failure in GPT-OSS MoE." + }, + { + "left": "issue:43525", + "right": "issue:46291", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config is unrelated to kernels dependency versioning." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT MLM prediction regression and Qwen3.5 recurrent-state reset are different model-specific bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 14, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:29159", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36032", + "issue:36331", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:43316", + "issue:43322", + "issue:43329", + "issue:43388", + "issue:43421", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43873", + "issue:43881", + "issue:44186", + "issue:44361", + "issue:44387", + "issue:44492", + "issue:44509", + "issue:44530", + "issue:44625", + "issue:44913", + "issue:45059", + "issue:45081", + "issue:45357", + "issue:45406", + "issue:45464", + "issue:45479", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:46291", + "issue:46402" + ], + "soft_pairs": [ + "issue:45081|issue:45812", + "issue:43531|issue:43572", + "issue:44186|issue:44361", + "issue:43421|issue:45812", + "issue:42915|issue:43322", + "issue:43540|issue:43782", + "issue:43526|issue:43881", + "issue:36331|issue:41720", + "issue:43316|issue:44913", + "issue:29159|issue:36032", + "issue:33357|issue:35707", + "issue:43388|issue:45059", + "issue:43550|issue:43653", + "issue:43606|issue:43720", + "issue:30064|issue:34634", + "issue:43873|issue:45704", + "issue:45406|issue:45464", + "issue:44530|issue:46291", + "issue:41720|issue:43746", + "issue:44387|issue:45701", + "issue:45357|issue:46402", + "issue:43606|issue:43653", + "issue:43531|issue:45759", + "issue:42490|issue:43329", + "issue:34634|issue:41720", + "issue:44492|issue:44509", + "issue:34567|issue:35707", + "issue:44625|issue:45479" + ], + "prepared_review_unit_hash": "dd94a886ad4eea7cc87d3708e98357cdd3a279bc0e3f92765785adac94389c04", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36311, + "estimated_input_tokens": 9078, + "estimated_eval_tokens": 18412 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pair looks like the same underlying bug/change. The items span distinct models, tokenizers, processors, configs, and runtime paths, so they should remain separate issues.", + "confidence": 0.97, + "canonical_issue_reason": "issue:43873 is the most central/high-activity report in this set (highest discussion plus multiple inbound references) and is the broadest bug description among the issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best representative issue because it has the strongest signal of impact and follow-on references, even though this set does not form a true duplicate cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45081", + "right": "issue:45812", + "accept": false, + "reason": "Different tokenizer families and failure modes: Mistral regex patching vs Granite token IDs regression." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Unrelated model/config bugs: Qwen3-MoE sliding window behavior vs StableLm pad_token_idx after upgrade." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both tokenizer-related, but they concern different classes and errors: LayoutLMv2 NER/padding crash vs MLuke AttributeError." + }, + { + "left": "issue:43421", + "right": "issue:45812", + "accept": false, + "reason": "Runtime post-processor update bug is distinct from Granite AutoTokenizer generating wrong token IDs." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Different domains: Qwen3Moe FP8 config failure vs Llava Next segmentation fault on load." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Different Qwen multimodal paths and symptoms: video input ValueError vs weight_only load error." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "Beit image-processor label reduction bug is unrelated to glm-4v-9b loading failure." + }, + { + "left": "issue:36331", + "right": "issue:41720", + "accept": false, + "reason": "Trainer API signature mismatch is unrelated to Qwen3 auto device mapping cudaErrorAssert." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Both are config issues, but different fields/models: Gemma3TextConfig API discrepancy vs GPTNeoX rotary_pct reload regression." + }, + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Tokenizer inconsistency vs T5Tokenizer method-name conflict are different bugs with different root causes." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "MacOS CLIP bus error is unrelated to progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Metrics batch-label dropping and SAM3 text/bbox behavior are different code paths and symptoms." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "torch.compile SDPA failure for Bamba is unrelated to BigBirdTokenizer special-token registration." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both involve loading/offloading, but they are separate bugs: CPU-offload device mismatch vs packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Image segmentation-map processing bug is unrelated to BarkProcessor voice_preset handling." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior is a different bug from T5's apex FusedRMSNorm memory leak." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Different API surfaces and failures: Gemma4Processor missing _tokenizer vs Qwen3.5 streaming chat/completions failure." + }, + { + "left": "issue:44530", + "right": "issue:46291", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is unrelated to kernels package version/revision import failure." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3 auto device mapping cuda error is unrelated to GraniteSpeech PEFT local checkpoint loading." + }, + { + "left": "issue:44387", + "right": "issue:45701", + "accept": false, + "reason": "CUDA memory growth under int4 quantization is unrelated to a broad tokenization-version change report." + }, + { + "left": "issue:45357", + "right": "issue:46402", + "accept": false, + "reason": "Both are multimodal save/load issues, but one is wrong visual encoder keys on save and the other is missing text encoder weights on load." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "Different tokenizer/model subsystems and unrelated failure modes." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to AutoModelForCausalLM from_config text_config unwrapping for Qwen multimodal configs." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Trailing-slash import warning is unrelated to multimodal token counting video-branch undefined names." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset bug is unrelated to Qwen3 device-mapping cuda assertions." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategy docs/code is not the same as stale docs listing removed pipeline tasks." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token counter update bug is unrelated to progressive generation with embeddings and cache reuse." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Qwen3.5 num_labels propagation is a config wiring bug, while the other is a classification loss degeneracy across models." + } + ] + }, + "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": 21, + "review_unit_index": 15, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:39692", + "issue:41720", + "issue:42886", + "issue:43381", + "issue:43421", + "issue:43531", + "issue:43596", + "issue:43645", + "issue:43653", + "issue:43746", + "issue:43782", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44077", + "issue:44186", + "issue:44220", + "issue:44368", + "issue:44451", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44610", + "issue:44625", + "issue:44671", + "issue:44843", + "issue:44898", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45859", + "issue:45923", + "issue:46032", + "issue:46050", + "issue:46463" + ], + "soft_pairs": [ + "issue:44451|issue:44991", + "issue:44514|issue:44545", + "issue:43531|issue:44589", + "issue:44898|issue:45072", + "issue:45759|issue:45859", + "issue:44077|issue:44625", + "issue:44062|issue:44451", + "issue:45588|issue:45759", + "issue:43421|issue:43927", + "issue:44368|issue:45357", + "issue:44987|issue:45042", + "issue:39692|issue:44610", + "issue:44918|issue:44936", + "issue:43782|issue:45381", + "issue:43746|issue:46050", + "issue:43421|issue:43653", + "issue:43381|issue:43844", + "issue:42886|issue:44843", + "issue:45704|issue:45923", + "issue:46032|issue:46463", + "issue:45704|issue:45859", + "issue:41720|issue:45704", + "issue:43421|issue:44186", + "issue:45663|issue:45923", + "issue:43531|issue:43645", + "issue:43994|issue:44671", + "issue:45081|issue:45356", + "issue:43746|issue:43873", + "issue:39692|issue:43873", + "issue:43596|issue:44220" + ], + "prepared_review_unit_hash": "d0a423acdfe6f5fd1717a0e707e3d7cc9e2dd59c3d949df2f7e4544f67971ce3", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37114, + "estimated_input_tokens": 9279, + "estimated_eval_tokens": 18814 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous, with two clear duplicate pairs standing out: 44514/44545 and 42886/44843. The Mistral-regex pair 45081/45356 should be rejected as related tokenizer regressions, not proven the same bug.", + "confidence": 0.69, + "canonical_issue_reason": "If one issue must anchor the cluster, 44545 is the best local canonical because it is an exact duplicate of 44514 and has slightly more inbound context; the set is otherwise mixed, so this is only a local choice.", + "canonical_pr_reason": null, + "best_issue_reason": "44545 is the strongest representative issue in the cluster because it sits in the clearest exact-duplicate pair (44514/44545) and is well-supported by the packet.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading problems, but they affect different models and the packet does not show the same underlying bug." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Near-identical title and symptom; these appear to be the same batched apply_chat_template crash." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Different failures: Qwen3-MoE sliding-window behavior vs a Float8 storage TypeError." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Different subsystems and symptoms: Perceiver resize/interpolation vs dtype mismatch in inference." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Multimodal config unwrapping and CPU-offload incompatibility are different code-path problems." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "A post_init restriction on patchtsmixer is unrelated to Qwen3.5 num_labels propagation." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "An AddedToken keyword collision is unrelated to a model-specific tokenizer load failure." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink handling is unrelated to Qwen multimodal config unwrapping." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Both are tokenizer-related, but runtime special-token post-processing is not the same bug as save/load loss of custom token IDs." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "A training warning about tie_word_embeddings does not match a visual-encoder save_pretrained regression." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "A model-loading regression and a PIL image processor dependency issue are different problems." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Both mention processor/model mismatches, but the concrete failures are different models and different symptoms." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "TRl embedding unpacking failure is unrelated to trainer.evaluate() failing after trainer.train()." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Different Qwen-VL bugs: weight_only load failure vs wrong vision_position_ids in video input." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "PEFT local checkpoint loading and sinq quantized save/load corruption are different issues." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both involve tokenizer state, but the concrete bugs differ: runtime post-processor updates vs BigBird mask-token registration." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Gradient checkpointing in eval mode is unrelated to ZeRO-3 gradient growth." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": true, + "reason": "Same offline tokenizer-loading regression; 44843 identifies the concrete _patch_mistral_regex model_info() call that breaks HF_HUB_OFFLINE." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "A T5 RMSNorm memory leak is unrelated to a missing gradient-checkpointing flag." + }, + { + "left": "issue:46032", + "right": "issue:46463", + "accept": false, + "reason": "Mamba2 use_cache correctness and a T5 decoder-input error message are not the same issue." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "A normalization memory leak is unrelated to Qwen3_5Moe CPU-offload behavior." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "Auto device mapping cuda asserts and a T5 RMSNorm leak have no common underlying change." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Different tokenizer failures: special-token post-processing vs LayoutLMv2 NER padding/truncation crashes." + }, + { + "left": "issue:45663", + "right": "issue:45923", + "accept": false, + "reason": "Gemma-4 FSDP2 shared_kv_states bug is unrelated to a missing capability flag on Nemotron-3-Nano-Omni." + }, + { + "left": "issue:43531", + "right": "issue:43645", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior is unrelated to Jupyter custom-model initialization in Transformers 5.0.0." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different model regressions with different symptoms: SigLIP2 pipeline nonsense vs CamemBERT masked-LM prediction errors." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "They share the Mistral-regex area, but the packet does not prove the same underlying bug rather than two related tokenizer regressions." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Local PEFT checkpoint loading and quantization/offloading behavior are distinct issues." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "SigLIP2 docs/model-processor mismatch plus quantization failure is not the same as general offloading with quantization." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "A ZeRO-3/BertModel indexing error is unrelated to _torch_extract_fbank_features()." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Most of the conservative false/true calls are grounded, but one accepted edge looks overstated: issue:45081 vs issue:45356 share the general _patch_mistral_regex area, yet the packet does not show they are the same underlying bug rather than two related tokenizer regressions. Also, the canonical-issue rationale for 44545 leans on \u2018clearer report,\u2019 which is not directly supported by the packet. The cluster otherwise looks mixed, so I\u2019d reject on precision grounds." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 16, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:36331", + "issue:41720", + "issue:42915", + "issue:43295", + "issue:43322", + "issue:43335", + "issue:43388", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43618", + "issue:43643", + "issue:43653", + "issue:43697", + "issue:43716", + "issue:43720", + "issue:43746", + "issue:43756", + "issue:43819", + "issue:43950", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44448", + "issue:44488", + "issue:44661", + "issue:44671", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45183", + "issue:45356", + "issue:45479", + "issue:45685", + "issue:45834", + "issue:45907", + "issue:45987", + "issue:46032", + "issue:46294" + ], + "soft_pairs": [ + "issue:45356|issue:45907", + "issue:42915|issue:43746", + "issue:43540|issue:46294", + "issue:30064|issue:36010", + "issue:43720|issue:43819", + "issue:43653|issue:43720", + "issue:44220|issue:44661", + "issue:43550|issue:45072", + "issue:43295|issue:45834", + "issue:43618|issue:43697", + "issue:44062|issue:44488", + "issue:43550|issue:43819", + "issue:43540|issue:44008", + "issue:44186|issue:44898", + "issue:43653|issue:43819", + "issue:45183|issue:45479", + "issue:43322|issue:45685", + "issue:43531|issue:43643", + "issue:44843|issue:45081", + "issue:43335|issue:43756", + "issue:43388|issue:46032", + "issue:43716|issue:44008", + "issue:43643|issue:44877", + "issue:30064|issue:36331", + "issue:41720|issue:44991", + "issue:44448|issue:44671", + "issue:43950|issue:45987", + "issue:44821|issue:45084" + ], + "prepared_review_unit_hash": "e5386a338e95e8f253d297c84940dfb5c2f8c727d1b1f2e7db5259e51ee46cf1", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35898, + "estimated_input_tokens": 8975, + "estimated_eval_tokens": 18206 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: it mixes unrelated tokenizer, multimodal, config, inference, and training bugs across different models and code paths. I do not see a single canonical issue or PR that would safely represent the whole set.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Different subsystems: tokenizer regression vs Hub file-listing revision forwarding. No shared underlying bug." + }, + { + "left": "issue:42915", + "right": "issue:43746", + "accept": false, + "reason": "Different models and failures: Qwen3Moe FP8 config vs GraniteSpeech PEFT checkpoint loading." + }, + { + "left": "issue:43540", + "right": "issue:46294", + "accept": false, + "reason": "Both are Qwen video issues, but one is a video-processing ValueError and the other is chat-template feature/token mismatch; separate defects." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Unrelated: image processor segmentation-map handling vs GenerationMixin import error." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "Different code paths: accelerate loading/unpacked BitNet weights vs DAC forward/from_latents STE mismatch." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "Tokenizer special-token registration issue vs BitNet weight unpacking bug; unrelated." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio feature extraction bug vs tokenizer mapping lookup failure; no same underlying change." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Both are model inference bugs, but in different models and mechanisms: SDPA/torch.compile vs bf16 dtype mismatch." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Both are multimodal regressions, but one is processor.tokenizer/image-to-tokenizer breakage and the other is OCR long-input index error." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "Different regressions: CLIP output attentions field vs RTDetrV2 output changes." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "One is AddedToken constructor misuse; the other is loading a specific model/tokenizer failure. Symptoms and fixes differ." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Unrelated model bugs: SDPA compile failure vs DAC latent reconstruction mismatch." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Different multimodal paths: Qwen3OmniMoe video processing vs Gemma3n audio mask variable collision." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer padding/NER crash vs image-classification resolution/interpolation bug; no common root cause." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "BigBird special-token registration is unrelated to DAC latent/STE behavior." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Both touch sequence classification, but one is input-validation messaging and the other is zero-loss behavior; distinct bugs." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Completely different areas: Llava Next load crash vs MPS histogram implementation in MoE." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Qwen3-MoE sliding-window issue vs AutoConfig trust_remote_code field-loss regression; unrelated." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Both involve _patch_mistral_regex, but one is offline model_info access and the other is backend_tokenizer AttributeError; separate failure modes." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs Smollm3 RoPE-layer-count regression; different model internals." + }, + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Evaluation gather/truncation bug vs Mamba2 cache correctness bug; not the same issue." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Different models and mismatches: Mistral image-preprocessor dtype vs Gemma3n audio mask variable collision." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-loading regressions, but one is missing fields from trust_remote_code and the other is strict config rejection for granite_speech." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Image processor segmentation maps vs CustomTrainer.compute_loss signature change; unrelated." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "Qwen3 auto device-map CUDA assert vs tokenizer loading failure for a different model; no shared bug." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Both are v5 output regressions, but they affect different models and likely different code paths." + }, + { + "left": "issue:43950", + "right": "issue:45987", + "accept": false, + "reason": "from_pretrained buffer restoration bug vs StaticCache return-type contract bug; unrelated." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading vs Jinja/template compilation error; different subsystems." + } + ] + }, + "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": 21, + "review_unit_index": 17, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36010", + "issue:39692", + "issue:42175", + "issue:43421", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43644", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43818", + "issue:43825", + "issue:43827", + "issue:43994", + "issue:44186", + "issue:44291", + "issue:44442", + "issue:44479", + "issue:44485", + "issue:44509", + "issue:44561", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44811", + "issue:44857", + "issue:44936", + "issue:44977", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45372", + "issue:45381", + "issue:45406", + "issue:45584", + "issue:45701", + "issue:45834", + "issue:45987", + "issue:46192", + "issue:46294", + "issue:46326", + "issue:46463" + ], + "soft_pairs": [ + "issue:43577|issue:43818", + "issue:43421|issue:44811", + "issue:39692|issue:45834", + "issue:43540|issue:45381", + "issue:45584|issue:45834", + "issue:44186|issue:44442", + "issue:43782|issue:44977", + "issue:43644|issue:43749", + "issue:43606|issue:45072", + "issue:44479|issue:45834", + "issue:43825|issue:43994", + "issue:46294|issue:46463", + "issue:34634|issue:36010", + "issue:44805|issue:44936", + "issue:28218|issue:29159", + "issue:44485|issue:44509", + "issue:43577|issue:43582", + "issue:44561|issue:45020", + "issue:45987|issue:46192", + "issue:43720|issue:45072", + "issue:44610|issue:45072", + "issue:44671|issue:45701", + "issue:43825|issue:43827", + "issue:43782|issue:45325", + "issue:43749|issue:44291", + "issue:45081|issue:45406", + "issue:42175|issue:45081", + "issue:45372|issue:46326", + "issue:43582|issue:44857" + ], + "prepared_review_unit_hash": "75328e0aaf8484a3e7736861c176582eaf177cfe4b7f97c1bf2bc571f77c4a3d", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36616, + "estimated_input_tokens": 9154, + "estimated_eval_tokens": 18564 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: most pairs are clearly unrelated bugs across different models, tokenizers, loaders, and docs. Only a couple of tokenizer/pipeline-related pairs look plausibly like duplicates or the same upstream change.", + "confidence": 0.76, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No single issue cleanly represents the whole cluster. If forced to pick a local representative, issue:43825 is the most concrete of the near-matches, but the cluster is too mixed to treat it as a true canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Different bugs and subsystems: BLIP2 dtype handling vs Video-LLaVA weight/temporal-attention issues." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Different tokenizer behaviors: runtime special-token/post-processor sync vs Whisper batch_decode skip_special_tokens." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "Docs/example problems for SigLIP2 vs a Kosmos2.5 OCR index error; not the same bug." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both are video-input issues, but they affect different Qwen models and likely different code paths." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Whisper empty-transcription failure after align_special_tokens is unrelated to Kosmos2.5 OCR indexing." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "LayoutLMv2 tokenizer NER/padding crash is unrelated to AutoTokenizer loading FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3VL weight_only loading error vs Qwen3.5 flash-attention generation bug; different failure modes." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Non-persistent buffer corruption and FSDP CPU RAM efficient loading are different loading-time bugs." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "CPU offload device mismatch in Bark differs from dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Qwen video-input regression is unrelated to Kosmos2.5 long-OCR index errors." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "Pipeline-task messaging/docs around v5 removals vs SigLIP2 example/model mismatch; not the same issue." + }, + { + "left": "issue:46294", + "right": "issue:46463", + "accept": false, + "reason": "Qwen3.5 video chat-template mismatch and T5 decoder validation messaging are unrelated." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "BarkProcessor voice_preset failure is unrelated to the GenerationMixin import error." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Mask/tensor shape mismatch is unrelated to trainer.evaluate() failing after train()." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": true, + "reason": "Both point to the same tokenizer inconsistency around added tokens / slow-vs-fast behavior, with the extra-space symptom fitting the broader mismatch." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE implementation discussion is unrelated to docs still mentioning removed pipeline tasks." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "BLIP2 dtype retention and Apple Silicon allocator warmup TypeError are unrelated." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "The former is a specific missing-helper regression; the latter is a broad umbrella for remote_code breakages, not the same concrete bug." + }, + { + "left": "issue:45987", + "right": "issue:46192", + "accept": false, + "reason": "StaticCache seq-length contract issue is unrelated to return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "BitNet packed-weight unpacking during accelerate loading is unrelated to dtype mismatches in other models." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "OmDet-Turbo processor/model input-size mismatch is unrelated to bfloat16 dtype mismatches." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "CamemBERT masked-LM regression is not the same as a general tokenization-version-change report." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": true, + "reason": "Both concern the v5 pipeline-removal fallout: one is the runtime error/help text, the other is stale docs still referencing removed pipeline tasks." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3VL weight_only loading error and Qwen2.5-VL rope-index scaling are different issues." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "FSDP CPU RAM loading failure and init_empty_weights _is_hf_initialized TypeError are distinct model-loading regressions." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Mistral tokenizer regex patch crash is unrelated to Gemma4Processor missing _tokenizer in transformers serve." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Missing TensorFlow in a torch install extra is unrelated to the Mistral tokenizer backend crash." + }, + { + "left": "issue:45372", + "right": "issue:46326", + "accept": false, + "reason": "Gemma4 processor import failure due to ReasoningEffort is unrelated to continue_final_message behavior for mistral_common tokenizers." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError is unrelated to LwDetrImageLoss AMP/CUDA crashes." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the cluster is indeed mostly heterogeneous, and the two accepted pairs are conservative matches based on the titles alone. I don\u2019t see any overstatement of evidence in the soft-edge verdicts." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 18, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:34634", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43638", + "issue:43645", + "issue:43650", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43746", + "issue:43819", + "issue:43854", + "issue:43906", + "issue:43927", + "issue:44060", + "issue:44368", + "issue:44387", + "issue:44554", + "issue:44589", + "issue:44610", + "issue:44617", + "issue:44811", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45003", + "issue:45020", + "issue:45042", + "issue:45084", + "issue:45459", + "issue:45538", + "issue:45600", + "issue:45803", + "issue:46209", + "issue:46285", + "issue:46421", + "issue:46424" + ], + "soft_pairs": [ + "issue:43906|issue:46209", + "issue:43262|issue:43819", + "issue:45003|issue:46285", + "issue:34634|issue:36331", + "issue:43526|issue:43697", + "issue:43653|issue:44912", + "issue:44060|issue:44368", + "issue:45084|issue:45538", + "issue:43742|issue:45020", + "issue:43526|issue:44554", + "issue:43329|issue:45600", + "issue:44589|issue:45042", + "issue:43452|issue:44991", + "issue:43638|issue:44936", + "issue:43650|issue:43906", + "issue:43746|issue:44811", + "issue:43550|issue:44912", + "issue:43819|issue:43927", + "issue:44387|issue:44617", + "issue:36010|issue:43645", + "issue:43257|issue:43329", + "issue:43720|issue:44610", + "issue:43854|issue:44977", + "issue:38175|issue:44610", + "issue:36010|issue:36331", + "issue:46421|issue:46424", + "issue:43531|issue:45042", + "issue:39692|issue:44912", + "issue:45459|issue:45803", + "issue:43262|issue:44811" + ], + "prepared_review_unit_hash": "6901391969a54bcc8855f9fe4bb95bca5b429c6ad772fe16e031d5ec9fad5e75", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36825, + "estimated_input_tokens": 9207, + "estimated_eval_tokens": 18670 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the soft pairs are mostly false positives spanning unrelated regressions in loading, audio, vision, training, cache, and quantization. I would not merge any of the suggested pairs. If one issue has to anchor triage, the broad remote_code regression issue is the closest umbrella, but it does not truly duplicate the rest.", + "confidence": 0.43, + "canonical_issue_reason": "issue:45020 is the broadest umbrella in the set: it describes a general regression where recent Transformers versions break remote_code models, which is the most plausible triage sink among otherwise unrelated bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the most representative high-level regression issue in this cluster; it is broad enough to serve as an anchor for triage, even though it is not a true duplicate of the other items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43906", + "right": "issue:46209", + "accept": false, + "reason": "Both touch FP8/model loading, but one is an isolated reproduction and the other is a B200 inference failure; different bugs and code paths." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and DAC latent/forward mismatch are unrelated defects." + }, + { + "left": "issue:45003", + "right": "issue:46285", + "accept": false, + "reason": "Unsafe sys.modules access is unrelated to the CLIP text-model structure inconsistency." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "Bark voice_preset and CustomTrainer compute_loss signature breakage are completely different areas." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "BEiT label reduction and RTDetrV2 output divergence are different model/processor bugs." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and MXFP4 quantization loading are unrelated." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both mention tied-weights warnings, but they involve different Qwen models and likely different root causes." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template-node compilation error and CLIPTokenizer max_length constant are unrelated." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "MobileLLM key error is a specific loading failure, while remote_code breakage is a broad version regression; not the same bug." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "BEiT label reduction and MPS attention correctness are unrelated." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Undefined video-token helper usage and stale auto_mappings config references are distinct issues." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Float8 storage lookup failure is unrelated to PIL backend processors requiring torchvision." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "gguf_file handling for tokenizer/model loading is different from the est-roberta tokenizer regression." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "DeepSpeed ZeRO3 index error during training and evaluate-after-train failure are different trainer issues." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "The placeholder issue title has no concrete technical overlap with the FP8 reproduction issue." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "PEFT local checkpoint loading and Whisper batch_decode skip_special_tokens handling are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "Bamba torch.compile+SDPA failure is unrelated to MXFP4 quantization fallback." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "DAC latent reconstruction mismatch and DiaConfig save/load token-ID loss are different defects." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both are OOM-related symptoms, but one is an int4 memory regression and the other is a Sam3Video-specific exhaustion issue." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "GenerationMixin import breakage and Jupyter custom-model initialization failure are different release regressions." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed and multimodal token counting video-branch bugs are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "BitNet packed-weight loading and OmDet-Turbo processor size mismatch are separate problems." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "GLM-4.7-Flash test loading failure and Qwen3.5 flash-attention generation bug do not share a concrete code path." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 and OmDet-Turbo processor resolution mismatch are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "GenerationMixin import regression and CustomTrainer compute_loss signature mismatch are unrelated." + }, + { + "left": "issue:46421", + "right": "issue:46424", + "accept": false, + "reason": "Both concern cache behavior, but one is about lazy initialization causing recompilation and the other about max_cache_length semantics; not the same bug." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue and torchvision dependency regression are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "SigLIP2 documentation/example errors and MXFP4 quantization loading are different classes of problems." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer protobuf error handling and Fuyu bare-except swallowing errors are separate exception-handling bugs." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio processor chat-template sampling-rate default and Whisper batch_decode skip_special_tokens are different processor behaviors." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 19, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:28282", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42490", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43329", + "issue:43450", + "issue:43452", + "issue:43454", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43643", + "issue:43696", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43824", + "issue:43927", + "issue:43931", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44568", + "issue:44756", + "issue:44821", + "issue:44849", + "issue:44912", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45488", + "issue:45538", + "issue:45701", + "issue:45759", + "issue:45859" + ], + "soft_pairs": [ + "issue:42490|issue:43257", + "issue:43257|issue:43696", + "issue:43329|issue:44479", + "issue:43454|issue:43577", + "issue:43643|issue:45759", + "issue:43329|issue:43450", + "issue:34567|issue:40990", + "issue:29159|issue:34567", + "issue:44568|issue:45356", + "issue:43606|issue:45859", + "issue:43749|issue:44756", + "issue:43257|issue:43931", + "issue:42915|issue:43742", + "issue:44291|issue:45042", + "issue:42915|issue:44821", + "issue:44987|issue:45538", + "issue:44849|issue:45701", + "issue:43697|issue:45357", + "issue:43927|issue:44912", + "issue:42915|issue:45538", + "issue:43452|issue:43582", + "issue:33357|issue:34567", + "issue:44568|issue:45488", + "issue:42175|issue:43824", + "issue:41720|issue:43540", + "issue:44062|issue:44991", + "issue:43317|issue:45538", + "issue:44479|issue:45381", + "issue:28282|issue:34689" + ], + "prepared_review_unit_hash": "020b54f38b82471e73a85432321c971106df420d75c2e9d8aefd671803c1befd", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35929, + "estimated_input_tokens": 8983, + "estimated_eval_tokens": 18222 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issue reports; only one soft pair looks like the same video-input regression in the Qwen VL family. Everything else is too model-/path-specific to be treated as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "issue:44479 is the best umbrella issue: it describes the broader v5.3.0 video-input regression across multiple Qwen VL/MoE variants, and issue:45381 appears to be a narrower instance of that same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 is the most representative issue because it captures the shared regression across several affected models rather than one model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "Different areas: local-path import warning vs Qwen3 MOE weight conversion." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Qwen3 MOE loading/conversion bug vs GPT-oss OOM; different failure mode." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Both are video-related, but one is undefined helpers in token counting and the other is a broader video-input regression." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "AyaVision weight tying vs BLIP2 dtype casting; unrelated code paths." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "AutoConfig missing fields with trust_remote_code is not the same as AutoModelForCausalLM.from_config not unwrapping text_config." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Undefined video-branch helpers vs batched video processor shape bug; different bugs." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "Trainer token counting metric issue vs high perplexity on a model; no shared root cause." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency vs TrainerState token counter not updating." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both tokenizer regressions, but for different models and different tokenizer behaviors." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "Both involve CPU offload, but different models and likely different failing code paths." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading vs mmap OOM on Strix Halo; different problems." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3 MOE weight conversion vs Qwen3-VL shape mismatch; same family only at a high level." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Qwen3 MOE FP8 config failure vs MobileLLM key error; unrelated." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "init_empty_weights argument handling vs image processor torchvision requirement; unrelated." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3 MOE FP8 loading vs AutoImageProcessor URL loading; no shared bug." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Model-loading regression for one repo vs CLIPTokenizer max-length bug; unrelated." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 hidden-states bug vs a broad tokenization-version complaint; not the same concrete defect." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "RTDetrV2 output differences vs Qwen3.5 save_pretrained key regression; different subsystems." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "DiaConfig token ID persistence vs GPT-oss MXFP4 load fallback; unrelated." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Qwen3 MOE FP8 loading vs CLIPTokenizer model_max_length; not the same issue." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf_file loading bug vs Apple Silicon allocator TypeError; different paths." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "MacOS bus error on CLIP model vs TrainerState token counter update; unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Tokenizer regressions, but different root causes and model families; not mergeable into one fix." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Packaging/backend dependency issue vs missing Qwen2.5-VL symbol import; unrelated." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Auto device-mapping cuda assert vs video processing ValueError in Qwen3OmniMoe; different failures." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "AddedToken keyword duplication vs tokenizer loading for a specific model; different bugs." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Dequantized model offload failure vs tokenizer max-length constant bug; unrelated." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "45381 is a specific report of the same Qwen2.5-VL video-input regression described broadly in 44479; same code path and likely same fix." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "PyTorch missing import error vs Llama 3.2 Vision model-loading breakage; unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: issue:44479 is clearly the broader v5.3.0 video-input regression report, and issue:45381 reads like a narrower report of the same Qwen VL video-position-id bug. The other soft-edge rejections are conservative based on the issue titles, with no obvious overclaiming." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 20, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:24643", + "issue:27957", + "issue:28282", + "issue:29127", + "issue:29159", + "issue:34567", + "issue:36331", + "issue:38175", + "issue:42175", + "issue:42915", + "issue:43257", + "issue:43404", + "issue:43493", + "issue:43526", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43906", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44560", + "issue:44737", + "issue:44792", + "issue:44821", + "issue:44898", + "issue:44964", + "issue:44987", + "issue:45072", + "issue:45081", + "issue:45137", + "issue:45325", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45588", + "issue:45823", + "issue:45907", + "issue:45987", + "issue:46399", + "issue:46463", + "issue:46468" + ], + "soft_pairs": [ + "issue:45137|issue:45405", + "issue:43540|issue:44479", + "issue:43540|issue:44560", + "issue:44987|issue:45907", + "issue:24643|issue:29159", + "issue:43742|issue:44291", + "issue:28282|issue:29127", + "issue:42915|issue:43749", + "issue:38175|issue:44898", + "issue:44737|issue:46463", + "issue:42175|issue:44488", + "issue:43577|issue:45072", + "issue:43493|issue:43994", + "issue:43906|issue:46468", + "issue:43257|issue:43756", + "issue:44291|issue:44821", + "issue:43526|issue:45479", + "issue:34567|issue:36331", + "issue:44062|issue:45081", + "issue:43582|issue:43782", + "issue:27957|issue:28282", + "issue:43404|issue:45987", + "issue:44737|issue:45588", + "issue:44792|issue:44964", + "issue:43756|issue:46399", + "issue:43906|issue:45381", + "issue:43756|issue:45325", + "issue:45823|issue:46463" + ], + "prepared_review_unit_hash": "1882a07dfdbed8eba0c9b543a42bd3206753d28100034f63f5e50ebd2bce3722", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36319, + "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 mostly a mix of unrelated Transformers bug reports across different models, features, and code paths. The only plausible duplicate pair is the two SigLIP2 reports; the rest are too model-specific or symptom-only to safely merge.", + "confidence": 0.84, + "canonical_issue_reason": "No single canonical fits the whole cluster cleanly; if one must be chosen, issue 43493 is the best root-cause report because it describes the underlying SigLIP2 implementation discrepancy more directly than the user-facing symptom in 43994.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43493 is the strongest standalone issue here: concrete, technically specific, and closest to the shared underlying defect in the only plausible duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "Different problems: DeepSpeed ZeRO3 deque failure vs an unrelated PEFT version bump." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Both mention video inputs, but they target different Qwen model families and the titles point to separate regressions." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Same general area, but different model families and different failure modes (ValueError vs StopIteration)." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "One is a model load failure, the other is a repo-listing API regression; not the same underlying bug." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "DeepSpeed training error and tokenizer inconsistency are unrelated code paths." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Both are loading-time errors, but one is a key lookup for a specific model and the other is an init_empty_weights compatibility issue." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "ImportError about missing PyTorch is unrelated to a LayoutLMv3 error-message clarity issue." + }, + { + "left": "issue:42915", + "right": "issue:43749", + "accept": false, + "reason": "Different subsystems: FineGrainedFP8Config on Qwen3Moe vs FSDP CPU RAM efficient loading." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different models and different symptoms; no clear shared code-path bug." + }, + { + "left": "issue:44737", + "right": "issue:46463", + "accept": false, + "reason": "XLNet device placement bug and T5 decoder-input validation messaging are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Packaging/backend selection issue vs a model-loading failure; not the same bug." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both involve dtype, but they affect different models and likely different conversion paths." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "These both point to the same SigLIP2 defect: HF output quality/behavior diverging from the expected JAX behavior for the same model." + }, + { + "left": "issue:43906", + "right": "issue:46468", + "accept": false, + "reason": "One is a reproduction of an older issue; the other is a broken example. No evidence they share the same root cause." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to Smollm3 RoPE layer dropping." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "init_empty_weights compatibility and loading AutoImageProcessor from URL are different failure modes." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "A fast image-processor label reduction bug is not the same as the sequence-classification zero-loss issue." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting and CustomTrainer compute_loss signature mismatch are separate trainer issues." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "AddedToken keyword collision and Mistral regex patching crash are unrelated tokenizer bugs." + }, + { + "left": "issue:43582", + "right": "issue:43782", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and Qwen3VL weight_only loading error are different code paths." + }, + { + "left": "issue:27957", + "right": "issue:28282", + "accept": false, + "reason": "XLM-RoBERTa Flash Attention 2 and missing PyTorch ImportError are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45987", + "accept": false, + "reason": "A tied-weights bug in Mistral3 and a StaticCache return-type bug are not the same underlying issue." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU tensor allocation and flash_attention sink-less AttributeError are unrelated." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Janus test failure and Phi-4 multimodal loading failure are different models and different bugs." + }, + { + "left": "issue:43756", + "right": "issue:46399", + "accept": false, + "reason": "RoPE-layer dropping in Smollm3 and missing apply_rotary_pos_emb decorator in Qwen3.5 are unrelated." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "A repro of an older issue and a qwen2.5-vl video-position-id bug are not the same bug." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen2.5-VL position/rope problems; not enough evidence they are the same fix." + }, + { + "left": "issue:45823", + "right": "issue:46463", + "accept": false, + "reason": "Gemma4 device mismatch and T5 decoder validation messaging are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is well grounded in the packet, and the lone accepted soft pair (43493/43994) is a conservative, defensible duplicate judgment. The other pairwise rejections are also appropriately cautious and do not overstate evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 21, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:41762", + "issue:43295", + "issue:43452", + "issue:43531", + "issue:43643", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43819", + "issue:44361", + "issue:44442", + "issue:44466", + "issue:44488", + "issue:44493", + "issue:44514", + "issue:44589", + "issue:44625", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44936", + "issue:45005", + "issue:45042", + "issue:45081", + "issue:45200", + "issue:45216", + "issue:45356", + "issue:45357", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45600", + "issue:45920", + "issue:46050", + "issue:46121", + "issue:46463" + ], + "soft_pairs": [ + "issue:45042|issue:45081", + "issue:43742|issue:44589", + "issue:43452|issue:44488", + "issue:45357|issue:45920", + "issue:43531|issue:44821", + "issue:45405|issue:45600", + "issue:44913|issue:46121", + "issue:43720|issue:44898", + "issue:43643|issue:45375", + "issue:43697|issue:45216", + "issue:45200|issue:46463", + "issue:24643|issue:29127", + "issue:44625|issue:44877", + "issue:41762|issue:44936", + "issue:44493|issue:46050", + "issue:44514|issue:45381", + "issue:44466|issue:45005", + "issue:43452|issue:44843", + "issue:43819|issue:44811", + "issue:30064|issue:33666", + "issue:43688|issue:43749", + "issue:28282|issue:32090", + "issue:28282|issue:33357", + "issue:44361|issue:44442", + "issue:43295|issue:45356" + ], + "prepared_review_unit_hash": "2ae273d47742d6eeada63284c0e19ea7fd7e8ec75b628116495fa96398210539", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 25, + "serialized_chars": 35423, + "estimated_input_tokens": 8856, + "estimated_eval_tokens": 17968 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly unrelated issues across different models and code paths. I only see one plausible duplicate pair: the v5 tied-weights regression reports (44466/45005).", + "confidence": 0.74, + "canonical_issue_reason": "issue:44466 is the most concrete tied-weights report: it names the exact lm_head.weight serialization symptom, the v5 regression context, and the device-dependent behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44466 is the best representative issue because it is the most specific, actionable description of the tied-weights bug and has the clearest regression framing.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Different subsystems and failures: image processor torchvision dependency vs Mistral regex tokenizer crash." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "Both are loading errors, but they describe different concrete failures and likely different root causes." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "One is a gguf_file/from_pretrained regression; the other is a specific model-loading failure. No clear same bug." + }, + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Qwen3.5 save_pretrained key serialization bug vs broad AutoTokenizer wrong-token-IDs regression; different code paths." + }, + { + "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-pin issue and removed-config CI breakage are unrelated maintenance problems." + }, + { + "left": "issue:44913", + "right": "issue:46121", + "accept": false, + "reason": "Rotary_pct reload defaulting and rope-validation TypeError are different config bugs." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "BitNet packed-weight loading bug vs Perceiver interpolation failure; different models and mechanisms." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both mention config fields, but one is generic trust_remote_code missing fields and the other is a specific strict-decorated config field drop." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "RTDetr output mismatch and Qwen3.5 save_pretrained corruption are distinct regressions." + }, + { + "left": "issue:45200", + "right": "issue:46463", + "accept": false, + "reason": "Gemma 4 mm_token_type_ids handling and T5 decoder-input validation messaging are unrelated." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "DeepSpeed weight shape RuntimeError vs LayoutLMv3 error-message clarity; different issues entirely." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and granite_speech strict config loading are different config-loading bugs." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "ZeRO-3 model-load index error and trainer.evaluate-after-train failure do not share the same code path." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Unexpected position-id keys and quantized save/load breakage are different symptoms." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "One is a batched apply_chat_template crash; the other is wrong vision_position_ids in video input. Different bugs." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": true, + "reason": "Both describe v5 tied-weights regressions in translation-style models and plausibly stem from the same save/load serialization path." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "gguf_file/from_pretrained breakage and unconditional model_info() in offline mode are distinct tokenizer-loading bugs." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "DAC latent reconstruction mismatch and Whisper batch_decode skip_special_tokens handling are unrelated." + }, + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Void segmentation-map processing and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Auxiliary-loss normalization and FSDP CPU-efficient loading are different training/loading issues." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "PyTorch import detection failure and NoneType GPU broadcast error are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Missing PyTorch import and MacOS bus error on CLIP are unrelated." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer classes and failures; no shared underlying bug is evident." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Processor/tokenizer API regression and Kimi-K2.5 codec handling regression are different bugs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analysis is grounded in the packet and stays conservative. The only accepted soft pair, 44466/45005, is plausibly the same tied-weights serialization regression; the rest are reasonably treated as unrelated based on their titles and described failure modes. The summary does not overstate the evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 22, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:24643", + "issue:27957", + "issue:29159", + "issue:32090", + "issue:33357", + "issue:34567", + "issue:38175", + "issue:41720", + "issue:43295", + "issue:43452", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43643", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43927", + "issue:44361", + "issue:44464", + "issue:44479", + "issue:44484", + "issue:44514", + "issue:44561", + "issue:44568", + "issue:44625", + "issue:44704", + "issue:44743", + "issue:44843", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45290", + "issue:45372", + "issue:45440", + "issue:45491", + "issue:45542", + "issue:45701", + "issue:45704", + "issue:45823", + "issue:46002", + "issue:46170", + "issue:46178", + "issue:46402" + ], + "soft_pairs": [ + "issue:45372|issue:46170", + "issue:43653|issue:44361", + "issue:44936|issue:45823", + "issue:38175|issue:43295", + "issue:44484|issue:44704", + "issue:43452|issue:43577", + "issue:45542|issue:46178", + "issue:24643|issue:27957", + "issue:29159|issue:33357", + "issue:43577|issue:43697", + "issue:43653|issue:44568", + "issue:44843|issue:44991", + "issue:44464|issue:44743", + "issue:43646|issue:43749", + "issue:43531|issue:43742", + "issue:44514|issue:45290", + "issue:43688|issue:45440", + "issue:43653|issue:43927", + "issue:43688|issue:45704", + "issue:45491|issue:46002", + "issue:43756|issue:46402", + "issue:43550|issue:43927", + "issue:43688|issue:45701", + "issue:43550|issue:44361", + "issue:44479|issue:44625", + "issue:43643|issue:44561", + "issue:44464|issue:44977", + "issue:32090|issue:34567", + "issue:41720|issue:44464" + ], + "prepared_review_unit_hash": "ccdfd091da38a4874a1c748820d9c2297e5a1ce13eb1e275288bcd97025054df", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36432, + "estimated_input_tokens": 9108, + "estimated_eval_tokens": 18472 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: it mixes unrelated tokenizer, trainer, multimodal, MoE, config, and loading regressions. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": "If a representative issue is required, #43646 is the broadest v5-regression report in the set; otherwise there is no true canonical duplicate for this cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "#43646 is the closest umbrella representative because it describes a broad Transformers 5.0 regression affecting custom model initialization, which is more general than the other very specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Different failure domains: Gemma processor loading/import dependency vs Windows ROCm torch.distributed import/type error." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Both are tokenizer bugs, but they affect different tokenizers and different concrete failures." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Trainer lifecycle failure vs Gemma4 device-map forward mismatch; unrelated code paths." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Different models and symptoms: zero probabilities in SigLIP2 vs processor.tokenizer/image-handling regression." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "A save_pretrained shard-size question is unrelated to AutoProcessor kwargs forwarding." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "GGUF loading breakage is not the same bug as Blip2 dtype remaining float32." + }, + { + "left": "issue:45542", + "right": "issue:46178", + "accept": false, + "reason": "Both mention backend/TF issues, but they are different backend resolution failures with different triggers." + }, + { + "left": "issue:24643", + "right": "issue:27957", + "accept": false, + "reason": "DeepSpeed weight-shape runtime error is unrelated to XLMRoberta Flash Attention 2." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer inconsistency and MacOS bus error on CLIP are different bugs." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Model dtype loading bug vs RTDetrV2 output divergence are distinct issues." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Different tokenizer special-token behaviors on different models; not the same bug." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Offline model_info call in tokenizer patch is unrelated to est-roberta tokenizer loading failure." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Both involve generation state, but one is compiled-forward inconsistency and the other is cache reset logic in a different model." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Custom model initialization regression is unrelated to FSDP CPU RAM efficient loading." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is not the same as a key error when loading MobileLLM-125M." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Same API surface, but different crashes: batched padding=False vs assistant tool-call messages with no content." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Auxiliary-loss normalization in MoE models is different from DeepSeekV3MoE implementation divergence." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "Tokenizer special-token registration bug and DiaConfig custom token-ID loss are unrelated." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "MoE auxiliary-loss normalization is unrelated to a T5 apex RMSNorm memory leak." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Both mention NaNs, but the model, locus, and root cause are different." + }, + { + "left": "issue:43756", + "right": "issue:46402", + "accept": false, + "reason": "RoPE-layer dropping in Smollm3 is unrelated to Sam3 text encoder weights not loading." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "torch.compile/SDPA failure is not the same bug as custom token IDs being lost on save/load." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Auxiliary-loss normalization is unrelated to broad tokenization changes." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Compiler/SDPA model failure and tokenizer AttributeError are different code paths." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression across Qwen variants is distinct from num_labels propagation in Qwen3.5 config." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Missing fields from trust_remote_code AutoConfig is different from the removed is_torch_fx_available regression." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency is not the same as flash-attention generation failure." + }, + { + "left": "issue:32090", + "right": "issue:34567", + "accept": false, + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are unrelated trainer bugs." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Auto device mapping cudaErrorAssert on A800 is a different failure than chunked generation inconsistency under compilation." + } + ] + }, + "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": 21, + "review_unit_index": 23, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:28218", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43688", + "issue:43720", + "issue:43742", + "issue:43825", + "issue:43844", + "issue:43927", + "issue:43994", + "issue:44361", + "issue:44451", + "issue:44466", + "issue:44488", + "issue:44554", + "issue:44568", + "issue:44610", + "issue:44625", + "issue:44821", + "issue:44843", + "issue:44849", + "issue:44964", + "issue:44977", + "issue:45357", + "issue:45405", + "issue:45588", + "issue:45812", + "issue:45823", + "issue:46002", + "issue:46050", + "issue:46192", + "issue:46209", + "issue:46365" + ], + "soft_pairs": [ + "issue:44977|issue:46209", + "issue:43504|issue:43720", + "issue:43504|issue:43550", + "issue:32090|issue:33666", + "issue:43504|issue:44610", + "issue:33666|issue:36010", + "issue:43577|issue:44554", + "issue:44466|issue:46050", + "issue:43742|issue:44821", + "issue:43577|issue:45588", + "issue:43720|issue:43927", + "issue:43262|issue:43295", + "issue:44554|issue:46365", + "issue:43452|issue:44451", + "issue:43606|issue:43927", + "issue:45823|issue:46050", + "issue:44849|issue:46209", + "issue:33666|issue:36331", + "issue:43994|issue:46192", + "issue:43278|issue:43688", + "issue:43720|issue:44361", + "issue:43525|issue:43572", + "issue:44625|issue:46002", + "issue:28218|issue:33357", + "issue:43493|issue:43825", + "issue:44964|issue:45405", + "issue:39692|issue:43493", + "issue:44849|issue:45357", + "issue:44568|issue:45812", + "issue:44488|issue:44843", + "issue:39692|issue:43295", + "issue:43493|issue:43844", + "issue:43278|issue:43844" + ], + "prepared_review_unit_hash": "8577d56ef3334bb1a421eae8a43c9d3e307adc7a6ca6ef6331357cdbf447c423", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37728, + "estimated_input_tokens": 9432, + "estimated_eval_tokens": 19120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft-edge pair looks like a true duplicate; most are unrelated bugs across different models, processors, or runtime paths. If a single issue must anchor the cluster, `issue:45812` is the strongest standalone regression, but the set is not one coherent duplicate group.", + "confidence": 0.87, + "canonical_issue_reason": "`issue:45812` is the clearest high-impact core regression here: broad `AutoTokenizer` token-id breakage across Granite models with multiple inbound references.", + "canonical_pr_reason": null, + "best_issue_reason": "`issue:45812` is the best representative issue because it is concrete, user-facing, and broader in impact than the other standalone reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44977", + "right": "issue:46209", + "accept": false, + "reason": "Different bugs: Qwen3.5 flash-attention generation vs FP8 inference failure on B200." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Different code paths: legacy-field model loading vs BitNet packed-weight unpacking during accelerate load." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Unrelated failures: BEiT preset loading vs Bamba torch.compile SDPA issue." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Both involve training, but the error types and affected models are different." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Legacy config-field load bug is unrelated to OmDet-Turbo processor image-size mismatch." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Different failures: multi-GPU training vs import-time `GenerationMixin` regression." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Different bugs: BLIP2 dtype casting vs MPS attention correctness when head dims differ." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "Both are save/load related, but one is tied-weight serialization and the other is quantization persistence." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "Loading a causal LM vs loading an image processor from URL are separate problems." + }, + { + "left": "issue:43577", + "right": "issue:45588", + "accept": false, + "reason": "BLIP2 dtype mismatch is unrelated to flash-attention crashing on `s_aux=None`." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "Packed-weight loading bug and DiaConfig token-id persistence bug are different underlying issues." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and processor/tokenizer API regression are distinct." + }, + { + "left": "issue:44554", + "right": "issue:46365", + "accept": false, + "reason": "MPS attention correctness bug is unrelated to `Trainer` import failing with datasets installed." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "Generic `from_pretrained`/gguf breakage is not the same as the ScandiBERT load failure." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to config token-id loss after save/load." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Different issues: Gemma4 device placement during forward vs sinq quantization persistence." + }, + { + "left": "issue:44849", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 hidden-states bug is separate from FP8 inference on B200." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Distributed training setup issue and custom-trainer signature mismatch are not the same bug." + }, + { + "left": "issue:43994", + "right": "issue:46192", + "accept": false, + "reason": "SigLIP2 correctness regression and `return_tensors` handling bug are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Embedding dtype drift in eval is not the same as auxiliary-loss normalization." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet packed-weight loading and MLukeTokenizer task AttributeError are different failures." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Similar theme, but different configs and attributes (`pad_token_id` vs `pad_token_idx`) and no shared concrete code path." + }, + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "`num_labels` propagation bug is distinct from NaNs in classification heads." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer added-space bug is unrelated to the MacOS CLIP bus error." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and pipeline error-message regression are different issues." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model loading failure and unreleased dependency version bump are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Both mention SigLIP2, but one is a doc example error set and the other is an implementation discrepancy." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Hidden-states bug and incorrect visual-encoder key serialization are separate regressions." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Different tokenizer regressions affecting different models and code paths." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Model load failure for a specific model is not the same as offline-mode breakage in `_patch_mistral_regex`." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Evaluation dtype mismatch is unrelated to ZeRO-3 gradient growth." + } + ] + }, + "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": 21, + "review_unit_index": 24, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:28218", + "issue:29942", + "issue:30333", + "issue:33357", + "issue:33666", + "issue:34567", + "issue:43065", + "issue:43257", + "issue:43278", + "issue:43299", + "issue:43334", + "issue:43504", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44112", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44464", + "issue:44466", + "issue:44493", + "issue:44589", + "issue:44734", + "issue:44792", + "issue:44964", + "issue:44977", + "issue:45200", + "issue:45229", + "issue:45405", + "issue:45520", + "issue:45584", + "issue:45588", + "issue:45759", + "issue:45823", + "issue:46050", + "issue:46285", + "issue:46369", + "issue:46424" + ], + "soft_pairs": [ + "issue:45200|issue:45229", + "issue:43540|issue:44977", + "issue:43646|issue:44466", + "issue:44464|issue:46424", + "issue:33357|issue:33666", + "issue:43334|issue:43572", + "issue:43824|issue:44589", + "issue:43257|issue:43299", + "issue:44589|issue:45520", + "issue:43927|issue:44361", + "issue:43504|issue:43606", + "issue:43334|issue:45759", + "issue:43278|issue:43701", + "issue:45229|issue:45823", + "issue:43575|issue:45405", + "issue:43572|issue:44291", + "issue:44792|issue:45405", + "issue:44493|issue:46285", + "issue:43504|issue:43653", + "issue:43606|issue:44112", + "issue:44464|issue:45584", + "issue:43994|issue:44493", + "issue:43873|issue:46050", + "issue:45588|issue:46050", + "issue:43065|issue:46369", + "issue:29942|issue:30333", + "issue:44423|issue:44734", + "issue:28218|issue:34567", + "issue:43844|issue:44964" + ], + "prepared_review_unit_hash": "3504df3302ee0c284cd2e8d18db9432d0df88cc5328e48eb90f8c62e714bef12", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36341, + "estimated_input_tokens": 9086, + "estimated_eval_tokens": 18428 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most candidate pairs are clearly unrelated and should not be merged. The only plausible duplicate pair is the Qwen3 multimodal loading/config bug pair (43334 and 45759), where the second issue describes the underlying config-unwrapping problem more directly.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45759 is the best canonical issue for the one genuine duplicate pair because it states the root cause (`from_config` not unwrapping `text_config`) more directly than the symptom report in issue:43334.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45759 is the strongest representative overall: it is specific about the failing code path and broad enough to cover the related Qwen3-VL loading symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Different bugs: defaulting mm_token_type_ids in text-only fine-tuning vs multi-GPU inference CUDA OOM." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Different failure modes and code paths: video input processing ValueError vs flash-attention generation behavior." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Unrelated: custom model initialization breakage vs inconsistent lm_head.weight serialization depending on device." + }, + { + "left": "issue:44464", + "right": "issue:46424", + "accept": false, + "reason": "Different cache issues: compiled forward chunked generation vs max_cache_length only honored for quantized cache." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "Different models and subsystems: CLIP bus error on MacOS vs Qwen2-VL multi-GPU training." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Both are config-related, but they affect different models and different missing fields; not the same bug." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Different errors and causes: missing import for a model class vs missing Float8 storage type." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Same family name, but different concrete bugs and loading paths: accelerate+deepspeed weight conversion vs Qwen3VL MoE load breakage." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Different runtime/environment issues: Float8 storage lookup vs flash_attn import key error on Python 3.13." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "Different token/config problems affecting different classes and symptoms; not the same underlying defect." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Different areas entirely: BEiT legacy field loading vs Bark CPU offload device mismatch." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": true, + "reason": "Same underlying Qwen3 multimodal loading bug: 43334 is the symptom, 45759 identifies the config-unwrapping root cause in the same code path." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Different issues: dtype drift between train/eval vs resume_from_checkpoint key mismatch." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Both involve Gemma4 and devices, but one is CUDA OOM and the other is a device mismatch; not the same bug." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "Completely unrelated: tensor-parallel OOM for a model load vs a dependency version bump." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Different compatibility problems: missing pad_token_idx vs unexpected _is_hf_initialized argument." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Different topics: a Janus image-generation test failure vs a PEFT minimum-version bump." + }, + { + "left": "issue:44493", + "right": "issue:46285", + "accept": false, + "reason": "Not the same bug: unexpected position-id keys across many models vs CLIP text model structure change." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Different models and defects: BEiT legacy field loading vs BigBirdTokenizer special-token registration." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Different issues: real CPU-offload device mismatch vs stale GraniteSpeech CI test." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Different cache/generation bugs with different triggers and symptoms." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "Different problems: SigLIP2 output quality issue vs position-id key regressions." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Both mention memory/offloading/quantization concerns, but the concrete failures are different and not mergeable as one fix." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "Different bugs: flash_attention sink-less model crash vs sinq save/load corruption." + }, + { + "left": "issue:43065", + "right": "issue:46369", + "accept": false, + "reason": "Different Sam3 topics: dummy Conv2d implementation vs memory overflow in sam3-litetext." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Completely unrelated CI and MLflow reporting failures." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Both are transformers serve crashes, but they hit different endpoints and have different root causes." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer spacing bug vs TrainerState token-count update bug; no shared underlying defect." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "Different training/loader issues: ZeRO-3 gradient growth vs latest-transformers model loading failure." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The only accepted duplicate pair, 43334/45759, is plausibly the same Qwen3 multimodal loading/config bug, and the rejected pairs are explained at a level that matches the visible titles without overclaiming." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 25, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:15354", + "issue:29159", + "issue:32090", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:42490", + "issue:43232", + "issue:43295", + "issue:43450", + "issue:43493", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43582", + "issue:43696", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43950", + "issue:43957", + "issue:44186", + "issue:44466", + "issue:44521", + "issue:44554", + "issue:44610", + "issue:44849", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45084", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45356", + "issue:45381", + "issue:45406", + "issue:45446", + "issue:45520", + "issue:45823", + "issue:46121", + "issue:46178", + "issue:46192", + "issue:46209", + "issue:46285" + ], + "soft_pairs": [ + "issue:42490|issue:43696", + "issue:29159|issue:32090", + "issue:45406|issue:46121", + "issue:45084|issue:45520", + "issue:44554|issue:45446", + "issue:43582|issue:44977", + "issue:15354|issue:38175", + "issue:43540|issue:45200", + "issue:45072|issue:45198", + "issue:45200|issue:45823", + "issue:43493|issue:43701", + "issue:44521|issue:45290", + "issue:37428|issue:43957", + "issue:43232|issue:44849", + "issue:44186|issue:46192", + "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:43572|issue:43957", + "issue:43550|issue:43844", + "issue:44466|issue:45127", + "issue:37428|issue:43824", + "issue:43950|issue:46285" + ], + "prepared_review_unit_hash": "fe0ef5677431d33bbb91e03702472eb6eba3a1631232d5a0a30f78d58e11be20", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36117, + "estimated_input_tokens": 9030, + "estimated_eval_tokens": 18316 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are not duplicates of one another; they span unrelated bugs across tokenizers, generation, model loading, video processing, quantization, and backend/import errors. The soft-similarity matches are mostly keyword overlap rather than the same underlying defect.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue fits this set: the issues describe different code paths and failure modes, so selecting one would misrepresent the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "No best issue candidate is suitable as a cluster representative because the set is heterogeneous and lacks a shared underlying bug/change.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Local-path import warning vs CUDA out-of-memory on GPT-oss; unrelated failures." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer behavior mismatch vs Trainer GPU broadcast NoneType error; different subsystems." + }, + { + "left": "issue:45406", + "right": "issue:46121", + "accept": false, + "reason": "Gemma4Processor `_tokenizer` attribute crash vs rope-validation list type error; unrelated." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Template-node compile error vs flash_attn import backend KeyError; different issues." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness bug vs PyTorch version check bug in flex_attention; not the same bug." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "Apple Silicon warmup TypeError vs Qwen3.5 flash-attention generation issue; unrelated." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp export problem vs unexpected zero probabilities in SigLIP2; unrelated." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video input processing ValueError vs Gemma4 multimodal token-type defaulting; different code paths." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "bfloat16 dtype mismatch in inference vs Wav2Vec2 save/tokenization failure; unrelated." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Missing mm_token_type_ids default vs PLE device mismatch in forward; not the same defect." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 implementation discrepancy vs resume_from_checkpoint key mismatch; unrelated." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both involve chat templates, but one is zero assistant masks for multimodal inputs and the other is tool-call/no-content crash; different bugs." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "Flash-attention import symbol error vs meta-device loading regression; unrelated loading issues." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs update after sync_gpus vs Qwen3.5 output_hidden_states bug; different generation paths." + }, + { + "left": "issue:44186", + "right": "issue:46192", + "accept": false, + "reason": "LayoutLMv2Tokenizer crash on NER/padding vs return_tensors being ignored when text_kwargs is passed; different tokenizer bugs." + }, + { + "left": "issue:45823", + "right": "issue:46192", + "accept": false, + "reason": "Gemma4 device mismatch in forward vs return_tensors argument handling; unrelated." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor.tokenizer regression in custom code vs batched video processor shape bug; different models and failures." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Processor/tokenizer regression vs OmDet-Turbo processor size mismatch; unrelated." + }, + { + "left": "issue:45381", + "right": "issue:46209", + "accept": false, + "reason": "Qwen2.5-VL video position ids bug vs FP8 inference failure on B200; unrelated." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor/tokenizer regression vs quantization offloading behavior; not the same issue." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness issue vs Kimi-K2.5 tokenizer regression; unrelated." + }, + { + "left": "issue:39692", + "right": "issue:46192", + "accept": false, + "reason": "SigLIP2 documentation/example errors vs return_tensors being ignored; docs and API behavior mismatch, not the same bug." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig vs model loading failure for physical-intelligence/fast; unrelated." + }, + { + "left": "issue:45520", + "right": "issue:46178", + "accept": false, + "reason": "flash_attn import KeyError on Python 3.13 vs tensorflow_text backend mapping error; different import paths." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression vs meta-device model loading breakage; unrelated." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "Bamba torch.compile SDPA bug vs gradient inflation under ZeRO-3; different training/runtime problems." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Serialization of tied lm_head weights vs LoRA merge collapse with extended vocab; superficially about weights, but different failure modes." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Different import failures: flash_attention_utils symbol missing vs Qwen2.5-VL class not exported." + }, + { + "left": "issue:43950", + "right": "issue:46285", + "accept": false, + "reason": "Non-persistent buffer corruption during from_pretrained vs CLIP text model structure inconsistency; unrelated 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": 21, + "review_unit_index": 26, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:28218", + "issue:35141", + "issue:38175", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43334", + "issue:43388", + "issue:43450", + "issue:43493", + "issue:43525", + "issue:43526", + "issue:43575", + "issue:43653", + "issue:43819", + "issue:43824", + "issue:43827", + "issue:43856", + "issue:43901", + "issue:43975", + "issue:44112", + "issue:44206", + "issue:44442", + "issue:44514", + "issue:44521", + "issue:44561", + "issue:44610", + "issue:44792", + "issue:44857", + "issue:44898", + "issue:45083", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45325", + "issue:45412", + "issue:45520", + "issue:45563", + "issue:45584", + "issue:45759", + "issue:46170", + "issue:46192", + "issue:46209", + "issue:46234", + "issue:46285", + "issue:46326" + ], + "soft_pairs": [ + "issue:44898|issue:45412", + "issue:43450|issue:43575", + "issue:45200|issue:45325", + "issue:42490|issue:43493", + "issue:43334|issue:43525", + "issue:43388|issue:43526", + "issue:42915|issue:43824", + "issue:44792|issue:45563", + "issue:43824|issue:43975", + "issue:38175|issue:43493", + "issue:28218|issue:35141", + "issue:41720|issue:46209", + "issue:43278|issue:43856", + "issue:43526|issue:46192", + "issue:43653|issue:44112", + "issue:44442|issue:45198", + "issue:43262|issue:43450", + "issue:38175|issue:45412", + "issue:45520|issue:46170", + "issue:45290|issue:46326", + "issue:45759|issue:46285", + "issue:44610|issue:44857", + "issue:44561|issue:46234", + "issue:43819|issue:44112", + "issue:44514|issue:44521", + "issue:44206|issue:45584", + "issue:43827|issue:43901", + "issue:43819|issue:45083" + ], + "prepared_review_unit_hash": "b6660799a45ced439a88a72f5f1783221efa2c7c1a1b4744e9310611e4b9f7b3", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36300, + "estimated_input_tokens": 9075, + "estimated_eval_tokens": 18406 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All candidate pairs look like near-topic neighbors rather than true duplicates; they share broad subsystems (tokenizers, multimodal chat templates, model loading, docs) but not the same concrete bug or fix path. I would reject every soft edge. For a representative issue, `issue:44521` is the strongest candidate because it is active, specific, and has the most evidence around a concrete multimodal `apply_chat_template` failure.", + "confidence": 0.72, + "canonical_issue_reason": "`issue:44521` is the most representative active bug report in the set: it has a concrete repro, multiple inbound refs, and a focused multimodal `apply_chat_template` failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "`issue:44521` is the best standalone issue to keep as the primary reference because it is open, well-scoped, and clearly actionable.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Different bugs: Perceiver resolution handling vs RT-DETR model memory release; no shared code path." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video processor batch-shape bug vs Qwen2 load OOM; unrelated symptoms and fixes." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Gemma 4 `mm_token_type_ids` defaulting vs Qwen2.5-VL rope indexing; both multimodal, but different code paths." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "Trailing-slash local path warning vs SigLIP2 HF/JAX discrepancy; unrelated issues." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Both mention missing `pad_token_id`, but they affect different model configs and likely need different fixes." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Metric-gathering tuple truncation vs BeitImageProcessorFast label reduction; not the same underlying bug." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3Moe FP8 failure vs Qwen2.5-VL import error; unrelated model-loading problems." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure vs stale `num_return_sequences` warning; different generate-path issues." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "Importing Qwen2.5-VL class vs DeepSeek detokenization regression; not the same bug." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Both are SigLIP2-related, but one is zero probabilities and the other is implementation mismatch vs JAX; too broad to merge." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added token vs embedding reinit on resize; unrelated." + }, + { + "left": "issue:41720", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3 auto device-map CUDA assert vs FP8 inference failure on B200; distinct failures." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Eval dtype regression vs MoE training memory usage; different layers of the stack." + }, + { + "left": "issue:43526", + "right": "issue:46192", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug vs `return_tensors` being ignored; unrelated APIs." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "BigBirdTokenizer special-token registration vs GraniteSpeech CI test staleness; not the same code-path problem." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizer/model loading failures for different architectures; no single fix would cover both." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate default vs video batching shape bug; unrelated processors." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 output issue vs RT-DETR memory leak; completely different bugs." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Python 3.13 `flash_attn` import handling vs Windows ROCm `torch.distributed` dependency; unrelated import issues." + }, + { + "left": "issue:45290", + "right": "issue:46326", + "accept": false, + "reason": "Both touch `apply_chat_template`, but one is tool-call/no-content crashes and the other is `continue_final_message` on mistral_common; not mergeable as one fix." + }, + { + "left": "issue:45759", + "right": "issue:46285", + "accept": false, + "reason": "Composite Qwen config unwrap bug vs CLIP text-model structure inconsistency; different models and code paths." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Processor input-size mismatch vs float16 AMP crash in a loss module; unrelated." + }, + { + "left": "issue:44561", + "right": "issue:46234", + "accept": false, + "reason": "Removal of `is_torch_fx_available` breaking remote-code models vs missing `is_falcon_mamba_ssm_available`; different import-utils gaps." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC latent/forward mismatch vs GraniteSpeech CI test failure; unrelated." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both involve `apply_chat_template`, but one crashes on batched padding=False input and the other returns all-zero assistant masks; different failure modes." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Unsupported `center` arg in LasrFeatureExtractor vs Whisper empty transcription after align_special_tokens; unrelated." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs regressions, but they refer to different pipeline docs and different deprecated behavior; not the same change." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "DAC consistency bug vs `_get_feat_extract_output_lengths` behavior in qwen3_omni_moe; 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": 21, + "review_unit_index": 27, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:39692", + "issue:42490", + "issue:43278", + "issue:43335", + "issue:43425", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43720", + "issue:43819", + "issue:43825", + "issue:43854", + "issue:44112", + "issue:44206", + "issue:44351", + "issue:44360", + "issue:44448", + "issue:44493", + "issue:44561", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:45083", + "issue:45127", + "issue:45200", + "issue:45245", + "issue:45290", + "issue:45310", + "issue:45381", + "issue:45459", + "issue:45478", + "issue:45588", + "issue:45593", + "issue:45823", + "issue:45859", + "issue:46234", + "issue:46402" + ], + "soft_pairs": [ + "issue:44857|issue:45593", + "issue:43335|issue:43550", + "issue:43611|issue:46402", + "issue:43606|issue:45083", + "issue:37428|issue:43425", + "issue:29942|issue:32090", + "issue:44448|issue:44743", + "issue:43335|issue:43653", + "issue:45200|issue:45588", + "issue:45127|issue:45245", + "issue:43550|issue:44112", + "issue:44855|issue:45083", + "issue:37428|issue:44351", + "issue:43819|issue:44493", + "issue:43335|issue:43504", + "issue:43278|issue:43825", + "issue:44206|issue:45823", + "issue:29127|issue:32090", + "issue:39692|issue:43526", + "issue:45588|issue:45859", + "issue:45290|issue:45381", + "issue:43335|issue:45083", + "issue:43720|issue:44112", + "issue:45459|issue:46234", + "issue:44448|issue:44779", + "issue:45310|issue:45478", + "issue:43335|issue:44112", + "issue:39692|issue:44493", + "issue:43335|issue:45593", + "issue:34689|issue:35141", + "issue:42490|issue:43854", + "issue:43606|issue:44857", + "issue:44360|issue:44561" + ], + "prepared_review_unit_hash": "90563a2ac5c0aa685dd70203b7e661839a6933ca17516ede6dea7a10b2e80504", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37210, + "estimated_input_tokens": 9303, + "estimated_eval_tokens": 18862 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a collection of unrelated issue reports with only one clear duplicate pair: the repeated Qwen3.5 MoE `from_pretrained` error (issue 45310 / 45478). The soft links are generally too broad to merge.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45310 is the earliest and more detailed report of the only clear duplicate pair in this set; issue:45478 repeats the same Qwen3.5 MoE `from_pretrained` failure.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the strongest canonical issue candidate because it matches the only exact duplicate and is the earlier report. The remaining issues are unrelated enough that none is a better global representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "Different models and failure modes: AMP/CUDA loss crash vs missing auxiliary losses in D-FINE." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Both mention model behavior, but the bugs are unrelated: SwitchTransformers config vs Bamba torch.compile/SDPA." + }, + { + "left": "issue:43611", + "right": "issue:46402", + "accept": false, + "reason": "Both are loading-related, but one is a generic v5 `base_model_prefix` regression and the other is Sam3Model text-encoder weights not loading." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "Device-offload mismatch in Bark vs a helper-function output-length issue in qwen3_omni_moe; no shared code-path." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Import error for a missing flash-attention helper vs torch 2.10 compatibility; too broad to be the same bug." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 test failures and a Trainer broadcast TypeError are different failures with no concrete overlap." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Different models and mechanisms: Pegasus output changes vs Qwen3 recurrent state reset with cache." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug vs BigBirdTokenizer special-token registration bug." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 mm_token_type_ids defaulting issue vs flash_attention.py crashing on `s_aux=None`; unrelated." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA/extended-vocab collapse and category-count limits are unrelated model-training constraints." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Both involve CI/model execution, but one is Bamba SDPA compile failure and the other is a GraniteSpeech stale test." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Python 3.13 import/`torch.jit.script` parsing problem vs qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols and different root causes." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC latent/forward mismatch is a different bug from unexpected `position_id` keys." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation and Beit legacy-field loading are separate config/loading issues." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype regression and pipeline translation messaging are unrelated v5 issues." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "Unsupported `center` arg in LasrFeatureExtractor vs Gemma4 PLE device mismatch; different code paths." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "Clarifying an error message in LayoutLMv3 is not the same bug as `_gpu_broadcast_one` rejecting NoneType." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "Documentation example errors vs BeitImageProcessorFast `reduce_labels` behavior; not the same underlying change." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Both touch flash/offload-related areas, but one is `s_aux=None` handling in flash_attention.py and the other is rotary embedding CPU-offload incompatibility." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Chat-template tool-call crash and Qwen2.5-VL video `vision_position_ids` are distinct failures." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "A SwitchTransformers sparse-layer config bug is unrelated to qwen3_omni_moe feature-length output." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "BitNet packed-weight loading and GraniteSpeech CI stale override tests do not share a concrete code path." + }, + { + "left": "issue:45459", + "right": "issue:46234", + "accept": false, + "reason": "Both involve import utilities, but one hides tokenizer errors without protobuf and the other is a missing availability helper." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer/model regressions; similar v5 timing is not enough to merge." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "These are the same Qwen3.5 MoE `from_pretrained` regression reported twice with matching titles and symptom." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "SwitchTransformers config and GraniteSpeech CI issues are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path trailing-slash warning and GLM-4.7-Flash unit-test loading failure are different loading bugs." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch in Bark vs float16 AMP crash in LwDetrImageLoss; no shared underlying bug." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer architecture comment and removal of `is_torch_fx_available` affecting trust_remote_code are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: issue 45310/45478 is the only clearly duplicated pair by title and symptom, and the soft-edge rejections stay appropriately conservative. I didn\u2019t see any unsupported overstatement or an implausibly asserted mergeable pair." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 28, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:27957", + "issue:28218", + "issue:29942", + "issue:32090", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:40444", + "issue:41720", + "issue:42490", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43701", + "issue:43746", + "issue:43873", + "issue:43927", + "issue:44448", + "issue:44466", + "issue:44493", + "issue:44561", + "issue:44568", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45083", + "issue:45229", + "issue:45372", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45698", + "issue:45823", + "issue:45920", + "issue:46192", + "issue:46209", + "issue:46399" + ], + "soft_pairs": [ + "issue:45920|issue:46192", + "issue:44857|issue:45072", + "issue:38175|issue:44493", + "issue:28218|issue:32090", + "issue:40444|issue:43526", + "issue:44448|issue:44977", + "issue:43701|issue:45083", + "issue:44977|issue:45229", + "issue:43746|issue:46209", + "issue:32090|issue:33357", + "issue:29942|issue:35141", + "issue:43550|issue:45083", + "issue:44855|issue:46399", + "issue:42490|issue:43575", + "issue:43927|issue:45698", + "issue:44561|issue:44779", + "issue:10000|issue:15354", + "issue:37428|issue:44987", + "issue:27957|issue:34689", + "issue:43278|issue:43329", + "issue:43873|issue:45229", + "issue:44857|issue:45588", + "issue:41720|issue:45229", + "issue:44568|issue:45491", + "issue:44977|issue:45563", + "issue:43278|issue:43381", + "issue:45372|issue:45588", + "issue:45588|issue:45823", + "issue:44448|issue:44466" + ], + "prepared_review_unit_hash": "05164501a961d68bd153844bbeacc4a381cde8e39d93a96db2f154afcb2ee06d", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36048, + "estimated_input_tokens": 9012, + "estimated_eval_tokens": 18280 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a duplicate cluster: the items span many unrelated bugs, regressions, and model-specific failures. The soft-similarity pairs mostly share buzzwords or subsystems, but they describe different code paths and symptoms, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the set is heterogeneous and the candidate pairs are only textually similar, not the same underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "#45920 is the strongest standalone issue to use as a representative report: it describes a broad, reproducible tokenizer regression affecting multiple models and has inbound references.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45920", + "right": "issue:46192", + "accept": false, + "reason": "Tokenizer ID regression across many models vs. `return_tensors` being ignored when `text_kwargs` is passed; different API behaviors and failure modes." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "`LwDetrImageLoss` fp16/CUDA crash vs dtype mismatches in `SwitchTransformers`/`TimmWrapperModel` bfloat16 inference; unrelated modules." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "`siglip2` zero probabilities vs unexpected position-id key warnings across models; different symptom and likely cause." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer extra-space bug vs Trainer `_gpu_broadcast_one` NoneType error; unrelated code paths." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "IterableDataset multi-image finetuning failure vs `BeitImageProcessorFast.reduce_labels` returning one label; different subsystems." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus v4/v5 output difference vs Qwen3.5 flash-attention generation crash; different model families and bugs." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint key mismatch on resume vs qwen3_omni_moe feature-length helper bug; unrelated behaviors." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Flash-attention generation failure vs Gemma4 multi-GPU CUDA OOM; different failure modes and code paths." + }, + { + "left": "issue:43746", + "right": "issue:46209", + "accept": false, + "reason": "PEFT local checkpoint load failure vs FP8 inference on B200; no shared underlying defect." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "`_gpu_broadcast_one` NoneType TypeError vs MacOS bus error with a CLIP model; different platforms and root causes." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures vs token-embedding resize reinitializing output embeddings; separate bugs." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure vs qwen3_omni_moe helper-function bug; unrelated model/runtime issues." + }, + { + "left": "issue:44855", + "right": "issue:46399", + "accept": false, + "reason": "Python 3.13 `@torch.jit.script` parsing/IndentationError vs missing decorator causing Qwen3.5 rotary registration error; superficially script-related but not the same bug." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "Import warning for local paths with trailing slash vs OOM loading `Qwen2-57B-A14B-Instruct` with tensor parallelism; different problem class." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "Custom token IDs lost on save/load vs wrong custom module loaded after `save_pretrained`; both involve serialization, but they affect different mechanisms." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Removal of `is_torch_fx_available` breaking `trust_remote_code` vs Deepseek tokenizer regression in v5; unrelated API and tokenizer issues." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel issue vs GeneratorExp not supported by `torch.jit.script` while exporting ViT; no common underlying bug." + }, + { + "left": "issue:37428", + "right": "issue:44987", + "accept": false, + "reason": "Missing `_flash_supports_window_size` import vs failure loading `physical-intelligence/fast`; different import/runtime causes." + }, + { + "left": "issue:27957", + "right": "issue:34689", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 support vs model loading broken for Llama 3.2 90B Vision Instruct; different models and symptoms." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype changes from BF16 to FP32 in eval vs undefined helper names in video token counting; unrelated bugs." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Quantization offloading not working as expected vs Gemma4 multi-GPU OOM; both memory-related, but not the same defect." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "`LwDetrImageLoss` fp16/CUDA crash vs `flash_attention.py` crashing on `s_aux=None`; distinct code paths." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3 auto device mapping `cudaErrorAssert` vs Gemma4 multi-GPU OOM; different model-specific failures." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer failing to add BOS/EOS with `add_special_tokens=True` vs Gemma3 NaN embeddings from sliding-window attention; tokenizer vs attention bug." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Flash-attention generation crash vs stale `num_return_sequences` warning in paged generate; not the same bug." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Embedding dtype regression vs gradient checkpointing being disallowed in eval mode; unrelated behavior." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "`ReasoningEffort` import failure for Gemma 4 processor loading vs `flash_attention.py` AttributeError on `s_aux=None`; different subsystems." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "`flash_attention.py` `s_aux=None` crash vs Gemma4 PLE device mismatch with `device_map='auto'`; distinct bugs." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Pegasus v4/v5 output differences vs device-dependent `lm_head.weight` serialization; both version-related, but not the same concrete 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": 21, + "review_unit_index": 29, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:25251", + "issue:29127", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43381", + "issue:43412", + "issue:43575", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43873", + "issue:43976", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44442", + "issue:44462", + "issue:44479", + "issue:44743", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45563", + "issue:45588", + "issue:45712", + "issue:45803", + "issue:46121", + "issue:46369" + ], + "soft_pairs": [ + "issue:43976|issue:45440", + "issue:45372|issue:45406", + "issue:43412|issue:45588", + "issue:25251|issue:32090", + "issue:45198|issue:45803", + "issue:43653|issue:44743", + "issue:45245|issue:45563", + "issue:32090|issue:33290", + "issue:42490|issue:43873", + "issue:44361|issue:45198", + "issue:44186|issue:45198", + "issue:33357|issue:39692", + "issue:29127|issue:34567", + "issue:43065|issue:45412", + "issue:44206|issue:44479", + "issue:43065|issue:43994", + "issue:43611|issue:46121", + "issue:44871|issue:44977", + "issue:43066|issue:44462", + "issue:45459|issue:45712", + "issue:42915|issue:42994", + "issue:43278|issue:43575", + "issue:45412|issue:46369", + "issue:29942|issue:30064", + "issue:45372|issue:45459", + "issue:43994|issue:45412", + "issue:44265|issue:45198", + "issue:43381|issue:43701", + "issue:43653|issue:44442", + "issue:43756|issue:45440" + ], + "prepared_review_unit_hash": "2cb5259acb03ecfbc896aad9a09b3bb16291ae191eadd5e0d329d3b3bb325cdd", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36616, + "estimated_input_tokens": 9154, + "estimated_eval_tokens": 18564 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated issue candidates. Several share broad themes like Gemma, tokenizer, quantization, or memory behavior, but none of the soft pairs look like the same concrete bug/change.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 43873 is the broadest and most active representative in the set (open, multiple inbound refs, substantial discussion) even though there is no clear duplicate issue to canonize against.", + "canonical_pr_reason": null, + "best_issue_reason": "43873 is the strongest global representative because it has the most discussion and a broadly actionable runtime/quantization scope.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Different problems: Python/version compatibility vs a DeepSeekV3MoE implementation divergence." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both involve Gemma 4, but one is an ImportError during processor loading and the other is a serve-time AttributeError for a missing tokenizer field." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Unrelated code paths: executorch export vs flash_attention integration crashing on sink-less models." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline top_k nesting bug is unrelated to Trainer NoneType broadcast failures." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Both touch input processing, but one is save/tokenization failure in Wav2Vec2 and the other is a bare-except bug swallowing errors in Fuyu tensor conversion." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBird tokenizer special-token decode bug is unrelated to Qwen recurrent-state caching behavior." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Categories-limit RuntimeError and stale generate warning are different issues with no shared failure path." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer broadcast NoneType error is not the same as Deepspeed Adafactor OOM." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Local-path import warning is unrelated to quantization/offloading behavior." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizers and different failure modes; no shared code bug." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2 NER/padding crash is unrelated to Wav2Vec2 save/tokenization failures." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "MacOS bus error in CLIP loading is not the same as a SigLIP2 doc example/model-processor mismatch." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and TrainerState token counting are unrelated." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder is unrelated to RT-DETR memory release/GC behavior." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Unsupported center arg in LasrFeatureExtractor is unrelated to the Qwen video regression." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Dummy layer cleanup and SigLIP2 nonsense outputs are different bugs." + }, + { + "left": "issue:43611", + "right": "issue:46121", + "accept": false, + "reason": "base_model_prefix loading regression is unrelated to rope-validation list handling." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Gemma eos_token_id mismatch and Qwen3.5 flash-attention generation failure are different model-specific issues." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer decoder-type regression in v5 is not the same as AutoTokenizer ignoring tokenizer.json." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer error masking due to missing protobuf is unrelated to leftover dummy classes in dummy_pt_objects.py." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "One is FP8 config loading for Qwen3Moe; the other is saving quantized models. Different lifecycle stages and failures." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Embedding dtype drift in eval is unrelated to tensor-parallel OOM on Qwen2-57B-A14B-Instruct." + }, + { + "left": "issue:45412", + "right": "issue:46369", + "accept": false, + "reason": "Both are memory-related, but RT-DETR GC leakage and Sam3-litetext memory overflow are not the same bug." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "FlashAttention 2 test failures are unrelated to void segmentation-map processing." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Gemma processor import failure and protobuf-related tokenizer error masking are separate issues." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 inference nonsense and RT-DETR memory retention are unrelated." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export failure with torch_compilable_check is unrelated to Wav2Vec2 tokenizer/save_pretrained bugs." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Gradient checkpointing in eval mode and resume_from_checkpoint key mismatch are different training-control issues." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "BigBird special-token decode bug is unrelated to AutoTokenizer failing to load a FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch is unrelated to DeepSeekV3MoE divergence." + } + ] + }, + "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": 21, + "review_unit_index": 30, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:27957", + "issue:29127", + "issue:29942", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43334", + "issue:43381", + "issue:43421", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43701", + "issue:43761", + "issue:43824", + "issue:43931", + "issue:43994", + "issue:44062", + "issue:44186", + "issue:44206", + "issue:44279", + "issue:44387", + "issue:44442", + "issue:44464", + "issue:44625", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45245", + "issue:45440", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45834", + "issue:45859", + "issue:45907", + "issue:46121", + "issue:46369", + "issue:46399" + ], + "soft_pairs": [ + "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:43824|issue:44279", + "issue:43526|issue:44186", + "issue:29127|issue:35141", + "issue:44625|issue:45584", + "issue:43575|issue:43701", + "issue:44442|issue:45072", + "issue:42490|issue:43381", + "issue:43575|issue:44387", + "issue:42915|issue:43931", + "issue:43761|issue:45859", + "issue:43526|issue:44857", + "issue:46121|issue:46399", + "issue:43334|issue:43531", + "issue:44977|issue:45245", + "issue:44464|issue:45491", + "issue:43262|issue:44206", + "issue:27957|issue:29127", + "issue:44743|issue:44811", + "issue:27957|issue:33290", + "issue:43452|issue:45588", + "issue:33290|issue:34689", + "issue:44062|issue:45907" + ], + "prepared_review_unit_hash": "1efeff33695d7ff2f23d4867eccd36d323767c7622ec6ca8e8f07aa62242f873", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35435, + "estimated_input_tokens": 8859, + "estimated_eval_tokens": 17974 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the soft pairs mostly connect unrelated bugs that only share broad Transformers vocabulary (models, tokenizers, attention, loading, OOM, etc.). I rejected every soft edge because the artifacts do not describe the same underlying defect or fixable change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime special-token post-processing bug vs Gemma3 NaN embeddings under sliding-window/all-padding; different subsystems and failure modes." + }, + { + "left": "issue:27957", + "right": "issue:29942", + "accept": false, + "reason": "One is XLM-RoBERTa Flash Attention 2 support, the other is failing FA2 tests; related area but not the same bug." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 box-validation clarity issue vs MacOS bus error loading a CLIP model; unrelated." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "DeepSeekV3 implementation divergence vs D-FINE auxiliary loss behavior; different models and defects." + }, + { + "left": "issue:43994", + "right": "issue:46369", + "accept": false, + "reason": "SigLIP2 incorrect outputs vs SAM3 LiteText memory overflow; not the same code path." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug vs Kosmos2.5 long OCR index error; different processors and symptoms." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Specific Qwen2.5-VL import error vs a generic dependency issue; no shared concrete bug." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "BeitImageProcessorFast label reduction vs LayoutLMv2Tokenizer NER/padding crash; unrelated tokenizer/processor bugs." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 error message concern vs token-embedding resize/post_init reinitialization bug; different functionality." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "Qwen3.5 num_labels propagation bug vs Whisper empty-transcription generation failure; different models and paths." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Tensor-parallel OOM when loading Qwen2-57B vs resume_from_checkpoint key mismatch; unrelated training/loading problems." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer loading failure vs dtype mismatch in bf16 inference; not the same bug." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "Import warning for local path with trailing slash vs gradient checkpointing in eval mode; unrelated." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory, but one is TP load OOM for a specific Qwen model and the other is increased reserved memory under int4 quantization; different causes." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3Moe FP8 config failure vs Qwen3-VL weight-shape mismatch; distinct model/config issues." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression vs Qwen3_5Moe rotary embedding CPU offload incompatibility; unrelated." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "BeitImageProcessorFast label reduction vs LwDetrImageLoss AMP/CUDA crash; different components." + }, + { + "left": "issue:46121", + "right": "issue:46399", + "accept": false, + "reason": "Rope-validation TypeError from JSON-loaded list vs missing decorator for apply_rotary_pos_emb registration; unrelated implementation bugs." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Missing pad_token_id in Qwen3-VL text config vs sliding_window issue in Qwen3-MoE; different config fields and failures." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Flash-attention generation bug for Qwen3.5 vs category-count RuntimeError; no shared code path." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency vs Gemma3 NaN embeddings with sliding-window attention; distinct bugs." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio chat-template sample-rate default bug vs LasrFeatureExtractor unsupported center argument; unrelated audio-processing issues." + }, + { + "left": "issue:27957", + "right": "issue:29127", + "accept": false, + "reason": "Flash Attention 2 support for XLM-RoBERTa vs LayoutLMv3 error-message clarity for box inputs; different models and concerns." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Qwen3.5 recurrent-state reset bug in modular code vs Whisper processor batch_decode skip_special_tokens issue; unrelated model/processor behavior." + }, + { + "left": "issue:27957", + "right": "issue:33290", + "accept": false, + "reason": "FA2 support issue vs Adafactor/deepspeed OOM; no overlap beyond generic training/inference context." + }, + { + "left": "issue:43452", + "right": "issue:45588", + "accept": false, + "reason": "gguf_file breaks model/tokenizer loading vs flash_attention.py crash on s_aux=None; different failure sites and causes." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Adafactor OOM in deepspeed vs Llama 3.2 Vision model-loading breakage; unrelated." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken special-arg duplication vs list_repo_files missing revision forwarding; different APIs and 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": 21, + "review_unit_index": 31, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:33290", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:42175", + "issue:42490", + "issue:42907", + "issue:43232", + "issue:43316", + "issue:43381", + "issue:43388", + "issue:43493", + "issue:43525", + "issue:43611", + "issue:43646", + "issue:43720", + "issue:43723", + "issue:43749", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44190", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44496", + "issue:44743", + "issue:44855", + "issue:44871", + "issue:44898", + "issue:44987", + "issue:45003", + "issue:45137", + "issue:45216", + "issue:45290", + "issue:45440", + "issue:45479", + "issue:45698", + "issue:45759", + "issue:45803", + "issue:46178", + "issue:46234", + "issue:46421" + ], + "soft_pairs": [ + "issue:43646|issue:44077", + "issue:33290|issue:35141", + "issue:33290|issue:34567", + "issue:43381|issue:43873", + "issue:44442|issue:44898", + "issue:45003|issue:45440", + "issue:43723|issue:45479", + "issue:42490|issue:44743", + "issue:43316|issue:43525", + "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:44987|issue:46178", + "issue:43873|issue:44871", + "issue:42907|issue:45698", + "issue:44442|issue:45803", + "issue:44464|issue:46421", + "issue:43749|issue:43931", + "issue:43316|issue:45759", + "issue:44479|issue:45290", + "issue:44855|issue:45137", + "issue:44190|issue:44496" + ], + "prepared_review_unit_hash": "c46279eb56890f3738842754a18e983ba353523fd8b68df001fa214aa1275d5d", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35766, + "estimated_input_tokens": 8942, + "estimated_eval_tokens": 18140 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a heterogeneous set of unrelated Transformers issues with only superficial overlap in themes like loading, generation, and config handling. I found no true duplicates among the soft edges, and there is no clear single canonical issue or PR for the cluster.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Both mention model initialization/post_init in v5, but one is a broad custom-model init regression and the other is a patchtsmixer-specific post_init policy change." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "OOM with Adafactor+DeepSpeed is unrelated to embedding resize/post_init reinitialization." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "Optimizer memory blowup and TrainerState token counter not updating are separate bugs." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing in eval mode and quantization offloading are different execution paths and failure modes." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer-loading failure and Perceiver image-classification resolution failure are unrelated subsystems." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to DeepseekV3MoE implementation divergence." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "AutoTokenizer v5 loading regression and single-label classification zero-loss bug do not share the same underlying code path." + }, + { + "left": "issue:42490", + "right": "issue:44743", + "accept": false, + "reason": "Trailing-slash local model loading warning is unrelated to recurrent-state reset in Qwen3.5 cache handling." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both involve config APIs, but one is a Gemma3TextConfig mismatch and the other is a missing pad_token_id on Llama4Config." + }, + { + "left": "issue:42175", + "right": "issue:46234", + "accept": false, + "reason": "Torch-only install backend inclusion and a missing import_utils availability helper are different packaging/import issues." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression is model-specific; ModelOutput key assignment bug is a generic container issue." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both are loading regressions, but they concern different releases and different causes (Llama 3.2 vision load vs base_model_prefix handling)." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "BitNet packed-weight loading failure is unrelated to a tokenizer class import/load failure." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs sync after sync_gpus and gather_for_metrics label trimming are distinct distributed-training bugs." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch on load and Qwen3.5 save_pretrained corruption are related only at a very broad model-family level." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "Hidden_states omission in CLIPVisionModel and attention-kernel divergence under torch.compile are different issues." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "patchtsmixer post_init policy and Qwen video-input regression are unrelated model families and code paths." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation mismatch and docs still referencing pipeline() are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:46178", + "accept": false, + "reason": "Loading a specific model under transformers>=5.1.0 and missing BACKENDS_MAPPING support for tensorflow_text are unrelated failures." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading behavior and Gemma-3 eos_token_id configuration mismatch are different bugs." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Saving dequantized Ministrals/Devstrals and loading the wrong custom module after save_pretrained are not the same defect." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer loading and Fuyu tensor conversion exception handling are unrelated." + }, + { + "left": "issue:44464", + "right": "issue:46421", + "accept": false, + "reason": "Chunked generation inconsistency with compiled forward and cache lazy initialization causing recompilation are related only superficially." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL shape mismatch are separate loading problems." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy and Qwen composite-config unwrapping are both config-related but not the same underlying bug." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Video-input regression in Qwen VL models and chat-template crashes on tool-call assistant messages are different features." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 JIT parsing/IndentationError and DeepSpeed ZeRO3 deque underflow are unrelated." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in an example script and an unrecognized custom model type 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": 21, + "review_unit_index": 32, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:25251", + "issue:27957", + "issue:28282", + "issue:32090", + "issue:41950", + "issue:42617", + "issue:43065", + "issue:43295", + "issue:43316", + "issue:43404", + "issue:43441", + "issue:43452", + "issue:43475", + "issue:43540", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43643", + "issue:43646", + "issue:43650", + "issue:43701", + "issue:43716", + "issue:43742", + "issue:43761", + "issue:43825", + "issue:44077", + "issue:44162", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44792", + "issue:44871", + "issue:44877", + "issue:44913", + "issue:45584", + "issue:45632", + "issue:45803", + "issue:46050", + "issue:46170", + "issue:46202", + "issue:46365", + "issue:46369", + "issue:46424" + ], + "soft_pairs": [ + "issue:43643|issue:44913", + "issue:43295|issue:43825", + "issue:43550|issue:43701", + "issue:25251|issue:28282", + "issue:43650|issue:46202", + "issue:43606|issue:43701", + "issue:45632|issue:46424", + "issue:42617|issue:43575", + "issue:43575|issue:43742", + "issue:16998|issue:22355", + "issue:27957|issue:32090", + "issue:43295|issue:44871", + "issue:43316|issue:43716", + "issue:44610|issue:46369", + "issue:43065|issue:44617", + "issue:43316|issue:44877", + "issue:43540|issue:44514", + "issue:44871|issue:45584", + "issue:43404|issue:43441", + "issue:44077|issue:44534", + "issue:45803|issue:46365", + "issue:44162|issue:44792", + "issue:45803|issue:46170", + "issue:41950|issue:43475", + "issue:44871|issue:46050", + "issue:43646|issue:44913", + "issue:43452|issue:44496", + "issue:43761|issue:44479" + ], + "prepared_review_unit_hash": "9f84122799a74d02b7a7e6e12f4cf5eeb3bc290dffbf3a738eb349525566281b", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 35731, + "estimated_input_tokens": 8933, + "estimated_eval_tokens": 18122 + }, + "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 issues spanning model loading, processor regressions, video/multimodal bugs, cache behavior, and a few v5 regressions. I don\u2019t see any soft pair that looks like the same concrete bug/change, so all candidate duplicate edges should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43646 is the broadest, most umbrella-like report in the set (`Transformers 5.0.0 breaks custom model initialization`) and is the closest thing to a cluster anchor, though this is not a true duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the best single issue to represent the cluster because it is the most general v5-regression report and could plausibly subsume nearby initialization-related breakages, but the overall set is still too heterogeneous for a strong canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Both are config/load behavior bugs, but one is about trust_remote_code dropping fields and the other about rotary_pct not round-tripping; different code paths and fixes." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Both mention v5 regressions, but one is processor.tokenizer / image passing behavior and the other is an incorrect pipeline error message for translation tasks; not the same bug." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "Unrelated failures: torch.compile with SDPA in Bamba vs resume_from_checkpoint key mismatch." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k output shape regression vs missing PyTorch import error; no shared underlying bug." + }, + { + "left": "issue:43650", + "right": "issue:46202", + "accept": false, + "reason": "Placeholder/body noise vs Gemma pipeline roundtrip issue; no meaningful technical overlap." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch vs checkpoint key mismatch; distinct subsystems and failure modes." + }, + { + "left": "issue:45632", + "right": "issue:46424", + "accept": false, + "reason": "Both involve cache, but one is a trust_remote_code path collision and the other is max_cache_length handling for quantized cache; different defects." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "3d_parallel.py launch failure vs tensor-parallel OOM on Qwen2-57B; not the same problem." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "OOM during TP loading of a large Qwen model vs KeyError loading MobileLLM-125M; unrelated." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "model_max_length behavior in DeBERTa-V3 vs missing transformers.onnx module; no duplicate relationship." + }, + { + "left": "issue:27957", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 support for XLMRoberta vs Trainer broadcast TypeError; different code paths." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "One is a processor/tokenizer regression in v5, the other is an eos_token_id config mismatch in Gemma-3; similar theme but not the same bug." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "API discrepancy in Gemma3TextConfig vs dtype mismatch between image preprocessor and model; unrelated." + }, + { + "left": "issue:44610", + "right": "issue:46369", + "accept": false, + "reason": "OmDet-Turbo preprocessing size mismatch vs Sam3-LiteText memory overflow; different models and failure modes." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Dummy Conv2d implementation issue vs CUDA OOM in Sam3Video; not the same underlying defect." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy vs strict config blocking granite_speech load; different config-validation problems." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Different Qwen multimodal bugs: video input ValueError in Qwen3OmniMoe vs batched chat template crash in Qwen2_5_VLProcessor." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma eos_token_id inconsistency vs Whisper empty-transcription failure after align_special_tokens; unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Weight tying bug in Mistral3ForConditionalGeneration vs FlashAttention failure in Ministral-3; same family, different concrete issues." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "Optional post_init policy issue vs non-persistent buffers filled with junk; unrelated v5 behavior changes." + }, + { + "left": "issue:45803", + "right": "issue:46365", + "accept": false, + "reason": "Bare except swallowing interrupts in Fuyu conversion vs Trainer import conflict with datasets; no shared bug." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 regression impacting workflows vs janus image-generation test failure; not the same code path." + }, + { + "left": "issue:45803", + "right": "issue:46170", + "accept": false, + "reason": "Error swallowing in Fuyu conversion vs Windows/ROCm import failure due to torch.distributed dependency; unrelated." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification pipeline using image processors vs missing fpn_position_embeddings on Sam3VisionEncoderOutput; different bugs." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "Gemma eos_token_id config mismatch vs sinq quantization save/load breakage; unrelated." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Broad v5 custom initialization breakage vs GPTNeoX rotary_pct not persisting on reload; different symptoms and fixes." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "gguf_file handling for tokenizer/model loading vs unrecognized model_type in config.json; both loading-related but not the same defect." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression vs Qwen video-input regression in v5.3.0; unrelated model families and code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 33, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:35141", + "issue:41720", + "issue:42490", + "issue:43065", + "issue:43388", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43572", + "issue:43673", + "issue:43688", + "issue:43761", + "issue:43827", + "issue:43931", + "issue:43994", + "issue:44060", + "issue:44190", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44521", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44913", + "issue:44933", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45200", + "issue:45245", + "issue:45362", + "issue:45381", + "issue:45406", + "issue:45542", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:46178", + "issue:46234", + "issue:46369", + "issue:46421" + ], + "soft_pairs": [ + "issue:43388|issue:43688", + "issue:45127|issue:45200", + "issue:43673|issue:44743", + "issue:43572|issue:46178", + "issue:44617|issue:45584", + "issue:43504|issue:45698", + "issue:44734|issue:45406", + "issue:45542|issue:46234", + "issue:43673|issue:46421", + "issue:43525|issue:43531", + "issue:45381|issue:46369", + "issue:44792|issue:44977", + "issue:45563|issue:45584", + "issue:43761|issue:44242", + "issue:44792|issue:45600", + "issue:43931|issue:44060", + "issue:44190|issue:44964", + "issue:44534|issue:44913", + "issue:35141|issue:42490", + "issue:44479|issue:44734", + "issue:43827|issue:43994", + "issue:41720|issue:45245", + "issue:44855|issue:45588", + "issue:43065|issue:44610", + "issue:44617|issue:46369", + "issue:44933|issue:45362", + "issue:44464|issue:45563", + "issue:44521|issue:45381" + ], + "prepared_review_unit_hash": "87b73cd22c6a783ba4a6150a7134fec1ade35ef397a47de61e81de5bc01efbeb", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36132, + "estimated_input_tokens": 9033, + "estimated_eval_tokens": 18322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster does not cohere around one underlying bug or change. The soft-similarity pairs are mostly unrelated regressions across different models, loaders, and runtime paths, so I reject all candidate duplicate links.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the items span unrelated bugs/regressions (generation/cache, multimodal input, config loading, backend/import errors, docs, and CI/test failures).", + "canonical_pr_reason": null, + "best_issue_reason": "No clear global best issue for deduping; the cluster is too heterogeneous to anchor on one representative bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Different bugs: batch-label gathering vs auxiliary-loss normalization." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "Different areas: LoRA/vocab merge collapse vs Gemma4 multimodal token defaults." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both involve cache/state, but the concrete failures differ (missing cache in chunked prefill vs recurrent state reset with seq_len>1)." + }, + { + "left": "issue:43572", + "right": "issue:46178", + "accept": false, + "reason": "Unrelated: StableLm config field regression vs TensorFlow backend mapping error." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Different models and failures: Sam3Video OOM vs Whisper empty-transcription generation crash." + }, + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Different load/save bugs: legacy preset field vs wrong custom module selection after save_pretrained." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Serve KV-cache indexing crash is unrelated to Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:45542", + "right": "issue:46234", + "accept": false, + "reason": "Different import/backend issues: undefined TF backend vs missing falcon-mamba availability helper." + }, + { + "left": "issue:43673", + "right": "issue:46421", + "accept": false, + "reason": "Both mention chunked prefill/cache, but one is missing cache creation and the other is lazy-init recompilation; not clearly the same bug." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Different Qwen/Llama config issues: missing pad_token_id vs sliding_window behavior." + }, + { + "left": "issue:45381", + "right": "issue:46369", + "accept": false, + "reason": "Different problems: Qwen2.5-VL video position IDs vs Sam3-LiteText memory overflow." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Unrelated model/test failure vs flash-attention generation bug." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "Stale generate warning is unrelated to Whisper's empty transcription failure." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "Hidden-states regression is unrelated to missing load-balancing loss." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to broken auto_mappings imports." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Both are Qwen3-family issues, but one is a weight-shape load error and the other is an incorrect tied-weights warning." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading failure is unrelated to multimodal model load mismatch." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk and rotary_pct reload loss are different config/runtime bugs." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Embedding reinit after resizing is unrelated to the trailing-slash local path import warning." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Both are runtime regressions, but one is video-input handling and the other is KV-cache continuation indexing." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs referencing pipeline() is unrelated to siglip2 producing nonsensical outputs." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3 device-mapping cuda assert is unrelated to the 2^24 category limit runtime error." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 decorator parsing error is unrelated to the flash_attention s_aux=None crash." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder is unrelated to OmDet-Turbo's input-size mismatch." + }, + { + "left": "issue:44617", + "right": "issue:46369", + "accept": false, + "reason": "Same broad SAM family, but different model variants and different memory issues." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Missing import from image_utils is unrelated to Qwen3.5-35B chat crashes." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Chunked-generation inconsistency with compiled forward is unrelated to a stale warning about num_return_sequences." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "All-zero assistant masks for multimodal inputs is a different bug from Qwen2.5-VL video position IDs." + } + ] + }, + "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": 21, + "review_unit_index": 34, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:25251", + "issue:27957", + "issue:36246", + "issue:38617", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:43116", + "issue:43295", + "issue:43335", + "issue:43404", + "issue:43540", + "issue:43550", + "issue:43644", + "issue:43976", + "issue:44016", + "issue:44077", + "issue:44112", + "issue:44186", + "issue:44190", + "issue:44246", + "issue:44279", + "issue:44462", + "issue:44683", + "issue:44734", + "issue:44737", + "issue:44855", + "issue:44877", + "issue:44913", + "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" + ], + "soft_pairs": [ + "issue:43644|issue:44913", + "issue:43976|issue:44855", + "issue:43550|issue:44190", + "issue:44462|issue:45907", + "issue:25251|issue:27957", + "issue:44077|issue:44877", + "issue:45127|issue:45357", + "issue:43540|issue:44734", + "issue:44112|issue:45600", + "issue:45632|issue:46234", + "issue:43335|issue:44190", + "issue:44737|issue:45803", + "issue:43116|issue:44190", + "issue:43295|issue:45276", + "issue:45276|issue:45584", + "issue:44938|issue:45542", + "issue:36246|issue:41628", + "issue:45632|issue:46421", + "issue:46082|issue:46121", + "issue:45245|issue:45584", + "issue:43404|issue:45216", + "issue:42175|issue:44279", + "issue:36246|issue:38617", + "issue:41950|issue:44016", + "issue:44246|issue:45003", + "issue:44186|issue:44734", + "issue:44683|issue:45446" + ], + "prepared_review_unit_hash": "c62e0eeee1d1b3660d6dcd1c962e5ab4486c2454299dca83ef3e8e46a7ee2e17", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35358, + "estimated_input_tokens": 8840, + "estimated_eval_tokens": 17936 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of distinct Transformers bugs and regressions, with only superficial similarity around imports, configs, or pipeline behavior. None of the proposed soft pairs look like the same underlying change, so I would not merge any of them as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "issue:41628 is the cleanest broad, user-facing import regression report in the set, but the overall cluster is too heterogeneous to treat as one duplicate group.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41628 is the best anchor issue because it describes a clear top-level API failure with broad impact and easy reproduction compared with the more model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Both involve config/state handling, but one is junk non-persistent buffers and the other is rotary_pct not persisting; different bugs and code paths." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Both are Python compatibility failures, but one is a version-loading problem and the other is a DebertaV2 torch.jit parsing IndentationError on 3.13." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "One is a Bamba torch.compile/SDPA failure; the other is a local dataset loading issue in an image-classification example script." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer.json precedence and list_repo_files revision forwarding are unrelated repository/file-access bugs." + }, + { + "left": "issue:25251", + "right": "issue:27957", + "accept": false, + "reason": "Top-k pipeline output shape and XLMRoberta Flash Attention 2 are unrelated pipeline/model issues." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "These are both config-related, but one is about disallowing optional post_init and the other about strict config loading for granite_speech." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "LoRA merge collapse with tied embeddings and incorrect visual encoder keys in save_pretrained are different save/merge regressions." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Qwen3OmniMoe video processing and serve KV-cache continuation indexing are different failures in different stacks." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "A stale GraniteSpeech CI test and broken auto_mappings references to removed configs are unrelated CI breakages." + }, + { + "left": "issue:45632", + "right": "issue:46234", + "accept": false, + "reason": "Cache-path collisions for trust_remote_code and a missing import_utils symbol are not the same bug." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers sparse-layer config creation and a local dataset loading failure are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet CPU device placement and a bare-except swallowing errors in FuyuBatchFeature are different code-path defects." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "A multi-label classification example returning empty results is unrelated to loading a local dataset in another example script." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Processor/tokenizer access regression and Gemma4 embedding resize propagation are different bugs." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 embedding resizing and Whisper empty-transcription generation are unrelated model behaviors." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 import failure and missing TensorFlow backend due to tensorboard-only install are separate environment issues." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols and different modules; no evidence of one underlying fix." + }, + { + "left": "issue:45632", + "right": "issue:46421", + "accept": false, + "reason": "Trust-remote-code cache collisions and chunked-prefill lazy cache initialization are different caching problems." + }, + { + "left": "issue:46082", + "right": "issue:46121", + "accept": false, + "reason": "LlamaConfig head_dim validation and rope-validation list handling are distinct config-validation bugs." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "A categories-limit runtime error and Whisper empty transcription are unrelated failures." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Mistral3 weight tying and Qwen3.5 checkpoint saving are different model serialization/tie-weights issues." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "Missing TensorFlow in the torch extra and a generic dependency issue are not the same concrete bug." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "These are different missing-import regressions in different modules, with no shared failing code path." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "A pipeline looking for image processors and a notebook syntax error are completely unrelated." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Slow import and unsafe sys.modules access are different import-time issues with different root causes." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "LayoutLMv2 tokenizer crashes and serve KV-cache continuation indexing are unrelated tokenizer/server bugs." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both mention flex_attention, but one is a compiled flex_attention failure on newer torch and the other is a version-check/import bug for AuxRequest." + } + ] + }, + "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": 21, + "review_unit_index": 35, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:27957", + "issue:28282", + "issue:35141", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43441", + "issue:43479", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43643", + "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:45216", + "issue:45335", + "issue:45372", + "issue:45375", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45600", + "issue:45759", + "issue:46082", + "issue:46123", + "issue:46234", + "issue:46369", + "issue:46399", + "issue:46424" + ], + "soft_pairs": [ + "issue:43673|issue:46424", + "issue:43650|issue:44220", + "issue:44964|issue:45600", + "issue:43643|issue:44496", + "issue:43854|issue:44964", + "issue:44112|issue:44964", + "issue:44016|issue:44220", + "issue:27957|issue:35141", + "issue:43479|issue:45375", + "issue:46123|issue:46369", + "issue:42907|issue:43441", + "issue:43531|issue:45600", + "issue:43976|issue:44987", + "issue:43295|issue:45335", + "issue:43526|issue:43901", + "issue:28282|issue:35141", + "issue:44496|issue:45042", + "issue:43441|issue:45588", + "issue:45372|issue:46234", + "issue:44016|issue:44062", + "issue:43299|issue:45216", + "issue:43673|issue:44336", + "issue:44792|issue:46424", + "issue:43976|issue:44279", + "issue:43381|issue:45593", + "issue:43824|issue:44933", + "issue:43901|issue:44496", + "issue:45335|issue:45584", + "issue:44496|issue:45759", + "issue:44075|issue:44368", + "issue:43577|issue:45216", + "issue:46082|issue:46399" + ], + "prepared_review_unit_hash": "aca53eca54d40d4ac5ba212c2c634db28875966f941987c182910290fda52c24", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37406, + "estimated_input_tokens": 9352, + "estimated_eval_tokens": 18960 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The soft-pair suggestions are mostly false positives: they share broad keywords like loading, config, cache, or model families, but the underlying bugs and code paths differ. No single issue stands out as a true canonical duplicate target for this heterogeneous set.", + "confidence": 0.88, + "canonical_issue_reason": "No clear canonical issue: the cluster spans many unrelated bugs across different models, processors, configs, and utilities, so none is a reliable duplicate anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue: there isn\u2019t a single issue here that represents the whole set or is clearly the central duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43673", + "right": "issue:46424", + "accept": false, + "reason": "Both mention cache behavior, but one is about chunked_prefill cache availability and the other about max_cache_length on quantized cache only; different bugs." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "Unrelated: one is a vague data request, the other is a specific _torch_extract_fbank_features() issue in audio preprocessing." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Different failures: Phi-4 multimodal loading vs auto_mappings referencing removed Sam3LiteText configs." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Both are config-loading problems, but one is about trust_remote_code dropping fields and the other about an unrecognized model missing model_type." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Different models and failure modes: Qwen2.5-VL import error vs Phi-4 multimodal load issue." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "CI stale test failure is unrelated to Phi-4 multimodal model loading." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error and audio feature extraction bug are unrelated." + }, + { + "left": "issue:27957", + "right": "issue:35141", + "accept": false, + "reason": "XLMRoberta FlashAttention support and token-embedding reinit on resize are different code paths." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Both are multimodal config issues, but they affect different fields and model families; not the same underlying bug." + }, + { + "left": "issue:46123", + "right": "issue:46369", + "accept": false, + "reason": "MaskGenerationPipeline batch-handling bug is unrelated to sam3-litetext memory overflow." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Same model family, but one is about saving dequantized checkpoints and the other about FlashAttention loading; different bugs." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Qwen3-MoE sliding_window bug is unrelated to auto_mappings referencing removed Sam3LiteText configs." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Python version support issue and a specific model-loading failure are not the same bug." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "One is a processor/tokenizer regression with images; the other is resize_token_embeddings not updating decoder.embed_tokens." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "BEiTImageProcessorFast label reduction bug and a docs mismatch for TextClassificationPipeline are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "PyTorch missing import error is unrelated to embedding resizing/reinitialization." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Unrecognized model config and PIL processor torchvision dependency regression are different problems." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both touch flash attention, but the concrete failures and code paths differ: Ministral loading vs s_aux=None crash in flash_attention.py." + }, + { + "left": "issue:45372", + "right": "issue:46234", + "accept": false, + "reason": "Different missing-import issues affecting different features and modules." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error is unrelated to AddedToken receiving duplicate special arguments." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Both involve Qwen multimodal models, but one is a loading regression and the other is a save_pretrained checkpoint issue." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Cache missing during chunked_prefill and stray ANSI codes in loading_report are unrelated." + }, + { + "left": "issue:44792", + "right": "issue:46424", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to max_cache_length handling." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Both mention dependency problems, but they are not the same concrete issue." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing in eval mode and missing auxiliary losses in D-FINE are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Both are import errors, but they concern different symbols and modules." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Docs for return_all_scores and an unrecognized model config are unrelated." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Different model-specific bugs: t5gemma embedding resize vs Whisper empty-transcription generation failure." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Both are config/model instantiation issues, but one is an unrecognized model and the other is failure to unwrap text_config." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Optimizer SGD args not being used is unrelated to the Qwen3.5 tie_word_embeddings warning." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "BLIP2 dtype handling and Qwen3.5 checkpoint saving are different bugs." + }, + { + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "LlamaConfig head_dim validation and missing apply_rotary_pos_emb decorator 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": 21, + "review_unit_index": 36, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41950", + "issue:43295", + "issue:43299", + "issue:43441", + "issue:43450", + "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:44623", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:45020", + "issue:45081", + "issue:45103", + "issue:45200", + "issue:45276", + "issue:45278", + "issue:45325", + "issue:45357", + "issue:45372", + "issue:45397", + "issue:45405", + "issue:45542", + "issue:45593", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45907", + "issue:46132", + "issue:46209" + ], + "soft_pairs": [ + "issue:43650|issue:46209", + "issue:45103|issue:45372", + "issue:44242|issue:45200", + "issue:44279|issue:45405", + "issue:38175|issue:44617", + "issue:44779|issue:45812", + "issue:43295|issue:44779", + "issue:43299|issue:45357", + "issue:44351|issue:45278", + "issue:44521|issue:45325", + "issue:41950|issue:45020", + "issue:39692|issue:43901", + "issue:41950|issue:45823", + "issue:44683|issue:45542", + "issue:43976|issue:45405", + "issue:44462|issue:45081", + "issue:44297|issue:44462", + "issue:43299|issue:43854", + "issue:43479|issue:43525", + "issue:43450|issue:44279", + "issue:45081|issue:45907", + "issue:44466|issue:45276", + "issue:43650|issue:46132", + "issue:44016|issue:44291", + "issue:43441|issue:45834", + "issue:44623|issue:45397", + "issue:44623|issue:44749", + "issue:43653|issue:45593" + ], + "prepared_review_unit_hash": "8599896de06695f65cebcfef329568bc7d5519bcb5ab0159c0fad2d082a82543", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35557, + "estimated_input_tokens": 8890, + "estimated_eval_tokens": 18036 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are mostly a heterogeneous grab-bag of tokenizer, processor, multimodal, config, CI, and documentation regressions. The soft pairs share broad themes, but I don\u2019t see two artifacts in any pair describing the same concrete bug/change closely enough to merge as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "No single canonical issue stands out: the cluster is too heterogeneous, and the overlapping titles reflect broad subsystem similarity rather than one underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue for deduping; any single issue would be arbitrary because the set spans unrelated failure modes and code paths.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43650", + "right": "issue:46209", + "accept": false, + "reason": "Completely unrelated: placeholder data report vs FP8/B200 inference failure." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "Different bugs: auto-docstring annotation handling vs Gemma 4 processor import failure." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "Separate training issues: router/load-balancing loss vs Gemma 4 mm_token_type_ids defaulting." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Only vaguely dependency-related; transformers compatibility issue is not the same as an unreleased peft version bump." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Unrelated model failures: SigLIP2 zero probabilities vs Sam3Video OOM." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Both tokenizer regressions, but different model families and likely different code paths." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Processor.tokenizer regression is distinct from Deepseek tokenization output changes." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Both Qwen multimodal issues, but loading failure and save_pretrained key serialization are different bugs." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "Import-related, but one missing symbol and the other a broad import-error report are not the same concrete defect." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Different multimodal logic: assistant_masks bug vs rope_index temporal position scaling." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video-classification processor lookup bug is unrelated to remote_code breakage." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both docs issues, but they concern different examples and behaviors." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "No shared bug: image-processor lookup in video pipeline vs Gemma4 device placement mismatch." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Different subsystems entirely: flex_attention compilation vs TensorFlow/tensorboard backend detection." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Unrelated versioning problems: Python support mismatch vs unreleased peft dependency." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Both tokenizer loading bugs, but one ignores tokenizer.json and the other crashes in Mistral regex patching." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Serialization/save_pretrained mismatch vs AutoTokenizer loading preference are different directions of the pipeline." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Different model-loading failures for different models and triggers." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Distinct config classes and fields: multimodal defaults vs missing pad_token_id." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "Video processor shape bug is unrelated to a generic dependency issue." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regex crash vs API forwarding regression in list_repo_files are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Both touch model weights, but serialization inconsistency and resize_token_embeddings behavior are different code paths." + }, + { + "left": "issue:43650", + "right": "issue:46132", + "accept": false, + "reason": "Placeholder issue vs attention registration behavior change are unrelated." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error is unrelated to init_empty_weights TypeError." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "Different model-specific failures: FlashAttention on Ministral vs long OCR index error in Kosmos2.5." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Both mention save/from_pretrained, but processor file omission and Gemma-4 zero3 loading are different defects." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Processor save_pretrained bug is unrelated to the reported performance regression." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer decode bug is unrelated to D-FINE auxiliary-loss behavior." + } + ] + }, + "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": 21, + "review_unit_index": 37, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:33290", + "issue:41628", + "issue:42175", + "issue:42994", + "issue:43262", + "issue:43317", + "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:44521", + "issue:44589", + "issue:44671", + "issue:44683", + "issue:44779", + "issue:45103", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45593", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45823", + "issue:46399" + ], + "soft_pairs": [ + "issue:42994|issue:43317", + "issue:45600|issue:45712", + "issue:33290|issue:45656", + "issue:44261|issue:44263", + "issue:44361|issue:45593", + "issue:44671|issue:44779", + "issue:42175|issue:43976", + "issue:44016|issue:45823", + "issue:43901|issue:43994", + "issue:43441|issue:44683", + "issue:43723|issue:44162", + "issue:43673|issue:43761", + "issue:44297|issue:45397", + "issue:43262|issue:44521", + "issue:43824|issue:43976", + "issue:43525|issue:43881", + "issue:43901|issue:44246", + "issue:43525|issue:43643", + "issue:44589|issue:45362", + "issue:41628|issue:43976", + "issue:44188|issue:44336", + "issue:43784|issue:45542", + "issue:44295|issue:45397", + "issue:44336|issue:44373", + "issue:45103|issue:46399", + "issue:45561|issue:45563", + "issue:44242|issue:45593", + "issue:44246|issue:44496" + ], + "prepared_review_unit_hash": "5dee2111a29cde0f93b2633c8a58a05f69a5463b797935521145995176e7999a", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 35889, + "estimated_input_tokens": 8973, + "estimated_eval_tokens": 18202 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pairs look like true duplicates; they mostly share broad subsystems or regression patterns but target different code paths, models, or packaging/docs issues.", + "confidence": 0.93, + "canonical_issue_reason": "issue:44521 is the strongest representative issue in this set: it is open, has the most visible user impact and corroboration, and describes a concrete multimodal `apply_chat_template` regression.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 is the best single issue to keep as the cluster representative because it is current, well-discussed, and reproducible; the rest are clearly separate bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Both involve quantization/offload, but one is about saving quantized models and the other about loading a dequantized model with `device_map=auto`; different failure modes." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are repo-maintenance/CI problems, but one is stale `auto_mappings.py` references and the other is leftover dummy classes leaking into `dir(transformers)`." + }, + { + "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:44261", + "right": "issue:44263", + "accept": false, + "reason": "Completely different code paths: missing `rms_norm_eps` in MLA q_a_layernorm versus `torch.split()` return handling in GlmMoeDsaIndexer." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "Tokenizer AttributeError on tasks is unrelated to D-FINE auxiliary losses when denoising is off." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 regressions, but one is CamemBERT masked-LM predictions and the other is Deepseek tokenizer behavior; different components." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "Packaging/install backend issue versus Python-version compatibility issue; no shared underlying bug." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "A notebook syntax error and a Gemma4 device-mismatch forward bug are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mentioning `return_all_scores` is not the same as bad `AutoModel`/pipeline outputs for SigLIP2." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Both are attention-kernel regressions, but they affect different kernels, versions, and code paths; too broad to merge." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading failure and ESM2 runtime breakage are different issues." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache/chunked-prefill bug versus CLIPVision hidden-states regression; distinct modules." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save metadata mismatch and Gemma4 zero3 `from_pretrained` crash are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Both touch `apply_chat_template` in multimodal contexts, but one is audio sampling-rate defaulting and the other is assistant-mask generation; different bugs." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Missing model import and Python-version support are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Llama4 missing `pad_token_id` and GLM-4V loading failure are different model-specific loading problems." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "A documentation mismatch and intermittent import slowness are not the same underlying issue." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Both involve config fields, but one is a specific model config missing `pad_token_id` while the other is `trust_remote_code` dropping fields on `AutoConfig`; different causes." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage support error versus Qwen3.5 chat crash; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "`AutoImageProcessor` import failure is unrelated to the Python-version compatibility bug." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "Torch.compile attention-kernel divergence and stray ANSI codes in loading_report are unrelated." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Both are import-time dependency issues, but they come from different missing backends and different code paths; not the same bug." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "Buffer read error for `position_ids` and Gemma4 zero3 loading failure are different problems." + }, + { + "left": "issue:44297", + "right": "issue:44373", + "accept": false, + "reason": "Tokenizer save config mismatch and a wrong docstring are clearly unrelated." + }, + { + "left": "issue:45103", + "right": "issue:46399", + "accept": false, + "reason": "Auto-docstring failure with future annotations and missing decorator registration for rotary embeddings are different issues." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "Pytest-xdist race in testing utils and a stale generate warning are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Both are losses gated by flags, but they affect different models and different loss terms; not the same code-path bug." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import slowness and unrecognized-model config errors 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": 21, + "review_unit_index": 38, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "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: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:45412", + "issue:45561", + "issue:45584", + "issue:45701", + "issue:45859", + "issue:46123", + "issue:46132", + "issue:46291", + "issue:46369", + "issue:46399", + "issue:46424" + ], + "soft_pairs": [ + "issue:42491|issue:43299", + "issue:43531|issue:45362", + "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:45003|issue:45561", + "issue:44336|issue:46424", + "issue:44779|issue:45292", + "issue:44295|issue:44749", + "issue:43475|issue:46369", + "issue:44242|issue:44857", + "issue:36246|issue:43596", + "issue:46291|issue:46399", + "issue:45859|issue:46399", + "issue:43673|issue:44561", + "issue:43901|issue:45003", + "issue:44162|issue:45200", + "issue:43352|issue:45250", + "issue:44496|issue:45003", + "issue:43257|issue:45137", + "issue:45292|issue:45584", + "issue:43262|issue:44246", + "issue:45412|issue:46123", + "issue:45292|issue:45701", + "issue:44561|issue:46399" + ], + "prepared_review_unit_hash": "7aa61f99bdb17fe0a6aff436ffb66c96174f113dadde1735fddf6bc0ce8d357f", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36021, + "estimated_input_tokens": 9006, + "estimated_eval_tokens": 18268 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous and mostly contains separate model-specific regressions and import/API breakages, with only weak topical overlap around v5 compatibility. I found no soft pair that is clearly the same underlying bug/change.", + "confidence": 0.62, + "canonical_issue_reason": "Issue 44561 is the broadest and most representative regression: a v5 API removal breaking trust_remote_code model loading. It best captures the cluster\u2019s repeated theme of version-upgrade breakages.", + "canonical_pr_reason": null, + "best_issue_reason": "44561 is the most general, actionable, and cluster-like issue among the set. It describes a concrete upstream compatibility regression rather than a niche model symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both mention Qwen3 MoE/v5 loading regressions, but one is LoRA compatibility and the other is Qwen3VL model loading; not the same concrete bug." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Different failures in different flows: sliding-window handling vs chat crash." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "Beit image processor label reduction vs unsafe sys.modules access are unrelated code paths." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline processor lookup and audio fbank feature extraction are different subsystems." + }, + { + "left": "issue:42886", + "right": "issue:46132", + "accept": false, + "reason": "Offline tokenizer cache loading and AttentionInterface.register behavior are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:44871", + "accept": false, + "reason": "Router-logits loss handling and Gemma eos_token_id configuration are distinct bugs." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI output in loading_report and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they are different missing symbols from different packages and likely different causes." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Missing Qwen2_5_VLImageProcessor export and missing HybridCache export are separate import failures." + }, + { + "left": "issue:44734", + "right": "issue:46424", + "accept": false, + "reason": "Both involve cache behavior, but one is KV-cache continuation tensor indexing and the other is max_cache_length semantics." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Unsafe sys.modules access and a pytest-xdist file race are unrelated." + }, + { + "left": "issue:44336", + "right": "issue:46424", + "accept": false, + "reason": "Terminal ANSI output and cache-length configuration are unrelated." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "Tokenizer correctness in v5 and resize_token_embeddings behavior are different bugs." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "A buffer/position_ids error and a performance regression are not the same issue." + }, + { + "left": "issue:43475", + "right": "issue:46369", + "accept": false, + "reason": "SAM3 vision encoder missing attribute vs SAM3 litetext memory overflow are different failures." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Both are loss-related, but they affect different models and code paths." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "ImportError for a model processor and a DeepSpeed/Bert index error are unrelated." + }, + { + "left": "issue:46291", + "right": "issue:46399", + "accept": false, + "reason": "Both mention kernels, but one is a dependency version requirement and the other is a missing registration decorator; different bugs." + }, + { + "left": "issue:45859", + "right": "issue:46399", + "accept": false, + "reason": "CPU offload incompatibility and use_kernels registration failure are different Qwen3.5 problems." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Chunked-prefill cache handling and removal of is_torch_fx_available are different regressions." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Docs mismatch for return_all_scores and unsafe sys.modules access are unrelated." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma 4 token-type-id handling are different model-specific issues." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "A model not supporting Flash Attention 2 yet is not the same as the broad Flash Attention 2 topic." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model config and sys.modules access are unrelated." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MOE weight conversion and DeepSpeed ZeRO3 deque/index errors are different loading failures." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "resize_token_embeddings/output_embeddings and Whisper empty-transcription generation are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio processor sampling-rate defaulting and import latency are unrelated." + }, + { + "left": "issue:45412", + "right": "issue:46123", + "accept": false, + "reason": "RT-DETR memory leaks and MaskGenerationPipeline final-batch dropping are different bugs." + }, + { + "left": "issue:44561", + "right": "issue:46399", + "accept": false, + "reason": "Both are v5 regressions, but one is trust_remote_code breakage from a removed helper and the other is a Qwen3.5 kernels registration crash." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Unsafe sys.modules access and captured_info.txt race conditions 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": 21, + "review_unit_index": 39, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:42886", + "issue:43065", + "issue:43408", + "issue:43475", + "issue:43540", + "issue:43653", + "issue:43701", + "issue:43906", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44188", + "issue:44297", + "issue:44361", + "issue:44368", + "issue:44466", + "issue:44479", + "issue:44483", + "issue:44623", + "issue:44749", + "issue:44869", + "issue:44877", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45335", + "issue:45356", + "issue:45375", + "issue:45397", + "issue:45459", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45656", + "issue:45685", + "issue:45759", + "issue:45987", + "issue:46082", + "issue:46123", + "issue:46170", + "issue:46202", + "issue:46421", + "issue:46431" + ], + "soft_pairs": [ + "issue:45375|issue:45759", + "issue:44877|issue:45375", + "issue:45685|issue:46431", + "issue:44987|issue:45459", + "issue:45397|issue:45656", + "issue:44466|issue:45335", + "issue:44995|issue:45200", + "issue:45137|issue:45600", + "issue:43475|issue:43540", + "issue:43906|issue:46202", + "issue:42886|issue:45229", + "issue:44479|issue:46170", + "issue:43653|issue:44162", + "issue:44060|issue:44077", + "issue:44483|issue:46202", + "issue:44368|issue:44977", + "issue:43065|issue:43408", + "issue:44995|issue:45588", + "issue:44368|issue:45563", + "issue:44188|issue:46202", + "issue:44877|issue:46170", + "issue:45987|issue:46082", + "issue:43931|issue:44368", + "issue:44188|issue:46421", + "issue:43065|issue:46123", + "issue:44297|issue:44623", + "issue:45584|issue:46170", + "issue:44361|issue:44749", + "issue:44623|issue:45656", + "issue:44749|issue:45656", + "issue:44869|issue:45356", + "issue:43701|issue:44749" + ], + "prepared_review_unit_hash": "827441067ed47529d2259b67bce9f1f4c57295c6ac23ced32903ec23801f9f29", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37441, + "estimated_input_tokens": 9361, + "estimated_eval_tokens": 18978 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: tokenizer/save_pretrained/config regressions, SAM3/Qwen video issues, Gemma/DeepSpeed loading problems, and a few cache/compile bugs. I do not see a safe duplicate set among the soft pairs.", + "confidence": 0.18, + "canonical_issue_reason": "No single canonical issue is safe here; the cluster spans several unrelated bug families, and the strongest similarities are only broad subsystem overlap.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue is a good canonical representative for the whole cluster; at most, a few pairs form narrow subthemes, but they are not the same underlying defect.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Same Qwen config area, but different concrete bugs: missing field vs text_config unwrapping." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both are config-loading problems, but they affect different models and code paths." + }, + { + "left": "issue:45685", + "right": "issue:46431", + "accept": false, + "reason": "Unrelated failures: MPS histogram support vs missing chat template." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Both touch loading/tokenizer behavior, but the root causes are different." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Both mention DeepSpeed-related behavior, but they are separate loading/optimizer bugs." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both involve embedding tying/resize behavior, but in different models and mechanisms." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Cache state corruption is unrelated to Gemma4 token-type defaults." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "Checkpoint resume failure is unrelated to auto_mappings cleanup." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Both are video/model issues, but they fail in different components and ways." + }, + { + "left": "issue:43906", + "right": "issue:46202", + "accept": false, + "reason": "Different model reproductions with unrelated decoding/tokenization behavior." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Tokenizer offline cache bug and Gemma4 OOM are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:46170", + "accept": false, + "reason": "Both are regressions, but one is video-input handling and the other is Windows/ROCm import dependency." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer special-token registration and ESM2 breakage are different bugs." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Different features and failures: tied-weights warning vs optional post_init policy." + }, + { + "left": "issue:44483", + "right": "issue:46202", + "accept": false, + "reason": "Chat API request rejection is unrelated to the Gemma pipeline roundtrip issue." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "These are separate warnings/issues in different code paths." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Both SAM3-related, but one is a dummy layer and the other is a wrong-model-type instantiation warning." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Cache/indexer state bug is unrelated to flash-attention s_aux handling." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "Both are warning-related, but they concern different models and behaviors." + }, + { + "left": "issue:44188", + "right": "issue:46202", + "accept": false, + "reason": "Compile-time attention branching and Gemma tokenizer roundtrip are unrelated." + }, + { + "left": "issue:44877", + "right": "issue:46170", + "accept": false, + "reason": "Config strictness and Windows ROCm import dependency are different problems." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "StaticCache return-type bug and LlamaConfig head_dim validation are unrelated." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Model weight-shape mismatch is unrelated to the tie_word_embeddings warning." + }, + { + "left": "issue:44188", + "right": "issue:46421", + "accept": false, + "reason": "Both mention compile/cache behavior, but the concrete failures differ." + }, + { + "left": "issue:43065", + "right": "issue:46123", + "accept": false, + "reason": "SAM3 pixel-decoder dummy layer and mask batching logic are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Similar save_pretrained surface, but tokenizer_class mismatch and missing processor files are distinct defects." + }, + { + "left": "issue:45584", + "right": "issue:46170", + "accept": false, + "reason": "Whisper empty-transcription failure is unrelated to the ROCm import issue." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "Tokenizer AttributeError and post-upgrade slowdown are unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Processor serialization and duplicate optimizer steps are different bugs." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "Training slowdown after upgrade does not match the deepspeed step-count bug." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Whisper timestamp decoding and Kimi tokenizer regression are unrelated." + }, + { + "left": "issue:43701", + "right": "issue:44749", + "accept": false, + "reason": "Checkpoint key mismatch is unrelated to the tokenizer/filtering slowdown 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": 21, + "review_unit_index": 40, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:36010", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:42831", + "issue:42994", + "issue:43232", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43704", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43873", + "issue:44038", + "issue:44162", + "issue:44188", + "issue:44222", + "issue:44351", + "issue:44623", + "issue:44871", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45406", + "issue:45478", + "issue:45753", + "issue:46077", + "issue:46123", + "issue:46132", + "issue:46202", + "issue:46326", + "issue:46399" + ], + "soft_pairs": [ + "issue:43646|issue:44162", + "issue:43232|issue:43673", + "issue:45230|issue:45397", + "issue:44871|issue:46132", + "issue:44038|issue:44623", + "issue:42994|issue:43482", + "issue:43475|issue:45381", + "issue:43475|issue:43643", + "issue:28282|issue:30990", + "issue:42371|issue:45092", + "issue:41628|issue:44933", + "issue:45375|issue:45478", + "issue:45406|issue:46202", + "issue:42994|issue:43525", + "issue:43723|issue:46326", + "issue:45310|issue:45375", + "issue:43784|issue:46399", + "issue:42673|issue:43704", + "issue:45081|issue:45103", + "issue:43873|issue:44945", + "issue:43232|issue:44188", + "issue:44964|issue:45137", + "issue:41628|issue:44351", + "issue:44945|issue:45753", + "issue:44222|issue:45310", + "issue:42994|issue:43824", + "issue:46077|issue:46123", + "issue:44038|issue:45310", + "issue:36010|issue:41628", + "issue:42831|issue:44222" + ], + "prepared_review_unit_hash": "d742f5fee4e7ad83b573b548343f587cdf9be1214229485a57abd5d18277cdc1", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36348, + "estimated_input_tokens": 9087, + "estimated_eval_tokens": 18430 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are only loosely related by broad Transformer/model-family themes, not true duplicates. The strongest confirmed duplicate in the soft edges is the Qwen3ForCausalLM VRAM leak pair (42673/43704). The packet also contains another exact-title duplicate pair, issue:45310 and issue:45478, so the cluster is not dominated by a single unique duplicate case.", + "confidence": 0.89, + "canonical_issue_reason": "issue:42673 is the earlier report in the exact duplicate pair with issue:43704, and both describe the same Qwen3ForCausalLM VRAM leak symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the strongest issue representative because it is the earliest exact duplicate match for the Qwen3ForCausalLM VRAM leak and has the most canonical wording.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Both are Transformers v5 regressions, but they affect different model paths and symptoms." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Generation-related, but sync_gpus handling and chunked_prefill cache behavior are different bugs." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Generic bug report vs Gemma4 ZeRO3 from_pretrained failure; not the same defect." + }, + { + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "EOS token config mismatch and AttentionInterface.register behavior change are different issues." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Different concrete failures: Qwen3-VL-Moe v5 breakage vs processor.save_pretrained missing files." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are distinct code paths." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Different model families and failure modes: SAM 3 output attribute error vs qwen2.5-vl video position IDs." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "SAM 3 Video attribute error and AutoConfig missing fields under trust_remote_code are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "Missing PyTorch ImportError and Sentence Transformers getting stuck are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API settings and remote-code meta-init incompatibility are different problems." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Both are import errors, but for different symbols and different export paths." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but missing config field vs from_pretrained error is not enough to treat as one bug." + }, + { + "left": "issue:45406", + "right": "issue:46202", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer and Gemma pipeline roundtrip failure are different code paths." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized model saving failure and Llama4Config missing pad_token_id are different bugs." + }, + { + "left": "issue:43723", + "right": "issue:46326", + "accept": false, + "reason": "Tokenizer loading in v5 and continue_final_message behavior are distinct issues." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Related Qwen3.5 issues, but one is a from_pretrained error and the other is a missing config field." + }, + { + "left": "issue:43784", + "right": "issue:46399", + "accept": false, + "reason": "Sentence-transformers import NameError and apply_rotary_pos_emb registration failure are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact same title and same VRAM leak symptom; clearly the same underlying bug." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Different functions and different contexts: mistral regex patch vs auto-docstring annotation handling." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Offloading with quantization and incorrect pipeline-parallel output are distinct failures." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Different generation/attention paths with no concrete evidence of the same defect." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Phi-4 multimodal load failure and DeepSpeed ZeRO3 deque error are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import/export regressions, but they concern different missing names and APIs." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Incorrect pipeline-parallel output and Qwen3_5 infinite loop on one image are different bugs." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "FP8 save_pretrained and Qwen3.5 MoE from_pretrained are opposite directions and not the same issue." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving and Qwen2.5-VL import failure are unrelated." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Gemma4 attention-mask construction and final-partial-batch handling in MaskGenerationPipeline are different components." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Both involve Qwen and Transformers v5, but the concrete failures differ." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "GenerationMixin import failure and AutoImageProcessor import failure are separate missing-symbol regressions." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues and FP8 save_pretrained are not the same bug." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the summary overstates uniqueness. The packet contains another obvious duplicate pair with the exact same title, issue:45310 and issue:45478, so it is not grounded to say the Qwen3ForCausalLM VRAM leak is the only clear duplicate. The 42673/43704 canonical choice is plausible, but the overall 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": 1, + "component_count": 21, + "review_unit_index": 41, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:42175", + "issue:42222", + "issue:42898", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43504", + "issue:43526", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43976", + "issue:44016", + "issue:44060", + "issue:44075", + "issue:44079", + "issue:44222", + "issue:44242", + "issue:44295", + "issue:44297", + "issue:44448", + "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:45310", + "issue:45356", + "issue:45405", + "issue:45561", + "issue:45656", + "issue:45712", + "issue:45812", + "issue:46231" + ], + "soft_pairs": [ + "issue:43232|issue:44792", + "issue:42175|issue:45712", + "issue:44496|issue:45561", + "issue:43504|issue:45561", + "issue:42175|issue:44933", + "issue:43526|issue:45561", + "issue:43976|issue:44933", + "issue:42898|issue:44568", + "issue:43377|issue:44242", + "issue:43278|issue:45137", + "issue:30990|issue:35141", + "issue:44623|issue:45310", + "issue:30333|issue:33290", + "issue:43479|issue:44869", + "issue:42222|issue:45200", + "issue:42898|issue:45812", + "issue:43479|issue:44079", + "issue:43723|issue:44568", + "issue:42898|issue:44779", + "issue:43784|issue:43824", + "issue:44016|issue:46231", + "issue:42898|issue:45356", + "issue:43874|issue:45092", + "issue:44734|issue:45103", + "issue:44295|issue:45656", + "issue:44075|issue:46231", + "issue:42898|issue:44448", + "issue:44222|issue:44297", + "issue:44060|issue:45125", + "issue:45405|issue:45712", + "issue:44297|issue:45656", + "issue:44933|issue:45712" + ], + "prepared_review_unit_hash": "6708dd8dc5db9cd34b315ba62e88025c5b6e4257c0281fd1f607c2ffab23d7f8", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37503, + "estimated_input_tokens": 9376, + "estimated_eval_tokens": 19008 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated Transformers issues grouped by lexical similarity; I found no true duplicate pairs and no PRs in this cluster.", + "confidence": 0.93, + "canonical_issue_reason": "issue:42898 is the clearest representative of the repeated tokenizer/v5 regression theme, but the cluster is otherwise heterogeneous.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42898 has the most broadly framed, user-visible bug report among the mixed issues, so it is the best fallback representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Different bugs: generation kwargs sync handling vs a Janus image-generation test failure." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Unrelated: missing TensorFlow backend in a torch extra vs leftover dummy class cleanup." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrelated: model config loading error vs a pytest-xdist race in test utilities." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Different areas and failure modes: BEiT preset loading vs captured_info.txt race." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "No shared bug: backend dependency packaging vs a nonexistent import from image_utils." + }, + { + "left": "issue:43526", + "right": "issue:45561", + "accept": false, + "reason": "Different defects: BEiT image processor label reduction vs testing race condition." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Different problems: Python version compatibility vs import error in image_utils." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different behaviors and likely different code paths." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Different bugs: MIMI padding-mask inconsistency vs MoE load-balancing loss gating." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Different failures: embedding dtype drift vs DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Unrelated: Sentence Transformers load hang vs token-embedding resize/post_init reinit bug." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Different code paths: processor save_pretrained file omission vs Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "Different operational failures: MLflow reporting on failed training jobs vs OOM with Adafactor in DeepSpeed." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Unrelated model/config init bug vs Whisper word-timestamp decode crash." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Different multimodal model issues: broken VitPose models vs Gemma 4 mm_token_type_ids defaulting." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different models and distinct symptoms." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Different artifacts: multimodal config initialization vs ModelOutput key assignment behavior." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Related only by tokenizer vocabulary of v5; one is AutoTokenizer loading, the other add_special_tokens output behavior." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 tokenizer regressions, but the affected models and failure mechanisms differ." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Different import errors in different packages/models: sentence-transformers import vs Qwen2.5-VL symbol export." + }, + { + "left": "issue:44016", + "right": "issue:46231", + "accept": false, + "reason": "Different scopes: notebook syntax error vs documentation parameter-name mismatch." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Both mention tokenizer regressions, but Kimi-K2.5 codec handling is a separate bug from clean_up_tokenization_spaces." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Different multimodal loading failures: missing image-processor method vs legacy remote-code/meta-init incompatibility." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Different failures: serve KV-cache continuation indexing vs auto_docstring parsing under future annotations." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Different bugs: buffer registration/reading issue vs optimizer step being called twice with DeepSpeed." + }, + { + "left": "issue:44075", + "right": "issue:46231", + "accept": false, + "reason": "Different areas: optimizer args ignored vs documentation parameter mismatch." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Different tokenizer/output regressions: Pegasus v4-v5 output mismatch vs clean_up_tokenization_spaces change." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Unrelated: FP8 MoE save_pretrained vs tokenizer_class mismatch in tokenizer_config.json." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Different Qwen3-related issues: tied-weights warning bug vs missing tensor-parallel plan." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "Unrelated: PEFT version pin bump vs dummy class cleanup." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Different bugs: tokenizer save metadata vs DeepSpeed optimizer step duplication." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Unrelated import/export issue vs leftover dummy classes in the repo." + } + ] + }, + "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": 21, + "review_unit_index": 42, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28282", + "issue:30333", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:43122", + "issue:43232", + "issue:43475", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:44190", + "issue:44206", + "issue:44222", + "issue:44263", + "issue:44279", + "issue:44295", + "issue:44297", + "issue:44373", + "issue:44448", + "issue:44484", + "issue:44485", + "issue:44623", + "issue:44734", + "issue:44871", + "issue:44977", + "issue:44991", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45245", + "issue:45310", + "issue:45381", + "issue:45834", + "issue:46082", + "issue:46132", + "issue:46209", + "issue:46305", + "issue:46328", + "issue:46421", + "issue:46424", + "issue:46431" + ], + "soft_pairs": [ + "issue:27957|issue:30333", + "issue:44222|issue:44623", + "issue:30333|issue:34689", + "issue:44448|issue:46132", + "issue:42831|issue:44623", + "issue:43479|issue:43644", + "issue:30333|issue:36331", + "issue:46209|issue:46431", + "issue:30333|issue:35141", + "issue:22355|issue:28282", + "issue:45245|issue:46132", + "issue:44297|issue:45310", + "issue:43122|issue:46209", + "issue:43475|issue:46305", + "issue:41628|issue:42757", + "issue:45071|issue:46082", + "issue:43723|issue:44977", + "issue:45381|issue:46305", + "issue:44206|issue:45103", + "issue:43232|issue:46424", + "issue:44734|issue:46421", + "issue:44295|issue:44623", + "issue:44222|issue:44484", + "issue:44263|issue:45092", + "issue:43232|issue:45834", + "issue:44190|issue:46082", + "issue:44373|issue:44485", + "issue:43723|issue:44871", + "issue:43723|issue:44991", + "issue:44279|issue:46328" + ], + "prepared_review_unit_hash": "a24255677f9903317e6b86c1e803bbfc444ff7c24d090402dcf2e2e8b1b15478", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35873, + "estimated_input_tokens": 8969, + "estimated_eval_tokens": 18194 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is not a true duplicate cluster: the soft pairs share only shallow wording (imports, save_pretrained, cache, tokenizer, etc.) but describe different bugs or features. I would reject all soft edges and avoid picking a canonical issue from this batch.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical issue fits this set; the items span unrelated regressions, feature requests, and dependency/import problems, and the soft links are only superficial lexical matches.", + "canonical_pr_reason": null, + "best_issue_reason": "There is no strong global best issue representative because there is no shared underlying bug/change across the listed issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27957", + "right": "issue:30333", + "accept": false, + "reason": "Different problems: Flash Attention 2/model support vs MLFlow not reporting failed jobs." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "Both mention save_pretrained, but one is FP8/MoE saving behavior and the other is processor file persistence." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "Unrelated failures: MLFlow job state reporting vs Llama 3.2 vision model loading regression." + }, + { + "left": "issue:44448", + "right": "issue:46132", + "accept": false, + "reason": "Different code paths: Pegasus v4/v5 output change vs AttentionInterface.register changing callback behavior." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FP8 accuracy regression is unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Config default initialization bug vs non-persistent buffer junk during load; different mechanisms." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "Both are trainer-related, but one is MLFlow reporting after failure and the other is a signature mismatch in compute_loss." + }, + { + "left": "issue:46209", + "right": "issue:46431", + "accept": false, + "reason": "FP8 inference on B200 is unrelated to chat template availability in processor.save_pretrained/apply_chat_template." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "Different areas: MLFlow reporting vs embedding resize/post_init reinitialization." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Both are import errors, but one is missing transformers.onnx and the other is missing PyTorch for AutoModel." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "Category-limit RuntimeError is unrelated to AttentionInterface.register behavior changes." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer save metadata mismatch is different from a Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:43122", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer behavior change is unrelated to FP8 inference failure on B200." + }, + { + "left": "issue:43475", + "right": "issue:46305", + "accept": false, + "reason": "SAM 3 missing attribute during vision encoding is unrelated to a streaming mask off-by-one bug." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols from different packages and likely different dependency mismatches." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "Type-checking regression in v5.4.0 is unrelated to LlamaConfig head_dim validation." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading failure in v5 and Qwen3.5 flash-attention generation failure are different bugs." + }, + { + "left": "issue:45381", + "right": "issue:46305", + "accept": false, + "reason": "Both mention positional/window logic, but they affect different models and code paths; not the same bug." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Feature extractor arg handling and auto_docstring annotation parsing are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:46424", + "accept": false, + "reason": "Both touch generation/cache, but one is sync_gpus model kwargs handling and the other is cache length configuration." + }, + { + "left": "issue:44734", + "right": "issue:46421", + "accept": false, + "reason": "Serve-time KV-cache tensor indexing crash is not the same as lazy cache init causing recompilation." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Position_ids buffer access bug is unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 MoE save_pretrained behavior and max_shard_size default are different save_pretrained concerns." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "torch.split return semantics in a specific indexer are unrelated to remote-code checkpoint meta-initialization compatibility." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Both are cache/generation related in a broad sense, but they are distinct failures with different roots." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Dataset loading in an example script is unrelated to LlamaConfig head_dim validation." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "Wrong docstring is documentation-only and not the same as the GLM-5 RoPE implementation bug." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Tokenizer loading failure and inconsistent eos_token_id configuration are separate issues." + }, + { + "left": "issue:43723", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-related, but they concern different model-specific loading failures rather than the same defect." + }, + { + "left": "issue:44279", + "right": "issue:46328", + "accept": false, + "reason": "Generic dependency issue is unrelated to the SequenceFeatureExtractor.pad performance regression." + } + ] + }, + "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": 21, + "review_unit_index": 43, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28218", + "issue:30990", + "issue:38175", + "issue:39401", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43232", + "issue:43408", + "issue:43479", + "issue:43550", + "issue:43644", + "issue:43646", + "issue:43761", + "issue:43824", + "issue:44038", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44368", + "issue:44373", + "issue:44483", + "issue:44568", + "issue:44829", + "issue:44913", + "issue:45092", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45593", + "issue:45753", + "issue:46002", + "issue:46050", + "issue:46082", + "issue:46132", + "issue:46305", + "issue:46369", + "issue:46402" + ], + "soft_pairs": [ + "issue:44483|issue:45406", + "issue:44263|issue:44360", + "issue:42907|issue:46050", + "issue:30990|issue:38175", + "issue:45092|issue:45357", + "issue:43644|issue:46305", + "issue:44829|issue:46002", + "issue:42831|issue:45310", + "issue:44261|issue:45092", + "issue:42831|issue:44038", + "issue:43408|issue:46402", + "issue:41628|issue:43824", + "issue:44263|issue:44373", + "issue:43479|issue:44913", + "issue:39401|issue:44568", + "issue:43761|issue:46082", + "issue:22355|issue:27957", + "issue:42886|issue:43066", + "issue:42757|issue:45278", + "issue:27957|issue:30990", + "issue:42886|issue:42994", + "issue:22355|issue:28218", + "issue:44368|issue:46305", + "issue:42831|issue:45753", + "issue:43408|issue:46369", + "issue:43232|issue:45593", + "issue:44295|issue:44297", + "issue:43550|issue:46132", + "issue:43646|issue:46002" + ], + "prepared_review_unit_hash": "ec974d5f551e1961e2182c91e5df479081bc836bef29e278d3ed9351c2af3a5e", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35477, + "estimated_input_tokens": 8870, + "estimated_eval_tokens": 17996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly lexical noise: the issues span unrelated bugs, import errors, tokenizer quirks, multimodal regressions, quantization, and config problems. None of the soft pairs look like the same underlying defect or change.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the cluster is heterogeneous and the soft pairs do not converge on one underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No single issue is a safe representative for deduping; the items target different subsystems and root causes.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat/completions request handling and Gemma4Processor _tokenizer import are unrelated bugs." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "torch.split return values in GlmMoeDsaIndexer is unrelated to the DSA ReLU discussion." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Both involve saving/loading quantized models, but they describe different failures and code paths." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Sentence Transformers loading hang and SIGLIP2 zero probabilities are different problems." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "InternVL2 meta-init incompatibility and Qwen3.5 visual-key save regression are distinct issues." + }, + { + "left": "issue:43644", + "right": "issue:46305", + "accept": false, + "reason": "Non-persistent buffer junk and MoonshineStreaming window off-by-one are unrelated." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Flash Attention 3 training degeneration and NaNs after checkout/init are different failure modes." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy and Qwen3.5 MoE from_pretrained errors are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "MLA layernorm epsilon precision bug is unrelated to InternVL2 meta initialization." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues do not match the Qwen3-VL-Moe bug." + }, + { + "left": "issue:43408", + "right": "issue:46402", + "accept": false, + "reason": "sam3_tracker warning and sam3 text-encoder weights not loading are different regressions." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but they concern different missing classes and likely different fixes." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "torch.split return values and a wrong docstring for position_ids are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Phi4MultimodalConfig default subconfigs and GPTNeoXConfig rotary_pct reload are different config bugs." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Wrong offset_mapping and missing BOS/EOS insertion are separate tokenizer behaviors." + }, + { + "left": "issue:43761", + "right": "issue:46082", + "accept": false, + "reason": "CLIPVision hidden_states regression and LlamaConfig head_dim validation are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:27957", + "accept": false, + "reason": "Missing transformers.onnx module and XLMRoberta Flash Attention 2 are unrelated." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline cache loading and tokenizer decoder type mismatch are different tokenizer issues." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "General import errors after upgrade are broader, but not the same concrete bug as is_offline_mode import failure." + }, + { + "left": "issue:27957", + "right": "issue:30990", + "accept": false, + "reason": "Flash Attention 2 support and Sentence Transformers loading hang are unrelated." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline tokenizer cache loading and quantized model saving failure are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "ONNX import failure and added-token spacing bug are different tokenizer topics." + }, + { + "left": "issue:44368", + "right": "issue:46305", + "accept": false, + "reason": "tie_word_embeddings warning and MoonshineStreaming mask off-by-one are unrelated." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FineGrainedFP8 accuracy and Qwen3_5 infinite-loop-on-image are unrelated." + }, + { + "left": "issue:43408", + "right": "issue:46369", + "accept": false, + "reason": "sam3_tracker warning and sam3-litetext memory overflow are different bugs." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs after sync_gpus and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "position_ids buffer read error and tokenizer_class mismatch in config are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:46132", + "accept": false, + "reason": "Bamba SDPA torch.compile failure and AttentionInterface.register behavior change are different code paths." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Non-persistent buffer junk and NaNs in classification heads after init are not the same defect." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 44, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:22355", + "issue:29159", + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42907", + "issue:43116", + "issue:43122", + "issue:43408", + "issue:43452", + "issue:43475", + "issue:43576", + "issue:43756", + "issue:43784", + "issue:43824", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44188", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44360", + "issue:44373", + "issue:44485", + "issue:44488", + "issue:44514", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44779", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45290", + "issue:45478", + "issue:45600", + "issue:46326" + ], + "soft_pairs": [ + "issue:43576|issue:43784", + "issue:42371|issue:44263", + "issue:44117|issue:44704", + "issue:30990|issue:33357", + "issue:44188|issue:44373", + "issue:43756|issue:44060", + "issue:42907|issue:44945", + "issue:44261|issue:44373", + "issue:43122|issue:44521", + "issue:39401|issue:44488", + "issue:41628|issue:45278", + "issue:43122|issue:46326", + "issue:42907|issue:43937", + "issue:43976|issue:44945", + "issue:44704|issue:45290", + "issue:42907|issue:44964", + "issue:43784|issue:45600", + "issue:44038|issue:44297", + "issue:39401|issue:44521", + "issue:43116|issue:45200", + "issue:44360|issue:44485", + "issue:44038|issue:44295", + "issue:44625|issue:44829", + "issue:43408|issue:43475", + "issue:43122|issue:44514", + "issue:43122|issue:44779", + "issue:42907|issue:43452", + "issue:44038|issue:44222", + "issue:43824|issue:44351", + "issue:45216|issue:45478", + "issue:22355|issue:29159" + ], + "prepared_review_unit_hash": "a9e3233c155ca08402033b9b2799e91f50832fc6781a20a5cf00e167523f6952", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36649, + "estimated_input_tokens": 9163, + "estimated_eval_tokens": 18582 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The soft pairs are mostly false positives: they span unrelated tokenizer regressions, import errors, model-loading bugs, chat-template issues, and model-specific architecture problems. I would not merge any of the proposed pairs as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": "No single strong canonical issue emerges because the set is heterogeneous. If a representative is required, issue 43122 is the cleanest tokenizer-regression example, but it does not cover the rest of the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43122 is the most general and clearly described tokenization regression among the candidates, so it is the best single representative; however, the overall set is too broad to treat as one duplicate cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Different bugs: one is the `transformers env` CLI in v5, the other is a `NameError` when importing sentence-transformers." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "Unrelated topics: TF32 settings/API behavior versus a `torch.split()` indexing/return-value issue." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "Both involve loader paths, but one is `TOKENIZER_MAPPING_NAMES` returning None and the other is `AutoProcessor.from_pretrained` not forwarding kwargs; not the same code-path bug." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Different failure modes: Sentence Transformers loading hangs versus a MacOS bus error with a CLIP model." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "One is a torch.compile branching issue in attention kernels; the other is a wrong docstring for `position_ids`." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Distinct model-specific issues: Smollm3 RoPE layer handling versus Qwen3-Next tied-weights warning miswiring." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Different symptoms and subsystems: dequantized save failures versus incorrect output under pipeline parallelism." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "MLA layernorm precision/config issue versus a documentation-only `position_ids` bug." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer regression versus multimodal assistant_mask generation in `apply_chat_template`; related area, but not the same bug." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Different tokenizer problems: wrong `offset_mapping` for Qwen3 versus a specific model load failure for `cjvt/sleng-bert`." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Both are import-related, but one is missing `AutoImageProcessor` and the other is many import errors after a version upgrade; too broad and not the same fix." + }, + { + "left": "issue:43122", + "right": "issue:46326", + "accept": false, + "reason": "Both are tokenizer-related, but one is tokenization output drift across versions and the other is `continue_final_message` behavior for mistral_common tokenizers." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Different features: dequantized model saving versus GLM-5 `GenerationConfig` validation." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python-version compatibility problem versus pipeline-parallelism output corruption." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "`AutoProcessor.from_pretrained` kwarg forwarding is unrelated to assistant messages with tool calls crashing in `apply_chat_template`." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Saving dequantized models and loading a multimodal Phi-4 model are different code paths and symptoms." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Importing sentence-transformers with a `NameError` is unrelated to `auto_mappings.py` referencing removed Sam3LiteText configs." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe load bug versus tokenizer class metadata mismatch on `save_pretrained`; different bugs." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Wrong `offset_mapping` for Qwen3 is not the same as assistant mask generation for multimodal chat templates." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Multi-label classification example script bug versus Gemma 4 `mm_token_type_ids` defaulting; different affected APIs." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "Indexer missing a ReLU is unrelated to GLM-5 RoPE implementation details." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Qwen3-VL-Moe loading issue versus a `position_ids` buffer access error; not the same bug." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "Qwen3.5 `num_labels` propagation and flash_attention_3 training degradation are unrelated." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Different SAM 3 problems: model-type warning on instantiation versus missing `fpn_position_embeddings` attribute." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Both touch template/tokenization flows, but one is version-dependent tokenization drift and the other is a batched chat-template crash with `padding=False`." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer regressions affecting different models and behaviors; not the same underlying defect." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Dequantized save failure and `gguf_file` breaking `from_pretrained` are separate loader/saver issues." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Qwen3-VL-Moe load bug versus FP8 `save_pretrained` for moe models; different code paths." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different symbols (`Qwen2_5_VLForConditionalGeneration` vs `HybridCache`) and different fixes." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Saved-checkpoint regression for Qwen3.5 is not the same as a Qwen3.5 moe `from_pretrained` error." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "Completely unrelated: missing `transformers.onnx` module versus inconsistent slow/fast tokenizer behavior." + } + ] + }, + "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": 21, + "review_unit_index": 45, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:33453", + "issue:36296", + "issue:38617", + "issue:39401", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:43064", + "issue:43066", + "issue:43116", + "issue:43278", + "issue:43441", + "issue:43493", + "issue:43611", + "issue:43673", + "issue:43825", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44188", + "issue:44222", + "issue:44351", + "issue:44355", + "issue:44485", + "issue:44704", + "issue:44779", + "issue:44936", + "issue:44938", + "issue:44945", + "issue:45072", + "issue:45083", + "issue:45278", + "issue:45446", + "issue:45478", + "issue:45563", + "issue:45588", + "issue:45753", + "issue:46002", + "issue:46231", + "issue:46328" + ], + "soft_pairs": [ + "issue:42757|issue:44351", + "issue:36296|issue:43116", + "issue:43278|issue:44945", + "issue:42175|issue:44938", + "issue:43066|issue:44779", + "issue:43493|issue:45753", + "issue:22355|issue:29127", + "issue:43992|issue:44704", + "issue:42673|issue:45072", + "issue:38617|issue:45278", + "issue:43937|issue:44485", + "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:44075|issue:44936", + "issue:44038|issue:45478", + "issue:43976|issue:45446", + "issue:43116|issue:43825", + "issue:44222|issue:45478", + "issue:45083|issue:45446", + "issue:43064|issue:43673", + "issue:33453|issue:39401", + "issue:43937|issue:45563", + "issue:39401|issue:44779", + "issue:44117|issue:45588" + ], + "prepared_review_unit_hash": "d61bdc5604ddd74c2994f246c520f232c4d26ab0f8dc4000def0e981e16596cd", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 35562, + "estimated_input_tokens": 8891, + "estimated_eval_tokens": 18038 + }, + "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. None of the proposed soft edges look like the same underlying bug or change, so all should be rejected. There are no PRs in the set.", + "confidence": 0.96, + "canonical_issue_reason": "issue:45278 is the broadest, most representative problem statement in the set: a general import-regression report after upgrading versions. It is still not a duplicate of the other issues, but it is the best single umbrella-style issue here.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the most suitable representative issue because it describes a clear, high-level regression affecting multiple imports, unlike the narrower model- or task-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports from different packages/symbols: `is_offline_mode` from `huggingface_hub` vs `HybridCache` from `transformers`." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor-parallel training bug versus an example-script multi-label classification output issue; different code paths and symptoms." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "BF16-to-FP32 dtype change in evaluate is unrelated to incorrect output under pipeline parallelism." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Packaging/backend dependency omission for `pip install '.[torch]'` is unrelated to a Python 3.14 load failure." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Both mention tokenizers, but one is a decoder-type regression and the other is incorrect Deepseek tokenizer results; not the same bug." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "Different models and failures: SigLIP2 HF/JAX discrepancy versus Qwen3_5 infinite-loop behavior on a specific image." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "Module import failure for `transformers.onnx` is unrelated to a LayoutLMv3 error-message clarity issue." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Different loading paths: missing `embed_tokens.weight` in UMT5Encoder versus kwargs propagation in `AutoProcessor.from_pretrained`." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM use is unrelated to bfloat16 dtype mismatch CI/inference failures." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "A single missing symbol import is not the same as a broad upgrade-induced wave of import errors." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "GenerationConfig validation failure and GLM-5 RoPE discussion are different GLM-5 issues." + }, + { + "left": "issue:44188", + "right": "issue:46328", + "accept": false, + "reason": "Torch.compile attention-kernel divergence is unrelated to SequenceFeatureExtractor padding performance." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled Python file errors and a documentation parameter-name mismatch are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "Base-model-prefix loading regression is unrelated to NaNs in classification heads after checkout/init." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Example-script multi-label classification empties and SigLIP2 implementation discrepancy are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 behavior warning/documentation issue is unrelated to a missing ImageProcessor method causing an AttributeError." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are different model-specific bugs." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "SGD optimizer arguments not being used is unrelated to `trainer.evaluate()` failing after `trainer.train()`." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Different Qwen3 MoE variants and different failures: one is a broad v5 bug report, the other is a from_pretrained error for Qwen3.5 Moe." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python version support regression and a PyTorch version-check bug for AuxRequest import are not the same issue." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Multi-label classification example output bug is unrelated to a pipeline translation-task support/error-message issue." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "FP8 save_pretrained behavior for MoE models is not the same as a Qwen3.5 Moe loading error." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Qwen3_omni_moe feature-length helper behavior is unrelated to the flex_attention AuxRequest version check." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "Wrong optimizer states under FSDP2/PEFT/cpu_ram_efficient_loading is unrelated to missing GenerationMixin cache in chunked prefill." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Tokenizer-loading regression is too broad to be the same as a specific wrong offset_mapping bug." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "Invalid GenerationConfig and a stale num_return_sequences warning in paged generate are distinct generation issues." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer correctness bugs, but they affect different models and different behaviors (`offset_mapping` vs general incorrect results)." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Tokenizer mapping fallback bug and flash_attention crashing on `s_aux=None` are unrelated code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 46, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42490", + "issue:42617", + "issue:42673", + "issue:43122", + "issue:43408", + "issue:43421", + "issue:43450", + "issue:43582", + "issue:43673", + "issue:43742", + "issue:43827", + "issue:43906", + "issue:43937", + "issue:43986", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44291", + "issue:44380", + "issue:44410", + "issue:44466", + "issue:44484", + "issue:44496", + "issue:44829", + "issue:44877", + "issue:44933", + "issue:44977", + "issue:45071", + "issue:45081", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45362", + "issue:45405", + "issue:45446", + "issue:45561", + "issue:46002", + "issue:46077", + "issue:46178", + "issue:46231", + "issue:46298", + "issue:46328" + ], + "soft_pairs": [ + "issue:44380|issue:46077", + "issue:43827|issue:45446", + "issue:42673|issue:43906", + "issue:43450|issue:43986", + "issue:43673|issue:46328", + "issue:44466|issue:46002", + "issue:44291|issue:45071", + "issue:43937|issue:44977", + "issue:45405|issue:45446", + "issue:44117|issue:45081", + "issue:45341|issue:45561", + "issue:44410|issue:45310", + "issue:43122|issue:46298", + "issue:45341|issue:46328", + "issue:38617|issue:41628", + "issue:44829|issue:45216", + "issue:42222|issue:46178", + "issue:42617|issue:45341", + "issue:45362|issue:46328", + "issue:43408|issue:43742", + "issue:39401|issue:43421", + "issue:42673|issue:43582", + "issue:42490|issue:42673", + "issue:44877|issue:46231", + "issue:42371|issue:44261", + "issue:44484|issue:45310", + "issue:38617|issue:44220", + "issue:43937|issue:44496", + "issue:44933|issue:45278" + ], + "prepared_review_unit_hash": "b788be39585c367334634831900814a5d6a2a675e1cf91401fe051e235bce57e", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36265, + "estimated_input_tokens": 9067, + "estimated_eval_tokens": 18390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of unrelated transformer issues: import-regression reports, tokenizer/model-specific bugs, video/model loading failures, and a few docs/testing/perf nits. I did not find any convincing duplicate pairs; the closest thing to a hub is the broad import-regression report #45278.", + "confidence": 0.63, + "canonical_issue_reason": "No single true duplicate stands out across this mixed set; #45278 is the broadest representative of the import-regression subgroup.", + "canonical_pr_reason": null, + "best_issue_reason": "#45278 is the most general, high-signal issue here and the best umbrella candidate if one issue had to represent the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "Different concrete bugs: GPT2 attention scaling under SDPA/FlashAttention vs Gemma4 building the wrong attention mask." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs cleanup about pipeline() references is unrelated to a PyTorch version check in flex_attention." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "One is a Qwen3 VRAM leak in multi-threaded loading; the other is only a reproduction of a different issue." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video-shape handling and missing-torchvision import failure are different code paths." + }, + { + "left": "issue:43673", + "right": "issue:46328", + "accept": false, + "reason": "Generation cache during chunked prefill is unrelated to feature-extractor padding conversion overhead." + }, + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "Serialization of tied weights is not the same bug as NaNs after checkout/init." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Unexpected init_empty_weights arguments and PretrainedConfig type-checking are distinct regressions." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "GLM-5 generation-config validation and Qwen3.5 flash-attention generation failure are different problems." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "A PEFT minimum-version bump is unrelated to the flex_attention import-version guard." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and Mistral regex patch crashes are different tokenizer-loading bugs." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "A generic testing_utils bug is not the same concrete race on captured_info.txt under xdist." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Qwen3Next missing projections and Qwen3.5 MoE from_pretrained failures are different model/config issues." + }, + { + "left": "issue:43122", + "right": "issue:46298", + "accept": false, + "reason": "Tokenization differences across versions are unrelated to a noisy logging message." + }, + { + "left": "issue:45341", + "right": "issue:46328", + "accept": false, + "reason": "testing_utils is unrelated to SequenceFeatureExtractor padding performance." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they point to different missing symbols and no shared code path is evident." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 training degeneration and Qwen3.5 checkpoint serialization are different failures." + }, + { + "left": "issue:42222", + "right": "issue:46178", + "accept": false, + "reason": "Vitpose model loading and a BACKENDS_MAPPING backend-name error are unrelated." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "3d_parallel.py runtime failure is not the same as a testing_utils issue." + }, + { + "left": "issue:45362", + "right": "issue:46328", + "accept": false, + "reason": "Qwen3.5 chat crashes and a padding-performance issue are different code paths." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "A model-type warning for sam3_* is unrelated to a key error loading MobileLLM-125M." + }, + { + "left": "issue:39401", + "right": "issue:43421", + "accept": false, + "reason": "Wrong offset_mapping and runtime post-processor synchronization after special-token edits are different tokenizer behaviors." + }, + { + "left": "issue:42673", + "right": "issue:43582", + "accept": false, + "reason": "A Qwen3 VRAM leak is unrelated to an AppleSilicon caching_allocator_warmup TypeError." + }, + { + "left": "issue:42490", + "right": "issue:42673", + "accept": false, + "reason": "Trailing-slash local-model import warnings and Qwen3 VRAM leaks do not match." + }, + { + "left": "issue:44877", + "right": "issue:46231", + "accept": false, + "reason": "Strict config loading and a documentation parameter-name mismatch are not the same bug." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 API guidance and RMS norm epsilon precision mismatch are different issues." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "A shard-size question is unrelated to a Qwen3.5 from_pretrained error." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "Configuration import failure and an fbank extraction issue are unrelated." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Invalid GenerationConfig and unrecognized model config are different loader/generation failures." + }, + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "Both mention imports, but one is a specific missing image_utils import while the other is a broad post-upgrade import-regression report; not enough to treat as the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 47, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:27957", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42831", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43335", + "issue:43352", + "issue:43441", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43673", + "issue:43874", + "issue:43931", + "issue:43986", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44360", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44829", + "issue:44933", + "issue:44987", + "issue:45003", + "issue:45071", + "issue:45341", + "issue:45372", + "issue:45440", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45656", + "issue:45753", + "issue:46123", + "issue:46172", + "issue:46231", + "issue:46369" + ], + "soft_pairs": [ + "issue:43066|issue:43122", + "issue:42371|issue:42831", + "issue:44410|issue:45478", + "issue:44075|issue:45656", + "issue:42222|issue:45542", + "issue:44483|issue:45464", + "issue:43097|issue:44493", + "issue:43012|issue:43577", + "issue:42175|issue:42222", + "issue:43673|issue:45341", + "issue:41628|issue:45071", + "issue:44263|issue:44485", + "issue:43874|issue:44261", + "issue:45071|issue:45542", + "issue:27957|issue:43352", + "issue:44263|issue:46172", + "issue:44360|issue:46172", + "issue:44484|issue:45478", + "issue:43931|issue:44410", + "issue:43441|issue:44829", + "issue:43931|issue:44038", + "issue:43097|issue:43452", + "issue:43335|issue:45071", + "issue:45003|issue:45341", + "issue:44933|issue:46231", + "issue:45753|issue:46369", + "issue:43986|issue:45341", + "issue:43673|issue:46123", + "issue:44117|issue:44987", + "issue:43986|issue:44484", + "issue:43874|issue:44263", + "issue:45440|issue:46172", + "issue:43531|issue:45341", + "issue:44261|issue:44485", + "issue:43986|issue:45372" + ], + "prepared_review_unit_hash": "744bbc464c8edf75794f5562988c9c8b324c10f7fefe275f81d091926eb2460e", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37770, + "estimated_input_tokens": 9443, + "estimated_eval_tokens": 19142 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are thematically related to Transformers regressions, but the soft pairs mostly point to different bugs, models, or code paths rather than true duplicates. I would reject every soft edge and treat this as a heterogeneous set, not a single dedup cluster.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Both concern tokenizer behavior in v5, but one is about an incorrect decoder type and the other about changed tokenization output; likely related, not the same underlying bug." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 API guidance and FineGrainedFP8 accuracy are different precision/configuration problems with no shared concrete failure mode." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "One is a missing layer-0 projection bug in qwen3next; the other is a from_pretrained error for Qwen3.5 MoE. Same family, different code-path issue." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "SGD arguments being ignored is about optimizer parameter handling; double optimizer.step() under DeepSpeed is a separate training loop bug." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "VitPose model breakage and missing TensorFlow backend when only tensorboard is installed are unrelated package/model loading issues." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Both involve chat/completions, but one is a generic request rejection and the other is a streaming inference failure on a specific Qwen3.5 model." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removing tie_embeddings_and_encoder_decoder and unexpected position-id keys are different breaking changes with different affected code paths." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both mention dtype/precision, but one is a PyTorch warning during bfloat16 compilation while the other is a BLIP2 loading bug leaving Qformer in float32." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "A TensorFlow backend/install issue is not the same bug as VitPose model breakage." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Generation cache missing during chunked prefill and a testing_utils bug are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "AutoImageProcessor import failure and PretrainedConfig type checking regression affect different APIs and code paths." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "A torch.split return-value issue in one GLM component is not the same as the GLM-5 RoPE implementation bug." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Missing get_number_of_image_patches in an image processor and rms_norm_eps precision mismatch are unrelated multimodal/model config bugs." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Type-checking regression in PretrainedConfig is unrelated to a missing TensorFlow backend dependency." + }, + { + "left": "issue:27957", + "right": "issue:43352", + "accept": false, + "reason": "Flash Attention 2 support issues for XLM-Roberta and Nemotron are separate model-specific capability gaps." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "DSA indexer output handling and deepseek-v4 comb.to(dtype).transpose(-1, -2) are different implementation bugs." + }, + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "A missing ReLU in the DSA indexer is unrelated to the deepseek-v4 tensor transpose/dtype issue." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "save_pretrained shard sizing and Qwen3.5 MoE loading failure are not the same bug." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Model weight-shape mismatch for Qwen3-VL-30B-A3B-Instruct and missing projections in qwen3next are different model-loading/model-architecture issues." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "FlashAttention failure in Ministral-3 and degenerate training with flash_attention_3 are different failures, even if both mention flash attention." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Both are Qwen3-VL-related, but one is a weight-shape mismatch and the other is a transformers 5.0 loading issue for Qwen3-VL-Moe; not the same concrete bug." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "A removed config attribute and gguf_file breaking from_pretrained are unrelated regression types." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction and PretrainedConfig type checking are separate configuration bugs." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils and a testing_utils bug do not share the same code path." + }, + { + "left": "issue:44933", + "right": "issue:46231", + "accept": false, + "reason": "A missing import from image_utils and a documentation parameter-name mismatch are unrelated." + }, + { + "left": "issue:45753", + "right": "issue:46369", + "accept": false, + "reason": "An infinite loop on a specific Qwen3_5 image and SAM3 LiteText memory overflow are distinct model/runtime failures." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "AutoProcessor crash without torchvision and a testing_utils bug are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Cache handling in GenerationMixin and MaskGenerationPipeline partial-batch dropping are different pipeline/generation bugs." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and a transformers>=5.1.0 model-loading failure are separate lookup/loading problems." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Video-model loading without torchvision and save_pretrained shard size behavior are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Different GLM/VL components and different failures; no shared concrete bug." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeekV3MoE divergence from remote implementation and deepseek-v4 tensor handling are different model-specific issues." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and a testing_utils bug are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Missing rms_norm_eps in MLA q_a_layernorm and GLM-5 RoPE implementation are separate precision/architecture bugs." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "A crash due to missing torchvision and a mistral_common import error affecting Gemma 4 processor loading are different dependency 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": 21, + "review_unit_index": 48, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:38617", + "issue:42757", + "issue:42890", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43352", + "issue:43502", + "issue:43550", + "issue:43824", + "issue:43825", + "issue:43873", + "issue:43906", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44077", + "issue:44117", + "issue:44162", + "issue:44315", + "issue:44351", + "issue:44368", + "issue:44380", + "issue:44410", + "issue:44484", + "issue:44655", + "issue:44704", + "issue:44936", + "issue:44938", + "issue:45103", + "issue:45375", + "issue:45479", + "issue:45584", + "issue:45588", + "issue:45712", + "issue:45753", + "issue:46050", + "issue:46132", + "issue:46172", + "issue:46178", + "issue:46298", + "issue:46365", + "issue:46421" + ], + "soft_pairs": [ + "issue:43097|issue:45588", + "issue:43097|issue:46050", + "issue:38617|issue:42757", + "issue:43064|issue:44315", + "issue:43502|issue:43992", + "issue:44062|issue:44117", + "issue:44410|issue:45375", + "issue:42890|issue:46298", + "issue:44038|issue:44410", + "issue:38617|issue:44351", + "issue:43097|issue:43873", + "issue:44936|issue:46365", + "issue:44704|issue:45103", + "issue:44380|issue:46421", + "issue:43986|issue:44117", + "issue:43824|issue:43986", + "issue:44938|issue:46178", + "issue:43066|issue:43906", + "issue:43992|issue:44484", + "issue:43208|issue:43550", + "issue:43012|issue:43550", + "issue:45712|issue:46298", + "issue:43825|issue:44655", + "issue:42907|issue:44655", + "issue:43012|issue:44368", + "issue:42907|issue:45753", + "issue:43097|issue:46298", + "issue:43352|issue:44380", + "issue:43352|issue:43976", + "issue:44162|issue:46172", + "issue:45584|issue:46365", + "issue:44380|issue:46132", + "issue:44380|issue:45479", + "issue:44077|issue:46365", + "issue:43064|issue:44380" + ], + "prepared_review_unit_hash": "de9d4bfc1473316aacd6b3568ab8ae3c80be3aefdb954aa1259ed9022b13b361", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38093, + "estimated_input_tokens": 9524, + "estimated_eval_tokens": 19304 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a false merge: the items span unrelated import errors, model-specific bugs, training regressions, and warnings. I found no soft-edge pair that looks like the same underlying fix/change, so all candidate links should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 43097 is the strongest representative only by prominence and external references; it is a broad Transformers v5 regression report, while the rest of the items are mostly unrelated bugs. There is no true duplicate canonical across the whole set.", + "canonical_pr_reason": null, + "best_issue_reason": "43097 is the best single issue to stand in for the cluster because it has the broadest user impact and the most linkage, but it should not be treated as a duplicate target for the other issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Different bugs: removed tie_embeddings API vs flash_attention crash on s_aux=None." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Different areas: API removal regression vs broken quantized save/load after sinq." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and modules; not the same failure." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Different training problems: FSDP2/PEFT optimizer-state mismatch vs Liger Kernel not applied with model_init." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Different code paths: local_files_only still making requests vs missing embed_tokens.weight on load." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "Tokenizer construction TypeError vs tokenizer mapping returning None; not the same bug." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen issues: missing projections in a model layer vs config field dropped by @strict." + }, + { + "left": "issue:42890", + "right": "issue:46298", + "accept": false, + "reason": "Unrelated: flaky test seeding issue vs noisy logging message." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Different Qwen model bugs with different symptoms and code paths." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Separate missing imports for different names; not a single underlying export bug." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "API removal regression vs quantization/offloading behavior; too different to merge." + }, + { + "left": "issue:44936", + "right": "issue:46365", + "accept": false, + "reason": "Trainer evaluation failure after train vs import failure for Trainer due to datasets; unrelated." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "AutoProcessor kwargs forwarding vs docstring annotation crash; different subsystems." + }, + { + "left": "issue:44380", + "right": "issue:46421", + "accept": false, + "reason": "Attention backend config ignored vs cache lazy-init recompilation; not the same fix." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "AutoProcessor/torchvision crash vs tokenizer mapping None handling; unrelated." + }, + { + "left": "issue:43824", + "right": "issue:43986", + "accept": false, + "reason": "Import error for a specific VL class vs video-model crash without torchvision; different causes." + }, + { + "left": "issue:44938", + "right": "issue:46178", + "accept": false, + "reason": "Python 3.14 load failure vs invalid BACKENDS_MAPPING backend error; unrelated." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Wrong tokenizer decoder type vs isolated reproduction of an older issue; not the same concrete bug." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing weight during load vs max_shard_size default question; different concerns." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training bugs vs Bamba compile/SDPA failure; different models and failures." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "PyTorch warning on bfloat16 compile vs model compile failure under SDPA; not duplicate." + }, + { + "left": "issue:45712", + "right": "issue:46298", + "accept": false, + "reason": "Dummy class cleanup vs noisy logging; different maintenance issues." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Pipeline error-message wording vs inability to save Pipeline objects; distinct problems." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Dequantized save failure vs Pipeline save_pretrained issue; unrelated save paths." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Different warnings with different triggers: bfloat16 precision warning vs tie_word_embeddings warning." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Different model-specific bugs: save/dequantization failure vs infinite loop on a specific image." + }, + { + "left": "issue:43097", + "right": "issue:46298", + "accept": false, + "reason": "API removal regression vs noisy logging; no shared bug." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Unsupported Flash Attention 2.0 for a model vs GPT2 attention scaling ignored under backends; different issues." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Model-level attention backend support vs Python-version compatibility failure; unrelated." + }, + { + "left": "issue:44162", + "right": "issue:46172", + "accept": false, + "reason": "ESM2 broken vs deepseek-v4 tensor transpose issue; different models and code paths." + }, + { + "left": "issue:45584", + "right": "issue:46365", + "accept": false, + "reason": "Whisper empty transcription bug vs Trainer import blocked by datasets; unrelated." + }, + { + "left": "issue:44077", + "right": "issue:46365", + "accept": false, + "reason": "patchtsmixer post_init policy vs Trainer import conflict; different problems." + }, + { + "left": "issue:44380", + "right": "issue:46132", + "accept": false, + "reason": "Ignored attention scaling config vs AttentionInterface.register behavior change; not the same code-path bug." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention backend config bug vs zero-loss classification problem; unrelated." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "Distributed optimizer-state corruption vs attention scaling backend bug; different subsystems 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": 21, + "review_unit_index": 49, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:36010", + "issue:36683", + "issue:38617", + "issue:42175", + "issue:42490", + "issue:42757", + "issue:42994", + "issue:43012", + "issue:43066", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43502", + "issue:43526", + "issue:43582", + "issue:43611", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43867", + "issue:43901", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44230", + "issue:44380", + "issue:44483", + "issue:44485", + "issue:44514", + "issue:44625", + "issue:44655", + "issue:44683", + "issue:44938", + "issue:45216", + "issue:45276", + "issue:45290", + "issue:45593", + "issue:45600", + "issue:45698", + "issue:46077", + "issue:46082", + "issue:46172", + "issue:46209", + "issue:46298", + "issue:46365" + ], + "soft_pairs": [ + "issue:36683|issue:43582", + "issue:44483|issue:44514", + "issue:42175|issue:46365", + "issue:43208|issue:43526", + "issue:42490|issue:43012", + "issue:43012|issue:44683", + "issue:38617|issue:43502", + "issue:42757|issue:43824", + "issue:45600|issue:46298", + "issue:44038|issue:44230", + "issue:45698|issue:46172", + "issue:43618|issue:44380", + "issue:43618|issue:43761", + "issue:44483|issue:45290", + "issue:43299|issue:43867", + "issue:43066|issue:46209", + "issue:36683|issue:43475", + "issue:44625|issue:46365", + "issue:42994|issue:44655", + "issue:43756|issue:44485", + "issue:45276|issue:46077", + "issue:43901|issue:44655", + "issue:36010|issue:42757", + "issue:43352|issue:43937", + "issue:43611|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" + ], + "prepared_review_unit_hash": "cb2a89f2931fcc35ec35a91483d27572086d280ebc09fe2e9ea9c6e5dbe71ae3", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36539, + "estimated_input_tokens": 9135, + "estimated_eval_tokens": 18526 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are unrelated standalone bugs, mostly in Transformers models, config, or packaging. The only plausible duplicate pair is the CLIP v5 regression around missing optional outputs in `CLIPVisionModel.forward`.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43761 is the clearest umbrella report for the CLIP v5 output regression: it explicitly ties the failure to `CLIPVisionModel.forward`, notes it is a v5 regression, and states the expected behavior when `output_hidden_states=True`.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43761 is the best representative issue among the plausible duplicates because it is explicit, regression-focused, and broad enough to cover the paired CLIP output symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "Different bugs in unrelated areas: Gemma3 config missing `vocab_size` vs Apple Silicon allocator warmup TypeError." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Both relate to chat/VL workflows, but one is API request validation and the other is a chat-template crash on batched input; not the same code-path bug." + }, + { + "left": "issue:42175", + "right": "issue:46365", + "accept": false, + "reason": "Separate packaging/import problems with different symptoms and triggers." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "Unrelated model/training bugs; no shared underlying change." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "Both mention warnings, but one is a local-path import warning and the other is a PyTorch precision warning; different fixes." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "Precision warning vs compiled flex-attention failure; distinct issues." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "ImportError for a missing utility vs incorrect network access despite `local_files_only=True`." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different symbols, modules, and breakages; not one bug." + }, + { + "left": "issue:45600", + "right": "issue:46298", + "accept": false, + "reason": "Breaking CI from removed config references vs a noisy logging message." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Same model family, but loading regression and FP8 inference support are different problems." + }, + { + "left": "issue:45698", + "right": "issue:46172", + "accept": false, + "reason": "save/load custom-module corruption vs a DeepSeek tensor operation bug; unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIP attentions regression vs GPT-2 attention scaling with SDPA/FlashAttention; different code paths." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": true, + "reason": "Both point to the same CLIP v5 forward-output regression where optional outputs are no longer populated; one PR could plausibly restore both attentions and hidden states." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Both involve chat-related APIs, but the failures are different: endpoint request rejection vs template crash with tool-call assistant messages." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Different model-loading failures with no clear shared root cause." + }, + { + "left": "issue:43066", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer decoder regression vs FP8 inference on B200; unrelated." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Different missing-attribute bugs in different models/configs." + }, + { + "left": "issue:44625", + "right": "issue:46365", + "accept": false, + "reason": "Qwen3.5 config propagation bug vs Trainer import issue; unrelated." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Both involve saving, but quantized model saving and Pipeline save_pretrained are distinct behaviors and fixes." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both mention RoPE, but they concern different models and different defects." + }, + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Same Gemma4 family, but token embedding resizing and bidirectional attention masks are separate bugs." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Docs mismatch for `return_all_scores` vs pipeline serialization; unrelated." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different packages and symbols; not the same underlying breakage." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Flash Attention support gap vs invalid GenerationConfig error; different code paths." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Model loading with `base_model_prefix` vs saving Pipeline objects; unrelated." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "Gemma3 config bug vs a patchtsmixer post_init validation rule; different issues." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "State-dict sorting/load failure and a Qwen3.5 save_pretrained regression are both serialization-adjacent but not the same bug." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version compatibility reports, but for different versions and likely different breakpoints." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "Qwen config propagation bug vs Gemma4 attention mask bug; unrelated." + }, + { + "left": "issue:43352", + "right": "issue:45593", + "accept": false, + "reason": "Flash Attention not supported for Nemotron vs D-FINE auxiliary-loss behavior; unrelated." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Different Gemma/Llama config behaviors and different failure modes." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet. The only accepted pair, issue:43618 and issue:43761, is a plausible single CLIPVisionModel.forward v5 regression affecting optional outputs, and the other soft-edge rejections stay conservative rather than overclaiming duplicate status. The overall characterization of the remaining items as unrelated standalone bugs is supported by their titles and descriptions." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 50, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:28282", + "issue:33453", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42890", + "issue:43010", + "issue:43064", + "issue:43065", + "issue:43502", + "issue:43525", + "issue:43575", + "issue:43618", + "issue:43723", + "issue:43867", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44075", + "issue:44077", + "issue:44112", + "issue:44222", + "issue:44246", + "issue:44355", + "issue:44360", + "issue:44387", + "issue:44492", + "issue:44556", + "issue:44625", + "issue:44855", + "issue:44857", + "issue:44938", + "issue:44964", + "issue:45071", + "issue:45125", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45561", + "issue:45600", + "issue:45850", + "issue:46202", + "issue:46285", + "issue:46365" + ], + "soft_pairs": [ + "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:46202", + "issue:43867|issue:45375", + "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", + "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:44387|issue:44556", + "issue:44222|issue:45125", + "issue:43010|issue:43065" + ], + "prepared_review_unit_hash": "0f58f8d9bcd99b4302e16e3a6d46af2a95d9320745371ef344f490edd90f00d7", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36822, + "estimated_input_tokens": 9206, + "estimated_eval_tokens": 18668 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a grab bag of unrelated Transformers issues. The only plausible duplicate pair is the import-time slowdown report (41669/44246); the rest are distinct model-, config-, loading-, training-, or test-specific bugs.", + "confidence": 0.82, + "canonical_issue_reason": "issue:41669 is the most concrete root-cause report in the set: it explains a specific import-time performance regression from `import *` usage, and it best matches the only plausible duplicate report here (issue:44246).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41669 is the best canonical issue for the small import-performance subcluster because it is specific, actionable, and closer to a root cause than the more generic symptom report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different model configs and different missing attributes (`vocab_size` vs `pad_token_id`)." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "Different bugs: missing `attentions` assignment in CLIP outputs vs Qwen3.5 save-pretrained checkpoint corruption." + }, + { + "left": "issue:36683", + "right": "issue:44625", + "accept": false, + "reason": "Both are config-related, but they concern different fields and model families (`vocab_size` vs `num_labels`)." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Different models and different loading failures (`UMT5Encoder` missing weights vs Qwen3.5 MoE from_pretrained error)." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Unrelated: distributed trainer optimizer-state bug vs a typo in cache strategies." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Different topics: `update()` no-grad decoration vs broad import errors after upgrade." + }, + { + "left": "issue:44075", + "right": "issue:46365", + "accept": false, + "reason": "Optimizer argument handling is unrelated to an import failure when `datasets` is installed." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Both involve import/parse failures, but one is a generic compiled-file error and the other is a Python 3.13 `torch.jit.script` parsing issue." + }, + { + "left": "issue:43867", + "right": "issue:46202", + "accept": false, + "reason": "Different loading/round-trip failures for different pipelines; no shared root cause is evident." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen/Gemma loading paths and different missing pieces (`state_dict` ordering vs a dropped vision config field)." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python 3.14 import/load breakage is not the same as `PretrainedConfig` type-checking regression." + }, + { + "left": "issue:36683", + "right": "issue:43064", + "accept": false, + "reason": "Different subsystems: config attribute error vs trainer/FSDP/PEFT optimizer-state corruption." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "Both describe the same import-time slowdown; 41669 identifies a concrete likely cause, while 44246 is the symptom report." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Separate test failures: missing `set_seed()` in SamHQ integration tests vs xdist race on `captured_info.txt`." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "One is a specific cache/layer update bug; the other is an overly broad complaint about many models breaking in v5." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Same general area of model loading, but different model families and no shared concrete failure mode." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "Different bugs in different Qwen3.5 save/load flows; not the same underlying issue." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Unrelated regression in `update()` handling vs a Python-version compatibility problem." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug vs tensor-parallel load OOM are related only at a broad subsystem level, not the same concrete defect." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "Different models and different issues: CLIP attentions output vs PatchTSMixer `post_init` API policy." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "One is a missing-PyTorch import error for AutoModel; the other is a tokenizer-loading regression." + }, + { + "left": "issue:43618", + "right": "issue:46285", + "accept": false, + "reason": "Different CLIP-related regressions with different symptoms; one is missing attentions, the other is a structural mismatch between CLIP text model classes." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "No shared root cause: CLIP output regression vs float16 AMP CUDA crash in another loss path." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "Different model/code-path problems: DSA indexer ReLU discussion vs DeepSeekV3MoE parity divergence." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "Unrelated: offline-loading API requests vs tokenizer `AddedToken` keyword collision." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Different failures in different tests/models; no evidence they are the same bug." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "CI test flakiness from missing seeds is unrelated to stale `auto_mappings.py` config references." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Both are CI/test issues, but they involve different tests and different failure causes." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Related only at a very high level (`from_pretrained`/reload regressions); the concrete failures differ." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Different problems: int4 quantization memory growth vs checkpoint reload compatibility." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "Same model family, but different concrete defects: FP8 save-pretrained behavior vs missing tensor-parallel plan." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Completely different bugs: cache/layer update semantics vs a dummy Conv2d in Sam3PixelDecoder." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The one accepted pair, issue:41669/issue:44246, is the only clearly supported duplicate-like match, and the remaining soft-edge rejections do not overclaim shared root causes." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 51, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:27957", + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:42548", + "issue:42757", + "issue:42831", + "issue:43010", + "issue:43232", + "issue:43299", + "issue:43502", + "issue:43526", + "issue:43576", + "issue:43673", + "issue:43824", + "issue:43827", + "issue:43874", + "issue:43901", + "issue:44222", + "issue:44230", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44655", + "issue:44749", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:45071", + "issue:45125", + "issue:45230", + "issue:45375", + "issue:45468", + "issue:45478", + "issue:46132", + "issue:46143", + "issue:46144", + "issue:46209" + ], + "soft_pairs": [ + "issue:43576|issue:46132", + "issue:43502|issue:44485", + "issue:42548|issue:44351", + "issue:45125|issue:45478", + "issue:44355|issue:44749", + "issue:30333|issue:33453", + "issue:45125|issue:45375", + "issue:43010|issue:43232", + "issue:43502|issue:44222", + "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:44222|issue:44230", + "issue:44908|issue:44961", + "issue:44485|issue:44961", + "issue:46143|issue:46144", + "issue:41628|issue:42548", + "issue:36296|issue:43901", + "issue:44315|issue:46209", + "issue:36296|issue:43526", + "issue:43673|issue:44492", + "issue:44623|issue:45230" + ], + "prepared_review_unit_hash": "b62d4a07638a52c2d8b554da8524cf1736dba8988c3b0734cfec74a40df71634", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 37696, + "estimated_input_tokens": 9424, + "estimated_eval_tokens": 19104 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated closed issues that happen to share broad Transformers/v5 or import/loading themes. I found no soft pair that looks like the same underlying bug or change, so all soft edges are rejected.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43576 is the best anchor for the cluster because it is a concrete v5 regression, has the most similarity links, and is representative of the broad \u201cTransformers v5 broke something\u201d subgroup.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43576 is the strongest representative issue in this set: it is specific enough to be actionable, yet broad enough to sit near several other v5-related loading/import regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Both touch attention-related behavior in Transformers v5, but one is a broken CLI/env command and the other is a change in AttentionInterface.register semantics; not the same bug." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "One is about unwanted network requests despite local_files_only=True; the other is a GLM-5 RoPE implementation issue. Different code paths and symptoms." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different symbols/modules (PreTrainedModel vs HybridCache). These are distinct API breakages." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Both concern Qwen3.5 MoE, but one is missing tensor-parallel planning and the other is a from_pretrained error in transformers>=5.4.0. Too different to merge." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Both mention loading/performance after upgrade, but one is compiled Python file errors and the other is data filtering slowdown. Not the same underlying defect." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLFlow reporting after failed training and tokenizer loading regression are unrelated subsystems and failure modes." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Both are Qwen3.5 MoE-related, but one is tensor-parallel support and the other is a config field being dropped by @strict. Different bugs." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "One is about no_grad decoration on cache/layer update methods; the other is generation kwargs handling after sync_gpus. Similar area, but not the same code-path problem." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "One is an unexpected network request issue; the other is an FP8 save_pretrained moe bug. Different operations and failure points." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Sentence Transformers loading hangs versus documentation confusion for input_ids/past_key_values. One is a runtime bug, the other is docs wording." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Both involve v5 model loading regressions, but they affect different model families and likely different serialization assumptions." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "The local_files_only network-request bug is unrelated to an issue titled only 'racoon'; there is no evidence of the same defect." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "Both are accuracy/processor-related, but one is FineGrainedFP8 accuracy and the other is a missing image-processor method causing AttributeError." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLFlow reporting on failure and tensor parallel training are separate concerns with no shared code path indicated." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled file loading errors and PretrainedConfig type-checking breakage are unrelated kinds of regressions." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs still referencing pipeline() and a one-word issue title do not indicate the same underlying bug." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled Python file errors and inability to save Pipeline objects are different problems in different parts of the API." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer loading regression and Llama 3.2 Vision model loading breakage are distinct loading failures affecting different components." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel not applied in model_init and a DSA indexer lacking ReLU are unrelated implementation issues." + }, + { + "left": "issue:27957", + "right": "issue:33453", + "accept": false, + "reason": "Flash Attention 2 support for XLMRoberta and a tokenizer loading regression are different feature/regression areas." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Importing is_offline_mode from huggingface_hub and a broken transformers env command are both v5-adjacent, but they are different missing-API vs CLI issues." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "A broken env command and a PretrainedConfig typing regression are unrelated user-facing regressions." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "The env command failure and a missing Qwen2.5-VL class import are different import/CLI breakages." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation confusion for model.forward and a missing layer_type_validation import are not the same bug." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "The 'racoon' issue provides no evidence of overlap with Gemma-4 audio positional encoding." + }, + { + "left": "issue:43576", + "right": "issue:44938", + "accept": false, + "reason": "A broken transformers env command and Python 3.14 load failures are both environment-related, but the concrete failures differ." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "Both mention FP8, but one is saving moe checkpoints and the other is inference support for specific Qwen3-VL/Qwen3.5 models." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Inverse_sqrt scheduler kwargs handling and the unrelated 'racoon' issue are not the same defect." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation and an unrelated placeholder title do not match." + }, + { + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "One is about **kwargs not being propagated in RoFormer methods; the other is that RoFormer attention does not use the attention interface. Related subsystem, but distinct bugs." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "AutoImageProcessor import failure and PreTrainedModel import failure are separate missing exports." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor parallel training bug and text-classification docs mentioning return_all_scores are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:46209", + "accept": false, + "reason": "Liger Kernel application in model_init and FP8 inference on B200 are different performance/runtime issues." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor parallel training and BeitImageProcessorFast reduce_labels behavior are different areas with no shared failing path." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Generation cache missing during chunked_prefill and a typo in cache strategies are not the same bug." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Missing files on processor.save_pretrained and a generic 'Bug report' have no evidence of being 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": 21, + "review_unit_index": 52, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:41669", + "issue:42371", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43827", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44230", + "issue:44246", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44393", + "issue:44492", + "issue:44556", + "issue:44655", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45362", + "issue:45468", + "issue:45478", + "issue:45542", + "issue:45753", + "issue:45850", + "issue:46095", + "issue:46218", + "issue:46328" + ], + "soft_pairs": [ + "issue:45362|issue:45850", + "issue:44273|issue:45542", + "issue:44230|issue:46095", + "issue:44038|issue:45125", + "issue:44393|issue:45071", + "issue:44273|issue:45003", + "issue:43576|issue:44556", + "issue:44393|issue:45753", + "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:44297|issue:45230", + "issue:44295|issue:45230", + "issue:44246|issue:44273", + "issue:43352|issue:44393", + "issue:43576|issue:44908", + "issue:43976|issue:44393", + "issue:44908|issue:45071", + "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:45230|issue:45310", + "issue:44360|issue:45468", + "issue:44655|issue:44861", + "issue:44273|issue:44908", + "issue:43576|issue:46218" + ], + "prepared_review_unit_hash": "61872d20b5b7ecafbe5b708a2081f3d2c850dcc65314b2a3e1989c14d7dd7445", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 42, + "serialized_chars": 39687, + "estimated_input_tokens": 9922, + "estimated_eval_tokens": 20100 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated issue reports. The only soft match I\u2019d accept is the import-time slowdown pair, which appears to describe the same performance regression from different angles.", + "confidence": 0.81, + "canonical_issue_reason": "issue:41669 is the most concrete, actionable report for the import slowdown, with a specific suspected cause and clear performance impact; issue:44246 is the broader symptom report.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41669 best represents the underlying bug because it identifies the likely root cause and the exact regression surface.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Specific Qwen3.5 chat crash vs a broad complaint that many models break; no shared concrete code path." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading malfunction and TensorFlow backend installation error are different failures." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 inference support request and DeepSeek weight-conversion question are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Same model family, but one is a Qwen3-VL-Moe compatibility bug and the other is a tensor-parallel planning gap." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Qwen3-VL bbox output problem vs a PretrainedConfig type-checking regression; different bugs." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Both touch import/lazy-loading behavior, but one is a lazy-loading correctness issue and the other is unsafe sys.modules access." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Broken env command and checkpoint reload incompatibility are separate v5 regressions." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "BBox hallucination/error and infinite-loop-on-one-image are not the same symptom or code path." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "A specific Llama4Config attribute error is not the same as a generic claim that many models break." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-time slowdown from import * usage is different from a missing GenerationMixin cache during chunked_prefill." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both are v5 reload incompatibilities, but they involve different loading scenarios and model-specific failures." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression is too specific to match the broad breakage complaint." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "A typo in cache-strategy docs/metadata is unrelated to a general model-breakage report." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "General import errors after upgrade and a broad 'many models break' complaint are not the same concrete bug." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 API guidance and a cache-strategy typo are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:46218", + "accept": false, + "reason": "Import slowdown and a specific missing Qwen3_5ForCausalLM import are different failures." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Wrong model-type warning and Liger Kernel not being applied with model_init are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Missing set_seed in tests and inverse_sqrt scheduler kwargs handling are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer save metadata mismatch is not the same as a generic bug report." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "position_ids buffer access error is unrelated to the generic bug report title." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Both concern import/lazy-loading behavior, but one is import slowness and the other is a lazy-loading correctness issue; not the same concrete bug." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention 2 support warning and Qwen3-VL bbox output failure are different issues." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "Broken env command and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python version support breakage and Qwen3-VL output errors are unrelated." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor kwargs passthrough and inverse_sqrt scheduler kwargs are different subsystems." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-time performance regression is not the same as failing to import HybridCache." + }, + { + "left": "issue:44492", + "right": "issue:46328", + "accept": false, + "reason": "Cache-strategy typo and SequenceFeatureExtractor.pad performance issue are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Import slowdown regression and a testing_utils bug are different code paths." + }, + { + "left": "issue:43976", + "right": "issue:45850", + "accept": false, + "reason": "Python 3.9/3.10 support mismatch is not the same as a broad post-v5 breakage complaint." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL bbox hallucination/error and Qwen3.5 Moe from_pretrained failure are different model bugs." + }, + { + "left": "issue:43976", + "right": "issue:46218", + "accept": false, + "reason": "Python-version compatibility and a Qwen3_5ForCausalLM import failure are unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Optional post_init policy and _get_tied_weight_keys crashing on list input are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing pipeline() and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion about forward args and a tokenizer-loading regression are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel not applied with model_init and degenerate training under flash_attention_3 are different bugs." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken env command and lazy loading not working are different issues, even if both are v5 regressions." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "A placeholder 'Bug report' title does not establish the same concrete failure as the Qwen3.5 Moe from_pretrained error." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer ReLU discussion and Gemma4AudioRelPositionalEncoding 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": 21, + "review_unit_index": 53, + "review_unit_count": 53, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-453", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:37934", + "issue:38617", + "issue:41669", + "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:46172", + "issue:46218" + ], + "soft_pairs": [ + "issue:46018|issue:46172", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:31515|issue:31795", + "issue:43704|issue:44485", + "issue:44485|issue:45468", + "issue:43298|issue:46095", + "issue:45850|issue:46218", + "issue:42371|issue:43704", + "issue:43502|issue:43519", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:44261|issue:45468", + "issue: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:44861|issue:45290", + "issue:44938|issue:44998", + "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": "f54ee0083abfc14deb9c31d9c977c5c440d58c72ba0348d30bb8254a352b7382", + "component_budget": { + "node_count": 453, + "item_count": 453, + "soft_pair_count": 1677, + "serialized_chars": 867752, + "estimated_input_tokens": 216938, + "estimated_eval_tokens": 434132 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 47, + "serialized_chars": 40309, + "estimated_input_tokens": 10078, + "estimated_eval_tokens": 20412 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated Transformer issues that happen to share broad loading/model/tokenizer language. I don\u2019t see a true duplicate set; the soft links are mostly spurious similarity matches.", + "confidence": 0.94, + "canonical_issue_reason": "Issue 45850 is the closest broad umbrella for the cluster because it describes widespread 5.0 breakage, but the underlying reports are still too diverse to be real duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "45850 is the best single representative of the overall theme because it is the broadest regression report in the set, though it is not a tight canonical duplicate anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "Different DeepSeek-V4 bugs: gated expert logic vs a tensor transpose/dtype issue." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are unrelated bugs." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Vision processor timestamp math is unrelated to scheduler kwargs handling." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import star cleanup and a multithreaded VRAM leak are different problems." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation confusion in forward args is not the same as a model-loading regression." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Slow checkpoint loading and forward-API documentation are unrelated." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM is unrelated to GLM-5 RoPE math." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "Both touch positional encoding, but they are different models and different concrete failures." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Audio continuous batching support and a DeepSeek-V4 weight-conversion question are unrelated." + }, + { + "left": "issue:45850", + "right": "issue:46218", + "accept": false, + "reason": "A broad 5.0 breakage complaint is not the same as a specific missing import for Qwen3_5ForCausalLM." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 settings guidance is unrelated to a VRAM leak." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "Local-files-only network leakage and timestamp calculation are different code paths." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Timestamp math and missing save_pretrained files are unrelated processor bugs." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Processor timestamp logic is unrelated to multithreaded memory leakage." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return handling is unrelated to Gemma4 audio positional encoding." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU-offload device mismatch and audio positional encoding are different model-specific issues." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "RMS norm epsilon precision mismatch is not the same as Gemma4 audio positional encoding." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "Different import failures for different symbols/models; not the same missing export." + }, + { + "left": "issue:43352", + "right": "issue:46218", + "accept": false, + "reason": "Flash Attention unsupported and missing Qwen3_5 import are unrelated." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Both are tokenizer-v5 reports, but one is broad behavior drift and the other is a specific decoder-type bug." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs bug and spam/off-topic issue are unrelated." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "Cache-strategy typo and unrelated spam issue are not duplicates." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions regression is unrelated to the off-topic item." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "Spam/off-topic text is not a duplicate of a broad 5.0 regression report." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weight key crash and apply_chat_template tool-call crash are different failure modes." + }, + { + "left": "issue:44938", + "right": "issue:44998", + "accept": false, + "reason": "Python 3.14 load failure is unrelated to the off-topic item." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id and tied-weight key handling are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow checkpoint loading and an ImportError for layer_type_validation are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "StableLmConfig pad token regression is unrelated to tied-weight key crashes." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 inference support and missing processor save files are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLflow reporting on failed jobs is unrelated to checkpoint loading speed." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Checkpoint loading performance and tensor-parallel training bugs are different issues." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 inference support and Qwen3.5 MoE from_pretrained failure are not the same bug." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Checkpoint loading slowness and tokenizer-loading regression are different problems." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Audio continuous batching support and FP8 support are unrelated feature gaps." + }, + { + "left": "issue:43717", + "right": "issue:46144", + "accept": false, + "reason": "Mamba init_weights usage and RoFormer attention-interface usage are unrelated architecture issues." + }, + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2former ignore_value handling is unrelated to DeepSeek-V4 gating." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "A type-hint typo in config is unrelated to tokenizer regex warnings." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Ignoring a configured loss value and missing kwargs in loss calculation are different bugs." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss kwargs handling and RoFormer attention interface are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2former ignore_value and tokenizer regex warnings do not share a code path." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "Different model subsystems: segmentation loss handling vs attention interface." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Ignore_value usage and a typo in quantization operation naming are unrelated." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Loss kwargs forwarding and a config type-hint mistake are unrelated." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "Type-hint correction and a typo in a conversion API name are different fixes." + }, + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Llama4 norm semantics and a ModernBERT type-hint issue are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Segmentation ignore_value logic is unrelated to a config annotation typo." + } + ] + }, + "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": 21, + "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": "d6b022fe048fb172de4c6c8f48e899062b498cdfe75f0986fc87a0f5d0d7d59a", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2875, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2875, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three items are separate feature requests in different areas: token truncation behavior, a learning-rate scheduler feature, and Wav2Vec2 word timestamp offsets. No pair looks like the same underlying bug or change.", + "confidence": 0.98, + "canonical_issue_reason": "issue:10519 is the most concrete and self-contained request, with a specific API behavior change described in the title.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:10519 is the clearest representative of the cluster because it is specific and narrowly scoped; the other issues are unrelated in topic and subsystem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:10519", + "right": "issue:11307", + "accept": false, + "reason": "Different domains and requested changes: truncation direction for tokenization vs. Wav2Vec2 word boundary timestamps." + }, + { + "left": "issue:10519", + "right": "issue:11005", + "accept": false, + "reason": "Unrelated features: tokenizer truncation behavior vs. ReduceLROnPlateau-like scheduler functionality." + }, + { + "left": "issue:11005", + "right": "issue:11307", + "accept": false, + "reason": "No shared underlying bug/change; one asks for a scheduler feature, the other for speech model timing offsets." + } + ] + }, + "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": 21, + "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": "213ce43d2eb4b8ba8cbecd04767d18b56af5b19550e2ee914386937bf2eb1290", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2479, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2479, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The three issues are only loosely related by model/tokenizer/HF-loading wording, but they describe different problems: Tapas tokenization mismatch, a private-model login error message, and a failure loading bert-base-uncased. I would not merge any of the soft pairs as duplicates.", + "confidence": 0.86, + "canonical_issue_reason": "Issue #16618 is the broadest and most general of the set, with the most discussion and a user-facing model-loading failure, so it is the best single representative if one is needed.", + "canonical_pr_reason": null, + "best_issue_reason": "#16618 is the best issue candidate because it is the clearest, most general symptom report in the cluster and most likely to attract related discussion.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "These appear to be different bugs: tokenizer behavior for Tapas versus a model-loading failure for bert-base-uncased. They do not share a concrete code path or fix." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Both concern model downloads/loading, but one is specifically a private-model authentication/error-message issue while the other is a generic load failure. Too broad to treat as the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 21, + "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": "2a16aa37975dffdadd38eea6ab1a77c6f6b8a57c84f1e350b674261500d37f9c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate community-contribution tracking issues for different features: Flash Attention 2 support vs GGUF support. They share a broad \u201cmore architectures\u201d theme but do not describe the same underlying bug or change.", + "confidence": 0.98, + "canonical_issue_reason": "Not applicable as a duplicate canonical: the issues track different feature requests (Flash Attention 2 support vs GGUF support), so neither can serve as the canonical for the other.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a global representative of the other; they are unrelated feature-tracking threads in the same general area but for different deliverables.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different requested features and likely different implementation work; same broad subsystem/theme is not enough to treat them as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 5, + "component_count": 21, + "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-26413-2", + "nodes": [ + "issue:26413", + "issue:31501" + ], + "soft_pairs": [ + "issue:26413|issue:31501" + ], + "prepared_review_unit_hash": "bf6f367a3d4a3b3d1f97cb00c97be0775c99036f58d79989e176e1846bf66add", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1593, + "estimated_input_tokens": 399, + "estimated_eval_tokens": 1054 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1593, + "estimated_input_tokens": 399, + "estimated_eval_tokens": 1054 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues both concern training failing to start, but the evidence is too thin to treat them as the same bug. The first is a specific `resume_from_checkpoint`/`epoch_iterator` failure; the second is a broader FineWeb SLM training startup problem with no explicit linkage to that code path.", + "confidence": 0.88, + "canonical_issue_reason": "Issue #26413 is the more specific, clearly framed bug report with an explicit failure message and likely reproducible code path.