diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index d6958a32fa6342ef682bf14ab43240d34c20ed3b..7e28c33b5b13dd1a9ecde8ea2b685360d796b07b 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": "20260520T060132Z", - "generated_at": "2026-05-20T06:12:48Z", + "snapshot_id": "20260520T120132Z", + "generated_at": "2026-05-20T12:09:56Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1709,273 +1709,45 @@ ] }, { - "cluster_id": "cluster-44018-2", + "cluster_id": "cluster-46100-2", "summary": "Cluster of 2 related pull requests linked by soft_similarity.", "status": "open", "confidence": 0.5, "canonical_issue_number": null, - "canonical_pr_number": 44068, + "canonical_pr_number": 46100, "issue_numbers": [], "pr_numbers": [ - 44018, - 44068 + 46100, + 46114 ], "evidence_types": [ "soft_similarity" ], "pr_comparisons": [ { - "left_pr_number": 44018, - "right_pr_number": 44068, - "code_similarity": 0.766, - "size_similarity": 0.939, + "left_pr_number": 46100, + "right_pr_number": 46114, + "code_similarity": 1.0, + "size_similarity": 1.0, "file_overlap": 1.0, - "area_overlap": 0.425, - "patch_similarity": 0.866, + "area_overlap": 1.0, + "patch_similarity": 1.0, "shared_filenames": [ - "src/transformers/models/gpt_neo/modeling_gpt_neo.py", - "tests/models/gpt_neo/test_modeling_gpt_neo.py" + "tests/models/vit/test_modeling_vit.py" ], "shared_file_areas": [ { - "filename": "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "filename": "tests/models/vit/test_modeling_vit.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 + 216, + 226 ] ], "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 + 216, + 226 ] ] } @@ -2039,15 +1811,6 @@ "target_issue_number": 43979, "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." }, - { - "cluster_id": "cluster-44018-2", - "canonical_pr_number": 44068, - "duplicate_pr_numbers": [ - 44018 - ], - "target_issue_number": null, - "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." - }, { "cluster_id": "cluster-45081-3", "canonical_pr_number": 45317, @@ -2083,6 +1846,15 @@ ], "target_issue_number": 46032, "reason": "PRs in cluster-46032-3 are treated as duplicates because they converge on issue #46032 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46100-2", + "canonical_pr_number": 46100, + "duplicate_pr_numbers": [ + 46114 + ], + "target_issue_number": null, + "reason": "PRs in cluster-46100-2 are treated as duplicates because they share soft_similarity evidence." } ], "best_issue": { diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 288374183884e12f51189a728ff997c6c0695368..3595b3b387dc200e8117a6aa66fa996d5cc3ecd7 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": "20260520T060132Z", - "generated_at": "2026-05-20T06:12:48Z", + "snapshot_id": "20260520T120132Z", + "generated_at": "2026-05-20T12:09:56Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -13,8 +13,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -140,7 +140,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -568,8 +568,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -667,7 +667,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -1035,12 +1035,12 @@ "issue:43881|issue:43883", "issue:43824|issue:43872" ], - "prepared_review_unit_hash": "bf949578801540b7549b606fe8b69096db7105588d04a3b88e78445b00c42cad", + "prepared_review_unit_hash": "2c53fb31ab352fd5406e8fb26dc1c10745e22839f880366414e9ecd6d80c0399", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -1057,226 +1057,223 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated Transformer issues were grouped together. The only clear duplicate subgroup is the non-persistent-buffer corruption regression around `from_pretrained`/Transformers 5.x, with one plausible SigLIP2 implementation-vs-quality pair. The rest are different models, code paths, or symptom classes and should stay separate.", - "confidence": 0.89, - "canonical_issue_reason": "issue:43950 is the clearest and most general statement of the non-persistent-buffer corruption regression, and it subsumes the broader 5.x buffer-junk behavior described by nearby duplicates.", + "summary": "The cluster is mostly heterogeneous Transformers bug reports, with one clear duplicate subcluster around non-persistent buffer corruption in v5. Most soft pairs are only loosely related by subsystem or version, not the same underlying bug.", + "confidence": 0.82, + "canonical_issue_reason": "issue:43950 is the strongest representative of the recurring v5 regression about non-persistent buffers being corrupted; it is the clearest, most general statement of that bug and subsumes the closer follow-ups (43643/44534).", "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the strongest representative issue in this set because it describes a concrete, reproducible regression with a clear failure mode and broad impact.", + "best_issue_reason": "issue:43950 has the broadest impact and best captures the repeated buffer-corruption regression theme seen in the cluster, making it the most suitable anchor issue.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "Different bugs: one is a DSA/indexing discussion, the other is stale docs about a removed command." + "reason": "One is a DSA/ReLU discussion; the other is a docs note about a removed run command. Same docs area, not the same bug." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "Both involve model/training failures, but the code paths and symptoms are different: FSDP weight shape error vs TRL embedding unpacking failure." + "reason": "FSDP weight-shape runtime error vs. Qwen3.5 embedding unpacking with TRL; different failing code paths." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization/offloading behavior is unrelated to the tied-weights regression in translation models." + "reason": "Quantization/offloading behavior vs. tied-weights issues in translation models; related theme, but not the same defect." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading and tied-weights translation-model issues are separate failures." + "reason": "NemotronH checkpoint loading failure is model-specific and not the same as tied-weights handling in v5." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "Both are docs-ish, but they concern different content and different defects." + "reason": "A cache-strategy typo and a documentation note about a removed command are unrelated." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "Different models and different failure mechanisms; not the same underlying bug." + "reason": "Qwen3.5 embedding unpacking with TRL is unrelated to GraniteMoEHybrid calling an invalid method." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Different model-specific inference issues with no shared concrete code path." + "reason": "Both involve model/runtime behavior, but one is detokenization and the other is an NVFP4 execution error; not the same bug." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Distinct model loading/runtime problems; not mergeable as one fix." + "reason": "Qwen-3-8B-NVFP4 runtime failure and GraniteMoEHybrid invalid method call are different concrete failures." }, { "left": "issue:43950", "right": "issue:44534", - "accept": true, - "reason": "Same underlying regression: non-persistent buffers get corrupted/junked in Transformers 5.x and during `from_pretrained`." + "accept": false, + "reason": "These are very close, but the wording suggests separate reports of the same regression rather than a clearly shared issue/PR target from the available data." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "A documentation/example mismatch is not the same as a SigLIP2 embedding-quality regression." + "reason": "SigLIP2 docs/example errors vs. text embedding quality regression; both about SigLIP2, but different problems." }, { - "left": "issue:43644", + "left": "issue:43643", "right": "issue:43950", - "accept": true, - "reason": "Both report the same non-persistent-buffer corruption regression in Transformers 5.x." + "accept": false, + "reason": "Both touch v5 configuration/buffer handling, but 43643 is remote-code config field loss while 43950 is buffer corruption." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Different numerical/attention failures: autocast dtype mismatch vs 3D position-id/SDPA collapse." + "reason": "Autocast dtype mismatch in Phi-tiny-MoE is not the same as Qwen3.5 position-id/SDPA issues." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "Trainer API keyword mismatch and ZeRO-3 loading index error are unrelated." + "reason": "Trainer compute_loss signature error and Gemma3 ZeRO-3 loading failure are unrelated." }, { "left": "issue:41628", "right": "issue:45800", "accept": false, - "reason": "Import surface breakage and torch/transformers incompatibility are different issues." + "reason": "Importing AutoImageProcessor and torch/transformers version incompatibility are different issues." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Checkpoint corruption and noisy logging/loading behavior are not the same bug." + "reason": "Corrupted checkpoint vs. generic loading noise; not the same root cause." }, { "left": "issue:43950", "right": "issue:45902", "accept": false, - "reason": "Related buffer materialization symptoms, but one is a general `from_pretrained` regression and the other is a specific Qwen3.5 MoE rotary-embedding memory issue." + "reason": "Both concern buffer/materialization state, but 45902 is Qwen3_5MoeTextRotaryEmbedding after to_empty(cuda), while 43950 is a broader v5 buffer corruption regression." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Different quantization/storage incompatibilities with different failure modes." + "reason": "bitsandbytes constructor-arg incompatibility and missing Float8 storage are separate backend errors." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Both concern SigLIP2 output quality, but the reports do not pin to the same concrete defect or code path." + "reason": "Unexpected zero probabilities and worse text embeddings for SigLIP2 are different symptoms with different likely causes." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad tokenizer error handling and a missing model attribute are unrelated." + "reason": "Bad AutoTokenizer error for Voxtral and Molmo tied-weights attribute error are unrelated." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality is not the same as a tokenization-version change." + "reason": "SigLIP2 text embedding quality and generic tokenization changes are not the same bug." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Both mention Qwen3.5/offload-ish behavior, but one is Flash Attention/position_ids and the other is CPU-offload compatibility in rotary embedding." + "reason": "Both involve Qwen3.5 and offload/position embeddings, but one is flash-attn crash from packed-sequence misinterpretation and the other is CPU-offload incompatibility." }, { "left": "issue:45663", "right": "issue:45902", "accept": false, - "reason": "Different model internals and different failure mechanisms." + "reason": "Gemma4 FSDP2 shared_kv_states bug and Qwen3_5Moe rotary-buffer initialization bug are different model-specific failures." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Tokenizer-loading diagnostics and glm-4v loading failure are separate issues." + "reason": "Voxtral tokenizer loading error and glm-4v-9b loading failure are distinct model-loading problems." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Training-memory inefficiency and int4 quantization OOM are different regressions." + "reason": "LoRA gradient checkpointing ineffectiveness and int4 CUDA reserved-memory growth are different training regressions." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "General loading noise vs inability to load a specific model are not the same bug." + "reason": "Generic loading noise versus inability to load a specific model; same area, not same bug." }, { "left": "issue:43054", "right": "issue:43493", - "accept": true, - "reason": "Likely the same SigLIP2 discrepancy: one reports worse text embeddings, the other explicitly compares HF vs original JAX implementation." + "accept": false, + "reason": "Both are SigLIP2-related, but one is embedding quality and the other is HF vs JAX implementation discrepancy." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "Both involve `remote_code`, but one is missing fields from `AutoConfig` and the other is a broader breakage report." + "reason": "Remote-code config field loss and broader breakage with recent transformers versions are not the same concrete failure." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Different model/output issues: detokenization bug vs Python 3.13 import error." + "reason": "Detokenization regression and Python 3.13 flash_attn import KeyError are unrelated." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Unnecessary loading noise and model load failure are distinct symptoms with no shared root cause shown." + "reason": "Both mention model loading problems, but they are for different checkpoints and likely different root causes." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "Both mention memory, but the mechanisms differ: ineffective gradient checkpointing vs MoE training memory usage." + "reason": "Gradient checkpointing/LoRA and Qwen3 MoE memory usage are separate training concerns." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "DeepSpeed ZeRO-3 loading failure and Qwen3 MoE FP8 failure are unrelated." + "reason": "Gemma3 ZeRO-3 loading failure and Qwen3Moe FineGrainedFP8 failure are different model/backend paths." }, { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "MPS histogram support and Gemma4 device-mismatch are different backend/model bugs." + "reason": "MPS histogram int error in MoE and Gemma4 PLE device mismatch under auto device map are unrelated." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "Different model loading failures with no shared code path." + "reason": "glm-4v-9b loading failure and Molmo missing all_tied_weights_keys are different model-specific bugs." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Import surface issue and bitsandbytes constructor incompatibility are unrelated." + "reason": "Qwen2.5-VL import failure and bitsandbytes _is_hf_initialized incompatibility are unrelated import/runtime issues." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The grouping is grounded in the packet, and the verdicts stay mostly conservative. The non-persistent-buffer regression cluster is well supported by the issue titles, and the SigLIP2 pairing is phrased cautiously enough to be acceptable. No mergeability problems are evident here." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -1288,8 +1285,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -1381,7 +1378,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -1614,8 +1611,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -1707,7 +1704,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -1943,8 +1940,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -2034,7 +2031,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -2261,8 +2258,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -2355,7 +2352,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -2684,12 +2681,12 @@ "issue:45084|issue:45507", "issue:43262|issue:43377" ], - "prepared_review_unit_hash": "ec121e942cb2344b673e593bce5ff5e6cef1e9d673ccafda9cbe3e0d157cf354", + "prepared_review_unit_hash": "7ccfce1e8308c630c09e85b4477a6e6dd10330864cccd7056869ba2c97ffa629", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -2706,204 +2703,204 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous set of unrelated Transformers regressions. The only plausible duplicate family is the GGUF-loading breakage; the rest are different subsystems, models, or failure modes and should stay separate.", - "confidence": 0.36, - "canonical_issue_reason": "issue 43452 is the broadest and most explicit report of the GGUF/from_pretrained regression, so it best represents the one plausible duplicate family in this cluster.", + "summary": "These issues are mostly unrelated model- or feature-specific regressions. I did not find any clear PRs, and nearly all soft pairs are too broad or touch different code paths to be treated as duplicates. The only plausible near-duplicate is the GGUF loading regression pair.", + "confidence": 0.44, + "canonical_issue_reason": "Issue 43644 is the broadest concrete regression in the set and is centered on core Transformers v5 initialization/state handling, making it the most reasonable anchor issue among these disparate reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue 43452 is the best overall representative because it describes the underlying GGUF loading breakage at the API level, rather than a model-specific symptom.", + "best_issue_reason": "Issue 43644 is the best representative because it describes a clear, upstream v5 regression with a core runtime symptom rather than a model-specific edge case.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper pipeline return_language and CamemBERT masked-LM predictions are unrelated bugs in different code paths." + "reason": "Whisper pipeline language return vs CamemBERT MLM predictions are different models and different failures." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "Qwen3 MoE training memory usage and mmap OOM on Strix Halo are different resource issues with no shared bug surface." + "reason": "MoE training memory usage and Strix Halo mmap OOM are unrelated causes and code paths." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Tokenizer loading failure for EMBEDDIA/est-roberta and a Mistral regex patch crash are different tokenizer regressions." + "reason": "Both are tokenizer-load regressions, but they involve different models and different tokenizer fix paths." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "CamemBERT masked-LM prediction errors and tokenizer-loading failure are distinct model/tokenizer behaviors." + "reason": "Incorrect masked-LM predictions is not the same bug as tokenizer loading failure." }, { "left": "issue:44737", "right": "issue:45859", "accept": false, - "reason": "XLNet CPU tensor placement and Qwen3.5 MoE CPU-offload incompatibility are different implementation bugs." + "reason": "Both mention device/offload behavior, but XLNet positional encoding and Qwen3.5 MoE rotary embedding are distinct implementations." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer GPU broadcast TypeError and Whisper return_language pipeline behavior are unrelated." + "reason": "Trainer GPU broadcast TypeError and Whisper return_language failure are unrelated." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "save/from_pretrained extra_state handling and Qwen video-input regression are different serialization vs multimodal-input issues." + "reason": "Save/from_pretrained extra_state handling and video input regression affect different subsystems and symptoms." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "Tokenizer mapping/model addition failure and flex_attention compilation on torch>=2.9 are unrelated changes." + "reason": "Tokenizer mapping change failure and compiled flex_attention failure are unrelated issues." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Loading sleng-bert and the Mistral regex patch crash are separate tokenizer-loading failures." + "reason": "Both are loading problems, but they concern different models and different tokenizer-loading logic." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and _is_hf_initialized regression do not share the same code path." + "reason": "Apple Silicon allocator warmup TypeError is unrelated to the _is_hf_initialized flag regression." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "base_model_prefix loading regression and non-persistent buffer junk are different v5 breakages." + "reason": "base_model_prefix loading breakage and non-persistent buffer junk are separate v5 regressions." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "High perplexity on gpt-oss-20b and runtime special-token post-processor updates are unrelated." + "reason": "High perplexity on a model and tokenizer post-processor update behavior are not the same bug." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Mistral image dtype mismatch and template-node compilation failure are different subsystems." + "reason": "Image preprocessor dtype mismatch and template compilation error are unrelated." }, { "left": "issue:44206", "right": "issue:44220", "accept": false, - "reason": "Both are audio-feature related, but the titles do not establish the same concrete failure; treat as separate until the exact call path matches." + "reason": "Both are audio-feature related, but the titles point to different failing functions and no shared concrete bug is evident." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Loading sleng-bert and transformers>=5.1.0 failing on physical-intelligence/fast are different model-loading regressions." + "reason": "Different model-loading failures; same broad theme is not enough to merge." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "AutoConfig trust_remote_code missing fields and PagedAttentionCache linear_attention crash are unrelated." + "reason": "AutoConfig trust_remote_code field loss and PagedAttentionCache linear_attention crash are unrelated." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "T5 tokenizer special-token conflict and Qwen3Moe FineGrainedFP8Config failure are different model-specific bugs." + "reason": "T5 tokenizer special-token conflict and Qwen3Moe FP8 failure are different code paths." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "TrainerState token counting and Qwen2.5-VL iterable multi-image finetuning are unrelated training issues." + "reason": "TrainerState token counting and multi-image iterable dataset finetuning are unrelated." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Whisper batch_decode skip_special_tokens and Whisper empty-transcription generation after align_special_tokens are related only by model family, not the same bug." + "reason": "Whisper batch_decode skip_special_tokens handling and empty-transcription generation after align_special_tokens are related only at a high level, not the same concrete bug." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Molmo tied-weights attribute error and sleng-bert loading failure are unrelated." + "reason": "Molmo tied-weights attribute error and BERT model loading failure are unrelated." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Loading vesteinn/ScandiBERT and loading physical-intelligence/fast are different model-specific failures." + "reason": "Different model-loading regressions with different affected models and likely different root causes." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Mixtral grouped_mm tracing error and template-node compilation error are unrelated code paths." + "reason": "Mixtral torch-dynamo grouped_mm tracing failure is unrelated to the template-node compile error." }, { "left": "issue:43452", "right": "issue:43482", "accept": true, - "reason": "Both describe transformers v5 GGUF/from_pretrained loading regressions and likely share the same gguf_file code path." + "reason": "Both reports are about GGUF loading breaking in transformers v5, and the titles suggest the same gguf_file/from_pretrained code path." }, { "left": "issue:44464", "right": "issue:45987", "accept": false, - "reason": "Chunked generation with compiled forward and StaticCache seq-length type mismatch are separate generation/cache issues." + "reason": "Chunked generation with compiled forward and StaticCache seq-length return type are different bugs." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "Qwen3Moe FP8 training failure and Qwen3.5 PagedAttentionCache crash are different model/runtime bugs." + "reason": "Qwen3Moe FP8 training failure and Qwen3.5 PagedAttentionCache crash are separate issues." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Non-persistent buffer junk and custom model initialization breakage are both v5 regressions but not the same concrete bug." + "reason": "Both are v5 init regressions, but one is about buffer initialization and the other about custom model initialization; the concrete bug is not the same." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "GPU+CPU offload dequantized-model loading and GGUF loading failure are different loading paths." + "reason": "gpu+cpu offload dequantized-model loading and GGUF loading failure are not the same specific bug." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Qwen3.5 output_hidden_states regression and Qwen2.5-VL rope-index scaling bug are unrelated model behaviors." + "reason": "Qwen3.5 hidden-states output bug and Qwen2.5-VL RoPE position-id scaling are unrelated." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "T5 tokenizer special-token conflict and notebook custom-model initialization failure are unrelated." + "reason": "Tokenizer special-token conflict and Jupyter custom-model initialization are different failure modes." }, { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Template compilation error and GraniteMoEHybrid invalid method call are different bugs." + "reason": "Template compilation error and GraniteMoEHybrid invalid method call are unrelated." }, { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Audio processor sampling-rate default and MIMI batched-vs-single padding-mask mismatch are different audio issues." + "reason": "Audio processor sample-rate default and MIMI encoder padding-mask behavior are different bugs." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The lone accepted pair, issue:43452 and issue:43482, is plausibly the same GGUF/from_pretrained loading regression, and the rest of the soft-edge verdicts are appropriately treated as separate bugs based on the available titles." + "feedback": "The assessment is conservative and grounded in the packet. The issue pair verdicts generally match the titles/scope, and the summary correctly avoids overclaiming by treating the GGUF pair as only a plausible near-duplicate rather than a definite match. No PR pair evidence is present in the packet, so the lack of PR conclusions is appropriate." }, "evaluator_used": true, "retried": false, @@ -3008,12 +3005,12 @@ "issue:44683|issue:45020", "issue:44403|issue:44991" ], - "prepared_review_unit_hash": "ecb6f9eb34e1bfaec12fd6b66801fafb6e4973526c2585d667b77aac01873997", + "prepared_review_unit_hash": "574dfc8082600ec93f5f9db9b5d228bbd78ddeb261a0411167f8339b86835b2d", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -3030,219 +3027,219 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated Transformers regression reports. The main correction is to be more conservative on the Qwen video/rope-position pair: the titles suggest thematic similarity, but not enough evidence that they are the same concrete bug or one mergeable fix.", - "confidence": 0.78, - "canonical_issue_reason": "issue:45020 is the broadest umbrella item here: it captures a recent Transformers-version regression breaking remote-code model loading, which matches the recurring v5 compatibility/load-failure theme across the set.", + "summary": "This cluster is mostly a grab bag of Transformers v5-era regression reports; only the Qwen3.5 save_pretrained pair looks like the same underlying bug. The rest are superficially similar but target different models, code paths, or failure modes.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43646 is the best canonical issue: it is a concrete, broadly relevant Transformers 5.0.0 regression about custom model initialization, and its wording matches several nearby init/loading breakage reports better than the more model-specific tickets.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the clearest representative of the recurring v5 load/serialization breakages and has the widest scope among the listed issues.", + "best_issue_reason": "issue:43646 is also the best standalone issue to keep because it is specific enough to be actionable while still serving as the cleanest anchor for closely related v5 initialization regressions.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43089", "right": "issue:45987", "accept": false, - "reason": "Different problems: generation runtime overhead vs. a StaticCache length type contract." + "reason": "Different symptoms and code paths: generation performance overhead vs a StaticCache return-type contract bug." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "Different failure modes: missing tied-weights attribute vs. tokenizer loading error for a different model." + "reason": "Unrelated failures: missing all_tied_weights_keys on Molmo vs tokenizer loading for EMBEDDIA/est-roberta." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Both involve tied weights, but one is a model-specific weight-tying bug and the other is a save_pretrained serialization inconsistency; not the same concrete bug." + "reason": "Both mention tied weights, but one is a model-specific untied lm_head bug and the other is save_pretrained serialization varying by device." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Both are model-load regressions, but for different models and different root causes." + "reason": "Different affected models and likely different regressions; both are loading failures but not the same bug." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Different code paths: bad model outputs vs. processor/model input-size mismatch." + "reason": "Different problem classes: nonsensical model outputs vs an image processor size mismatch." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Custom model init regression and junk non-persistent buffer initialization are distinct issues." + "reason": "Both are v5 regressions, but one is custom model initialization and the other is non-persistent buffer corruption; not the same underlying bug." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Both are multimodal loading/runtime failures, but they affect different models with different errors." + "reason": "Different multimodal failures affecting different model families and different stages of loading/processing." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "Generic loading noise is unrelated to the Mistral tokenizer patch crash." + "reason": "Generic loading noise issue vs a specific Mistral tokenizer regex crash." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "MIMI batching/padding bug is unrelated to DAC latent reconstruction mismatch." + "reason": "Different audio-model correctness bugs: batching/padding-mask mismatch vs DAC latent conversion mismatch." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "Qwen3-MoE FP8 config failure and template-node compilation error are different bugs." + "reason": "FP8 config failure on Qwen3Moe is unrelated to the template-node compilation error." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export exportability issue is unrelated to an OmDet processor input-size bug." + "reason": "torch.export/exportability bug vs processor output-size mismatch; different layers and failure modes." }, { "left": "issue:43645", "right": "issue:44534", "accept": false, - "reason": "Notebook custom-model init regression and non-persistent buffer corruption are not the same issue." + "reason": "Notebook-specific custom model init regression is not the same as junk-filled non-persistent buffers." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "Audio feature extraction bug is unrelated to compiled flex_attention failing on newer torch." + "reason": "Audio feature extraction bug vs compiled flex_attention failure on newer torch." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "ScandiBERT load failure does not share the same tokenizer regex crash path." + "reason": "Different models and different tokenizer/config loading failures." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Different affected models and different loading failures." + "reason": "Different regressions: model loading for one checkpoint vs a Mistral tokenizer regex patch crash." }, { "left": "issue:45216", "right": "issue:45357", "accept": true, - "reason": "Same underlying Qwen3.5 save_pretrained regression: 45357 is a later report of the same incorrect checkpoint/visual-key serialization problem." + "reason": "Same underlying Qwen3.5 save_pretrained regression; 45357 looks like a narrower follow-up about incorrect visual encoder keys." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Mistral3 tied-weight loading bug and LoRA merge collapse are different code paths." + "reason": "Both involve weights, but one is a missing tie in a specific model and the other is LoRA merge collapse after vocabulary extension." }, { "left": "issue:42915", "right": "issue:44291", "accept": false, - "reason": "Qwen3-MoE FP8 config failure is unrelated to init_empty_weights/_is_hf_initialized argument handling." + "reason": "Both are model-loading problems, but FP8 config failure is unrelated to the init_empty_weights/_is_hf_initialized argument bug." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "_is_hf_initialized flag regression is unrelated to mmap/OOM on Strix Halo." + "reason": "Different root causes: transformers v5 flag handling vs mmap-related OOM avoidance on Strix Halo." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "Tokenizer add_special_tokens conflict and MobileLLM key error are different loading bugs." + "reason": "Tokenizer constructor conflict vs a KeyError when loading a different model." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "AddedToken special-argument duplication is unrelated to model loading failure for physical-intelligence/fast." + "reason": "Generic AddedToken constructor issue is not the same as the physical-intelligence/fast loading failure." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "Qwen3-MoE FP8 config and Float8 storage deserialization are different issues." + "reason": "Both mention fp8/floating-point storage, but one is a config compatibility issue and the other is a missing storage type error." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL loading failure and Qwen3-MoE sliding_window handling are unrelated." + "reason": "Different model families and different failure modes: PaddleOCR-VL load error vs Qwen3-MoE sliding_window behavior." }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Sliding-window logic for Qwen3-MoE is unrelated to strict config loading for granite_speech." + "reason": "Sliding-window regression in Qwen3-MoE is unrelated to strict config blocking granite_speech loading." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch version compatibility and loading noise are not the same underlying bug." + "reason": "Torch version incompatibility is not the same as the transformer-loading noise report." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "base_model_prefix compatibility and non-persistent buffer junk are different v5 regressions." + "reason": "Base-model-prefix loading breakage and junk non-persistent buffers are distinct v5 regressions." }, { "left": "issue:44479", "right": "issue:45325", "accept": false, - "reason": "The titles suggest a Qwen vision/position-id theme, but there is not enough evidence they describe the same concrete bug or a single mergeable fix." + "reason": "Both touch Qwen2.5-VL/Qwen3-VL media position handling, but one is a broad video-input regression and the other is a specific still-image temporal position_ids scaling bug." }, { "left": "issue:36032", "right": "issue:41720", "accept": false, - "reason": "Tokenizer AttributeError and Qwen3 auto device map cudaErrorAssert are unrelated." + "reason": "Tokenizer load-time AttributeError and Qwen3 auto device mapping cudaErrorAssert are unrelated." }, { "left": "issue:42175", "right": "issue:45542", - "accept": true, - "reason": "Same backend packaging issue: the torch extra installs tensorboard without TensorFlow, leading to missing/undefined tf backend errors." + "accept": false, + "reason": "These are both backend/dependency complaints, but one is about missing TensorFlow in a torch install and the other is tensorboard installing without TensorFlow." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "Whisper runtime failure is unrelated to Float8 storage deserialization." + "reason": "Different issues: Whisper runtime failure vs missing Float8 storage type." }, { "left": "issue:40990", "right": "issue:44811", "accept": false, - "reason": "Perplexity regression and Whisper batch_decode skip_special_tokens behavior are unrelated." + "reason": "Perplexity quality issue is unrelated to Whisper batch_decode skipping special-token handling." }, { "left": "issue:44683", "right": "issue:45020", "accept": false, - "reason": "Compiled flex_attention torch-version breakage is not the same as the remote_code model-loading regression." + "reason": "Compiled flex_attention on new torch and remote_code loading breakages are different bugs." }, { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "Generic loading noise does not match the EMBEDDIA tokenizer loading failure." + "reason": "Unnecessary loading noise is not the same as tokenizer load failure for EMBEDDIA/est-roberta." } ] }, "evaluator_result": { - "accept": false, - "feedback": "Mostly grounded overall, but one soft-edge accept is too aggressive: issue:44479 vs issue:45325 is only a thematic Qwen vision/video-position similarity from titles, not enough evidence that they are the same concrete bug or mergeable into one fix. The other accepted pairs look better supported, but because the packet does not ground that Qwen pair strongly, the result is not conservative enough." + "accept": true, + "feedback": "The analysis stays grounded in the packet and the soft-edge calls are conservative. The selected canonical issue, issue:43646, is a reasonable generic anchor for the v5 custom-init regressions, and the only accepted pair (issue:45216 / issue:45357) is well supported by the titles. I don\u2019t see evidence of overstatement." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -3255,8 +3252,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -3347,7 +3344,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -3574,8 +3571,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -3666,7 +3663,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -3893,8 +3890,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -3986,7 +3983,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -4219,8 +4216,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -4307,7 +4304,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -4513,8 +4510,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -4604,7 +4601,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -4908,12 +4905,12 @@ "issue:45584|issue:45834", "issue:43825|issue:43994" ], - "prepared_review_unit_hash": "e6627d0da56686098dad564913625f8b9a5d36730f6d24ac723ff30ee141acf6", + "prepared_review_unit_hash": "dcc36a166ffc30ba8ffce8c037835a67b6c6fabf44c9ddf4e56909b5b382e330", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -4930,180 +4927,174 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly a grab-bag of unrelated Transformers bug reports. A few pairs are loosely thematically related (remote_code regressions, Mistral regex handling, quantization/offloading, cache behavior), but none look like the same concrete underlying bug. The broadest representative issue is the remote_code regression report.", - "confidence": 0.66, - "canonical_issue_reason": "issue:45020 is the broadest umbrella report here: it covers recent-version breakage affecting multiple remote_code models, which makes it the best cluster anchor among mostly specific regressions.", + "summary": "This is not a true duplicate cluster: the items span many unrelated regressions across tokenizers, vision, quantization, model loading, and training APIs. The soft pairs are mostly thematic similarities, not the same concrete bug or change.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the most globally representative because it is broad, user-facing, and plausibly subsumes several of the remote_code-related regressions in the set.", + "best_issue_reason": "issue:45020 is the broadest, most cluster-central report (remote_code regressions in recent transformers versions), so it is the best representative issue, but it does not actually cover the rest of the heterogeneous set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43388", "right": "issue:46032", "accept": false, - "reason": "Different subsystems and failures: metric gathering/label truncation vs. Mamba2 cache correctness." + "reason": "Different bugs: gather_for_metrics label truncation vs Mamba2 use_cache corruption." }, { "left": "issue:42915", "right": "issue:43746", "accept": false, - "reason": "Unrelated models and bugs: Qwen3Moe FP8 loading vs. GraniteSpeech PEFT checkpoint loading." + "reason": "Different model families and failures: Qwen3Moe FP8 config vs GraniteSpeech PEFT checkpoint loading." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "Both are model-specific bugs, but one is SwitchTransformers config construction and the other is Smollm3 RoPE layer handling." + "reason": "Different config/model behaviors: sparse encoder layer creation vs Smollm3 RoPE layer drop." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Different modalities and code paths: image preprocessor dtype mismatch vs. Gemma3n audio tensor name collision." + "reason": "Both are dtype/vision-adjacent, but one is image preprocessor mismatch and the other is Gemma3n audio mask name collision." }, { "left": "issue:43295", "right": "issue:45834", "accept": false, - "reason": "Different bug classes: processor/tokenizer regression vs. Kosmos2.5 long-OCR indexing error." + "reason": "Different regressions: processor.tokenizer/image forwarding vs long OCR index error." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "Unrelated failures: torch.compile+SDPA on Bamba vs. bfloat16 dtype mismatch in other models." + "reason": "Different code paths: torch.compile + SDPA bug vs bfloat16 inference dtype mismatch." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "No shared code path: compile/SDPA failure vs. DAC latent reconstruction mismatch." + "reason": "Different failure modes: SDPA/compile issue vs DAC.from_latents vs forward mismatch." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Local PEFT checkpoint loading is unrelated to quantization offloading behavior." + "reason": "Different root causes: PEFT local checkpoint loading vs quantization offloading behavior." }, { "left": "issue:43950", "right": "issue:45987", "accept": false, - "reason": "Different contracts: buffer loading corruption vs. cache sequence-length return type." + "reason": "Different API contracts: non-persistent buffer corruption vs StaticCache.get_seq_length return type." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "Tokenizer/padding crash in LayoutLMv2 is unrelated to Perceiver image-size interpolation failure." + "reason": "Different models and symptoms: LayoutLMv2 tokenizer crashes vs Perceiver image classification with non-default resolution." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "Tokenizers.AddedToken argument collision is not the same as failing to load a specific model/tokenizer." + "reason": "Different tokenizer issues: AddedToken special arg collision vs failure to load a specific model." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Both touch _patch_mistral_regex, but one is offline network access and the other is an attribute error on a different tokenizer object." + "reason": "Different offline/regression bugs: unconditional model_info call vs Mistral regex patch crashing on tokenizer internals." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "Tokenizer special-token registration bug is unrelated to DAC forward/from_latents mismatch." + "reason": "Different tokenizer/model areas: BigBirdTokenizer special-token registration vs DAC latent/forward mismatch." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Different areas: Qwen3-MoE sliding window vs. AutoConfig missing fields under trust_remote_code." + "reason": "Both involve loading/configuration, but one is sliding_window behavior in Qwen3-MoE and the other is missing fields from trust_remote_code AutoConfig." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "Both involve config loading, but one is missing fields from trust_remote_code and the other is strict config rejection for granite_speech." + "reason": "Related to config loading, but not the same bug: missing fields in AutoConfig return vs strict granite_speech config rejection." }, { "left": "issue:43540", "right": "issue:45381", "accept": false, - "reason": "Both are video-input issues, but they affect different model families and likely different preprocessing bugs." + "reason": "Different multimodal bugs: video input processing error in Qwen3OmniMoe vs incorrect vision_position_ids in Qwen2.5-VL." }, { "left": "issue:43577", "right": "issue:43818", "accept": false, - "reason": "Different model families and symptoms: BLIP2 dtype persistence vs. Video-LLaVA tower/temporal-attention issue." + "reason": "Different model components: BLIP2 dtype loading issue vs Video-LLaVA missing temporal attention/shared weights." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Qwen3VL from_pretrained weight_only failure is distinct from Qwen3.5 flash-attention generation behavior." + "reason": "Different generation failures in different models: Qwen3VL weight_only load error vs Qwen3.5 flash-attention generation issue." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Separate model regressions: Pegasus output drift vs. CamemBERT masked LM predictions." + "reason": "Both are output-quality regressions, but Pegasus v4/v5 differences and CamemBERT masked LM predictions are separate model-specific bugs." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Special-token post-processor update bug is different from Whisper batch_decode skip_special_tokens handling." + "reason": "Related to tokenization post-processing, but one is runtime special-token updates and the other is Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Image processor void segmentation maps have nothing to do with CustomTrainer compute_loss signature changes." + "reason": "Different domains and APIs: image processor void segmentation maps vs CustomTrainer compute_loss signature mismatch." }, { "left": "issue:44561", "right": "issue:45020", "accept": false, - "reason": "44561 is one specific removed helper causing remote_code breakage; 45020 is a broad umbrella report and not the same concrete bug." + "reason": "Both mention remote_code/trust_remote_code regressions, but they are different concrete breakages in different code paths." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "AutoImageProcessor URL loading and template-node compilation errors are unrelated." + "reason": "Different loader/template problems: AutoImageProcessor URL loading vs template compilation error." }, { "left": "issue:39692", "right": "issue:45834", "accept": false, - "reason": "SigLIP2 docs/quantization example errors are unrelated to Kosmos2.5 long OCR indexing." + "reason": "Different problems: SigLIP2 doc/example issues and quantization failure vs Kosmos2.5 long OCR input index error." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "Different topic scope: GLM-5 RoPE implementation discussion vs docs still mentioning removed pipeline tasks." }, { "left": "issue:45584", "right": "issue:45834", "accept": false, - "reason": "Whisper empty-transcription generation bug is unrelated to Kosmos2.5 OCR input length handling." + "reason": "Different multimodal failures: Whisper empty transcription after alignment vs Kosmos2.5 OCR input length indexing." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "Pipeline error messaging and SigLIP2 nonsensical outputs are different issues." - }, - { - "left": "issue:45987", - "right": "issue:46032", - "accept": false, - "reason": "Cache length return-type bug is distinct from Mamba2 use_cache correctness for seq_len>1." - }, - { - "left": "issue:44843", - "right": "issue:45081", - "accept": false, - "reason": "Offline model_info() call and backend_tokenizer attribute access are different Mistral-regex failure modes, not one bug." + "reason": "Different issue types: pipeline translation-task messaging vs SigLIP2 model/pipeline nonsensical results." } ] }, @@ -5122,8 +5113,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -5209,7 +5200,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -5406,8 +5397,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -5497,7 +5488,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -5806,12 +5797,12 @@ "issue:43646|issue:43749", "issue:36010|issue:42175" ], - "prepared_review_unit_hash": "081daf439202ab801a31cf93ff834ce6e23de1af0bbce07832772baef4bf5512", + "prepared_review_unit_hash": "59341eb4ec80f21001fbc787035abcac0d4a65ae029a8d4f7a704686697ad1f8", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -5828,210 +5819,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous issue cluster with two plausible duplicate sub-pairs: Qwen2.5-VL video regression and SigLIP2 output discrepancy. The rest are distinct bugs spanning unrelated models, loading paths, tokenizer behavior, and runtime environments.", - "confidence": 0.57, - "canonical_issue_reason": "Issue 44479 is the best representative because it describes the broad v5.3.0 video-input regression across multiple Qwen VL variants, and it has a strong near-duplicate in 45381.", + "summary": "This set is mostly heterogeneous, with two plausible duplicate families standing out: SigLIP2 output-correctness reports and Qwen video-input regression reports. Everything else appears to be different bugs touching different code paths.", + "confidence": 0.64, + "canonical_issue_reason": "Issue 44479 is the broadest concrete regression report and has the clearest duplicate-like companion (45381); it is the best single anchor if one issue must be chosen from this mixed set.", "canonical_pr_reason": null, - "best_issue_reason": "Same reason as canonical: it is the broadest, clearest regression report in the cluster and best captures the shared Qwen video-input failure mode.", + "best_issue_reason": "Issue 44479 is the most representative actionable issue here because it describes a concrete, versioned regression with multiple affected models and a clear fix surface.