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #26413 is the best representative because it names the concrete error and underlying training control-flow problem; #31501 is too broad from the title alone.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26413", + "right": "issue:31501", + "accept": false, + "reason": "Both mention training not starting, but one is specifically about `resume_from_checkpoint` hitting an empty `epoch_iterator` while the other is a general FineWeb SLM startup failure. No evidence they are the same underlying bug or fixable in one change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 21, + "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": "57e5ee589e06df80630543a675117e169abf39e6aa102a568e80e2930631eba3", + "component_budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13526, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13526, + "estimated_input_tokens": 3382, + "estimated_eval_tokens": 7020 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Using only titles, body lengths, and soft-similarity scores, this looks like two tight duplicate-like subgroups rather than one big duplicate set: three identical generic placeholder translation issues, and one identical Thai-titled pair. The language-specific translation requests (Turkish, Spanish, Romanian) appear separate and should not be merged based on this evidence alone.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 43834 is the best canonical anchor for the largest high-similarity subgroup: it has the generic placeholder title shared by the other two placeholder issues and sits in the densest cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "43834 is the strongest overall issue representative because it belongs to the largest near-duplicate cluster and connects to the other placeholder-title copies with very high similarity.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Turkish translation request vs. generic placeholder issue; related topic, but not the same underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "Different language-specific request vs. generic placeholder issue; similarity is only boilerplate-level." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Turkish request and placeholder-template issue are not the same concrete task." + }, + { + "left": "issue:27088", + "right": "issue:38435", + "accept": false, + "reason": "Different language requests (Turkish vs. Romanian); separate issues." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "Different language requests (Turkish vs. Spanish); separate issues." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "Spanish translation request vs. generic placeholder issue; too little evidence of the same underlying task." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "Spanish request and placeholder-template issue look related only at the template level, not duplicate." + }, + { + "left": "issue:28936", + "right": "issue:38435", + "accept": false, + "reason": "Spanish vs. Romanian translation requests are distinct issues." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Spanish request vs. placeholder-template issue; not the same change or request." + }, + { + "left": "issue:38435", + "right": "issue:44370", + "accept": false, + "reason": "Romanian request is language-specific; placeholder issue is a different artifact." + }, + { + "left": "issue:38435", + "right": "issue:44365", + "accept": false, + "reason": "Romanian request vs. generic placeholder issue; related, but not a clear duplicate." + }, + { + "left": "issue:38435", + "right": "issue:43834", + "accept": false, + "reason": "Romanian translation request is distinct from the placeholder-template issue." + }, + { + "left": "issue:38435", + "right": "issue:46222", + "accept": false, + "reason": "Romanian issue and Thai-titled issue do not look like the same translation request." + }, + { + "left": "issue:38435", + "right": "issue:46226", + "accept": false, + "reason": "Romanian issue and Thai-titled issue do not look like the same underlying task." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": true, + "reason": "Same generic placeholder title and extremely high similarity; highly likely duplicate." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": true, + "reason": "Same generic placeholder title and extremely high similarity; highly likely duplicate." + }, + { + "left": "issue:43834", + "right": "issue:46222", + "accept": false, + "reason": "Very high similarity, but different titles and no body text to show the same underlying issue." + }, + { + "left": "issue:43834", + "right": "issue:46226", + "accept": false, + "reason": "Very high similarity, but different titles and insufficient evidence to merge." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": true, + "reason": "Identical placeholder-template title with near-perfect similarity; highly likely duplicate." + }, + { + "left": "issue:44365", + "right": "issue:46222", + "accept": false, + "reason": "High textual similarity alone is not enough here; the titles are different and the evidence is too thin." + }, + { + "left": "issue:44365", + "right": "issue:46226", + "accept": false, + "reason": "High similarity, but different titles and no direct evidence of the same issue." + }, + { + "left": "issue:44370", + "right": "issue:46222", + "accept": false, + "reason": "Placeholder-template issue vs. Thai-titled issue; not the same concrete request." + }, + { + "left": "issue:44370", + "right": "issue:46226", + "accept": false, + "reason": "Placeholder-template issue vs. Thai-titled issue; not the same concrete request." + }, + { + "left": "issue:46222", + "right": "issue:46226", + "accept": true, + "reason": "Identical Thai title and perfect similarity; highly likely duplicate pair." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the summary overstates what the packet supports. The packet only shows titles, body lengths, and soft-similarity scores; it does not expose the bodies, and every pair is marked deterministic_accept=false. Saying the placeholder issues or Thai-titled issues are \"exact\" or even clearly duplicate is stronger than the evidence here. The cross-language separations are conservative, but the accepted soft-edge pairs need weaker wording (e.g., \"highly similar\" / \"possible duplicates\")." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 7, + "component_count": 21, + "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": "f1ad9f7ce3672a3f9491d76b92071c71f657be7778fad29eed149934a3a5ca6b", + "component_budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21657, + "estimated_input_tokens": 5415, + "estimated_eval_tokens": 11086 + }, + "budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21657, + "estimated_input_tokens": 5415, + "estimated_eval_tokens": 11086 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a loose collection of model-support requests across different architectures and modalities. Despite textual similarity, the items are not the same underlying bug/change, so the soft edges should not be merged.", + "confidence": 0.97, + "canonical_issue_reason": "If a single anchor is needed, issue 42738 is the broadest conceptual request in the set, but it is still only a rough umbrella and not a true duplicate of the model-specific requests.", + "canonical_pr_reason": null, + "best_issue_reason": "42738 is the closest representative issue because it is the most general architecture-style request; however, this cluster is heterogeneous and lacks a real canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Different model families and different implementation targets; not the same change." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert config support and Qwen3-TTS support are unrelated model additions." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert config vs RF-DETR: different architectures and code paths." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are separate vision models, not one bug or patch." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR is a detection model; LFM2.5 Audio is an audio model request." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "Different model classes; no shared concrete fix path." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR and Qwen3-ASR are unrelated model support requests." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "Different modalities and model families; not mergeable as one PR." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are unrelated requests." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT support and weight-sparse transformer support are different changes." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "BERT-like encoder work vs audio model support; distinct code paths." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are different model additions with no shared fix." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are separate model requests, not duplicates." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are unrelated." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 and sparse-transformer support are different concrete changes." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 vs ModernVBERT are distinct model-support requests." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 and Voxtral TTS address different models and modalities." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated model support changes." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT and Qwen3-ASR are different model requests." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are not the same underlying issue." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT is a specific model request; 42738 is a broad BERT-like/RoPE family request." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "BERT-like models with RoPE and Voxtral TTS are different model domains." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "TTS and ASR are related modality families but still distinct model support requests." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Both are TTS-related, but they concern different concrete models and fixes." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and implementation paths." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Audio generation support and sparse-transformer support are unrelated." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio are different audio-model requests." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio 1.5B are different speech/audio model requests." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "ASR support and EUPE encoder support are unrelated." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Speech recognition support and sparse-transformer support are not the same code change." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio and SAM3-LiteText are different multimodal model requests." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Sparse-transformer support and EUPE model support are distinct changes." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral TTS are separate speech-model requests." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral TTS are unrelated model support requests." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "ASR support and OCR support are different modalities and code paths." + }, + { + "left": "issue:45322", + "right": "issue:46133", + "accept": false, + "reason": "EUPE and TIPSv2 are different vision-language model requests with no shared concrete fix." + } + ] + }, + "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": 21, + "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": "f0b3266543caebdedb1a6f0b0b7a1214fe1a62ad63a81a62c990734b56707193", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1503, + "estimated_input_tokens": 376, + "estimated_eval_tokens": 1008 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1503, + "estimated_input_tokens": 376, + "estimated_eval_tokens": 1008 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are unrelated: one is a docs request for adding Pruna as an optimization framework, while the other is a Vision RoPE refactor. They should not be treated as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue: these are about different topics and do not describe the same underlying problem or change.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 46443 is the more substantive standalone code-related item, but it is not a duplicate of issue 38740.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38740", + "right": "issue:46443", + "accept": false, + "reason": "Different subjects and scopes: documentation/framework listing vs Vision RoPE refactor. No shared underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 9, + "component_count": 21, + "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": "28133fcd8bcfac0d513f43b8f08eeb2065c2961cf47db7db52d45294c7543d83", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are not duplicates: one asks for Set Block Decoding, while the other requests DeepSeek v3.2 support. They target different features and should not be merged.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 43037 is the more specific and actionable request; 41084 is a separate, broader feature request.", + "canonical_pr_reason": null, + "best_issue_reason": "43037 is the best issue to keep as the representative item because it names a concrete model support request, making it easier to route and implement.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "Different underlying requests: block decoding vs. model support for DeepSeek v3.2. They do not describe 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": 21, + "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": "5e5ad7dd4c48c7a8bafe05e002358f766a42843259ed2d87aaff7d056bfca893", + "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 issues are only loosely related through Trainer customization, but they target different behaviors: one is about `remove_unused_column` handling, the other about FLOPS estimation for custom trainers. They do not look like the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "If one must be surfaced as the representative issue, `#43598` is the more concrete Trainer API/customization request; `#43600` is a separate FLOPS-estimation problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "Reject: both mention custom trainers, but they affect different code paths and outcomes (`remove_unused_column` behavior vs FLOPS estimation). This is broad thematic overlap, not a duplicate underlying problem." + } + ] + }, + "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": 21, + "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": "da18fee54d1845ec5e5d6d30049a40e382a3e47b542d55b27214a4ae7543ead9", + "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 not duplicates: one asks for Qwen3-Omni auto-registration for generation, while the other requests sequence-classification support for Qwen3.5. They target different model families and different AutoModel heads.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the requests are about different Qwen variants and different auto-class targets.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43684 is the broader, more central model-registration request, so it is the better representative if one must be chosen.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different model families and different tasks (conditional generation vs sequence classification); not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 21, + "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": "9bacf7033ca970dac859ace97a1bd90280f26f9b3f7568b2148b96e760e61c81", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are low-signal issues with generic or empty titles and no links, filenames, or targets. There is not enough evidence to treat any pair as the same underlying bug, though the three \"huggingface\" issues look superficially similar.", + "confidence": 0.35, + "canonical_issue_reason": "Issue 43859 is the most recent and the only one with any discussion activity, so it is the least empty representative among otherwise sparse reports.", + "canonical_pr_reason": null, + "best_issue_reason": "43859 is the best issue candidate only by marginal metadata signal; none of the issues contains enough substantive description to establish a clearly better canonical report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Titles are unrelated/generic and there is no body detail or linkage showing the same bug." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "No substantive evidence connects \"Fett\" to the \"huggingface\" report beyond weak similarity." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "Insufficient evidence that these describe the same underlying issue." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": false, + "reason": "Both titles are identical, but the reports are empty enough that identical wording alone is not enough to confirm a duplicate bug." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "Same generic title, but no descriptive content, links, or targets to prove the same underlying problem." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "Same generic title, but no concrete issue details to establish a true duplicate relationship." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 21, + "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": "debbb935d687b4dc0314ed21df69244119f189dc9b7dea445bc9d0a96689fc63", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are low-signal issue reports (mostly spam-like titles) with no bodies or issue targets. They do not show evidence of a shared underlying bug, so none of the soft pairs should be merged 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:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are generic spam-like issues with no concrete description; same title text alone is not enough to establish the same underlying problem." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Same as above: no actionable body or shared target, so this looks like separate spam rather than a duplicate bug report." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam-like and the other is a vague documentation suggestion; different problem types and no evidence of a shared issue." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Titles are unrelated and there is no body detail or target to indicate the same underlying issue." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "A documentation enhancement request is not the same underlying issue as an empty spam report." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Different artifact types and no shared concrete bug/change to merge." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Even though both titles are identical spam placeholders, there is no substantive content tying them to one bug or change." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "No shared concrete subject matter; one is spam-like, the other has a distinct title." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "No evidence of the same underlying problem; the reports are too vague and unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 21, + "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": "532ca6ccf3a0a227d8b4776db4a7de0875ea7beb06af7378d40e61ee854f991b", + "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": "The two closed issues have the same title and highly similar body length, so they appear to be duplicate reports of the same energy-reduction proposal. Issue #45145 is the better canonical because it was created first and already has an inbound reference.", + "confidence": 0.98, + "canonical_issue_reason": "#45145 is the earliest report and already has inbound reference activity, while #45151 appears to be the later duplicate of the same issue content.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best representative because it is the first-created instance of the identical issue and is likely the original thread for discussion.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Same title and very similar body structure/length; these look like duplicate reports of the same underlying issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The conclusion is grounded in the packet: the two closed issues have identical titles, very similar body lengths, and a soft-similarity match. Choosing #45145 as canonical is also conservative because it was created first and has one inbound reference. The soft-edge accept looks reasonable, though the confidence is a bit high given the evidence is still mostly similarity-based." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 21, + "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": 16, + "component_count": 21, + "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": "7b76c4d324e633500176b38c88aa3974c16636b2454b3e95158ff4267cdaf8b6", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1673, + "estimated_input_tokens": 419, + "estimated_eval_tokens": 1094 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1673, + "estimated_input_tokens": 419, + "estimated_eval_tokens": 1094 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are not duplicates: one reports a Gemma4 audio-module failure from `torch.finfo()` on `uint8` weights during 4-bit NF4 inference, while the other reports a `BitsAndBytesConfig.llm_int8_skip_modules` interaction that drops the default `lm_head` exclusion and triggers an `AssertionError` in 4-bit inference.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Different root causes and code paths: dtype handling in Gemma4 audio modules versus BitsAndBytesConfig skip-module logic affecting `lm_head` exclusion. They share the same broad 4-bit inference area 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": 17, + "component_count": 21, + "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": "6a707e5e3448a544ad2b41e5935385a589effb8a31fa7f934e27c53bf68d01f0", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two closed issues with personal-name/date titles and no body content or technical context; they do not appear to describe the same bug or request.", + "confidence": 0.98, + "canonical_issue_reason": "No strong canonical issue exists; if one must be chosen, issue #45762 is marginally earlier, but it has no meaningful technical content to serve as a true duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good cluster anchor because both are effectively empty and unrelated, but #45762 is the least arbitrary choice due to earlier creation time.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "Reject: the titles reference different people and different dates, and there is no shared technical description, issue target, or evidence of 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": 18, + "component_count": 21, + "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": "3a2a237c243cb1d88df9a47d67bf10f49cbb026ff974555c919ceb6b60b1e6dc", + "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 closed issues do not look like duplicates of the same bug or change; they are just sparse title-only issues for different named people, with no shared context beyond a surname/date pattern.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the titles refer to different people and contain no shared problem statement or actionable issue context.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a meaningful representative for a duplicate cluster; both are unrelated, body-less entries.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "Reject: only superficial similarity (same surname and date-like text). The titles identify different people and there is no evidence of the same underlying bug or request." + } + ] + }, + "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": 21, + "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": "7543f6da070c70847fe812244cb6ba089674f6ce0df8e5960366590f8b1c0970", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1542, + "estimated_input_tokens": 386, + "estimated_eval_tokens": 1028 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1542, + "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/value-model class, while the other asks for FP8 training support. They do not describe the same underlying bug or change.", + "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 feature requests in different areas (model class addition vs FP8 training support); no shared code-path or duplicated underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 20, + "component_count": 21, + "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": "42b0bfa2c742e9248f855f7bcf0e2981a6633f062df11639c2920f1abd55205c", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster most plausibly splits into two likely duplicate issue pairs (45870/45871 and 45872/45873) plus one outlier (45869). The evidence supports duplication at the pair level, but not a single canonical issue for the whole cluster.", + "confidence": 0.88, + "canonical_issue_reason": "No single canonical issue cleanly covers the whole cluster. The strongest reading is two likely duplicate pairs\u201445870/45871 and 45872/45873\u2014with 45869 as an unrelated outlier. Without body-text evidence, I would not collapse the two topical pairs into one issue.", + "canonical_pr_reason": null, + "best_issue_reason": "As a global representative, 45870 is the most defensible anchor only because it is the earlier item in one of the two likely duplicate pairs. That said, 45872 is equally plausible for the second pair, so the cluster is better handled by splitting than by picking one overall canonical issue.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Financial dataset bundle vs. a prediction example request. The shared financial wording is too broad to treat them as the same issue." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Financial dataset bundle vs. a prediction example request. Similar theme, but clearly different underlying requests." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Strong duplicate candidate: same title, same body length, and the closest similarity signal in the set. I cannot prove exact textual duplication, but they look like the same issue report." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Both mention DataForge and financial time series, but one is an example request and the other adds a scraping service. Different scope, so not the same underlying issue." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Different request scope: the prediction example and the scraping-service variant are not the same issue." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Different request scope: example/prediction versus prediction plus scraping service. Too different to merge as duplicates." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Different request scope: example/prediction versus prediction plus scraping service. Not the same underlying issue." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Strong duplicate candidate: same title, same body length, and the strongest similarity signal for this pair. There is no evidence of a separate underlying request." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The split into two topical pairs plus one outlier is plausible, but the writeup is too certain for the evidence shown. The packet only gives soft-similarity, identical titles, and body lengths; it does not prove the issues are exact textual duplicates. Also, 45870/45871 and 45872/45873 have identical titles, so the claim that one has the \"cleaner, descriptive title\" is not grounded. Please downgrade the duplicate language to a more conservative phrasing unless you have body-text evidence." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 21, + "component_count": 21, + "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": "5505c40d92acf2d9e764bb68ac0c13d8590b835a908af386f9c1ce781e5752a6", + "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": "Two related static-cache crash reports, but they appear to describe different failure modes: one is linear-attention shape corruption during early initialization, the other is a Qwen3.5 `KeyError` for `linear_attention`.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 46439 is the better canonical issue because it describes the underlying static-cache/linear-attention corruption more generally, rather than one model-specific symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "46439 is the strongest representative for the cluster because its title points to the broader mechanism and likely root cause affecting linear-attention layers under static cache.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46439", + "right": "issue:46441", + "accept": false, + "reason": "Both involve static cache and linear attention, but one reports shape corruption in early initialization while the other is a `KeyError` in Qwen3.5; these do not clearly look like the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "0774dc55518e9ce607d8a143712671358f4ecabf8d2e3e6c5e6dddf7e9369317", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 38088, + "estimated_input_tokens": 9522, + "estimated_eval_tokens": 19300 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 38088, + "estimated_input_tokens": 9522, + "estimated_eval_tokens": 19300 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a false-positive similarity group of unrelated pull requests. The only true duplicate-like pair is the two `fixed_cross_entropy` PRs (#43251 and #43254), which implement the same loss-function kwargs support in the same file and target the same issue. All other pairs are broad-subsystem or surface-level lookalikes and should stay separate.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43251 is the best canonical representative because it has the clearest, most explicit scope (`Fix(43240): pass kwargs to nn.functional.cross_entropy`) and fully covers the same concrete change as #43254.", + "best_issue_reason": null, + "best_pr_reason": "PR #43251 is the best fit for the cluster only because it is the sole legitimate duplicate anchor; it is more precise than #43254 and directly describes the code-path fix in `fixed_cross_entropy`.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "Unrelated changes: VideoPrism model/docs/Auto wiring vs an ExecuTorch dynamic-shape test." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Tokenizer API additions for MistralCommon vs adding a new MobileLLM model." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Missing tokenizer methods vs an ExecuTorch export test; no shared bug/change." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor/pretrained robustness fixes vs a new model implementation; different code paths." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Loss-function kwargs support vs adding MobileLLM; unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Loss utility kwargs fix vs Mistral4 test/model adjustments; not the same change." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "ExecuTorch export test vs Mistral4 fixes; only superficial similarity around testing." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Dynamic-shape export test vs new MobileLLM model; different artifacts and intent." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "Model addition vs `cls.config` resolution fix; unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "New MobileLLM model vs Falcon output-capture refactor; different concerns." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactor vs model-creation skill docs; not the same underlying bug/change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon output-interface refactor vs Mistral4 test fixes; different code paths." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test/model changes vs a conversion-mapping typo fix; only broad model-maintenance similarity." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46267", + "accept": false, + "reason": "Mistral4 changes vs chat response parsing/serve updates; unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 changes vs TDT loss kernel integration; different subsystem and goal." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 fixes vs `cls.config` resolution order fix; unrelated implementation areas." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 fixes vs model-creation skill draft; not the same change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "MobileLLM model addition vs model-creation skill docs and scaffolding; not duplicates." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs change `src/transformers/loss/loss_utils.py` to add the same `cross_entropy` kwargs support (`weight`, `label_smoothing`) for the same issue target #43240; these could plausibly be merged into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well-grounded in the packet and the only accepted pair (#43251/#43254) is supported by the shared file and shared issue target. The rejections stay conservative: the other pairs are indeed only soft-similarity matches across unrelated subsystems. No overstatement stands out." + }, + "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": "048c241623a37ccab685e6cf0820bf4958a309a4a7ca2a7ce5045442cadbe6ac", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10868, + "estimated_input_tokens": 2717, + "estimated_eval_tokens": 5690 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10868, + "estimated_input_tokens": 2717, + "estimated_eval_tokens": 5690 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster contains one base VibeVoice implementation PR, one separate VibeVoice Realtime extension, and one unrelated Parakeet-RNNT PR. The VibeVoice PRs are not duplicates; the Parakeet PR is clearly separate.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 40546 is the clearest representative because it is the foundational VibeVoice implementation and the earliest, broadest change set in the cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 40546 is the best cluster representative because it introduces the core VibeVoice model support; the other PRs either extend a different VibeVoice variant or belong to a different model family.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Both target VibeVoice, but 42781 is a separate VibeVoice Realtime addition rather than the same concrete bug/change as the base implementation. The scope is overlapping, not duplicate." + }, + { + "left": "pull_request:42781", + "right": "pull_request:46331", + "accept": false, + "reason": "These are different model families and code paths: VibeVoice Realtime versus Parakeet-RNNT. Shared auto-registration files are too broad to indicate 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": 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": "b2e223d7701d36f399031f0681aa6ab79df050b91868752fc9207a963e16b5f7", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are unrelated feature additions: one adds TimesFM covariate forecasting support, and the other adds DINOv3 image classification support. They should not be clustered as duplicates or merged as the same change.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: the PRs address different models, different code paths, and different user-facing features with no meaningful overlap.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a good global best representative for this cluster because they are independent, unrelated feature PRs rather than duplicates of one another.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different subsystems and different changes: TimesFM covariate forecasting versus DINOv3 image classification. They do not appear to fix the same underlying bug or implement 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": 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": "815535a6a04af0fa15091711233a3517a145b1b46cd256c6fdbb067740de30ce", + "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 are distinct model-addition pull requests (DeepSeek OCR, Isaac, Kimi2-6) with separate model directories, docs, and auto-registration changes. They are superficially similar template-style PRs, but not duplicates of the same underlying bug or feature.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: these are separate feature PRs for different models, not the same change.", + "best_issue_reason": null, + "best_pr_reason": "No PR is a true duplicate target; 45630 is the most extensive/reviewed, but it still implements a different model than the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "DeepSeek OCR vs Isaac: different model families, different new files under src/transformers/models/deepseek_ocr/ vs src/transformers/models/isaac/, so not the same code-path change." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Isaac vs Kimi2-6: both add new model scaffolding, but for unrelated models with different implementations, configs, and tests; not mergeable as one duplicate PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "d16b2045a30c238cb12f6443652aa2a2473f90e529de9548343cc26332e402e6", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two unrelated workflow maintenance PRs: one upgrades many GitHub Actions for Node 24 compatibility, while the other bumps docker/build-push-action in several Docker-related workflows. They overlap in some files but address different actions and different change scopes.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 42975 is the broader and more representative workflow-actions upgrade, spanning many files and explicitly targeting Node 24 compatibility; PR 42976 is a narrower Docker action version bump.", + "best_issue_reason": null, + "best_pr_reason": "PR 42975 is the better cluster representative because its title and diff align with the overarching maintenance theme and it touches the widest set of workflows.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "They are both CI/workflow dependency upgrades, but they change different actions (`actions/*` vs `docker/build-push-action`) and do not fix the same concrete problem, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "a438447a3081ee8804976067bb52c18364402eb4edea75d8c70d164c8de5adab", + "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 PRs (ViT NEPA, Molmo, Molmo2, and CTSM), not duplicate fixes. The overlap is mostly standard docs/auto-registration boilerplate, not the same underlying change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No true duplicate PR exists in this set; if a single representative is needed, 43451 (Molmo2) is the broadest and most reviewed model-addition PR.", + "best_issue_reason": null, + "best_pr_reason": "43451 is the strongest representative by scope and review activity, but it is still a distinct model addition rather than a duplicate of the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different model additions: ViT NEPA vs Molmo. Shared auto/docs files are boilerplate, not the same code-path or fix." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "ViT NEPA and Molmo2 introduce different models with different configs/tests; no same underlying bug or mergeable single change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Completely different models and subsystems: vision model addition vs timeseries CTSM. Shared registration files are incidental." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are distinct model families; overlapping docs/auto mappings do not indicate one underlying change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo vs CTSM are unrelated model additions across different domains; no shared concrete bug or code-path." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM touch different model implementations and tests; similarity is limited to generic model-registration boilerplate." + } + ] + }, + "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": "7aa242bc75b2c851c7564819643451a95426f3435f12dba0c3f44639b725689d", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 11142, + "estimated_input_tokens": 2786, + "estimated_eval_tokens": 5828 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 11142, + "estimated_input_tokens": 2786, + "estimated_eval_tokens": 5828 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These four PRs are not duplicates of one another. PR 43291 is a Whisper decoding fix, PR 43665 is mostly test-file noise with arbitrary failing asserts, PR 44815 is an FP8 dequantization fix, and PR 46380 is a generation/watermarking validation fix. They touch different code paths and solve different problems.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44815 is the best canonical representative because it is a substantive, self-contained bug fix with real implementation changes and targeted tests; the others are either unrelated or appear to be test-only noise.", + "best_issue_reason": null, + "best_pr_reason": "PR 44815 is the strongest example of a genuine fix: it targets a concrete runtime path in FP8 dequantization, spans the relevant implementation files, and is more coherent than the other PRs.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Whisper tokenizer decoding changes are unrelated to the CLIP/ViT test edits; 43665 looks like test-noise rather than the same bug fix." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "One is a Whisper decode/tokenization fix, the other is an FP8 dequantization fix; different subsystems and different concrete bugs." + }, + { + "left": "pull_request:43291", + "right": "pull_request:46380", + "accept": false, + "reason": "Whisper decoding and generation watermarking validation are separate code paths with unrelated failure modes." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "43665 only injects arbitrary failing asserts into tests, while 44815 fixes FP8 dequantization logic; not the same change or bug." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46380", + "accept": false, + "reason": "The bogus test-only edits in 43665 do not match the generation/watermarking validation fix in 46380." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46380", + "accept": false, + "reason": "FP8 dequantization and generation configuration watermarking validation are unrelated concrete issues and would not be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "08248353b78bca26a3bdcce85fbd0c06dff274849370872add1e09bd1e0d2f7f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4563, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4563, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two pull requests do not look like duplicates: one is a narrow repo-bot formatting/check change, while the other is a broader workflow/testing diff. I would not merge them as the same underlying fix.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43488 is the cleaner representative: it is a small, focused change around dependency/version formatting and repo-bot checks, with a limited diff and clear intent.", + "best_issue_reason": null, + "best_pr_reason": "PR 43488 is the best PR to keep as the canonical example because it is narrowly scoped and self-contained; PR 43532 touches multiple workflows and test-related files, making it less representative of a single concrete fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Different scopes and likely different purposes: 43488 is a small repo-bot/version-formatting change, while 43532 changes CI/workflow and test infrastructure across several files. They do not appear to fix the same concrete code-path problem." + } + ] + }, + "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": "7aab7591b25c775b52a198ecc4171b3a69fe9ceade3aa67ba67646fa996c4c9e", + "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": "Two distinct feature PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share infrastructure but target different tasks, models, tests, and docs, so they are not duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR for a duplicate cluster: the two PRs implement different pipelines and should remain separate changes.", + "best_issue_reason": null, + "best_pr_reason": "PR #43613 is the broader representative of the shared multimodal-segmentation theme, but it is still not a duplicate of #43612.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Different pipeline features with different concrete targets: concept segmentation vs visual segmentation. Shared docs/registry scaffolding does not make them 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": "2bec84ad19f2ef6eee7c7cfcc7c4037fa487063c3531914cb75dac36fa30523c", + "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": "These are two separate trainer.py enhancement PRs: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering for custom trainers/PEFT. They share a file and broad subsystem but not the same underlying change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43651 is the better canonical representative because it has more discussion and a clearly scoped functional change in the training loss path.", + "best_issue_reason": null, + "best_pr_reason": "PR 43651 is the strongest standalone PR in this cluster due to higher discussion activity and a concrete, user-visible behavior change; 43663 is a different trainer extensibility refactor.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Reject: they touch the same file but fix different code paths\u2014loss scaling during training vs. signature-column filtering for input preprocessing. They would not plausibly be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "64d3c9b0be0ac24b2f646f8a90d308a3c6804d2dbbb899f5cd5883773d6d9d6b", + "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": "Both PRs target the same import-time `TypeAdapter` NameError in `src/transformers/cli/serve.py`; they differ mainly in implementation style, not in the underlying bug. PR 43842 is the more focused fix.", + "confidence": 0.77, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43842 is the clearest representative: it directly guards the `TypeAdapter` reference with `TYPE_CHECKING`/string annotations and is narrowly scoped to the import-time failure.", + "best_issue_reason": null, + "best_pr_reason": "PR 43842 is the best fit for the cluster because it targets the concrete `TypeAdapter` NameError with minimal collateral changes and an explicit issue link.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": true, + "reason": "Both PRs fix the same `TypeAdapter` NameError in the CLI serve module. 43656 uses a broader `from __future__ import annotations` approach, while 43842 uses a direct `TYPE_CHECKING`/quoted-annotation guard; these are alternate implementations of the same underlying fix." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs touch `src/transformers/cli/serve.py`, both are titled around the `TypeAdapter` NameError, and the soft-edge reasoning stays conservative by describing them as alternate implementations of the same fix. I don\u2019t see an overstatement or a mergeability issue here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "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: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:44018|pull_request:44066", + "pull_request:44066|pull_request:44068", + "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:44059|pull_request:44722", + "pull_request:44098|pull_request:44101", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:44071|pull_request:44086", + "pull_request:43997|pull_request:44073", + "pull_request:44017|pull_request:44071", + "pull_request:44073|pull_request:44101", + "pull_request:44059|pull_request:44071", + "pull_request:44073|pull_request:44161", + "pull_request:44018|pull_request:44059", + "pull_request:44056|pull_request:44101", + "pull_request:44059|pull_request:44068", + "pull_request:44101|pull_request:44161", + "pull_request:44017|pull_request:44018", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:43997|pull_request:44018", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:44018|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44018|pull_request:44129", + "pull_request:44010|pull_request:44073", + "pull_request:43997|pull_request:44074", + "pull_request:44071|pull_request:44129", + "pull_request:44010|pull_request:44018", + "pull_request:43997|pull_request:44129", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154" + ], + "prepared_review_unit_hash": "06c18c038c8dea7dca482fa7227add62ee3956a53f8ba470b4e790bb9006d782", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84890, + "estimated_input_tokens": 21223, + "estimated_eval_tokens": 42702 + }, + "budget": { + "node_count": 27, + "item_count": 27, + "soft_pair_count": 72, + "serialized_chars": 75085, + "estimated_input_tokens": 18772, + "estimated_eval_tokens": 37800 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of independent model-by-model output-tracing refactors under the same umbrella issue (#43979). The only clear duplicate subgroup is the GPT-J / CodeGen set, with a smaller duplicate pair in GPT-Neo.", + "confidence": 0.81, + "canonical_issue_reason": null, + "canonical_pr_reason": "No issue artifact here; for the PR side, 44722 is the best canonical representative because it is the broadest/latest GPT-J/CodeGen output-tracing refactor and overlaps the clear duplicate subgroup.", + "best_issue_reason": null, + "best_pr_reason": "44722 is the strongest single PR to represent the duplicate subgroup: it covers the same GPT-J output-tracing change as the other GPT-J PRs and also includes the CodeGen copy, making it the broadest/revised version.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different models/files (CVT/FNet vs GPT-J); same tracking issue is not enough." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Different model families and code paths; not the same change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Same GPT-J output-tracing refactor in `modeling_gptj.py`; these look like overlapping versions of the same change and could be merged." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "RegNet vs EfficientNet; separate model-specific migrations." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet vs GPT-J/CodeGen; unrelated code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen vs EfficientNet; different models." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 vs DeBERTa-v2; unrelated refactors." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen vs MGP-STR; different model implementations." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen vs MPT; different model families." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Same GPT-Neo file and same output-tracing migration; one looks like a later variant of the same refactor." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "GPT-Neo vs GPT-J/CodeGen; different underlying model code." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen vs GPT-Neo; not the same model or code path." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet vs GPT-J/CodeGen; unrelated model changes." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR vs GPT-J/CodeGen; different code paths." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet vs GPT-J; separate model-specific refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo vs ResNet; different model families." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet vs MPT; unrelated changes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo vs Flaubert; different model implementations." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo vs GPT-J/CodeGen; not the same refactor." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo vs Flaubert; unrelated model code." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT vs GPT-J/CodeGen; different code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo vs GPT-J; different underlying model." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo vs MGP-STR; unrelated." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo vs MGP-STR; different models." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44085", + "accept": false, + "reason": "MPT vs GPT-J; unrelated." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 vs MGP-STR; different code paths." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J vs MGP-STR; not the same change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT2 vs GPT-J/CodeGen; different transformer blocks." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT vs Flaubert/XLM; separate model migrations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo vs MPT; different model families." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "SegFormer vs GPT-J/CodeGen; unrelated code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "SegFormer vs LongT5; separate refactors." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "SegFormer vs VisualBERT; different model code." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo vs MPT; unrelated model changes." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT vs ImageGPT; different implementations." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet vs Flaubert/XLM; not the same change." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet vs GPT-Neo; different model-specific refactors." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet vs LongT5; unrelated code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet vs LongT5; separate model migrations." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet vs LongT5; different models." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44071", + "accept": false, + "reason": "RegNet vs MPT; unrelated." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert vs MPT; different model families." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo vs SpeechT5; not the same code path." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT vs SpeechT5; unrelated model changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert vs VisualBERT; separate refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet vs TextNet; different model code." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44101", + "accept": false, + "reason": "MPNet vs Flaubert/XLM; unrelated." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT2 vs MPT; different model families." + }, + { + "left": "pull_request:44101", + "right": "pull_request:44161", + "accept": false, + "reason": "Flaubert/XLM vs LongT5; not a duplicate." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44018", + "accept": false, + "reason": "SegFormer vs GPT-Neo; different underlying model." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet vs MPNet; unrelated changes." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT vs ImageGPT; different code paths." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert vs DeBERTa-v2; unrelated refactors." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa-v2 vs TextNet; different model-specific changes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "Both are GPT-J/CodeGen output-tracing refactors over the same copied block; 44722 looks like the broader/revised version." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Same GPT-J output-tracing change in `modeling_gptj.py`; 44722 subsumes it and also updates the CodeGen copy." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-J vs SpeechT5; different model families." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": false, + "reason": "CVT/FNet vs RegNet/ResNet; unrelated model refactors." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": false, + "reason": "CVT/FNet vs SpeechT5; different code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": false, + "reason": "DeBERTa-v2 vs SpeechT5; not the same underlying change." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: most PRs are separate model-specific output-tracing refactors under the same umbrella issue, and the only clearly mergeable overlap clusters are the GPT-J/CodeGen subgroup and the GPT-Neo pair. The soft-edge accept/reject calls are conservative and match the diff scope shown." + }, + "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": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44024", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44024|pull_request:44076", + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "2ad8556b96deb4e5a57acbba9808c6af32da4e120d15bb7267621818ff112291", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84890, + "estimated_input_tokens": 21223, + "estimated_eval_tokens": 42702 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 6, + "serialized_chars": 17126, + "estimated_input_tokens": 4282, + "estimated_eval_tokens": 8820 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These PRs are all model-specific refactors that migrate different architectures to standardized output tracing/capture_outputs patterns. They are related in theme, but they do not describe the same bug or the same code-path change, so they should not be deduplicated together.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "44129 is the strongest standalone representative because it has the clearest intent, an explicit issue link, and a concrete output-tracing refactor in a single model.", + "best_issue_reason": null, + "best_pr_reason": "44129 is the best single PR to keep as the representative example of this refactor pattern: it is self-contained, specific, and tied to an explicit tracking issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "Different models and different implementation paths; both are similar output-tracing refactors, but they are not the same underlying fix." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "VisualBert and TextNet are unrelated model codepaths, so these are parallel refactors rather than duplicate changes." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "Same broad standardized-output theme, but MPNet and VisualBert have distinct model-specific tracing logic and are not mergeable as one duplicate PR." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "These touch different models and different output-capture setups; similarity is thematic only." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are separate architectures with separate output-handling code, so this is not the same change." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "Both are output-tracing refactors, but they affect different models and code paths; the shared concept is too broad to treat as a duplicate." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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-43998-14", + "nodes": [ + "pull_request:43998", + "pull_request:43999", + "pull_request:44000", + "pull_request:44001", + "pull_request:44002", + "pull_request:44003", + "pull_request:44004", + "pull_request:44025", + "pull_request:44026", + "pull_request:44027", + "pull_request:44028", + "pull_request:44029", + "pull_request:44030", + "pull_request:46438" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030", + "pull_request:44000|pull_request:46438", + "pull_request:44027|pull_request:46438", + "pull_request:44029|pull_request:46438", + "pull_request:44028|pull_request:46438", + "pull_request:43998|pull_request:46438" + ], + "prepared_review_unit_hash": "55990ded056047f06528309ba3f83e760ed52f3c5b6e32f14c9b66587765df17", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 44, + "serialized_chars": 43563, + "estimated_input_tokens": 10891, + "estimated_eval_tokens": 22038 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 44, + "serialized_chars": 43563, + "estimated_input_tokens": 10891, + "estimated_eval_tokens": 22038 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All items are separate model-specific output-tracing refactors (or one utility speedup), not duplicates of the same bug/change. They share a common theme, but each PR targets different model implementations and code paths, so none should be merged as duplicate PRs.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44003 is the best representative of the cluster because it is a broader output-tracing refactor touching two related model files (FalconMamba and Mamba) and clearly reflects the common pattern in the set.", + "best_issue_reason": null, + "best_pr_reason": "PR 44003 is the strongest single representative of the shared refactor pattern, with the broadest model coverage among the model PRs shown.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Same refactor theme, but different model families and files (timm_backbone vs upernet); not the same underlying change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different model implementations and code paths; both are output-tracing refactors only at a high level." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Separate model families (timm_backbone vs univnet); no shared concrete bug or mergeable fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and files; similar refactor pattern does not make them duplicates." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Timm backbone refactor vs CodeGen attention/cache refactor are distinct code paths." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Unrelated model families and implementations; only the generic tracing refactor theme matches." + }, + { + "left": "pull_request:43998", + "right": "pull_request:46438", + "accept": false, + "reason": "Model refactor vs modular-conversion utility speedup are unrelated changes." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "MobileNetV1 and UnivNet are unrelated model code paths; not the same bug/change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models and outputs; same broad tracing refactor only." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "MobileNetV1 vs Depth Anything are distinct model implementations; not duplicates." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Different model families and files; only similar refactoring style." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "MobileNetV1 and Speech Encoder-Decoder fix different code paths." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Vision/Text dual encoder and UnivNet are unrelated models; no common concrete defect." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families and implementations; same tracing refactor theme only." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model stacks and files; not plausibly one PR." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Vision/Text dual encoder vs CodeGen are separate code paths and fixes." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families and output semantics; not the same underlying change." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "UnivNet and UperNet are unrelated implementations; no duplicate bug/change." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and model files; same refactor motif only." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "UnivNet vs Depth Anything are distinct model-specific refactors." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "UperNet and Mamba/FalconMamba changes are unrelated code paths." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "Different models and mechanics; one is segmentation, the other CodeGen cache/output handling." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Mamba/FalconMamba output tracing and CodeGen cache/output refactor address different model internals." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Different model families and concrete changes; same high-level theme is insufficient." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "Separate model implementations with different forward paths and return behavior." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "Different concrete model code paths; not mergeable as a single duplicate fix." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Depth Anything and Vision/Encoder-Decoder are unrelated models; similar refactor pattern only." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Different encoder/decoder model families and files; not the same underlying issue." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "Depth Anything vs SuperPoint target unrelated model code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families and refactor scopes; only generic tracing work overlaps." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "Depth Anything and DPR are distinct model implementations; not duplicates." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "Vision Encoder-Decoder and Speech Encoder-Decoder are separate models with different code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "Different model families and concrete refactors; not one underlying bug." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "Unrelated model internals despite similar output-tracing theme." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "Different architectures and files; cannot plausibly be merged into one PR." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "Speech Encoder-Decoder and SuperPoint are unrelated changes." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families and separate output handling code paths." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "Different model implementations; only the general refactor style matches." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "SuperPoint and Mamba/FalconMamba are unrelated model code paths." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "Different models and different forward/output semantics; not duplicates." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "Mamba/FalconMamba and DPR are unrelated model implementations and fixes." + }, + { + "left": "pull_request:44000", + "right": "pull_request:46438", + "accept": false, + "reason": "Vision/Text dual encoder refactor is unrelated to the modular-conversion pool reuse utility change." + }, + { + "left": "pull_request:44027", + "right": "pull_request:46438", + "accept": false, + "reason": "Speech Encoder-Decoder model refactor and modular conversion utility speedup are unrelated." + }, + { + "left": "pull_request:44029", + "right": "pull_request:46438", + "accept": false, + "reason": "Model output-tracing refactor vs utility process-pool optimization are different changes." + }, + { + "left": "pull_request:44028", + "right": "pull_request:46438", + "accept": false, + "reason": "SuperPoint model refactor and modular-conversion utility speedup are unrelated." + }, + { + "left": "pull_request:43998", + "right": "pull_request:46438", + "accept": false, + "reason": "Model refactor vs utility performance change; no shared concrete bug or 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": 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": "32ae3f77219dce503b473f82903f8b05ebbdeb9fda6aa8867cd00bfbd6211061", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6551, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6551, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests are new, unrelated model integrations: one adds PenguinVL and the other adds Voxtral TTS. Their overlap is limited to shared Transformers registration/docs scaffolding, which is not enough to treat them as the same underlying change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: these are separate model additions with different code paths, tests, and targets.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a global best representative for this cluster because the cluster is not a true duplicate set.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Reject: PenguinVL implementation and Voxtral TTS support are distinct model ports with different architecture, files, and goals; shared auto-registration/docs edits are boilerplate and do not indicate the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "2d86ac1456ca38e68f3406fbbdb96c0fc66e5156e407b7a55026deccc8c9d2ce", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 3, + "serialized_chars": 9635, + "estimated_input_tokens": 2409, + "estimated_eval_tokens": 5074 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 3, + "serialized_chars": 9635, + "estimated_input_tokens": 2409, + "estimated_eval_tokens": 5074 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster splits into two groups: a foundational Module Fusion API PR, and a sequence of n-to-1 KernelConfig fusion PRs that look like iterative revisions of the same feature work. I would keep the Module Fusion API separate and treat the later KernelConfig PRs as the closest duplicates.", + "confidence": 0.7, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46339 is the broadest and most mature KernelConfig fusion change: it refactors core loading paths, updates hub kernel integration, touches modeling utilities, and adds kernel tests. It looks like the best representative of the kernel-fusion line.", + "best_issue_reason": null, + "best_pr_reason": "PR 46339 is the best standalone candidate because it covers the most complete end-to-end n-to-1 kernel fusion flow and appears to supersede earlier incremental versions.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "44979 introduces the new Module Fusion API foundation, while 45363 implements KernelConfig-based n-to-1 fusion on top of it. Related, but not the same concrete change." + }, + { + "left": "pull_request:45363", + "right": "pull_request:46339", + "accept": true, + "reason": "Both target the same n-to-1 KernelConfig fusion feature and overlap on the same integration paths. 46339 looks like a later, broader revision of the same change stream." + }, + { + "left": "pull_request:45666", + "right": "pull_request:46339", + "accept": true, + "reason": "These are very close revisions of the same KernelConfig n-to-1 fusion work, with strong filename overlap and the later PR reading like an extension/simplification of the earlier one." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 44979 is clearly a separate Module Fusion API PR, while 45363/45666/46339 all sit on the same KernelConfig n-to-1 fusion line and share overlapping files/titles. The soft-edge verdicts are conservative enough overall, and the accepted pairs look plausibly mergeable as iterative revisions of the same feature work." + }, + "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": "46432e2da72908acd0f4ef90bfb5ba17a7639d010d79e765da202fa23756f777", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4413, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4413, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs target the same Mistral regex AttributeError in the same file; 45317 is the more complete fix because it includes the code change plus test coverage. They are duplicates at the PR level rather than separate changes.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45317 is the better canonical PR: it fixes the same AttributeError as 45086, but also adds tests and has a clearer, more specific description of the bug and trigger condition.", + "best_issue_reason": null, + "best_pr_reason": "PR 45317 is the best representative because it covers the same underlying code-path fix and validates it with tests, making it the strongest standalone record of the change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same concrete bug in _patch_mistral_regex, same touched code path and file, and the second PR is essentially a superset of the first with tests added." + } + ] + }, + "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": "3f1025c6886fca19dbf229b46ae530cb503a67acdcdba1aaa763bd25d5c0be86", + "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": "Both PRs are about the NemotronH model family, but they target different changes: #45115 is a smaller refactor/inheritance and checkpoint-conversion mapping update, while #45569 is a broader feature split introducing NemotronHDense/NemotronHSparse plus docs/tests. They overlap in shared model files, but not in the same concrete fix.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45569 is the better canonical representative because it is the larger, more user-visible NemotronH family change and includes the new dense/sparse model split, docs, and tests.", + "best_issue_reason": null, + "best_pr_reason": "PR #45569 is the strongest overall artifact in this cluster: it covers the full architecture decomposition and documentation updates, whereas #45115 is a narrower refactor focused on inheritance and conversion mapping.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Related NemotronH work, but not the same underlying change: #45115 is a refactor/conversion-mapping update; #45569 introduces new dense/sparse model variants and supporting docs/tests. Shared files alone are not enough to merge them as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "5b5c8fa5594d7567097e1b083d27787f31796e99e9e1f65d3081017ddc4a6104", + "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": "These two PRs are related but not duplicates: one adds heterogeneous config plumbing, while the other adds broader heterogeneous modeling/runtime support that builds on that config work. They share files because of the shared infrastructure, but they do not appear to be the same concrete change.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45332 is the broader umbrella change: it covers heterogeneous model support across cache/masking/modeling paths, not just config plumbing. It is the better representative of the overall feature set in this cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR #45332 is the strongest standalone representative because it implements the end-to-end heterogeneous model behavior, whereas PR #45333 is a narrower per-layer configuration piece that could plausibly be a prerequisite or companion change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Related feature work, but not the same underlying bug/change. #45333 is focused on per-layer configuration support; #45332 extends heterogeneous support into modeling/cache/runtime code paths and is materially broader." + } + ] + }, + "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": "dc5a9bf184ffb31d52947b80ff803d0856af62b385c7e08112ed42aed74dfdc6", + "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-focused, but they are different changes: one is a broad type-checking sweep across many transformer modules, while the other adds specific typing protocols/helpers and related fixes. The overlap is incidental rather than the same underlying fix.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "No issue link exists; if a representative PR is needed, PR 45425 is the more focused typing-refinement change with substantial review activity.", + "best_issue_reason": null, + "best_pr_reason": "PR 45425 is the better cluster representative because it targets a narrower, more concrete typing API/stub adjustment, whereas PR 45415 is a broad type-checking pass across many files.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Reject: both are typing-related, but 45415 is a wide sweep to add type checking across many files, while 45425 introduces specific typing protocols and utility annotations. They do not appear to fix the same concrete code-path problem and would not naturally collapse 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": "9d926012c44f28379a6b5a05efee576e5597676d63cbaf8f756440c366260d57", + "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": "Two open PRs were clustered only by superficial cache/multi-token similarity, but they address different model families and different bugs: Qwen3.5/OLMo hybrid cached forward behavior vs Mamba2 use_cache handling for seq_len > 1.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46084 is the more self-contained and explicit fix: it targets a single Mamba2 cached-forward bug, includes an issue reference, and has a narrow code-path change. PR #45513 spans multiple model files and appears broader and less directly tied to one concrete bug.", + "best_issue_reason": null, + "best_pr_reason": "#46084 is the strongest representative because it has a clear, singular failure mode (use_cache with seq_len > 1), an explicit issue target, and a focused patch. #45513 is broader and less specific.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45513", + "right": "pull_request:46084", + "accept": false, + "reason": "Different underlying bugs and code paths: #45513 fixes multi-token cached forward in Qwen3.5/OLMo hybrid linear attention, while #46084 fixes Mamba2 use_cache behavior when seq_len > 1. Shared caching language 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": 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": "418ba20040f77fc6b1de481edb94ff86349e9be09f52f2d286785470541fad0a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4507, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4507, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are both EP-related, but they fix different distributed-loading bugs on different code paths: one for DeepSpeed ZeRO-3/accelerate config handling, the other for FSDP2 expert parameter overwrites during sharded loading. They should not be deduplicated.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45662 is the more concrete representative of a distinct bugfix: it targets expert parameters being overwritten during distributed loading, with code changes in model loading and MoE/tensor-parallel integration.", + "best_issue_reason": null, + "best_pr_reason": "PR #45662 is the strongest standalone PR here because it addresses a specific runtime failure mode in the loading path and contains the broader, more direct fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Related only by broad EP/distributed-loading theme. #45548 fixes EP + DeepSpeed ZeRO-3 loading via accelerate launch and config serialization; #45662 fixes EP + FSDP2 experts being overwritten by rank-0 broadcast. Different bug, different code path, not 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": 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": "eaf6d84a0003c8a6499fbbb20d754a7932901973d53fe71688207eb734b9e9f4", + "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 xdist-related capture of debug artifacts in `testing_utils.py`, but they are not the same change: #45639 makes the log write path worker-safe, while #45645 also updates CI artifact collection and notification handling. The overlap is thematic, not a single mergeable fix.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "#45645 is the better canonical PR because it includes the core xdist-safe file handling plus the downstream CI changes needed to consume multiple captured-info files; it is the more complete end-to-end fix.", + "best_issue_reason": null, + "best_pr_reason": "#45645 is the strongest representative of the underlying problem because it addresses the same xdist collision issue in `testing_utils.py` and extends it through the CI path that surfaces those artifacts.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": false, + "reason": "They share the same affected helper files and issue target, but #45645 is broader and changes CI artifact aggregation plus notification logic, so this is not a single same-scope code-path fix that could plausibly be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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:46320|pull_request:46401" + ], + "prepared_review_unit_hash": "3b66a3da56305d968031eadec299630de43903458318f5d81f14b5012cf77e62", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7030, + "estimated_input_tokens": 1758, + "estimated_eval_tokens": 3772 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7030, + "estimated_input_tokens": 1758, + "estimated_eval_tokens": 3772 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three separate PRs: one adds GLM-ASR to auto mappings, one adds attn_implementation support for RecurrentGemma, and one adds SDPA dispatch support for BLIP-2/InstructBLIP variants. The two attn_implementation PRs are thematically similar but fix different model code paths and should not be merged as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46320 is the strongest representative of the cluster because it is a focused, substantive model-support change with clear code, conversion, docs, and tests; however, it is still not a duplicate of the BLIP-2 dispatch PR.", + "best_issue_reason": null, + "best_pr_reason": "PR #46320 is the best single representative of the attn_implementation theme, but the cluster does not contain a true duplicate PR pair.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45670", + "right": "pull_request:46320", + "accept": false, + "reason": "Different changes: auto-model registration for GLM-ASR vs RecurrentGemma attn_implementation dispatch. They touch unrelated code paths and would not be merged into one PR." + }, + { + "left": "pull_request:46320", + "right": "pull_request:46401", + "accept": false, + "reason": "Both are attn_implementation support PRs, but for different architectures and model stacks (RecurrentGemma vs BLIP-2/InstructBLIP). The underlying code paths are distinct, so this is not a duplicate." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "3336b5bfe11d55a0da6feb563fb00b4f3e9818af3c45e663f81828da27eab9c0", + "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 open pull requests: one implements TRF003 for model forward/return_tuple handling across text/audio models, and the other adds TRF018 to detect dead do_* flags in image/video processor overrides. They share some model filenames, but not the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: the two PRs address different lint/modeling-rule problems in different code paths, so they should not be merged as duplicates.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a good global representative of the other; each is the primary change for its own distinct rule and subsystem.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45720", + "right": "pull_request:46259", + "accept": false, + "reason": "Different underlying changes: TRF003 updates model forward signatures/tuple handling, while TRF018 adds a rule about unused do_* flags in image/video processor overrides. Shared files are incidental and not a single concrete fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "3f72d0e198383a34e1b5c03855bc4bd1f9469d103079291b7f6218a3de180009", + "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 open pull requests, but they are not duplicates: one is a broad model/workflow change PR, the other is an isolated test-migration placeholder with intentional failing asserts. They do not fix the same code-path problem and should not be merged together.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45932 is the more substantive representative of the cluster: it changes actual implementation/workflow files across multiple model paths, while PR 46100 is just a tiny test-only migration stub with obvious placeholder failures.", + "best_issue_reason": null, + "best_pr_reason": "PR 45932 is the best single PR representative because it contains the real code changes; PR 46100 is not a meaningful duplicate candidate and looks like a throwaway test edit.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "Reject: these are unrelated changes. PR 45932 modifies CI/model code, while PR 46100 only edits a ViT test file with deliberate failing asserts. They do not appear to fix the same bug or belong in one merged PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 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": "8c38e1de8a414f0d204b84543ea43e798edf151f4ccb08750b560c4e163104c7", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 8083, + "estimated_input_tokens": 2021, + "estimated_eval_tokens": 4298 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 8083, + "estimated_input_tokens": 2021, + "estimated_eval_tokens": 4298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three separate model-specific SDPA dispatch PRs, not duplicates of one another. They share a broad theme and similar wording, but they target different code paths and model families.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "No true canonical PR: each change is a distinct model-family implementation of SDPA dispatch, so they should not be deduplicated into one artifact.", + "best_issue_reason": null, + "best_pr_reason": "PR #46196 is the broadest representative, covering multiple related model families and the largest implementation scope, but it is still not a duplicate of the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46073", + "right": "pull_request:46154", + "accept": false, + "reason": "CTRL and T5 are separate architectures with different attention code paths; these are similar feature additions, not the same fix/change." + }, + { + "left": "pull_request:46073", + "right": "pull_request:46196", + "accept": false, + "reason": "This is CTRL-only vs wav2vec2_conformer/bert and SeamlessM4T/v2; the implementations and affected code paths are different and would not be merged into one PR." + }, + { + "left": "pull_request:46154", + "right": "pull_request:46196", + "accept": false, + "reason": "T5 SDPA dispatch and the audio/seq2seq model updates in #46196 are distinct model-specific changes, not one underlying bug or patch." + } + ] + }, + "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": "26176033073ffced1b5f25762be9e71da8d998b70527da563f12e6d620cf4d89", + "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 two PRs are not duplicates. PR #46126 is a focused feature/integration change around expert parallelism and Trainer wiring, while PR #46269 is a much broader distributed-branch base touching many unrelated distributed and model files plus CI. The shared filenames indicate subsystem overlap, not the same concrete fix or mergeable patch.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "#46126 is the more coherent, narrowly scoped PR with a clear user-facing purpose and concrete code path changes; #46269 is too broad and looks like an integration/base branch rather than a single fix.", + "best_issue_reason": null, + "best_pr_reason": "#46126 is the best representative PR for this cluster because it has a specific, well-defined change set and a clear title describing the feature being integrated.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46126", + "right": "pull_request:46269", + "accept": false, + "reason": "They overlap in distributed code, but the diffs are not the same underlying change: #46126 adds EP/Trainer integration, while #46269 is a broad distributed branch base with many extra files and CI changes. They 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": 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-46138-2", + "nodes": [ + "pull_request:46138", + "pull_request:46432" + ], + "soft_pairs": [ + "pull_request:46138|pull_request:46432" + ], + "prepared_review_unit_hash": "3da5e3ebfb97fcbf4839d8a7f1a4e62834ea63156ef215012c63c11127b7bad9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3863, + "estimated_input_tokens": 966, + "estimated_eval_tokens": 2188 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3863, + "estimated_input_tokens": 966, + "estimated_eval_tokens": 2188 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are unrelated workflow PRs: one changes how self-comment CI passes message input, the other hardens SSH runner setup by pinning a downloaded script and unsetting HF_TOKEN. They do not fix the same underlying problem.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46432 is the stronger representative because it makes the more substantive workflow hardening change, including supply-chain pinning and secret cleanup in ssh-runner.", + "best_issue_reason": null, + "best_pr_reason": "PR 46432 is the best overall PR in the cluster as a representative change: it is more concrete and security-relevant than the small CLI invocation adjustment in 46138.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46138", + "right": "pull_request:46432", + "accept": false, + "reason": "Different workflows and different concerns: stdin handling for self-comment CI versus pinning a remote script and clearing HF_TOKEN in ssh-runner. They are not the same underlying bug/change and would not 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": 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": "efc6146224f0a56f2e416449991d5e1f0574958a8b81dd251570a041d1b47a08", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4276, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4276, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs with a superficial similarity around avoiding file-open/source-reading patterns, but they target different code paths and different bugs: one in import/resource traversal, the other in attention-implementation capability detection.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46207 is the better representative of this cluster because it is a focused, self-contained fix in a single utility path with accompanying tests, and its title/body clearly describe the concrete behavior change.", + "best_issue_reason": null, + "best_pr_reason": "PR 46207 is the strongest standalone candidate overall: it addresses a specific source-reading heuristic in modeling_utils.py and appears easier to review/merge than the more infrastructure-oriented resource-resolution change in 46152.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46152", + "right": "pull_request:46207", + "accept": false, + "reason": "They both touch source-reading/open() patterns, but the underlying problems differ: 46152 is about resolving package resources without reading self source via file paths, while 46207 changes how model classes detect support for dynamic attention implementation. These are not the same bug 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": 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-46337-3", + "nodes": [ + "pull_request:46337", + "pull_request:46362", + "pull_request:46368" + ], + "soft_pairs": [ + "pull_request:46337|pull_request:46368", + "pull_request:46337|pull_request:46362", + "pull_request:46362|pull_request:46368" + ], + "prepared_review_unit_hash": "d6c4c5df68c32bf631d55c5a20f09278b880179e2fed0c4c4b0517c369b21413", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 5962, + "estimated_input_tokens": 1491, + "estimated_eval_tokens": 3238 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 5962, + "estimated_input_tokens": 1491, + "estimated_eval_tokens": 3238 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are not true duplicates. PRs 46362 and 46368 both skip a quantized-cache generation test, but they target different model families and likely different model-specific cache implementations. PR 46337 is a separate test-environment fix for Cohere2Moe device mapping/offload behavior.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46362 is the clearest representative of the quantized-cache test-skip pattern: it adds a focused skip for DeepSeek V4 due to an explicit cache incompatibility. The others are either for a different model family or a different failure mode entirely.", + "best_issue_reason": null, + "best_pr_reason": "PR 46362 is the best single representative because it is a minimal, explicit fix for a concrete quantized-cache incompatibility in one model test. PR 46368 is similar in shape but model-specific, and PR 46337 addresses a different device-map/offload test setup issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46337", + "right": "pull_request:46368", + "accept": false, + "reason": "Different concrete problems: 46337 tweaks Cohere2Moe test configuration to satisfy device-map/offload behavior, while 46368 simply skips a quantized-cache test for NemotronH. Same broad testing area, but not the same underlying bug or mergeable change." + }, + { + "left": "pull_request:46337", + "right": "pull_request:46362", + "accept": false, + "reason": "Not the same bug. 46337 is about Cohere2Moe failing hf_device_map/offload tests; 46362 skips a DeepSeek V4 quantized-cache test because its compressor/cache design is incompatible. Different models and different failure mechanisms." + }, + { + "left": "pull_request:46362", + "right": "pull_request:46368", + "accept": false, + "reason": "Both skip quant_cache tests, but for different model families with distinct cache implementations (DeepSeek V4 vs NemotronH). They are similar test workarounds, not the same concrete code-path problem, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 34, + "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": "2be7108f54a818583de6cb58a02df6fdff2c275dc1f1307d11e76c15f622a02e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5960, + "estimated_input_tokens": 1490, + "estimated_eval_tokens": 3236 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5960, + "estimated_input_tokens": 1490, + "estimated_eval_tokens": 3236 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The pair is not a duplicate cluster: PR #46376 adds multimodal support to continuous batching infrastructure, while PR #46417 adds the HunYuan VL model with docs, processors, modeling, and tests. They touch different code paths and represent separate changes.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "If one PR must represent the cluster, #46376 is the more central infrastructure change; #46417 is a separate model addition and does not overlap on the underlying bug/change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46376", + "right": "pull_request:46417", + "accept": false, + "reason": "Different scopes and code paths: continuous-batching multimodal support vs adding a new HunYuan VL model. They are not plausibly the same 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": 35, + "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-46465-2", + "nodes": [ + "pull_request:46465", + "pull_request:46470" + ], + "soft_pairs": [ + "pull_request:46465|pull_request:46470" + ], + "prepared_review_unit_hash": "850fb9d20bc32b5b1cf6f3167ba69738eb8afd6cd6671dcc48df3d5e22e5cc8e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4995, + "estimated_input_tokens": 1249, + "estimated_eval_tokens": 2754 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4995, + "estimated_input_tokens": 1249, + "estimated_eval_tokens": 2754 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests address the same decoder input-validation error-message bug in T5-family models. PR 46470 is the cleaner, test-backed fix; PR 46465 is a broader variant that changes more model files but targets the same underlying message formatting issue.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46470 is the best representative because it fixes the concrete decoder error-message wording in the core T5-family code path and includes tests, making it the more complete and precise change.", + "best_issue_reason": null, + "best_pr_reason": "PR 46470 is the strongest standalone PR: it directly targets the validation-message typo, is scoped to the affected T5-family models, and adds tests that confirm the behavior.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46465", + "right": "pull_request:46470", + "accept": true, + "reason": "Same underlying bug: decoder input-validation error messages use the wrong prefix formatting. The fixes operate on the same code path and could plausibly be combined into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs target the same shared issue target and make closely related decoder error-message wording fixes in overlapping T5-family code paths, with PR 46470 also adding tests. The soft-edge accept is conservative enough because the changes are variants of the same bugfix and could be consolidated into one PR." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/manifest.json b/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..5132d36e61177432ccda87179dd9bda946b263fa --- /dev/null +++ b/snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260607t120016z", + "artifacts": { + "hybrid": "snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-06-07T12:08:27Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260607T120016Z", + "variant": "hybrid" +} diff --git a/snapshots/20260607T120016Z/manifest.json b/snapshots/20260607T120016Z/manifest.json index 18201af76ed5f76b9a2e209e918949844dcdc134..05307c021877ce92d38f45c0029c058bddd212fe 100644 --- a/snapshots/20260607T120016Z/manifest.json +++ b/snapshots/20260607T120016Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 414 }, "extracted_at": "2026-06-07T12:00:16Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260607t120016z", + "runs": { + "hybrid-model-20260607t120016z": { + "analysis_id": "hybrid-model-20260607t120016z", + "artifacts": { + "hybrid": "snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260607T120016Z/analysis-runs/hybrid-model-20260607t120016z/manifest.json", + "model": null, + "published_at": "2026-06-07T12:08:27Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260607T120016Z", "watermark": {