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Different problems: Qwen3 MoE weight conversion with accelerate+deepspeed vs Qwen3-VL weight-shape mismatch on load." + "reason": "Different Qwen3-related loading failures: MOE weight conversion under accelerate+deepspeed vs Qwen3-VL shape mismatch." }, { "left": "issue:44291", "right": "issue:45042", "accept": false, - "reason": "Unrelated failures: init_empty_weights/_is_hf_initialized vs PIL image processors incorrectly requiring torchvision." + "reason": "Unrelated bugs: init_empty_weights argument handling vs PIL image processors incorrectly requiring torchvision." }, { "left": "issue:43452", "right": "issue:43582", "accept": false, - "reason": "GGUF tokenizer/model loading issue vs Apple Silicon caching_allocator_warmup TypeError; different code paths." + "reason": "Tokenizer/model gguf_file loading breakage and Apple Silicon allocator warmup TypeError are different code paths." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Different bugs on different models: FineGrainedFP8Config for Qwen3Moe vs MobileLLM load key error." + "reason": "Different failures on different models: FineGrainedFP8Config on Qwen3Moe vs a key error loading MobileLLM-125M." }, { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Qwen3 auto device-mapping CUDA assert and Qwen3OmniMoe video processing error are distinct failures." + "reason": "Both are Qwen-family failures, but one is a device-mapping assert on A800 and the other is a video-input ValueError in Qwen3OmniMoe; not the same bug." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "MacOS CLIP bus error and TrainerState token counter not updating are unrelated." + "reason": "Completely different symptoms and subsystems: MacOS bus error on CLIP vs TrainerState token counter not updating." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "FP8 quantization failure vs AutoImageProcessor URL loading; no shared underlying bug." + "reason": "Qwen3Moe FP8 loading and AutoImageProcessor URL loading are unrelated." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "DiaConfig save/load token-ID loss is unrelated to MXFP4 quantization fallback." + "reason": "Config save/load custom token IDs and MXFP4 quantization fallback are different issues." }, { "left": "issue:44987", "right": "issue:45538", "accept": false, - "reason": "Model-loading regression vs CLIPTokenizer model_max_length behavior; different components." + "reason": "Tokenizer loading regression for a specific model and CLIPTokenizer max_length bug are unrelated." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Qwen3Moe FP8 loading and CLIPTokenizer max length are unrelated." + "reason": "Different model families and failures; no shared code path." }, { "left": "issue:44479", "right": "issue:45381", "accept": true, - "reason": "Both describe the same 5.3.0 Qwen video-input regression, with 45381 reading like a narrower repro of the same vision_position_ids bug." + "reason": "Both describe the same Qwen video-input regression in 5.3.0; 45381 is a narrower report of the wrong vision_position_ids path covered by 44479." }, { - "left": "issue:44987", - "right": "issue:45907", + "left": "issue:43906", + "right": "issue:45381", "accept": false, - "reason": "Model loading failure vs repo-file revision forwarding in a CI helper; unrelated." + "reason": "One is an isolated reproduction of issue 38071; the other is a Qwen2.5-VL video-position bug. Not the same underlying bug." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Both involve video inputs, but they target different model families and different concrete failures." + "reason": "Both involve video inputs, but the models and reported failures differ (Qwen3OmniMoe ValueError vs broader Qwen2.5/Qwen3 video regression)." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "Dequantized-model offload/load failure is unrelated to CLIPTokenizer max_length." - }, - { - "left": "issue:43257", - "right": "issue:43696", - "accept": false, - "reason": "Qwen3 MoE weight conversion bug and GPT-oss OOM are different issues." - }, - { - "left": "issue:43906", - "right": "issue:45381", - "accept": false, - "reason": "One is only a reproduction of issue 38071; the titles do not indicate the same Qwen2.5-VL regression." + "reason": "Device-map/offload loading failure and CLIPTokenizer model_max_length bug are unrelated." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "SigLIP2 zero probabilities and Perceiver non-default-resolution failure are unrelated model bugs." + "reason": "Different models and failure modes: SigLIP2 zero probabilities vs Perceiver resolution/interpolation failure." }, { "left": "issue:43577", "right": "issue:45072", "accept": false, - "reason": "Different dtype/mismatch issues in different models: Blip2 vs SwitchTransformers/TimmWrapperModel." + "reason": "BLIP2 dtype propagation and SwitchTransformers/TimmWrapperModel bfloat16 mismatch are different bugs." }, { "left": "issue:43404", "right": "issue:45987", "accept": false, - "reason": "Mistral3 lm_head tying bug vs StaticCache sequence-length contract issue; unrelated." - }, - { - "left": "issue:45137", - "right": "issue:45405", - "accept": false, - "reason": "DeepSpeed ZeRO3 deque error is unrelated to a PEFT version bump." + "reason": "Mistral3 lm_head tying bug and StaticCache seq-length contract issue are separate code paths." }, { "left": "issue:24643", "right": "issue:29159", "accept": false, - "reason": "DeepSpeed training RuntimeError and tokenizer inconsistency are unrelated." + "reason": "DeepSpeed training weight-shape runtime error and tokenizer slow/fast inconsistency are unrelated." }, { "left": "issue:44737", "right": "issue:45588", "accept": false, - "reason": "XLNet CPU arange bug and flash-attention sink-less crash are different code paths." + "reason": "XLNet CPU tensor creation bug and flash-attention sink handling crash are unrelated." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "MobileLLM load key error and init_empty_weights unexpected-arg error are distinct loading bugs." + "reason": "Model loading key error vs init_empty_weights argument mismatch; different underlying problems." }, { "left": "issue:43257", "right": "issue:43756", "accept": false, - "reason": "Qwen3 MoE weight conversion and Smollm3 RoPE-layer count mismatch are unrelated." + "reason": "Qwen3 MoE deepspeed weight-conversion bug and Smollm3 RoPE-layer drop are different issues." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading failures, but the symptoms and implicated code paths are too different to treat as the same bug." + "reason": "AddedToken constructor regression and tokenizer loading failure for EMBIDDIA/est-roberta are not the same bug." }, { "left": "issue:43540", "right": "issue:45823", "accept": false, - "reason": "Qwen3OmniMoe video processing and Gemma4 PLE device mismatch are unrelated." + "reason": "Different model families and failure modes: Qwen3OmniMoe video ValueError vs Gemma4 PLE device mismatch." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Weight-conversion bug vs CUDA OOM; not the same underlying issue." + "reason": "Qwen3Moe MOE conversion failure and GPT-oss OOM are unrelated." }, { "left": "issue:43906", "right": "issue:45381", "accept": false, - "reason": "No direct evidence they share the same root cause beyond both mentioning video-related behavior." + "reason": "Reproduction of an unrelated upstream issue vs a Qwen2.5-VL video-position bug." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "TrainerState token tracking and custom compute_loss signature mismatch are unrelated." + "reason": "TrainerState token-count update bug and custom compute_loss signature breakage are different training issues." }, { "left": "issue:44062", "right": "issue:45081", "accept": false, - "reason": "AddedToken special-argument collision and Mistral regex/backend_tokenizer crash are different tokenizer failures." + "reason": "AddedToken keyword conflict and Mistral tokenizer regex patch crash are different tokenizer bugs." }, { "left": "issue:43493", "right": "issue:43994", "accept": true, - "reason": "Both point to the same SigLIP2 base model behaving incorrectly; 43994 reads like the user-facing manifestation of the implementation discrepancy in 43493." + "reason": "Both point to the same SigLIP2 correctness problem on the same model family; one is a JAX-vs-HF discrepancy, the other reports clearly wrong outputs." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Custom model initialization breakage in v5 and FSDP CPU RAM-efficient loading are different regressions." + "reason": "Transformers 5.0 custom initialization breakage and FSDP CPU RAM efficient loading are distinct issues." + }, + { + "left": "issue:36010", + "right": "issue:42175", + "accept": false, + "reason": "GenerationMixin import regression and missing TensorFlow backend in a torch install are unrelated packaging/import issues." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque error and MIN_PEFT_VERSION bump are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Model-loading regression in transformers>=5.1.0 and list_repo_files revision forwarding are different issues." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Same general area only superficially; the concrete failures and affected models differ." + }, + { + "left": "issue:43742", + "right": "issue:45538", + "accept": false, + "reason": "Loading key error for MobileLLM-125M and CLIPTokenizer max_length bug are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative overall. The two accepted near-duplicates are both supported by the titles: the Qwen video-input regression pair shares the same v5.3.0 / vision_position_ids failure mode, and the SigLIP2 pair shares the same model family with a plausible implementation-vs-pipeline manifestation. The remaining soft-edge rejections are mostly well-justified by clearly different model families or code paths." + "feedback": "The summary is grounded in the packet and stays mostly conservative. The two accepted duplicate families (Qwen video-input regression and SigLIP2 correctness reports) are plausibly same-bug pairs from the titles alone, and the remaining rejections are framed cautiously rather than overclaimed. No obvious mergeability or evidence overstatement issues stand out." }, "evaluator_used": true, "retried": false, @@ -6130,12 +6133,12 @@ "issue:44514|issue:45290", "issue:43295|issue:45356" ], - "prepared_review_unit_hash": "45f6eb0ea5486d24d232efcfb74c199a75952a4e6f1712579fd7c2471dad49d3", + "prepared_review_unit_hash": "2d7cd37159edac5068f34adda1fe1045ee6d358868c02ddc392224a66cf9a2dc", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -6143,183 +6146,177 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 36227, - "estimated_input_tokens": 9057, - "estimated_eval_tokens": 18370 + "serialized_chars": 36229, + "estimated_input_tokens": 9058, + "estimated_eval_tokens": 18372 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is loose and mostly heterogeneous: the items are all Transformers issues, but they span unrelated bugs in loading, tokenization, configs, vision models, and packaging. None of the soft edges look like true duplicates; they only share broad surface area.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 45920 is the best single representative only because it is the broadest active regression here: it describes a tokenizer-wide failure across multiple model families, whereas the other issues are narrower and more model-specific.", + "summary": "These are mostly distinct Transformers bug reports clustered by broad theme (loading, tokenization, config, and vision/multimodal regressions). No soft-edge pair looks like the same concrete bug/change, so I\u2019d keep them separate. If forced to pick an umbrella representative, the broadest recent regression is the AutoTokenizer token-id issue.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45920 is the broadest issue in the set: it describes a v5 regression affecting AutoTokenizer across multiple model families, which makes it the best umbrella representative for the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45920 is the most suitable global issue anchor for this cluster because it captures a widespread, user-facing tokenizer regression and is still open; however, it should be treated as a representative, not evidence that the cluster is truly duplicate-heavy.", + "best_issue_reason": "issue:45920 is the strongest global issue candidate because it is recent, regression-focused, and spans several affected models rather than a single narrow edge case.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43749", "right": "issue:44756", "accept": false, - "reason": "Both are loading/memory related, but one is about FSDP CPU RAM-efficient loading and the other is disabling mmap on Strix Halo to avoid OOM; different failure modes and fixes." + "reason": "Both are loading/memory-related, but one is FSDP CPU RAM efficient loading and the other is mmap/OOM on Strix Halo; different failure modes and code paths." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Packaging/backend omission vs missing exported class import; they involve different subsystems and root causes." + "reason": "TensorFlow backend packaging and a missing Qwen2.5-VL class import are unrelated issues; no shared underlying bug is evident." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "`init_empty_weights` type error is a model-loading API regression, while the other is loading an image processor from a URL; not the same bug." + "reason": "`init_empty_weights` with an unexpected argument and AutoImageProcessor URL loading are different APIs and different failures." }, { "left": "issue:45357", "right": "issue:45920", "accept": false, - "reason": "One is a Qwen3.5 save_pretrained key-mapping regression; the other is a tokenizer ID regression across many models. Different code paths." + "reason": "Qwen3.5 save_pretrained key mapping is a checkpoint-writing bug, while 45920 is tokenizer output regression; not the same code path." }, { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "Different models and different failures: a janus test regression versus inability to load Phi-4 multimodal instruct." + "reason": "Both involve multimodal model loading, but they affect different models and likely different loaders; too broad to merge." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "Both are model-loading complaints, but one is specifically about gguf_file handling for AutoTokenizer/AutoModelForCausalLM, while the other is a model-specific load failure for cjvt/sleng-bert." + "reason": "Both are model-loading failures, but 43452 is about gguf_file handling in from_pretrained while 44488 is a specific model load failure with no concrete overlap shown." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both mention config fields, but one is a trust_remote_code/from_pretrained field-loss issue and the other is a strict config dropping a specific Qwen3_5MoeVision field." + "reason": "Both mention missing config fields, but one is a trust_remote_code/AutoConfig issue and the other is a strict-config field-drop regression in a specific Qwen config." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Image processor torchvision requirements and Mistral regex patch crashes are unrelated bugs in different parts of the stack." + "reason": "Image processor torchvision dependency and Mistral regex patch crash are unrelated subsystems and unrelated bugs." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "Generic PyTorch import error vs Llama 3.2 vision model-loading breakage in a specific release; not the same underlying issue." + "reason": "Generic import-error-from-missing-dependency and a Llama 3.2 Vision load regression are not the same underlying issue." }, { "left": "issue:43742", "right": "issue:44589", "accept": false, - "reason": "KeyError while loading a model and missing Float8 storage object are distinct load-time failures with different causes." + "reason": "A key error loading MobileLLM-125M and a Float8 storage lookup failure are unrelated loading errors." }, { "left": "issue:43526", "right": "issue:45479", "accept": false, - "reason": "One is a processor label-reduction bug; the other is a training loss bug for single-label classification. Same broad area, different bug." + "reason": "One is an image processor label-reduction bug; the other is a sequence-classification loss/num_labels bug. Same word 'labels' but different functionality." }, { "left": "issue:43756", "right": "issue:45325", "accept": false, - "reason": "Both concern vision/position encoding, but they are different models and different mistakes: RoPE layer dropping vs incorrect temporal position scaling." + "reason": "Both involve RoPE/position ids in vision models, but they target different model families and distinct computations; too speculative to unify." }, { "left": "issue:45405", "right": "issue:45600", "accept": false, - "reason": "One is an unreleased dependency version bump; the other is stale auto_mappings references. Different maintenance issues." + "reason": "A dependency version bump and stale auto_mappings references are unrelated maintenance/CI issues." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE and AutoImageProcessor URL loading are unrelated bugs." + "reason": "Qwen3-MoE sliding_window behavior and AutoImageProcessor URL loading are unrelated bugs." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "RTDetrV2 output drift in v5 and Qwen3.5 save_pretrained corruption are different model regressions with different code paths." - }, - { - "left": "issue:28282", - "right": "issue:29127", - "accept": false, - "reason": "PyTorch missing ImportError and unclear LayoutLMv3 box-validation messaging are not the same bug." + "reason": "Different regressions in different model flows: RTDetrV2 output drift vs Qwen3.5 save_pretrained checkpoint corruption." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "BitNet packed-weight loading and Perceiver interpolation failure involve different models and different loading logic." + "reason": "BitNet packed-weight loading and Perceiver interpolation failure are different model-specific bugs." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Batched chat-template padding crash vs Qwen2.5-VL video position-id bug; both are chat/vision adjacent but not the same defect." + "reason": "Both are Qwen2.5-VL, but one is chat template batching/padding and the other is video position-id scaling; not the same concrete bug." }, { "left": "issue:43582", "right": "issue:43782", "accept": false, - "reason": "Apple Silicon caching_allocator_warmup TypeError and Qwen3VL weight_only load error are unrelated." + "reason": "Apple Silicon caching allocator warmup TypeError and Qwen3VL weight_only loading error are unrelated." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "DeepSpeed ZeRO-3 loading failure and trainer.evaluate-after-train failure do not share the same root cause." + "reason": "DeepSpeed ZeRO-3 model loading IndexError and trainer.evaluate() after train() are different training/runtime failures." }, { "left": "issue:44484", "right": "issue:44704", "accept": false, - "reason": "`max_shard_size` documentation/behavior question and AutoProcessor kwargs forwarding bug are different issues." + "reason": "A save_pretrained shard-size question and missing kwargs in AutoProcessor/cached_file are not the same issue." }, { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Torch extra backend packaging problem vs Python 3.14 load failure; different environments and causes." + "reason": "Backend packaging for torch install and Python 3.14 load failure are unrelated environment/setup reports." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "Qwen3.5 num_labels propagation and strict config blocking granite_speech are both config-related but not the same underlying bug." + "reason": "Both touch config handling, but one is num_labels propagation in Qwen3.5 and the other is strict config blocking granite_speech loading." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "gguf_file support breaks loading, while the BLIP-2 issue is dtype propagation; different subsystems." + "reason": "gguf_file loading breakage and BLIP2 dtype propagation are different model-loading problems with no shared code-path evidence." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Model loading with gguf_file is unrelated to the offline model_info call in `_patch_mistral_regex`." + "reason": "gguf_file handling and Mistral regex/offline model_info behavior are separate tokenizer-loading concerns." }, { "left": "issue:44514", "right": "issue:45290", "accept": false, - "reason": "Both touch `apply_chat_template`, but one is a batched-padding crash and the other is a tool-call/no-content assistant-message crash; separate failure modes." + "reason": "Both are apply_chat_template crashes, but one is batched padding=False and the other is assistant tool-call/no-content handling; likely separate edge cases, so too risky to merge." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "A regression in processor/tokenizer access with image inputs is unrelated to Kimi-K2.5 codec handling and the Mistral-regex warning." + "reason": "A regression around processor.tokenizer/images passing and a Kimi-K2.5 codec/mistral-regex regression are unrelated tokenizer issues." } ] }, @@ -6338,8 +6335,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -6430,7 +6427,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -6742,12 +6739,12 @@ "issue:43756|issue:45083", "issue:44488|issue:44843" ], - "prepared_review_unit_hash": "e9e8cc9eb70871f9eaafc825df502399b82fee21d0cc8be1348f50578efd800a", + "prepared_review_unit_hash": "230d7471aae9932714e99e75633a036b9557e4c41865278922953cb2e2f37f57", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -6764,186 +6761,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of unrelated Transformers bugs/regressions around loading, tokenizers, configs, and multimodal model behavior. The soft pairs share broad themes but not the same concrete failure mode, code path, or fix.", + "summary": "These items are a heterogeneous mix of unrelated bug reports across loading, tokenizers, configs, multimodal models, training, and docs; none of the soft pairs look like the same underlying issue, so there is no clear canonical duplicate in the cluster.", "confidence": 0.95, - "canonical_issue_reason": "Issue 45812 is the strongest anchor for the cluster: it describes a broad, well-reproduced tokenizer regression affecting all Granite models, with clear impact and multiple references.", + "canonical_issue_reason": "No single issue is a credible canonical representative: the reports span many distinct subsystems and failure modes rather than one shared bug.", "canonical_pr_reason": null, - "best_issue_reason": "45812 is the most actionable and high-signal issue here: broad blast radius, specific symptoms, and a concrete regression narrative. If one issue should represent the set, this is the best choice.", + "best_issue_reason": "No issue is a strong global best candidate; any one would be an arbitrary pick because the cluster is not a duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "Both are loading failures, but one is a key error for MobileLLM and the other is an AutoImageProcessor-from-URL issue; different code paths and root causes." + "reason": "Different loading paths and failures: MobileLLM key error vs AutoImageProcessor URL loading." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Both mention multi-GPU/training, but one is a Trainer broadcast NoneType error and the other is a Qwen2-VL training request; not the same bug." + "reason": "Both involve training, but one is a `_gpu_broadcast_one` NoneType error and the other is Qwen2-VL multi-GPU training; not the same bug." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "Both are model-loading complaints, but the affected APIs/models and failure modes differ; no evidence of a shared underlying defect." + "reason": "Both are model-loading related, but one is `gguf_file` handling in auto classes and the other is a specific model that cannot load; too different." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "One is a device-map/cuda assert on Qwen3, the other is inconsistent outputs with compiled forward during chunked generation; unrelated code paths." + "reason": "Unrelated failures: device-mapping CUDA assert vs inconsistent chunked generation with compiled forward." }, { "left": "issue:45823", "right": "issue:46050", "accept": false, - "reason": "Both involve model loading/forward behavior, but one is a device mismatch in Gemma4 and the other is sinq save/load quantization breakage; distinct bugs." + "reason": "Different issues: Gemma4 device mismatch on `device_map=\"auto\"` vs quantized model save/load breakage." }, { "left": "issue:33666", "right": "issue:36010", "accept": false, - "reason": "A multi-GPU training issue and an import error for GenerationMixin are not the same failure." + "reason": "Multi-GPU training support vs an import error for `GenerationMixin`; no shared underlying bug." }, { "left": "issue:43493", "right": "issue:43825", "accept": false, - "reason": "One is a SigLIP2 implementation discrepancy and the other is an inaccurate pipeline error message; same area only at a very broad level." + "reason": "SigLIP2 implementation discrepancy vs pipeline error messaging for translation tasks; unrelated." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Both are Qwen3.5 regressions, but one affects output_hidden_states and the other save_pretrained visual encoder keys; different bugs." + "reason": "Both are Qwen3.5 regressions, but one is `output_hidden_states` behavior and the other is incorrect save_pretrained visual keys." }, { "left": "issue:43606", "right": "issue:43927", "accept": false, - "reason": "A CPU offload device mismatch and a DiaConfig token-id persistence bug are unrelated." + "reason": "Device mismatch during CPU offload vs DiaConfig token ID persistence; different code paths." }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Both concern missing pad-token fields after v5, but they hit different configs/models and different attribute names; too broad to treat as the same bug." + "reason": "Both mention missing padding-related config fields after v5, but they affect different model configs and are not clearly the same bug." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "BitNet accelerate-loading unpacking and DiaConfig token-id persistence do not share the same root cause." + "reason": "BitNet packed-weight loading failure vs DiaConfig save/load token IDs; unrelated." }, { "left": "issue:43756", "right": "issue:44913", "accept": false, - "reason": "Both mention RoPE-related config behavior, but one is a Smollm3 layer-drop issue and the other is GPTNeoX rotary_pct reload behavior; distinct defects." + "reason": "Both involve RoPE/rotary settings, but one is a Smollm3 architecture bug and the other is GPTNeoX config reload behavior." }, { "left": "issue:43262", "right": "issue:43295", "accept": false, - "reason": "One is an audio sampling-rate default mismatch, the other is a processor.tokenizer regression; different processor behaviors." + "reason": "Audio chat-template sampling-rate default vs processor/tokenizer access regression; different functionality." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "An ImportError for missing PyTorch and a MacOS bus error with CLIP are unrelated environment/runtime failures." + "reason": "PyTorch-missing ImportError vs MacOS bus error on a CLIP model; not the same issue." }, { "left": "issue:43688", "right": "issue:43749", "accept": false, - "reason": "Aux-loss normalization and FSDP CPU RAM efficient loading are different training/runtime issues." + "reason": "Auxiliary-loss normalization bug vs FSDP CPU RAM efficient loading; different training internals." }, { "left": "issue:44568", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but one is about add_special_tokens on a specific tokenizer while the other is wrong token IDs across Granite models; not the same change or code path." + "reason": "Both are tokenizer regressions, but they affect different model families and concrete tokenization bugs." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "Missing PyTorch at import time and a Trainer _gpu_broadcast_one NoneType error are unrelated." + "reason": "ImportError for missing PyTorch vs Trainer broadcast NoneType error; unrelated." }, { "left": "issue:43720", "right": "issue:44361", "accept": false, - "reason": "Packed-weight unpacking during loading and a tokenizer AttributeError on tasks are not the same bug." + "reason": "BitNet accelerated loading bug vs MLukeTokenizer attribute error; no shared code-path." }, { "left": "issue:33666", "right": "issue:36331", "accept": false, - "reason": "Multi-GPU training for Qwen2-VL and an unexpected num_items_in_batch argument in CustomTrainer are separate Trainer-related issues." + "reason": "Multi-GPU training support vs custom trainer `compute_loss` signature change; distinct trainer issues." }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "A SigLIP2 documentation/example error and a model-vs-JAX implementation discrepancy are different kinds of bugs." + "reason": "SigLIP2 docs/example problems vs a model implementation discrepancy; same model family but not the same bug." }, { "left": "issue:43577", "right": "issue:44554", "accept": false, - "reason": "One is a dtype propagation issue in Blip2 loading; the other is an MPS attention correctness problem. No shared code-path evidence." + "reason": "Blip2 dtype propagation bug vs MPS attention correctness issue; unrelated." }, { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Model loading failure and a dependency version bump issue are unrelated." + "reason": "Model load failure vs dependency version bump; no common underlying defect." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Both are Qwen-family inference bugs, but one is video input handling in Qwen3OmniMoe and the other is flash-attention generation in Qwen3.5; different failure modes." + "reason": "Qwen3OmniMoe video-processing error vs Qwen3.5 flash-attention generation bug; different failure modes." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "A multimodal video-input regression and a num_labels propagation bug are unrelated." + "reason": "Video-input regression across Qwen variants vs `num_labels` propagation into Qwen3.5 text config; unrelated." }, { "left": "issue:43688", "right": "issue:45704", "accept": false, - "reason": "Aux-loss normalization and an apex RMSNorm memory leak are different model/component issues." + "reason": "Aux-loss normalization bug vs T5 apex RMSNorm memory leak; different model families and symptoms." }, { "left": "issue:45803", "right": "issue:45987", "accept": false, - "reason": "Swallowing KeyboardInterrupt in Fuyu tensor conversion and StaticCache.get_seq_length returning the wrong type are unrelated." + "reason": "Fuyu tensor conversion exception swallowing vs StaticCache return-type contract; unrelated." }, { "left": "issue:43334", "right": "issue:43572", "accept": false, - "reason": "Both are missing pad-token-related config attributes, but they affect different model families and different attribute names; not enough to merge." + "reason": "Qwen3-VL missing `pad_token_id` vs StableLmConfig missing `pad_token_idx`; similar shape, but different models and fields." }, { "left": "issue:43756", "right": "issue:45083", "accept": false, - "reason": "Both are qwen3_omni_moe-related, but one is a RoPE-layer count bug and the other is a helper function length-calculation issue; different defects." + "reason": "Smollm3 RoPE-layer issue vs a helper length-calculation bug in qwen3_omni_moe; not the same bug." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "A model load failure for sleng-bert and an offline-mode breakage in _patch_mistral_regex are unrelated loading issues." + "reason": "Model loading failure for a specific BERT model vs offline-mode breakage from an unconditional hub call; different root causes." } ] }, @@ -7046,12 +7043,12 @@ "issue:43844|issue:44964", "issue:28218|issue:34567" ], - "prepared_review_unit_hash": "263144f3c370052589de3f62aced61d8d80fef6319f882990c5bed802d0853d4", + "prepared_review_unit_hash": "a0221b611560d3ecb53fa8a51871e7a168fbf77f58489374c9126923a67ea54c", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -7068,180 +7065,180 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is highly heterogeneous: the paired items only share broad surface themes like tokenization, multimodal models, or loading regressions, but they describe different concrete failures and code paths. I would not merge any of the soft-edge pairs as duplicates.", - "confidence": 0.95, - "canonical_issue_reason": "No true duplicate cluster is present; if a representative is needed, #45701 is the broadest and most central open tokenizer-regression report, but it should not be treated as a duplicate of the others.", + "summary": "Loose similarity cluster of mostly unrelated Transformers issues; the soft pairs are thematic lookalikes, not true duplicates or mergeable PR equivalents.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Best overall representative for triage: #45701 is open, relatively high-activity, and broad enough to anchor the mostly tokenizer-related noise in this cluster.", + "best_issue_reason": "issue:45759 is the most concrete, reproducible loading/config regression in the set and is a reasonable representative of the broader model-loading subset, though the overall cluster is not duplicate-tight.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Different bugs: one is a Gemma 4 token-type default issue, the other is a multi-GPU OOM report." + "reason": "Same family (Gemma 4) but different failures: missing mm_token_type_ids default vs multi-GPU CUDA OOM." }, { "left": "issue:43493", "right": "issue:43844", "accept": false, - "reason": "Different subsystems and failures: SigLIP2 implementation mismatch vs. DeepSpeed/ZeRO-3 gradient scaling." + "reason": "Both relate to model correctness, but one is SigLIP2 implementation mismatch and the other is DeepSpeed/ZeRO gradient blow-up." }, { "left": "issue:43334", "right": "issue:45759", "accept": false, - "reason": "Both involve Qwen multimodal configs, but the concrete problems differ: missing pad_token_id vs. unwrapping text_config in from_config." + "reason": "Both touch Qwen multimodal loading, but one is a missing pad_token_id attribute and the other is AutoModel.from_config not unwrapping text_config." }, { "left": "issue:43824", "right": "issue:44589", "accept": false, - "reason": "Unrelated errors: an import failure for Qwen2.5-VL versus a Float8 storage lookup error." + "reason": "ImportError for a missing class vs a storage-type lookup failure; different code paths and symptoms." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Both touch multimodal/processor usage, but one is a SigLIP2 doc example issue and the other is a regression in custom code using processor.tokenizer." + "reason": "Docs/example errors and a regression in processor/tokenizer API access are related only at a very high level." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Same model family, different bug: accelerate+deepspeed conversion failure vs. Qwen3VL MoE loading regression." + "reason": "Both mention Qwen3 MoE, but one is accelerate+deepspeed conversion and the other is a dev-version loading regression for Qwen3VL." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Different code paths: custom model initialization breakage vs. device-dependent serialization of tied weights." + "reason": "Custom model initialization breakage is not the same bug as device-dependent lm_head.weight serialization." }, { "left": "issue:45229", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma4-related, but one is CUDA OOM and the other is a device-mismatch during forward." + "reason": "Both are Gemma 4-related, but one is inference OOM and the other is a device mismatch in forward with device_map=auto." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "Both are loading/API regressions, but they are separate failures: missing StableLmConfig field vs. unexpected _is_hf_initialized argument." + "reason": "Both involve init/load-time compatibility, but one is a missing pad_token_idx in StableLmConfig and the other is an unexpected _is_hf_initialized argument." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Different problems: evaluation dtype drift vs. gradient explosion under ZeRO-3." + "reason": "Different bugs: dtype drift between train/eval vs gradient normalization under HfDeepSpeedConfig." }, { "left": "issue:43504", "right": "issue:43606", "accept": false, - "reason": "Completely different issues: legacy config field loading for Beit vs. Bark CPU offload device mismatch." + "reason": "Legacy-field load failure in Beit semantic segmentation is unrelated to CPU offload device mismatch in bark-small." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Different failures: missing Float8 storage type vs. Python 3.13 flash_attn import key error." + "reason": "Both are type/compatibility problems, but one is missing Float8 storage handling and the other is a Python 3.13 flash_attn import key error." }, { "left": "issue:43688", "right": "issue:45701", "accept": false, - "reason": "Different topics: auxiliary-loss normalization in MoE models vs. a broad tokenization regression report." + "reason": "Aux-loss normalization and version-dependent tokenization are unrelated bugs." }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "Unrelated model/runtime bugs: MacOS bus error in CLIP vs. Qwen2-VL multi-GPU training." + "reason": "Both are multimodal/model-scale issues, but MacOS bus error on CLIP and Qwen2-VL multi-GPU training are different failure modes." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "Different tokenizer/config bugs: lost DiaConfig token IDs after save/load vs. MLukeTokenizer AttributeError on tasks." + "reason": "Both mention tokenizer/config state, but DiaConfig token IDs and MLukeTokenizer AttributeError are distinct problems." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "No shared underlying defect: Beit legacy-field loading vs. BigBirdTokenizer special-token registration." + "reason": "Loading a legacy BEiT field and BigBirdTokenizer special-token registration are not the same underlying bug." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Both are server-side crashes, but the failing paths differ: continuous batching multimodal input handling vs. KV-cache continuation indexing." + "reason": "Serving crashes with multimodal .to() handling are different from KV-cache continuation indexing bugs." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "One is a real runtime device mismatch bug; the other is a stale CI test failure in GraniteSpeech." + "reason": "Both are CI/device-related, but one is a real offload mismatch and the other is a stale GraniteSpeech test." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "Different concerns: tensor-parallel OOM on a Qwen model vs. an unreleased PEFT version bump." + "reason": "Tensor-parallel OOM is unrelated to a version pin bump for PEFT." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "Unrelated: a Janus image-generation test failure versus dependency versioning for PEFT." + "reason": "A failing janus generate-images test is not the same issue as the unreleased PEFT version constraint." }, { "left": "issue:44625", "right": "issue:45759", "accept": false, - "reason": "Related Qwen multimodal config area, but not the same bug: num_labels propagation vs. text_config unwrapping." + "reason": "Both involve Qwen multimodal config propagation, but num_labels plumbing and text_config unwrapping are different API bugs." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "Different model issues: SigLIP2 pipeline quality regression vs. unexpected key warnings around position ids." + "reason": "SigLIP2 bad outputs and unexpected position-id keys are separate correctness problems." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "Different defects: flash_attention sink-less s_aux handling vs. sinq quantization save/load breakage." + "reason": "Flash-attention sinkless-model AttributeError and sinq save/load quantization breakage are unrelated." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "No overlap beyond being general test/runtime failures: Flash Attention 2 CI vs. MLflow reporting after training failure." + "reason": "CI Flash Attention 2 failures and MLflow reporting after training failure are different subsystems and symptoms." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Both are generation-related, but the concrete bugs are unrelated: compiled-forward chunked generation vs. empty-transcription whisper handling." + "reason": "Chunked generation with compiled forward and Whisper empty-transcription failures have no concrete shared code path." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both involve apply_chat_template, but one is all-zero assistant masks for multimodal input and the other is a crash on tool-call assistant messages with no content." + "reason": "Both are chat-template edge cases, but all-zero assistant masks and tool-call content crashes are distinct failures." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "Different multimodal/model-loading failures: DeepSpeed gradient anomaly vs. inability to load Phi-4 multimodal." + "reason": "DeepSpeed gradient growth and Phi-4 multimodal load failure are unrelated." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Completely unrelated: tokenizer spacing after added tokens vs. TrainerState token-count tracking." + "reason": "Tokenizer spacing regression and TrainerState token counter not updating are unrelated trainer/tokenizer bugs." } ] }, @@ -7349,12 +7346,12 @@ "issue:37428|issue:43824", "issue:38175|issue:43493" ], - "prepared_review_unit_hash": "cc31c4056bdfd746b6493cb9d5111e03ac1ed0ea4bc0c52e0fbcd15b4f288025", + "prepared_review_unit_hash": "4173defa13453ab17c1c3689fcd065a399600c22be3a186a555e326d91d57bc9", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -7371,210 +7368,210 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated transformer bug reports with a few loose topical similarities (tokenizers, multimodal processors, loading/config regressions). I do not see any true duplicates among the soft pairs; all should be kept separate.", - "confidence": 0.58, - "canonical_issue_reason": "Issue 43295 is the most central concrete regression in the set and sits near the processor/tokenizer-related subtheme that has the most nearby soft links, making it the best representative issue for the cluster.", + "summary": "Loose similarity cluster of unrelated issue reports across tokenization, generation, quantization/offloading, multimodal processors, and import/runtime regressions. No soft pair looks like a safe duplicate/one-PR merge candidate.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the cluster is a broad topical bucket, not a duplicate set, so any representative would be misleading.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43295 is the clearest user-facing regression among the candidates and the best single anchor for this heterogeneous group, even though the surrounding items are not duplicates.", + "best_issue_reason": "If one issue must anchor the cluster, #43873 is the broadest and most active item in the closest quantization/offloading subgroup, but it is still not a true duplicate representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "Different bugs and code paths: MPS attention correctness vs a PyTorch version gate for AuxRequest import." + "reason": "Different bugs: MPS attention correctness vs a flex_attention import/version check. Same general area only, not the same code-path." }, { "left": "issue:43688", "right": "issue:43873", "accept": false, - "reason": "Auxiliary-loss normalization in MoE models is unrelated to quantization/offloading behavior." + "reason": "Aux-loss normalization in specific MoE models is unrelated to offloading/quantization behavior. Different failure mode and fix." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Template-node compile failure and flash_attn import KeyError are unrelated failures." + "reason": "One is a compile/template-node error, the other a flash_attn import KeyError on Python 3.13. Unrelated runtime paths." }, { "left": "issue:29159", "right": "issue:32090", "accept": false, - "reason": "Tokenizer slow/fast inconsistency is unrelated to a Trainer GPU broadcast TypeError." + "reason": "Tokenizer slow/fast inconsistency and Trainer GPU broadcast NoneType error are separate subsystems and symptoms." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Both involve multimodal processing, but they affect different model families and different concrete paths." + "reason": "Both are multimodal-related, but one is a Qwen3OmniMoe video-processing error and the other is Gemma 4 text-only token-type defaults. Not the same bug." }, { "left": "issue:43582", "right": "issue:44977", "accept": false, - "reason": "Apple Silicon allocator TypeError and Qwen3.5 flash-attention generation issues are distinct." + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation issues are unrelated." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "TorchScript GeneratorExp export failure is unrelated to SigLIP2 zero-probability outputs." + "reason": "JIT scripting support for generator expressions vs SigLIP2 zero probabilities are different issues entirely." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "Generation kwargs sync handling and Qwen3.5 hidden-state output behavior are different bugs." + "reason": "Both involve generation/model outputs, but one is a sync_gpus kwargs update bug and the other is output_hidden_states handling in Qwen3.5. Different code paths." }, { "left": "issue:45200", "right": "issue:45823", "accept": false, - "reason": "Gemma4 mm_token_type_ids defaults and device_map auto mismatch are separate issues." + "reason": "Same model family, but different concrete defects: missing mm_token_type_ids defaults vs device placement mismatch with device_map=\"auto\"." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "dtype mismatch in inference and Wav2Vec2 save/tokenization failures do not share the same bug." + "reason": "dtype mismatch in inference is not the same as Wav2Vec2 save/tokenization failure. No shared concrete bug." }, { "left": "issue:43873", "right": "issue:46050", "accept": false, - "reason": "Quantization/offloading behavior and sinq save/load breakage are not the same code-path problem." + "reason": "Quantization offloading failure and sinq save/load breakage are both quantization-adjacent, but they target different mechanisms and code paths." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "A BF16-to-FP32 dtype change in eval is unrelated to checkpoint key mismatches." + "reason": "Embedding dtype drift during eval and checkpoint key mismatch are unrelated training-state bugs." }, { "left": "issue:44315", "right": "issue:44849", "accept": false, - "reason": "Liger Kernel application in model_init is unrelated to Qwen3.5 output_hidden_states behavior." + "reason": "Liger Kernel application via model_init and Qwen3.5 hidden-state output behavior are separate issues." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer added-space behavior and a MacOS bus error in CLIP loading are unrelated." + "reason": "Tokenizer spacing after added tokens and a MacOS bus error loading CLIP are not the same defect." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "Both are processor-related regressions, but one is about processor.tokenizer/image forwarding and the other is about incorrect input shape." + "reason": "Processor/tokenizer API regression and OmDet-Turbo image-size mismatch both affect processors, but the concrete bugs differ." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "MPS attention correctness and tokenizer codec/regression behavior are unrelated." + "reason": "MPS attention correctness and Kimi-K2.5 tokenizer codec regression are unrelated." }, { "left": "issue:44898", "right": "issue:45412", "accept": false, - "reason": "Perceiver interpolation failure and RT-DETR memory release are different problems." + "reason": "Perceiver interpolation failure and RT-DETR memory retention are different model-specific bugs." }, { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "Missing pad_token_id in a config and a loading failure for a specific model are not the same bug." + "reason": "Missing pad_token_idx in StableLmConfig and a loading failure for a different model are separate regressions." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 implementation discrepancy is unrelated to resume-from-checkpoint key handling." + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." }, { "left": "issue:43572", "right": "issue:43957", "accept": false, - "reason": "Both concern model/config loading, but one is a missing config attribute and the other is meta-device loading breakage." + "reason": "Both are model-loading regressions, but one is a missing config attribute and the other is a meta-device loading breakage. Not the same bug." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Serialization of tied weights and LoRA merge collapse after vocab extension are distinct issues." + "reason": "Serialization of tied lm_head weights and LoRA vocabulary-merging collapse are distinct weight-handling problems." }, { "left": "issue:45200", "right": "issue:45325", "accept": false, - "reason": "Default mm_token_type_ids and Qwen2.5-VL rope index scaling are unrelated multimodal bugs." + "reason": "Different multimodal token/index bugs in different models; no shared concrete code-path." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Flash-attention import failure and meta-device model loading failures are different code paths." + "reason": "Flash-attention import failure and meta-device loading breakage are different import/loading regressions." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Processor.tokenizer regression and batched video shape errors are separate processor bugs." + "reason": "Processor.tokenizer regression and batched video-shape bug are separate processor issues." }, { "left": "issue:43550", "right": "issue:43844", "accept": false, - "reason": "torch.compile SDPA failure and DeepSpeed gradient growth are unrelated training/runtime issues." + "reason": "torch.compile/SDPA failure and ZeRO-3 gradient growth are unrelated training/runtime bugs." }, { "left": "issue:43334", "right": "issue:43525", "accept": false, - "reason": "These are both missing pad_token_id-style errors, but they affect different model families and are not clearly the same underlying defect." + "reason": "Both mention missing pad_token_id on config objects, but they affect different model families and are not enough evidence for the same underlying fix." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to offloading behavior under quantization." + "reason": "Processor tokenizer regression and quantization offloading behavior are unrelated." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "Importing Qwen2_5_VLForConditionalGeneration and Kimi tokenizer detokenization are unrelated." + "reason": "Import error for Qwen2.5-VL and DeepSeek detokenization regression are unrelated." }, { "left": "issue:38175", "right": "issue:45412", "accept": false, - "reason": "SigLIP2 zero-probability outputs and RT-DETR memory retention are different issues." + "reason": "SigLIP2 zero-probability outputs and RT-DETR memory release issues are completely different bugs." }, { "left": "issue:42490", "right": "issue:43696", "accept": false, - "reason": "Local-path import warning and GPT-oss CUDA OOM are unrelated." + "reason": "Path-with-slash import warning and GPT-oss OOM are unrelated." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Fine-grained FP8 config failure and a missing class import are distinct problems." + "reason": "Qwen3Moe FP8 config failure and Qwen2.5-VL import error are different model-load issues." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "Flash-attention helper import failure and missing Qwen2.5-VL export are unrelated." + "reason": "Both are import errors, but they involve different symbols and different failure modes." }, { "left": "issue:38175", "right": "issue:43493", "accept": false, - "reason": "SigLIP2 probability behavior and SigLIP2 implementation discrepancy are related only by model family, not the same concrete bug." + "reason": "Both concern SigLIP2, but one reports zero probabilities while the other is a broader HF-vs-JAX discrepancy; not enough to treat as the same concrete bug." } ] }, @@ -7593,8 +7590,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -7685,7 +7682,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -7994,12 +7991,12 @@ "issue:45198|issue:45803", "issue:45459|issue:45712" ], - "prepared_review_unit_hash": "668b3050c7862eaeee72f3522a97d77ee445d517e92ab96d7fbb939aa007758f", + "prepared_review_unit_hash": "739570aa18f35f07765d534acd5c0fada620074f66a84e6fcfef69abcb905a70", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -8016,216 +8013,216 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous; nearly all soft pairs are unrelated bugs. The only clear duplicate is the two identical Qwen3.5 MoE `from_pretrained` issues (45310/45478).", - "confidence": 0.95, - "canonical_issue_reason": "issue:45310 is the earlier, more established report of the duplicate Qwen3.5 MoE `from_pretrained` failure and is the better canonical issue.", + "summary": "This cluster is mostly noisy: the titles span many unrelated Transformers bugs. Only one clear duplicate pair stands out: the two Qwen3.5 MoE from_pretrained issues (45310/45478), which describe the same failure mode.", + "confidence": 0.92, + "canonical_issue_reason": "issue:45310 is the best representative because it is the earlier, more detailed report of the same Qwen3.5 MoE from_pretrained failure that 45478 repeats.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 best represents the only true duplicate pair in the cluster and has the stronger provenance (earlier report).", + "best_issue_reason": "issue:45310 is the strongest global issue candidate: it is a real duplicate target, specific enough to anchor the cluster, and has the fuller bug context.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "Different failures: DebertaV2 import/IndentationError on Python 3.13 vs a qwen3_omni_moe helper function output-length issue." + "reason": "Different problems: Python 3.13 import/torch.jit parsing vs a Qwen3 Omni helper output-length behavior issue." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "Different models and failure modes: SwitchTransformers config sparse-layer construction vs BEiT pretrained preset loading with a legacy field." + "reason": "Unrelated models and code paths: SwitchTransformers config creation vs BEiT preset loading with a legacy field." }, { "left": "issue:45310", "right": "issue:45478", "accept": true, - "reason": "Same underlying bug: both report `transformers>=5.4.0` / Qwen3.5 MoE `from_pretrained` failures with essentially the same title and scope." + "reason": "Same underlying bug: both report Qwen3.5 MoE from_pretrained failures on transformers>=5.4.0, with the later issue repeating the same error." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "Unrelated model/config bugs; no shared concrete code path." + "reason": "Both are model/config bugs, but they affect different components and failure modes." }, { "left": "issue:45588", "right": "issue:45859", "accept": false, - "reason": "One is a flash-attention crash on `s_aux=None`; the other is CPU-offload incompatibility in a rotary embedding forward pass." + "reason": "Both touch flash-attention/CPU-offload-adjacent areas, but one is an s_aux=None AttributeError and the other is a rotary embedding incompatibility." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Flash Attention 2 test failures are not the same as a Trainer `_gpu_broadcast_one` TypeError." + "reason": "Different failures: Flash Attention 2 test breakage vs Trainer/_gpu_broadcast_one NoneType type error." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "Documentation/example errors for SigLIP2 are unrelated to the position-id key regression." + "reason": "One is a SigLIP2 documentation example issue; the other is a broad position-id regression affecting many models." }, { "left": "issue:43720", "right": "issue:44112", "accept": false, - "reason": "Different CI/model-loading problems: BitNet packed-weight unpacking vs GraniteSpeech stale device override test." + "reason": "Different CI failures and root causes: unpacking packed weights during accelerate loading vs a stale device override test." }, { "left": "issue:43335", "right": "issue:45072", "accept": false, - "reason": "Different issues: sparse-layer config construction vs bfloat16 dtype mismatch in inference." + "reason": "Both are dtype-related in different models, but they are distinct bugs in separate code paths." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "Both are version-regression reports, but for different models and different failure modes." + "reason": "Pegasus v4/v5 output drift is not the same bug as Deepseek tokenizer regressions in v5." }, { "left": "issue:43335", "right": "issue:44112", "accept": false, - "reason": "No concrete overlap beyond both being CI/model-related." + "reason": "SwitchTransformers config creation and GraniteSpeech CI device override are unrelated." }, { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "Different symptoms and code paths: zero probabilities in SigLIP2 vs unexpected `position_id` key handling." + "reason": "Different symptom classes: zero probabilities in SigLIP2 vs unexpected position-id keys across many models." }, { "left": "issue:44292", "right": "issue:45520", "accept": false, - "reason": "Qwen NVFP4 runtime issue is unrelated to the Python 3.13 `flash_attn` import-key problem." + "reason": "One is a Qwen-3-8B-NVFP4 runtime problem; the other is a Python 3.13 flash_attn import_utils KeyError." }, { "left": "issue:28218", "right": "issue:35141", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization during resize/post_init." + "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization after resize when tie_word_embedding is false." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Flash-attention generation bug vs multi-GPU CUDA OOM; different concrete failures." + "reason": "Both are generation/inference problems, but they involve different models and distinct failure modes (flash-attention vs multi-GPU OOM)." }, { "left": "issue:42490", "right": "issue:43854", "accept": false, - "reason": "Local-path trailing-slash warning is unrelated to GLM-4.7-Flash unit-test loading failure." + "reason": "Local-path import warning and GLM-4.7-Flash unit-test load failure are not the same issue." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "CPU-offload device mismatch and AMP float16 loss crash are different bugs." + "reason": "CPU offload device mismatch in Bark is unrelated to float16 AMP crashes in LwDetrImageLoss." }, { "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": "A v5 API removal breaking trust_remote_code models is a different regression than Deepseek tokenizer behavior changes." }, { "left": "issue:44462", "right": "issue:44561", "accept": false, - "reason": "Tokenizer.json loading bug is unrelated to a v5 API removal breaking remote-code models." + "reason": "AutoTokenizer ignoring tokenizer.json and the is_torch_fx_available removal are separate loader/API issues." }, { "left": "issue:44857", "right": "issue:45072", "accept": false, - "reason": "Different numeric-stability/dtype issues with different models and code paths." + "reason": "Both are numeric-precision problems, but they affect different models and different code paths." }, { "left": "issue:44448", "right": "issue:44977", "accept": false, - "reason": "Different models and symptoms; both mention version/runtime differences but are not the same bug." + "reason": "Pegasus output differences and Qwen3.5 flash-attention generation failures are unrelated." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "Checkpoint key mismatch is unrelated to the qwen3_omni_moe helper-function issue." + "reason": "Checkpoint key mismatch and Qwen3 Omni helper behavior are distinct issues." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "torch.compile/SDPA Bamba failure is unrelated to qwen3_omni_moe feature-length handling." + "reason": "Bamba SDPA torch.compile failure is unrelated to the Qwen3 Omni feature-length helper bug." }, { "left": "issue:34689", "right": "issue:35141", "accept": false, - "reason": "Model loading breakage for Llama 3.2 Vision is unrelated to embedding resize/post_init reinitialization." + "reason": "Llama 3.2 Vision model loading regression and output embedding reinitialization are separate bugs." }, { "left": "issue:43927", "right": "issue:45698", "accept": false, - "reason": "Both involve save/load, but one is DiaConfig custom token IDs and the other is wrong custom-module resolution; different root causes." + "reason": "DiaConfig save/load token-ID loss and wrong custom module loading after save_pretrained are different serialization bugs." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "LayoutLMv3 error-message clarity issue is unrelated to Trainer `_gpu_broadcast_one` TypeError." + "reason": "LayoutLMv3 error-message clarity and Trainer broadcast type errors are unrelated." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "IterableDataset multi-image finetuning failure is unrelated to BEiT Fast image-label reduction bug." + "reason": "Qwen2.5-VL iterable-dataset finetuning failure and BEiT reduce_labels returning the wrong shape/value are different issues." }, { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Tokenizer special-tokens regression is unrelated to Gemma3 NaN embeddings from mixed-length sliding-window batching." + "reason": "mDeBERTa special-token insertion regression is not the same as Gemma3 NaN embeddings with mixed-length batching." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma 4 processor import failure from `mistral_common` is unrelated to flash-attention `s_aux=None` crash." + "reason": "Gemma 4 processor loading ImportError and flash_attention s_aux=None crash are different failure points." }, { "left": "issue:43412", "right": "issue:45588", "accept": false, - "reason": "ExecuTorch export guard / `erfinv` support issue is unrelated to the flash-attention sink-less-model crash." + "reason": "Executorch export problems for gemma3n are unrelated to the flash_attention AttributeError on sink-less models." }, { "left": "issue:44977", "right": "issue:45563", "accept": false, - "reason": "Flash-attention generation bug is unrelated to a stale warning about `num_return_sequences` in paged generate()." + "reason": "Qwen3.5 flash-attention generation failure and a stale paged-generate warning are not the same bug." }, { "left": "issue:45198", "right": "issue:45803", "accept": false, - "reason": "Wav2Vec2 save/tokenization failure is unrelated to a broad bare-except swallowing errors in Fuyu tensor conversion." + "reason": "Wav2Vec2 save_pretrained/tokenization failure and a bare except swallowing errors in FuyuBatchFeature are unrelated." }, { "left": "issue:45459", "right": "issue:45712", "accept": false, - "reason": "Tokenizer protobuf error masking is unrelated to leftover dummy PT classes leaking into `dir(transformers)`." + "reason": "Tokenizer protobuf error masking and leftover dummy classes in dummy_pt_objects.py are different maintenance/failure issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The only accepted soft pair, issue:45310 vs issue:45478, is strongly supported by the identical title and matching Qwen3.5 MoE from_pretrained failure. The rejected pairs are justified by clearly different titles/symptoms, and there is no unsupported claim beyond the cluster being mostly heterogeneous." + "feedback": "The summary is grounded in the packet: the cluster is broad and the identical Qwen3.5 MoE from_pretrained titles (45310/45478) are the only clear duplicate pair. The soft-edge verdicts are generally conservative and do not overstate similarity beyond what the titles support." }, "evaluator_used": true, "retried": false, @@ -8325,12 +8322,12 @@ "issue:29127|issue:34567", "issue:44610|issue:45412" ], - "prepared_review_unit_hash": "e054db0923b9ecb3174a9c7c6dd654ccd382382d0e9a7768b4830dc63b472869", + "prepared_review_unit_hash": "7db875d41f119fe783f590566776c35c55256c0df840cd2b5a5491fa58350443", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -8347,174 +8344,174 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a duplicate set: the soft-similar pairs span unrelated bugs across tokenizers, generation, vision processors, serving, quantization, and import compatibility. I rejected every soft edge.", + "summary": "This is a heterogeneous set of Transformers bug reports, not a tight duplicate cluster. The soft pairs mostly share only broad subsystem similarity (tokenizers, serving, vision models, flash-attn, quantization) but point to different code paths and failure modes.", "confidence": 0.96, - "canonical_issue_reason": "No single canonical issue is appropriate here; the items are heterogeneous and do not describe one underlying bug or change.", + "canonical_issue_reason": "Issue 44462 is the clearest standalone regression report here: it names a specific tokenizer-loading problem, has a concrete symptom, and is more actionable than the other broadly similar tokenizer issues.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue stands out because there is no cohesive duplicate cluster to canonically represent.", + "best_issue_reason": "Issue 44462 is the strongest representative issue in the set because it is specific, reproducible, and directly tied to user-facing loading behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36010", "right": "issue:37428", "accept": false, - "reason": "Both are import errors, but they affect different symbols and subsystems with no shared code path." + "reason": "Both are import errors, but they involve different symbols, modules, and root causes." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Flash Attention 2 test failures and embedding reinit are unrelated bugs." + "reason": "Flash Attention test failures and embedding reinitialization are unrelated bugs." }, { "left": "issue:42490", "right": "issue:43575", "accept": false, - "reason": "Local-path warning handling and tensor-parallel OOM are entirely different failure modes." + "reason": "A local-path import warning and a tensor-parallel OOM are completely different issues." }, { "left": "issue:44857", "right": "issue:45588", "accept": false, - "reason": "AMP/CUDA loss crash and flash-attention sink handling are different model/runtime issues." + "reason": "Both are runtime crashes, but in different components and code paths." }, { "left": "issue:43065", "right": "issue:45412", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder and RT-DETR memory retention are unrelated." + "reason": "Dummy Conv2d initialization and RT-DETR memory retention are unrelated model behaviors." }, { "left": "issue:45245", "right": "issue:45563", "accept": false, - "reason": "Category-count runtime limit and stale generation warning do not share a concrete bug." + "reason": "Category-count overflow and a stale generate warning are not the same bug." }, { "left": "issue:45372", "right": "issue:45459", "accept": false, - "reason": "Different import/tokenizer error causes; one is mistral_common compatibility, the other protobuf masking errors." + "reason": "Processor loading import failure and protobuf-error masking are different tokenizer-loading problems, not one defect." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "Both involve version-v4/v5 behavior, but one is generation output and the other serialization of tied weights." + "reason": "Different output behavior and serialization inconsistency are separate regression classes." }, { "left": "issue:45588", "right": "issue:45823", "accept": false, - "reason": "Flash-attention auxiliary-state crash and Gemma4 device mismatch are distinct issues." + "reason": "Flash-attention crash on s_aux=None and Gemma4 PLE device mismatch are unrelated." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "Tokenizer special-token registration and recurrent cache-state reset are unrelated." + "reason": "Tokenizer special-token registration and recurrent-state reset are unrelated." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Embedding dtype drift and eval-mode checkpointing restrictions are different problems." + "reason": "Eval-time dtype change and gradient checkpointing in eval are distinct issues." }, { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Python version compatibility and DeepSeek implementation divergence are not the same bug." + "reason": "Python version compatibility and DeepSeek implementation divergence are unrelated." }, { "left": "issue:43994", "right": "issue:45412", "accept": false, - "reason": "SigLIP2 wrong-output behavior and RT-DETR memory leak are unrelated." + "reason": "SigLIP2 inference mismatch and RT-DETR memory release are different model bugs." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Pipeline top_k nesting change and Trainer broadcast NoneType error have no shared code path." + "reason": "Pipeline output nesting and Trainer broadcast NoneType errors are unrelated." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 NER/padding crash and Wav2Vec2 save/tokenization failure are separate tokenizer bugs." + "reason": "Both affect tokenization, but the failing models and failure modes differ." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Tokenizer decoder type mismatch and AutoTokenizer ignoring tokenizer.json are different loading/serialization issues." + "reason": "Tokenizer decoder typing and ignoring tokenizer.json are related broadly, but not the same underlying bug." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Unsupported feature-extractor arg and Qwen video regression are unrelated regressions." + "reason": "Unsupported image-processor arg and video-input regression are separate regressions." }, { "left": "issue:44451", "right": "issue:45520", "accept": false, - "reason": "Model load failure for ScandiBERT and flash_attn import key error are unrelated." + "reason": "Model loading failure and a flash_attn import KeyError are unrelated." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "FineGrainedFP8Config failure and quantized model saving failure are both quantization-related but not the same bug." + "reason": "Quantization config failure and quantized saving failure are different code paths." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "EOS-token config inconsistency and flash-attention generation failure are different concrete problems." + "reason": "EOS-token config inconsistency and flash-attention generation issues are unrelated." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "MLukeTokenizer attribute errors and Wav2Vec2 save/tokenization failures do not share the same defect." + "reason": "Different tokenizers and different crash points; not the same bug." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "torch.export with torch_compilable_check and tokenizer/save failures are unrelated." + "reason": "torch.export with torch_compilable_check and Wav2Vec2 save/tokenization failure are unrelated." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "MacOS bus error in a community CLIP model and SigLIP2 documentation errors are not the same underlying issue." + "reason": "MacOS bus error for a community CLIP model and SigLIP2 documentation errors are not the same issue." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 test breakage and void segmentation-map processing are different bugs." + "reason": "Flash Attention test failures and void segmentation map processing are unrelated." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "KV-cache continuation indexing and Gemma4Processor missing _tokenizer are separate serving issues." + "reason": "Serving KV-cache indexing and Gemma4Processor missing _tokenizer are different serve-time failures." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 box-validation messaging and num_input_tokens_seen tracking are unrelated." + "reason": "LayoutLMv3 error-message clarity and num_input_tokens_seen updates are unrelated." }, { "left": "issue:44610", "right": "issue:45412", "accept": false, - "reason": "Processor output-size mismatch and RT-DETR memory release behavior are different problems." + "reason": "Processor size mismatch and memory release after deletion are unrelated issues." } ] }, @@ -8533,8 +8530,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -8623,7 +8620,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -8844,8 +8841,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -8932,7 +8929,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -9135,8 +9132,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -9227,7 +9224,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -9454,8 +9451,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -9542,7 +9539,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -9843,12 +9840,12 @@ "issue:42907|issue:43441", "issue:43388|issue:43688" ], - "prepared_review_unit_hash": "66e74064a42f1832d354a4368968ed994c423ea8ec4af1f0487bb1560f1b6594", + "prepared_review_unit_hash": "0e5aded3bb19f96d684a911a4c7bedfa738247bf64cbba9744f2887558566440", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -9856,207 +9853,207 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36555, - "estimated_input_tokens": 9139, - "estimated_eval_tokens": 18534 + "serialized_chars": 36557, + "estimated_input_tokens": 9140, + "estimated_eval_tokens": 18536 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The soft pairs are mostly superficial matches on broad Transformers terms (imports, multimodal, trust_remote_code, generation, CI) rather than the same concrete bug. I\u2019d reject every candidate edge; the set is too heterogeneous to form one duplicate cluster.", - "confidence": 0.78, - "canonical_issue_reason": "No single underlying bug cleanly covers this set; the issues span unrelated import failures, model/config regressions, generation bugs, docs/CI problems, and serialization quirks.", + "summary": "The cluster is highly heterogeneous: it mixes unrelated import errors, model/config regressions, pipeline bugs, and docs/test issues. The soft edges are driven by superficial lexical similarity (for example, many mentions of imports or model loading), but they do not describe the same underlying bug or change, so none should be merged.", + "confidence": 0.89, + "canonical_issue_reason": "If forced to pick a representative issue, issue 41628 is the broadest user-facing regression in the set: a straightforward import failure on a public API. It is closer to the recurring import/API-failure theme than the more specialized model or pipeline bugs.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45276 is the strongest standalone report here: it is concrete, model-specific, and describes a clear resize_token_embeddings/output-embedding propagation bug with a plausible fix target.", + "best_issue_reason": "Issue 41628 is the clearest, most general, and most likely to attract duplicate reports among this mixed set. That said, the cluster is not a true duplicate group, so this is only a representative choice.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:35141", "right": "issue:42490", "accept": false, - "reason": "Different bug classes: embedding resize/tie-word-weight behavior vs local-path slash warning handling." + "reason": "One is a token-embedding resize/tied-weights bug; the other is a local-model path trailing-slash import warning. Different code paths and failure modes." }, { "left": "issue:44112", "right": "issue:45600", "accept": false, - "reason": "Both mention CI, but they fail in different components and on different models; not the same code-path bug." + "reason": "CI stale-device-override test failure vs. auto_mappings referencing removed Sam3LiteText configs. Both are CI-related but clearly different breakages." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "Unrelated subsystems: SwitchTransformers config creation vs image-classification script dataset loading." + "reason": "SwitchTransformers sparse-layer construction bug vs. local dataset loading in an example script. No shared underlying defect." }, { "left": "issue:43650", "right": "issue:44220", "accept": false, - "reason": "No shared underlying bug; one is a vague placeholder issue, the other is an audio feature extraction problem." + "reason": "Placeholder-style issue title vs. audio feature extraction bug. No substantive overlap." }, { "left": "issue:39692", "right": "issue:44190", "accept": false, - "reason": "Documentation/model-processor mismatch and quantization issues do not match a local dataset loading failure." + "reason": "SigLIP2 docs example errors vs. local dataset loading in an image-classification script. Different problems, different surfaces." }, { "left": "issue:44496", "right": "issue:45803", "accept": false, - "reason": "Unrecognized model/config loading vs a bare except swallowing errors in Fuyu tensor conversion are different problems." + "reason": "Unrecognized model/config error vs. bare-except swallowing errors in Fuyu batch conversion. Different failure mechanisms." }, { "left": "issue:45276", "right": "issue:45584", "accept": false, - "reason": "Resize-token-embeddings weight propagation and Whisper empty-transcription handling are separate code paths." + "reason": "Gemma4 resize_token_embeddings behavior vs. Whisper empty transcription failure. Both involve generation/model behavior, but not the same bug." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols and different modules." + "reason": "Different import targets and likely different missing exports: Qwen2_5_VLImageProcessor vs. AutoImageProcessor." }, { "left": "issue:43479", "right": "issue:45375", "accept": false, - "reason": "Default config initialization when None is passed is not the same as a missing nested field being dropped by strict serialization." + "reason": "Phi4 multimodal default config initialization vs. Qwen3_5MoeVisionConfig missing a field. Both are config issues, but not the same one." }, { "left": "issue:45245", "right": "issue:45584", "accept": false, - "reason": "Category-count limit error in categorical encoding is unrelated to Whisper's empty-transcription failure." + "reason": "Category-count runtime error is unrelated to Whisper's empty-transcription generation failure." }, { "left": "issue:43295", "right": "issue:45276", "accept": false, - "reason": "Processor/tokenizer regression is a different bug from resize_token_embeddings not updating output embeddings." + "reason": "Processor/tokenizer regression in v4.57.5 vs. Gemma4 embedding resize bug. Different API surfaces and symptoms." }, { "left": "issue:44464", "right": "issue:45563", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency is not the same as a stale paged-generation warning." + "reason": "Chunked generation inconsistency with compiled forward vs. stale warning in paged generate. Different issues, one correctness and one warning behavior." }, { "left": "issue:44186", "right": "issue:44734", "accept": false, - "reason": "Tokenizer crash on NER/batched padding is unrelated to KV-cache continuation indexing in transformers serve." + "reason": "Tokenizer padding/truncation crash vs. serve KV-cache continuation indexing bug. Both are runtime bugs, but on different components." }, { "left": "issue:38175", "right": "issue:43065", "accept": false, - "reason": "SigLIP2 zero-probabilities and Sam3PixelDecoder dummy Conv2d are unrelated model bugs." + "reason": "SigLIP2 zero-probabilities model behavior vs. dummy Conv2d in Sam3PixelDecoder. Unrelated model internals." }, { "left": "issue:44683", "right": "issue:45446", "accept": false, - "reason": "Compiled flex_attention runtime failure on torch>=2.9 is not the same as an incorrect version check for AuxRequest import." + "reason": "Compiled flex_attention failure on torch>=2.9 vs. incorrect torch version gate for AuxRequest import. Related to torch checks only at a very broad level." }, { "left": "issue:44466", "right": "issue:44913", "accept": false, - "reason": "Serialization of lm_head.weight varies by device, while GPTNeoX rotary_pct reloads incorrectly; different defects." + "reason": "v5 serialization of lm_head.weight vs. GPTNeoX rotary_pct reload regression. Both are config/serialization related, but separate bugs." }, { "left": "issue:44933", "right": "issue:45362", "accept": false, - "reason": "A missing image_utils import and a Qwen3.5 chat crash are different failure modes." + "reason": "Nonexistent import from image_utils vs. Qwen3.5-35B chat crash. Import surface vs. model chat runtime failure." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Both involve weights, but one is Mistral3 lm_head tying and the other is Qwen3.5 save_pretrained checkpoint correctness; not the same concrete bug." + "reason": "Mistral3 lm_head tying bug vs. Qwen3.5 save_pretrained regression. Different models and different persistence/weight-tying problems." }, { "left": "issue:44246", "right": "issue:45003", "accept": false, - "reason": "Import slowdown is not the same as unsafe sys.modules access." + "reason": "Import taking long sometimes vs. unsafe sys.modules access in modeling_utils. Both touch import-time behavior, but one is performance and the other is unsafe access." }, { "left": "issue:44734", "right": "issue:45632", "accept": false, - "reason": "KV-cache continuation indexing and trust_remote_code cache-path collisions are unrelated." + "reason": "Serve KV cache indexing crash vs. trust_remote_code cache-path collision. Different subsystems and failure modes." }, { "left": "issue:44561", "right": "issue:45632", "accept": false, - "reason": "Breaking trust_remote_code by removing a helper is a different code path from local-model cache-path collision." + "reason": "Removal of is_torch_fx_available breaking trust_remote_code models vs. cache-path collision for local models. Both affect remote/local model loading, but not the same bug." }, { "left": "issue:36246", "right": "issue:38617", "accept": false, - "reason": "Both are import-related, but they target different missing names and different modules; not a single duplicate bug." + "reason": "Two separate import errors for different symbols in different modules. Same broad area, but not the same underlying missing export." }, { "left": "issue:45563", "right": "issue:45584", "accept": false, - "reason": "Paged generate warning cleanup and Whisper empty-transcription failure are separate issues." + "reason": "Paged generate warning mismatch vs. Whisper empty-transcription failure. Warning bug and functional generation bug are different." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "Missing fields from AutoConfig with trust_remote_code is not the same as failing to recognize a model in config.json." + "reason": "trust_remote_code AutoConfig missing fields vs. unrecognized model/config.json error. Both involve config loading, but one loses fields after loading and the other cannot resolve the model." }, { "left": "issue:44964", "right": "issue:45600", "accept": false, - "reason": "Phi-4 multimodal loading failure and stale auto_mappings references to removed Sam3LiteText configs are different problems." + "reason": "Phi-4 multimodal loading failure vs. Sam3LiteText auto_mappings CI breakage. Different models and different breakpoints." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "A CI test failure in GraniteSpeech is not the same as a model-loading regression for Phi-4 multimodal." + "reason": "CI stale device override test vs. model loading failure for Phi-4 multimodal. No shared code path." }, { "left": "issue:41950", "right": "issue:44016", "accept": false, - "reason": "Pipeline processor lookup bug vs a notebook syntax error; no shared underlying change." + "reason": "Video-classification pipeline using image processors vs. syntax error in a notebook. Completely unrelated." }, { "left": "issue:42490", "right": "issue:45632", "accept": false, - "reason": "Trailing-slash path warning and cache-path collision for trust_remote_code are distinct path-handling bugs." + "reason": "Trailing-slash warning on local model paths vs. cache-path collision between local models. Both concern path handling, but one is warning behavior and the other is cache key collision." }, { "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "Two different multimodal loading failures affecting different models and likely different code paths." + "reason": "Unit-test loading failure for GLM-4.7-Flash vs. Phi-4 multimodal loading failure. Different models and likely different root causes." }, { "left": "issue:42907", "right": "issue:43441", "accept": false, - "reason": "Dequantized save failure for Ministral/Devstral is unrelated to FlashAttention crashes in v5 RC." + "reason": "Saving dequantized Ministral/Devstrals vs. FlashAttention failure in Ministral-3. Same family name, but different operations and bug class." }, { "left": "issue:43388", "right": "issue:43688", "accept": false, - "reason": "Metric-gathering label truncation and auxiliary-loss normalization are unrelated training bugs." + "reason": "Metric gathering drops tuple labels in the last batch vs. auxiliary-loss normalization in OLMoE/GPT Oss. Both training-related, but not the same defect." } ] }, @@ -10075,8 +10072,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -10165,7 +10162,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -10380,8 +10377,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -10471,7 +10468,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -10698,8 +10695,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -10790,7 +10787,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -11101,12 +11098,12 @@ "issue:43931|issue:44368", "issue:44933|issue:45459" ], - "prepared_review_unit_hash": "ff794b14eae3d8728159a111e084c89edd466a8619cf619bfa46513f2852234d", + "prepared_review_unit_hash": "4139387bc5b2dce812fadf4a2be39656d607aeb503fa5fefed6737e54e484b90", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -11114,7 +11111,7 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35937, + "serialized_chars": 35939, "estimated_input_tokens": 8985, "estimated_eval_tokens": 18226 }, @@ -11123,186 +11120,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The proposed pairs are mostly superficial or subsystem-level overlaps rather than true duplicates. They should all be rejected. The item set is heterogeneous enough that there is no safe canonical issue or PR for the cluster.", - "confidence": 0.93, - "canonical_issue_reason": null, + "summary": "Most candidate pairs are clearly unrelated. The only soft duplicate-like pair is the resize_token_embeddings bug in the generic issue and the T5Gemma-specific manifestation.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45292 is the broader, model-agnostic report of resize_token_embeddings not updating output embeddings, while issue:45335 is a narrower T5Gemma-specific instance of the same behavior.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45292 is the best representative issue because it states the underlying bug in the most general form and can subsume the model-specific follow-up report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45987", "right": "issue:46082", "accept": false, - "reason": "Both concern config/value validation, but one is about StaticCache.get_seq_length() returning the wrong type and the other about LlamaConfig head_dim validation; different code paths and fixes." + "reason": "Different problems: StaticCache seq length contract vs LlamaConfig head_dim validation." }, { "left": "issue:36246", "right": "issue:43596", "accept": false, - "reason": "Both are import/loading failures, but one is a missing Qwen2.5-VL import and the other is a DeepSpeed ZeRO3/BertModel index error; unrelated bugs." + "reason": "ImportError for Qwen2.5-VL image processor is unrelated to DeepSpeed ZeRO3/BertModel indexing." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Both mention strict config handling, but one is about loading granite_speech and the other is a missing Qwen3_5MoeVisionConfig field being dropped; not the same defect." + "reason": "Both mention config handling, but they are different model configs and different failure modes; not the same concrete bug." }, { "left": "issue:45292", "right": "issue:45584", "accept": false, - "reason": "These both involve token/embedding handling, but one is about resize_token_embeddings not updating output embeddings and the other is a Whisper empty-transcription generation failure after align_special_tokens; distinct behavior." + "reason": "resize_token_embeddings/output embeddings bug is unrelated to Whisper empty transcription after align_special_tokens." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "Both touch testing/runtime warnings, but one is a pytest-xdist race in patched testing utils and the other is a stale warning for paged generate(); different locations and fixes." + "reason": "Testing-utils file race and paged generate stale warning are unrelated." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "Both involve incorrect handling of values in utilities/model code, but one is BeitImageProcessorFast reduce_labels returning one label and the other is unsafe sys.modules access in modeling_utils; not the same issue." + "reason": "BeitImageProcessorFast label reduction bug is unrelated to sys.modules access in modeling_utils." }, { "left": "issue:45292", "right": "issue:45701", "accept": false, - "reason": "The overlap is only around tokenization/embedding-related symptoms; resize_token_embeddings bugs are unrelated to version-dependent tokenization changes." + "reason": "Embedding resize bug is unrelated to tokenization changing across versions." }, { "left": "issue:43232", "right": "issue:44336", "accept": false, - "reason": "One is a generation kwargs/cache bug after sync_gpus, the other is ANSI codes emitted by loading_report when not attached to a terminal; no shared code-path." + "reason": "Generation kwargs update after sync_gpus is unrelated to ANSI output formatting in loading_report." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "Both involve loading/tokenizer-related errors, but one is a model load failure for physical-intelligence/fast under v5.1.0 and the other is protobuf error handling masking tokenizer exceptions; distinct bugs." + "reason": "Model loading failure and protobuf-related tokenizer error masking are different issues." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "One is an unrecognized model/config issue, the other is unsafe sys.modules access in modeling_utils; they do not describe the same underlying failure." + "reason": "Unrecognized model config and unsafe sys.modules access do not describe the same bug." }, { "left": "issue:43065", "right": "issue:43408", "accept": false, - "reason": "Both are SAM 3 related, but one reports a dummy Conv2d in Sam3PixelDecoder and the other a model-type mismatch warning for sam3_video vs sam3_tracker; different problems." + "reason": "SAM3 dummy Conv2d and SAM3 tracker-vs-video warning are related only by subsystem, not by the same bug." }, { "left": "issue:45375", "right": "issue:45759", "accept": false, - "reason": "Both reference Qwen multimodal config handling, but one is a missing deepstack_visual_indexes field in vision config and the other is AutoModelForCausalLM.from_config not unwrapping text_config; separate fixes." + "reason": "Both involve composite Qwen configs, but one is missing a field under strict parsing and the other is text_config unwrapping in from_config." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "One is audio processor chat-template sampling-rate defaulting incorrectly, the other is slow/occasional import behavior; no duplicate symptom or code path." + "reason": "Audio sampling-rate default bug is unrelated to import latency." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "One is Qwen3 MOE weight conversion with accelerate+deepspeed, the other is a Zero3 deque underflow; both involve distributed loading but are different failures." + "reason": "Both involve DeepSpeed, but one is MOE weight conversion during loading and the other is a deque IndexError in ZeRO3." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "Both mention index/label-related model behavior, but one is a GlmMoeDsa stale indexer cache crash and the other is Gemma 4 requiring mm_token_type_ids defaults; unrelated." + "reason": "Stale indexer cache crash and mm_token_type_ids defaults are different issues." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "Both are Qwen3 Omni/Video issues, but one is a missing fpn_position_embeddings attribute in SAM 3 Video and the other is a video-input ValueError in Qwen3OmniMoe; not the same bug." + "reason": "SAM3 vision encoder attribute error and Qwen3OmniMoe video-processing ValueError are unrelated." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "One is BigBirdTokenizer special-token registration/empty decode, the other is a broad ESM2 breakage report; no concrete shared defect." + "reason": "BigBirdTokenizer special-token registration is unrelated to ESM2 breakage." }, { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "Both are cache/indexing related, but one is a stale indexer cache crash in GlmMoeDsa and the other is a flash_attention AttributeError on s_aux=None; different code paths." + "reason": "Second-forward stale cache crash is unrelated to flash_attention s_aux=None AttributeError." }, { "left": "issue:45397", "right": "issue:45656", "accept": false, - "reason": "One is gemma-4 ZeRO3 from_pretrained loading, the other is optimizer.step being called twice with deepspeed; both involve deepspeed but are separate bugs." + "reason": "Gemma-4 ZeRO3 from_pretrained bug and duplicated optimizer.step under deepspeed are not the same failure." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "Both concern resize_token_embeddings not propagating everywhere, but one is a generic lm_head serialization/tied-weights inconsistency and the other is t5gemma decoder.embed_tokens not updating; related theme, not same fix." + "reason": "Both mention embedding-related behavior, but serialization of tied weights and resize_token_embeddings not updating decoder embeddings are different codepaths." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "One is a warning about tie_word_embeddings in Qwen3.5 LoRA fine-tuning, the other is a stale num_return_sequences warning in paged generate(); different model/setup and message." + "reason": "Qwen3.5 tied-weights warning is unrelated to a stale num_return_sequences warning." }, { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "Both are Qwen3.5-related warnings/behavior, but one is tie_word_embeddings warnings during fine-tuning and the other is flash-attention generation failure; distinct code paths." + "reason": "Warning about tie_word_embeddings is not the same as flash-attention generation failure." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "One is offline cache loading for tokenizers, the other is multi-GPU CUDA OOM for Gemma 4 31B; unrelated." + "reason": "Offline cache tokenizer loading and Gemma4 CUDA OOM are unrelated." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "One is unsafe sys.modules access in modeling_utils, the other is a testing race in captured_info.txt; different subsystems." + "reason": "sys.modules access bug and captured_info.txt race are unrelated." }, { "left": "issue:43673", "right": "issue:44734", "accept": false, - "reason": "Both touch generation cache behavior, but one is missing GenerationMixin cache in chunked_prefill and the other is a serve-time KV cache continuation tensor indexing crash; not the same underlying bug." + "reason": "Chunked-prefill cache missing and serve KV-indexing crash are different generation bugs." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "Both are DeepSpeed/CI-related failures, but one is a Zero3 deque underflow and the other is broken auto_mappings references to removed Sam3LiteText configs; unrelated." + "reason": "ZeRO3 deque error is unrelated to stale Sam3LiteText references in auto_mappings.py." }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "One is a Qwen3-Next tied-weights warning bug, the other is an optional post_init policy for patchtsmixer; different models and issues." + "reason": "Qwen3-Next tied-weights warning and patchtsmixer post_init policy are unrelated." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Both are Qwen model-loading-ish reports, but one is a weight-shape mismatch for Qwen3-VL-30B and the other is tie_word_embeddings warnings for Qwen3.5 fine-tuning." + "reason": "Qwen3-VL weight-shape mismatch is unrelated to a tied-weights warning in Qwen3.5 fine-tuning." }, { "left": "issue:44933", "right": "issue:45459", "accept": false, - "reason": "One is a missing/nonexistent import from image_utils, the other is tokenizer exception masking when protobuf is absent; different modules and symptoms." + "reason": "Removed image_utils import and protobuf decode-error masking are unrelated import issues." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "These describe the same underlying resize_token_embeddings bug: the generic report says output embeddings are not updated, and the T5Gemma report is a model-specific manifestation on decoder.embed_tokens." } ] }, @@ -11410,12 +11413,12 @@ "issue:44496|issue:45561", "issue:44964|issue:45137" ], - "prepared_review_unit_hash": "c041bdf49b4587e6f9e7c1bc83679b107f32bfe2a5cff3c555217125cccba793", + "prepared_review_unit_hash": "e23beb15d72be912bf9473f0e30e7552b85786d9fba9a7b2b56625f8aa5015a0", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -11423,228 +11426,228 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37420, - "estimated_input_tokens": 9355, - "estimated_eval_tokens": 18966 + "serialized_chars": 37422, + "estimated_input_tokens": 9356, + "estimated_eval_tokens": 18968 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most items are unrelated Transformers regressions, import errors, or model-specific bugs. There are at least two clear exact-duplicate issue pairs in the packet: issue:42673/issue:43704 (same Qwen3ForCausalLM VRAM leak symptom) and issue:45310/issue:45478 (identical title about Qwen3.5 MoE from_pretrained errors). Beyond those, the remaining soft edges are mostly conservative and should stay split.", - "confidence": 0.68, - "canonical_issue_reason": "issue:42673 is a strong canonical representative for its duplicate pair because it is the earlier, more detailed report of the Qwen3ForCausalLM multi-dataloader VRAM leak, matching issue:43704.", + "summary": "This cluster is mostly heterogeneous Transformers regression reports (imports, model loading, generation, quantization, and save/load issues). Only one soft pair looks like a true duplicate; the rest are related only at a broad subsystem/theme level and should stay separate.", + "confidence": 0.35, + "canonical_issue_reason": "Issue 42673 is the cleanest canonical issue because it is a concrete, well-scoped bug report and has an exact duplicate companion (43704) with the same title and failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the best representative overall among the clearly duplicated issues because it has the strongest duplicate evidence and richer discussion context; issue:45310 and issue:45478 are also exact duplicates, but represent a separate Qwen3.5 MoE from_pretrained problem.", + "best_issue_reason": "Issue 43646 is the best cluster representative overall because it most clearly summarizes the broader theme of Transformers v5 breaking existing model initialization paths, which matches the dominant pattern across the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Different models and failure modes: Whisper timestamp decode crash vs Kimi-K2.5 tokenizer regression." + "reason": "Both are tokenizer-related regressions, but one is a Whisper timestamp decode crash and the other is a Kimi tokenizer codec/regex regression; different code paths and fixes." }, { "left": "issue:44297", "right": "issue:44623", "accept": false, - "reason": "Both involve save_pretrained, but one is tokenizer metadata mismatch and the other is missing processor files." + "reason": "Both involve save_pretrained, but one is tokenizer metadata mismatch and the other is processor file omission; not the same underlying bug." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "Unrelated issues: MLukeTokenizer task attribute error vs a Chinese performance regression after upgrading Transformers." + "reason": "These are unrelated: one is an MLukeTokenizer AttributeError during tasks, the other is a broad performance regression after upgrading Transformers." }, { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Both are broad breakages, but the affected models and code paths differ; not the same bug." + "reason": "Both are Transformers regressions, but one is custom model initialization and the other is ESM2 breakage; too different to merge." }, { "left": "issue:43452", "right": "issue:45003", "accept": false, - "reason": "Loading gguf models/tokenizers is a different path from unsafe sys.modules access in modeling_utils." + "reason": "Both touch model-loading internals, but one is gguf/AutoModel loading and the other is unsafe sys.modules access in modeling_utils; different failures." }, { "left": "issue:43673", "right": "issue:43827", "accept": false, - "reason": "Generation cache behavior and deprecated pipeline docs are unrelated changes." + "reason": "One is a generation cache bug in chunked prefill, the other is a docs issue about pipeline() references; not the same change." }, { "left": "issue:44623", "right": "issue:45656", "accept": false, - "reason": "Processor save_pretrained missing files is unrelated to DeepSpeed causing optimizer.step to run twice." + "reason": "save_pretrained missing files and optimizer.step being called twice are unrelated bugs." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "A data-filtering slowdown and a DeepSpeed optimizer-step bug are not the same underlying issue." + "reason": "A data-filtering slowdown and a double optimizer step under deepspeed are different issues." }, { "left": "issue:44749", "right": "issue:45310", "accept": false, - "reason": "Different Qwen model issue types: performance regression versus from_pretrained loading error." + "reason": "Different symptoms and code paths: performance regression in data filtering vs Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model save failure and GGUF load failure are opposite directions and different code paths." + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are separate serialization/loading problems." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Both concern generation, but one is sync_gpus kwargs handling and the other is missing cache in chunked_prefill." + "reason": "Both are generation-related, but one is _update_model_kwargs_for_generation after sync_gpus and the other is missing GenerationMixin cache in v5.0.0; not the same bug." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF backend install packaging and remote-code meta-init incompatibility are unrelated." + "reason": "TF32 behavior settings and old InternVL2 meta-initialization incompatibility are unrelated." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "Missing PyTorch import error is not the same as Sentence Transformers hanging during load." + "reason": "AutoModel missing PyTorch and Sentence Transformers hanging on load are not the same underlying problem." }, { "left": "issue:43475", "right": "issue:45381", "accept": false, - "reason": "Different multimodal models and different bugs: missing attribute vs wrong vision_position_ids." + "reason": "Different multimodal model bugs: missing fpn_position_embeddings vs wrong vision_position_ids." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "Generic bug report does not match the specific Gemma-4 ZeRO3 from_pretrained issue." + "reason": "Generic bug report vs gemma-4 ZeRO3 from_pretrained issue; no shared failure mode." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "SAM 3 video encoder attribute error and trust_remote_code field-loss are unrelated." + "reason": "A missing attribute on a SAM3 output object is unrelated to trust_remote_code causing missing fields in AutoConfig." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Qwen3-VL-Moe loading issues and processor save_pretrained missing files are different problems." + "reason": "Qwen3-VL-Moe v5 regression and processor.save_pretrained missing files are different bugs." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Both are import-related, but the missing symbols and affected modules do not match closely enough." + "reason": "Both are import errors, but they concern different symbols and likely different export paths." }, { "left": "issue:43475", "right": "issue:44661", "accept": false, - "reason": "A SAM 3 video encoder attribute error is unrelated to tokenizer-mapping model registration." + "reason": "SAM3 output attribute failure and add-new-model-like/tokenizer mapping failure are unrelated." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Tokenizer regex patch crash and auto_docstring annotation parsing crash are different code paths." + "reason": "These are unrelated attribute errors in different utilities; they do not share the same code path." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized save failure and Llama4Config missing pad_token_id are unrelated." + "reason": "Quantized save failure and Llama4Config missing pad_token_id are separate model/config bugs." }, { "left": "issue:45375", "right": "issue:45478", "accept": false, - "reason": "Both are in the Qwen3.5 MoE area, but one is a missing config field and the other is a generic from_pretrained error; not enough to treat as the same bug." + "reason": "Both are Qwen3.5 MoE-related, but one is a missing config field silently dropped by strict validation and the other is a from_pretrained error; not the same concrete bug." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "A Qwen3.5 MoE loading error is not the same as a missing deepstack_visual_indexes config field." + "reason": "A from_pretrained loading error and a config-field dropping issue are related only by model family, not by the same fix." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "Missing AutoImageProcessor import and missing HybridCache import are different symbol regressions." + "reason": "Both are import failures, but they are for different names and likely separate API/export regressions." }, { "left": "issue:42673", "right": "issue:43704", "accept": true, - "reason": "Same exact title and symptom: Qwen3ForCausalLM leaks VRAM across multiple dataloader threads." + "reason": "Exact same bug report and title about Qwen3ForCausalLM leaking VRAM across multiple dataloader threads; clearly a duplicate." }, { "left": "issue:42831", "right": "issue:44222", "accept": false, - "reason": "FineGrainedFP8 accuracy degradation is different from FP8 save_pretrained failing for MoE models." + "reason": "FP8 accuracy degradation and FP8 save_pretrained failure are different issues." }, { "left": "issue:44222", "right": "issue:45310", "accept": false, - "reason": "FP8 MoE save_pretrained and Qwen3.5 MoE from_pretrained errors are different concrete failures." + "reason": "Different Qwen3.5 MoE problems: saving FP8 checkpoints vs loading from_pretrained." }, { "left": "issue:42994", "right": "issue:43824", "accept": false, - "reason": "A quantized model save issue is unrelated to a missing Qwen2.5-VL import." + "reason": "Quantized model saving and a missing Qwen2_5_VLForConditionalGeneration import are unrelated." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "GenerationMixin import error and AutoImageProcessor import error are separate missing-symbol regressions." + "reason": "Both are import errors after a version change, but they involve different symbols and different affected paths." }, { "left": "issue:42175", "right": "issue:45712", "accept": false, - "reason": "Torch-only install backend packaging and dummy_pt_objects leakage are related only broadly, not the same bug." + "reason": "Packaging/backend dependency omission and leftover dummy class cleanup are separate maintenance issues." }, { "left": "issue:44038", "right": "issue:45310", "accept": false, - "reason": "Qwen3-VL-Moe and Qwen3.5 MoE loading failures are different model families and likely different fixes." + "reason": "Both mention Qwen MoE and v5, but the reported failures are different model variants and different loading errors." }, { "left": "issue:44496", "right": "issue:45561", "accept": false, - "reason": "Unrecognized model/config loading and a pytest-xdist race in testing utils are unrelated." + "reason": "An unrecognized model/config error and a pytest-xdist race condition are unrelated." }, { "left": "issue:44964", "right": "issue:45137", "accept": false, - "reason": "Phi-4 multimodal loading and DeepSpeed ZeRO3 deque underflow are unrelated failures." + "reason": "A model loading failure and a DeepSpeed ZeRO3 deque IndexError are unrelated runtime bugs." } ] }, "evaluator_result": { - "accept": false, - "feedback": "Reject: the summary is not fully grounded because it misses another clear exact-duplicate pair in the packet, issue:45310 and issue:45478, which have the same title (\u2018[BUG] transformers>=5.4.0, Qwen3.5 Moe from_pretrained error\u2019). Because of that, the claim that issue:42673/issue:43704 is the only clear duplicate pair is overstated. The soft-edge verdicts are mostly conservative, but the overall clustering summary is incomplete." + "accept": true, + "feedback": "Grounded and conservative overall. The canonical duplicate callout for 42673/43704 is supported by the packet, and the summary appropriately keeps the rest of the pairs separate as broad-theme similarities rather than true duplicates." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -11743,12 +11746,12 @@ "issue:43976|issue:44933", "issue:44190|issue:46082" ], - "prepared_review_unit_hash": "e1518bb2967f084bdb457ab6442883bef2bd03c0daeaa2ae7c5011b06e577df9", + "prepared_review_unit_hash": "155917dd5282e710ea2f0f364a36a5d8bea880bc806081c1d70acfcab74d67ff", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -11765,192 +11768,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of distinct regressions across tokenizers, multimodal models, generation, DeepSpeed, docs, and packaging. The soft links are mostly superficial subsystem similarity rather than the same underlying bug, so I rejected all of them.", - "confidence": 0.93, - "canonical_issue_reason": "issue:42898 is the broadest, most central report in the tokenizer/regression family and best represents the cluster\u2019s strongest theme, even though the cluster overall is heterogeneous.", + "summary": "This cluster is mostly a noisy mix of unrelated issues, with a loose concentration around tokenizer/v5 regressions. None of the proposed soft edges look like true duplicates of the same underlying bug/change, so all should be rejected.", + "confidence": 0.9, + "canonical_issue_reason": "No single duplicate set is evident across the whole cluster. If one issue must represent the strongest canonical candidate, issue:45812 is the clearest broad, concrete tokenizer regression report with a silent v4\u2192v5 impact and actionable scope.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42898 is the clearest and most general issue among the set, with a broad v5 tokenizer behavior regression framing that can anchor related reports.", + "best_issue_reason": "issue:45812 stands out as the best representative issue because it describes a widespread, user-visible regression ('wrong token IDs' across Granite models) with a clear version boundary and concrete failure mode.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "Different bugs: MIMI batching/padding-mask mismatch vs MoE load-balancing loss gating." + "reason": "Different bugs: MIMI batched-vs-single encoder output mismatch vs MoE load-balancing loss gating." }, { "left": "issue:45561", "right": "issue:45600", "accept": false, - "reason": "Unrelated: xdist file race in testing utils vs stale Sam3LiteText mappings in auto_mappings.py." + "reason": "Unrelated: xdist file race in testing utils vs stale auto_mappings references to removed configs." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "Different models and failures: vitpose import/runtime breakage vs Gemma 4 multimodal token defaults." + "reason": "Different areas: broken VitPose model loading vs Gemma 4 text-only mm_token_type_ids defaulting." }, { "left": "issue:43901", "right": "issue:45561", "accept": false, - "reason": "Docs mismatch for TextClassificationPipeline vs parallel test-file race; not the same change." + "reason": "Docs/API mismatch for TextClassificationPipeline vs test utility race; not the same bug." }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation cache/sync_gpus bug vs janus image-generation test failure; no shared code-path." + "reason": "Generation kwargs/cache issue vs Janus image-generation test failure; no shared code-path evidence." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer regressions, but one is clean_up_tokenization_spaces and the other is add_special_tokens/BOS-EOS handling." + "reason": "Both tokenizer-related, but one is clean_up_tokenization_spaces behavior and the other is add_special_tokens/BOS-EOS handling; too different to merge." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "Stuck loading issue vs token embedding reinitialization on resize; different subsystems." + "reason": "Sentence Transformers loading hang is unrelated to token embedding reinitialization on resize." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation kwargs propagation vs attention-kernel divergence under torch.compile; unrelated code paths." + "reason": "Generation cache update issue vs attention kernel divergence under torch.compile; different subsystems and failure modes." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Processor save_pretrained file omission vs Qwen3.5 MoE from_pretrained error; not the same bug." + "reason": "Processor.save_pretrained missing files is unrelated to Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "Runtime kernel divergence vs a docstring correction for position_ids." + "reason": "Runtime kernel divergence is unrelated to a docstring typo for position_ids." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Import NameError in sentence-transformers vs missing Qwen2.5-VL symbol; different import failures." + "reason": "Importing sentence-transformers with latest transformers vs missing Qwen2.5-VL export are separate import-time problems." }, { "left": "issue:42898", "right": "issue:45812", "accept": false, - "reason": "Both tokenizer-related, but one is a generic v5 cleanup-space change and the other is Granite-specific token ID regression." + "reason": "Both are tokenizer regressions, but they affect different behaviors and model families; not the same underlying defect." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "Dtype drift during evaluate vs DeepSpeed ZeRO3 deque IndexError; unrelated." + "reason": "Embedding dtype drift during evaluation is unrelated to DeepSpeed ZeRO3 deque underflow." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "Serve KV-cache tensor indexing bug vs auto-docstring crash from future annotations; different areas." + "reason": "Serving KV-cache tensor indexing crash vs auto-docstring annotation parsing crash; unrelated code paths." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "Failed MLFlow reporting vs DeepSpeed adafactor OOM; no common root cause." + "reason": "MLFlow reporting on failed training jobs is unrelated to Adafactor OOM with DeepSpeed." }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Missing image-processor method vs remote-code meta-init incompatibility; distinct multimodal issues." + "reason": "Missing image-patch method in GLM46V processor is unrelated to InternVL2 meta-initialization incompatibility." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "Tokenizer loading regression vs special-token insertion behavior; separate tokenizer problems." + "reason": "Tokenizer loading failure vs tokenizer add_special_tokens behavior regression; not the same concrete bug." }, { "left": "issue:42898", "right": "issue:45356", "accept": false, - "reason": "Different tokenizer regressions: clean_up_tokenization_spaces behavior vs Kimi-K2.5 codec/warning handling." + "reason": "Different tokenizer regressions in v5 with different symptoms and model-specific handling." }, { "left": "issue:42898", "right": "issue:44779", "accept": false, - "reason": "Generic tokenizer cleanup change vs DeepSeek tokenizer output regression; not the same defect." + "reason": "Both concern tokenizer correctness, but the reported failures and affected models differ enough to avoid merging." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Incorrect tied-weights warning vs missing tensor-parallel plan; different model config/parallelism bugs." + "reason": "Incorrect tied-weights warning vs missing _tp_plan for tensor parallelism are distinct Qwen3-related issues." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "Dependency version bump problem vs leftover dummy-class cleanup; unrelated." + "reason": "PEFT version pin issue is unrelated to leftover dummy classes leaking into dir(transformers)." }, { "left": "issue:43673", "right": "issue:44188", "accept": false, - "reason": "GenerationMixin chunked-prefill cache issue vs attention kernel divergence; different failures." + "reason": "Chunked prefill cache missing in GenerationMixin is unrelated to torch.compile attention-kernel branching." }, { "left": "issue:42898", "right": "issue:44448", "accept": false, - "reason": "Both are v5 regressions, but one is tokenizer cleanup behavior and the other is Pegasus output drift." + "reason": "Different output between v4 and v5 for Pegasus is not shown to be the same root cause as clean_up_tokenization_spaces behavior change." }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "position_ids buffer access error vs optimizer.step being called twice under deepspeed; different bugs." + "reason": "Position_ids buffer read error and double optimizer.step under DeepSpeed are different runtime bugs." }, { "left": "issue:44933", "right": "issue:45712", "accept": false, - "reason": "Missing import from image_utils vs leftover dummy classes in dummy_pt_objects.py; unrelated cleanup/import issues." + "reason": "Missing import from image_utils and leftover dummy classes are separate repo hygiene/import issues." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Packaging/backend dependency omission vs image_utils import problem; no shared underlying bug." + "reason": "Packaging backend selection for pip extras is unrelated to an import from image_utils." }, { - "left": "issue:43479", - "right": "issue:44079", + "left": "issue:43976", + "right": "issue:44933", "accept": false, - "reason": "Phi4MultimodalConfig default-init bug vs ModelOutput key assignment bug; different object models." + "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to image_utils import issues." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "Both involve save_pretrained, but FP8 MoE serialization and tokenizer_class metadata mismatch are distinct problems." - }, - { - "left": "issue:43976", - "right": "issue:44933", - "accept": false, - "reason": "Python version incompatibility vs nonexistent image_utils import; unrelated." + "reason": "FP8 MoE save_pretrained behavior and tokenizer_class metadata mismatch are unrelated save/load bugs." }, { "left": "issue:44190", "right": "issue:46082", "accept": false, - "reason": "Local dataset loading failure in an example script vs LlamaConfig head_dim validation; different code paths." + "reason": "Local dataset loading in an example script is unrelated to LlamaConfig head_dim validation." } ] }, @@ -12056,12 +12053,12 @@ "issue:44829|issue:46002", "issue:42907|issue:46050" ], - "prepared_review_unit_hash": "6b7f748c04a766b796cb18b4a781103c939e9aa1c6dca383a1b58a48662c6198", + "prepared_review_unit_hash": "ff92a12ede27969af87fb7c95a4f06fa5683418a5c78edbc0dea9ceb0631a953", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -12078,192 +12075,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is a noisy mix of unrelated issue reports. The suggested soft pairs share only broad keywords or subsystem areas, not the same concrete bug, so none should be merged as duplicates.", - "confidence": 0.96, - "canonical_issue_reason": "No single issue is a defensible canonical representative here; the items span many unrelated regressions, import errors, serialization problems, and example-script bugs.", + "summary": "This cluster is heterogeneous: most items are distinct Transformers bugs across imports, training, save/load, generation, and model-specific regressions. None of the soft edges look like the same underlying fix, so they should stay separate.", + "confidence": 0.92, + "canonical_issue_reason": "No single canonical issue: these are mostly unrelated bugs spanning imports, training loops, save/load, generation, and model-specific regressions, so none cleanly absorbs the others.", "canonical_pr_reason": null, - "best_issue_reason": "No issue stands out as the best cluster anchor because the pairwise similarities are superficial rather than indicating one underlying defect.", + "best_issue_reason": "If one issue must anchor the cluster, issue 43723 (tokenizer loading in v5) is the most general user-facing regression, but the cluster is too mixed for it to be a true duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Both are example-script problems, but they affect different scripts and different failure modes; not the same bug." + "reason": "Both involve example scripts, but one is empty multi-label classification output and the other is local dataset loading failure in a different script and code path." }, { "left": "issue:44297", "right": "issue:45656", "accept": false, - "reason": "Tokenizer save metadata mismatch and deepspeed optimizer double-step are unrelated issues." + "reason": "Tokenizer save_pretrained metadata mismatch and a deepspeed optimizer double-step bug are unrelated problems." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "FP8 accuracy drift and missing processor save files are different serialization/runtime problems." + "reason": "FP8 accuracy loss and missing files from processor.save_pretrained affect different features and failure modes." }, { "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "Both mention save_pretrained, but one is FP8 MoE saving and the other is processor file omission; different code paths." + "reason": "Both mention save_pretrained, but one is FP8 MoE checkpoint saving and the other is processor file export; not the same bug." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Multimodal config initialization and Whisper timestamp decoding are unrelated." + "reason": "Phi4 multimodal config initialization and Whisper timestamp decode are separate model-specific issues." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "A feature extractor API crash and an auto-docstring annotation bug are not the same underlying defect." + "reason": "LasrFeatureExtractor center-arg crash and auto_docstring AttributeError with future annotations are unrelated subsystems." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols from different libraries and likely different causes." + "reason": "Both are import errors, but they involve different missing symbols and different dependency problems." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "MLflow job state reporting and custom trainer loss signature incompatibility are unrelated." + "reason": "MLflow failure reporting and a Trainer compute_loss signature break are different training-stack bugs." }, { "left": "issue:45071", "right": "issue:46082", "accept": false, - "reason": "Generic PretrainedConfig type checking and LlamaConfig head_dim validation are separate config-breaking changes." + "reason": "PretrainedConfig type-checking regression and LlamaConfig head_dim validation are distinct config issues." }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "Training-job tracking and Llama 3.2 vision model loading are unrelated." + "reason": "Failed training-job reporting in MLflow and a model-loading regression for Llama 3.2 Vision are unrelated." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer class metadata mismatch and Qwen3.5 MoE from_pretrained failure are not the same bug." + "reason": "Tokenizer_class metadata mismatch on save and a Qwen3.5 from_pretrained error are different save/load failures." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "FP8 MoE save_pretrained behavior and max_shard_size default discussion are different issues." + "reason": "FP8 MoE save_pretrained bug and the 50GB max_shard_size default are not the same change or bug." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "Torch.split return handling in a DSA indexer and remote-code checkpoint meta-init incompatibility are unrelated." + "reason": "A torch.split return-value issue in GlmMoeDsaIndexer is unrelated to InternVL2 meta-initialization incompatibility." }, { "left": "issue:28282", "right": "issue:30333", "accept": false, - "reason": "PyTorch import requirement and MLflow reporting after failed training are unrelated." + "reason": "Missing PyTorch import and MLflow reporting on failed training jobs are completely different issues." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "Tokenizer loading regression and flash-attention generation quality are different problems." + "reason": "Tokenizer loading regression and Qwen3.5 flash-attention generation failure are different code paths." }, { "left": "issue:43232", "right": "issue:45834", "accept": false, - "reason": "Generation kwargs sync behavior and long OCR index errors are unrelated." + "reason": "Generation state update after sync_gpus and a Kosmos2.5 OCR index error are unrelated runtime bugs." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "MLflow failure reporting and embedding resize/post_init reinitialization are unrelated." + "reason": "MLflow job reporting and embedding resizing/post_init reinitialization are unrelated concerns." }, { "left": "issue:43761", "right": "issue:46082", "accept": false, - "reason": "CLIP hidden_states regression and LlamaConfig head_dim validation are not the same issue." + "reason": "CLIPVisionModel hidden_states regression and LlamaConfig head_dim validation are different problems." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization offloading and pipeline-parallel output correctness are different distributed-inference bugs." + "reason": "Quantization offloading behavior and pipeline-parallel LLM output corruption are separate infra bugs." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "Docstring wording for position_ids and GLM-5 RoPE implementation are not the same defect." + "reason": "A docstring mistake for position_ids and GLM-5 RoPE implementation are not the same underlying issue." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "Chat-completions API request rejection and Gemma4Processor attribute crash are different serve-side failures." + "reason": "Chat-completions request rejection and Gemma4Processor missing _tokenizer crash are different server failures." }, { "left": "issue:44263", "right": "issue:44360", "accept": false, - "reason": "Both involve the DSA indexer, but one is about split return values and the other about a missing ReLU; different code changes." + "reason": "Torch.split return values and a missing ReLU in the DSA indexer describe different code defects." }, { "left": "issue:44295", "right": "issue:44623", "accept": false, - "reason": "Reading position_ids after buffer registration and processor.save_pretrained file omission are unrelated." + "reason": "Reading position_ids after buffer registration and processor.save_pretrained missing files are unrelated bugs." }, { "left": "issue:44514", "right": "issue:45103", "accept": false, - "reason": "Batched chat template padding crash and annotation-introspection docstring crash are unrelated." + "reason": "Batched apply_chat_template crashing and auto_docstring failing with future annotations are different features." }, { "left": "issue:44261", "right": "issue:45092", "accept": false, - "reason": "MLA rms_norm_eps precision issue and InternVL2 meta-init incompatibility are different configuration bugs." + "reason": "Missing rms_norm_eps precision issue in MLA q_a_layernorm and InternVL2 meta-init incompatibility are not the same bug." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Default multimodal config initialization and non-persistent buffer junk values are unrelated." + "reason": "Phi4MultimodalConfig None-handling and non-persistent buffer junk in v5 are distinct regressions." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "InternVL2 meta-init compatibility and Qwen3.5 visual-encoder key saving are different save/load issues." }, { "left": "issue:43723", "right": "issue:44871", "accept": false, - "reason": "Tokenizer loading regression and Gemma eos_token_id mismatch are different model/config issues." + "reason": "Tokenizer loading failure and Gemma3 eos_token_id inconsistency affect different parts of model initialization." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Sentence-transformer loading stall and zero-probabilities in SigLIP are unrelated." + "reason": "Sentence Transformers loading hangs and SigLIP2 zero-probability outputs are unrelated." }, { "left": "issue:44829", "right": "issue:46002", "accept": false, - "reason": "Flash-attention-3 training degeneration and NaNs after checkout/init are different classification-head problems." + "reason": "Flash-attention training degeneration and NaNs in classification heads after checkout/init are different bugs." }, { "left": "issue:42907", "right": "issue:46050", "accept": false, - "reason": "Dequantized model save failure and sinq save/load corruption are both quantization-related but not the same concrete bug." + "reason": "Dequantized model save failure and quantized model save/load breakage are related by theme but not the same concrete bug or fix." } ] }, @@ -12282,8 +12285,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -12371,7 +12374,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -12586,8 +12589,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -12676,7 +12679,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -12897,8 +12900,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -12987,7 +12990,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -13196,8 +13199,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -13288,7 +13291,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -13612,12 +13615,12 @@ "issue:42907|issue:44655", "issue:44483|issue:44514" ], - "prepared_review_unit_hash": "3edb7453502972e0c7d4483fd2d464a5c332bcd48ece3ffde9d4a3da1e4b0b2e", + "prepared_review_unit_hash": "7c7d875a084d8d3f121b5fffd2abe1a17aac54b3270f28cbd68b66fb0cbb018a", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -13625,231 +13628,231 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38015, - "estimated_input_tokens": 9504, - "estimated_eval_tokens": 19264 + "serialized_chars": 38017, + "estimated_input_tokens": 9505, + "estimated_eval_tokens": 19266 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: import/symbol errors, processor/loading regressions, tokenizer issues, model-specific architecture bugs, and unrelated training/offload problems. The soft pairs mostly share only broad Transformers vocabulary, not the same concrete failure mode, so none look safe to merge as duplicates.", - "confidence": 0.19, - "canonical_issue_reason": "Issue 43986 is the closest representative of the dominant user-facing theme here: a loading/processor crash with a clear reproduction and broad impact. Still, the cluster does not form a true duplicate set.", + "summary": "These items are mostly unrelated Transformers bug reports that only share broad topical similarity (imports, tokenizers, model loading, offload/quantization, training, and API behavior). None of the soft pairs look like the same underlying bug or change, so I reject all duplicate links.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical duplicate emerges: the issues span distinct failure modes and subsystems, so there isn't one issue that cleanly represents the whole set as a duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": "43986 is the best anchor for this mixed cluster because it is well-scoped, reproducible, and broadly representative of the loading/import failures that appear most often among the candidates.", + "best_issue_reason": "issue:43873 is the most active open report with the broadest concrete impact (quantization/offloading), so it is the best representative if one issue must be chosen; even so, it is not a duplicate of the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Different problems: Python version compatibility vs incorrect outputs under pipeline parallelism." + "reason": "Different problems: a Transformers/Python compatibility issue versus incorrect pipeline-parallel LLM output." }, { "left": "issue:43986", "right": "issue:44484", "accept": false, - "reason": "Processor-loading crash vs a save_pretrained shard-size question; no shared bug path." + "reason": "Different bug classes: optional torchvision-dependent processor crash versus a question about save_pretrained shard size." }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different missing symbols in different dependencies." + "reason": "Both are import errors, but for different missing symbols from different modules (`layer_type_validation` vs `is_offline_mode`)." }, { "left": "issue:43064", "right": "issue:44315", "accept": false, - "reason": "Different training failures: optimizer-state corruption under FSDP2/PEFT vs Liger Kernel not applied with model_init." + "reason": "Different training issues: FSDP2/PEFT optimizer state corruption versus Liger Kernel not applied with `model_init`." }, { "left": "issue:43097", "right": "issue:46050", "accept": false, - "reason": "Removed tie_embeddings API vs quantized model save/load corruption; unrelated root causes." + "reason": "Different serialization/config topics: removed tie-embeddings API warning versus broken quantized save/load after `sinq`." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Different Qwen3 family bugs: missing projections in a model layer vs a config field being dropped." + "reason": "Different model/config bugs: missing QKV/gate projections in qwen3next versus a Qwen3_5MoeVisionConfig field being dropped." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "Tokenizer constructor TypeError vs mapping lookup returning None; separate code paths." + "reason": "Different tokenizer failures: `AddedToken(special=...)` argument conflict versus `TOKENIZER_MAPPING_NAMES` being `None` unexpectedly." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Two distinct missing imports from transformers, with different symbols and likely different compatibility breaks." + "reason": "Different missing imports from different packages (`layer_type_validation` in configuration utils vs `HybridCache` from transformers)." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "One is about a removed config option; the other is a flash_attention crash on sink-less models." + "reason": "Unrelated: tie-embeddings deprecation/removal warning versus an `integrations/flash_attention.py` crash on `s_aux=None`." }, { "left": "issue:42673", "right": "issue:45200", "accept": false, - "reason": "VRAM leak in multi-threaded dataloading vs a Gemma 4 multimodal token-id defaulting issue." + "reason": "Different areas: VRAM leak with multiple dataloader threads versus Gemma 4 text-only fine-tuning needing `mm_token_type_ids` defaulting." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "Related model family, but different concrete failures and no evidence of the same code path." + "reason": "Different Qwen3-family bugs in different model variants: Qwen3-VL-Moe loading versus qwen3next missing projections." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "AutoProcessor kwargs forwarding bug vs an auto-docstring crash with future annotations." + "reason": "Different layers: `AutoProcessor.from_pretrained` kwarg forwarding versus an `auto_docstring` crash under future annotations." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Model weight-shape mismatch vs missing layer projections; not the same defect." + "reason": "Different load-time failures: Qwen3-VL-30B-A3B weight shape mismatch versus missing qwen3next projections." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "local_files_only network leak vs missing embed_tokens weight in UMT5Encoder; unrelated." + "reason": "Different code paths: offline/local_files_only network leakage versus UMT5Encoder missing `embed_tokens.weight` on load." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "Layernorm epsilon precision bug vs GLM-5 RoPE implementation discussion; different issues." + "reason": "Different numerical/model issues: missing `rms_norm_eps` precision sensitivity versus GLM-5 RoPE implementation concerns." }, { "left": "issue:44964", "right": "issue:45341", "accept": false, - "reason": "Model loading regression vs a small testing_utils bug; no overlap in underlying behavior." + "reason": "Different failures: Phi-4 multimodal model loading versus a small testing_utils bug." }, { "left": "issue:45341", "right": "issue:45859", "accept": false, - "reason": "Testing utility bug vs CPU offload incompatibility in Qwen3_5MoeTextRotaryEmbedding." + "reason": "Different domains: testing utility bug versus Qwen3_5 rotary embedding incompatibility with CPU offload." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "API deprecation/removal warning vs quantization offloading not working; too broad to merge." + "reason": "Related broadly to v5 model behavior, but one is an API/config removal warning and the other is a quantization/offloading bug." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "Different failures: missing torchvision during processor loading vs tokenizer mapping assumptions." + "reason": "Different issues: missing torchvision for video processor loading versus tokenizer mapping returning `None`." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type vs an isolated reproduction of another issue; no shared concrete bug." + "reason": "Different tokenizer/model concerns: wrong decoder type in v5 versus an isolated reproduction of another issue." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel.py execution issue vs a testing_utils bug; unrelated." + "reason": "Different problems: failure running `3d_parallel.py` versus a minor testing utility bug." }, { "left": "issue:43986", "right": "issue:45290", "accept": false, - "reason": "Both involve chat/processor behavior, but one is a missing-dependency loading crash and the other is a tool-call templating bug." + "reason": "Different processor/template issues: missing torchvision crash versus batched chat-template handling with tool calls." }, { "left": "issue:43986", "right": "issue:45372", "accept": false, - "reason": "Loading crash due to missing torchvision vs Gemma 4 processor break from a missing mistral_common symbol." + "reason": "Different missing-dependency/import problems: torchvision absence versus `ReasoningEffort` import failure in mistral_common." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "Different testing-related problems: a small utility bug vs xdist race on captured_info.txt." + "reason": "Different test infra bugs: a small testing_utils issue versus an xdist race on `captured_info.txt`." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "Processor-loading crash vs testing_utils bug; no same underlying defect." + "reason": "Different scope: video-model processor dependency crash versus a testing_utils bug." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Different API surface issues: pipeline error message semantics vs inability to save Pipeline objects." + "reason": "Different API regressions: pipeline task support/error messaging versus inability to save Pipeline objects." }, { "left": "issue:45341", "right": "issue:45588", "accept": false, - "reason": "Testing utility bug vs flash_attention AttributeError; unrelated code paths." + "reason": "Different layers: testing utility bug versus flash-attention handling of `s_aux=None`." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing encoder weight on load vs save shard size default; different functionality." + "reason": "Different concerns: missing checkpoint weight on UMT5 load versus save_pretrained shard-size default behavior." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training blockers vs Bamba torch.compile+SDPA failure; not the same bug." + "reason": "Different model/back-end bugs: xLSTM training blockers versus Bamba SDPA/torch.compile failure." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "bfloat16 compile warning vs model-specific torch.compile/SDPA failure." + "reason": "Different symptoms: a bfloat16 compile precision warning versus an actual model compile failure with SDPA." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "Precision warning during compile vs a tie_word_embeddings warning in fine-tuning; different symptoms and causes." + "reason": "Both mention warnings, but one is a PyTorch precision warning and the other is a model config warning about `tie_word_embeddings`." }, { "left": "issue:42886", "right": "issue:43381", "accept": false, - "reason": "Offline cache loading regression vs gradient checkpointing not usable in eval mode." + "reason": "Different runtime constraints: offline cache loading bug versus gradient checkpointing being unavailable in eval mode." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "Saving dequantized models vs an infinite-loop generation bug on a specific image." + "reason": "Different model behaviors: save/load of dequantized models versus an infinite loop on a specific image." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Dequantized save failure vs inability to save Pipeline objects; different save paths and causes." + "reason": "Different serialization paths: dequantized model save failure versus saving Pipeline objects." }, { "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "Chat/completions endpoint behavior vs Qwen2_5_VLProcessor batching crash; not the same code-path problem." + "reason": "Different request-processing bugs: v5 chat/completions schema rejection versus batched `apply_chat_template` crashing on tool-call assistant messages." } ] }, @@ -13868,8 +13871,8 @@ "review_unit_count": 48, "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-10000-410", "nodes": [ @@ -13962,7 +13965,7 @@ "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -14279,12 +14282,12 @@ "issue:44222|issue:45125", "issue:42548|issue:44351" ], - "prepared_review_unit_hash": "2f4d83ec0ac10ebba2e7c0e3527a5b5b95b73d26ccd2ab3e31e09ce82f2b007c", + "prepared_review_unit_hash": "f8cb859e7eaf4a7dfe3d42a9789409bf3c7b1f34817348abe2e2014a43a435a3", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -14292,226 +14295,241 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37455, - "estimated_input_tokens": 9364, - "estimated_eval_tokens": 18984 + "serialized_chars": 37457, + "estimated_input_tokens": 9365, + "estimated_eval_tokens": 18986 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is largely heterogeneous: most items are unrelated import/load, model config, flash-attention, tensor-parallel, or test-flake reports. The only obvious exact duplicate in the set is the Qwen3.5 MoE from_pretrained regression (issue:45310 / issue:45478), but none of the supplied soft edges are strong enough to merge.", - "confidence": 0.84, - "canonical_issue_reason": "issue:45310 is the cleaner canonical for the duplicated Qwen3.5 MoE from_pretrained regression: it is earlier and has a follow-up duplicate (issue:45478) with the same title and symptom.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers issues (imports, model loading, FlashAttention, tensor-parallel, tests). The only strong duplicate-looking pair is the import-time slowdown reports around `transformers` import performance.", + "confidence": 0.68, + "canonical_issue_reason": "issue:41669 is the best canonical issue because it is the broadest, most clearly scoped performance regression in the set and has the closest duplicate-style neighbor (`issue:44246`).", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the best representative issue because it captures the recurring Qwen3.5 MoE from_pretrained bug clearly and is the earliest linked report; the rest of the set is mostly unrelated noise.", + "best_issue_reason": "issue:41669 is the strongest global issue candidate: clear title, concrete symptom, and it represents the only clearly related duplicate cluster in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "Different failures: missing embed_tokens.weight in UMT5Encoder vs Qwen3.5 MoE from_pretrained error." + "reason": "Both are loading errors, but for different models and causes (`UMT5Encoder` missing weights vs Qwen3.5 MoE `from_pretrained` failure). Not the same bug." }, { "left": "issue:43441", "right": "issue:44380", "accept": false, - "reason": "Both involve attention backends, but the concrete bugs differ: Ministral-3 FlashAttention failure vs GPT2 attention scaling being ignored." + "reason": "Both mention FlashAttention/backend behavior, but one is a model-support error for Ministral-3 and the other is GPT2 attention scaling being ignored. Different code paths." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "Unrelated model/code-paths: GPT2 attention scaling vs a sequence-classification zero-loss bug across multiple models." + "reason": "GPT2 backend scaling bug is unrelated to the sequence-classification zero-loss bug. Same broad modeling area, not the same issue." }, { "left": "issue:43352", "right": "issue:44380", "accept": false, - "reason": "Nemotron FlashAttention support error is not the same underlying bug as GPT2 attention scaling being ignored." + "reason": "Nemotron FlashAttention 2.0 support error is a model-specific capability issue; GPT2 attention scaling under SDPA/FA is a different backend bug." }, { "left": "issue:43867", "right": "issue:45375", "accept": false, - "reason": "Both are load/model issues, but one is state_dict sorting and the other is a missing config field; different root causes." + "reason": "One is a state-dict loading/order failure, the other is a config field being dropped by `@strict`. Different underlying failures." }, { "left": "issue:43352", "right": "issue:45478", "accept": false, - "reason": "Different models and symptoms; Nemotron FlashAttention support is unrelated to Qwen3.5 MoE from_pretrained." + "reason": "Different models and symptoms: Nemotron FA2 unsupported vs Qwen3.5 MoE `from_pretrained` error." }, { "left": "issue:41669", "right": "issue:44246", - "accept": false, - "reason": "Related to import performance in a broad sense, but not the same concrete bug or fix scope." + "accept": true, + "reason": "Both report the same underlying regression: `transformers` import becoming slow, with the latter describing the same import-time slowdown symptom in broader terms." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Both are Python import errors, but one is about compiled Python files and the other is a torch.jit.script parsing/indentation issue." + "reason": "Both fail while importing/using Python files, but one is about compiled Python files and the other is a PyTorch JIT parsing/indentation issue on Python 3.13. Different root causes." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "CLIPOutput attentions assignment and GPT2 attention scaling are different attention-related regressions." + "reason": "CLIPOutput attentions assignment regression is unrelated to GPT2 attention scaling under SDPA/FlashAttention." }, { "left": "issue:36683", "right": "issue:44117", "accept": false, - "reason": "Gemma3Config missing vocab_size and TOKENIZER_MAPPING_NAMES returning None are separate from_pretrained/config issues." + "reason": "Gemma3Config missing `vocab_size` and `TOKENIZER_MAPPING_NAMES` returning `None` are distinct `from_pretrained` failures." }, { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "A missing set_seed in a model integration test is not the same as a pytest-xdist race in captured_info.txt." + "reason": "Both are test flakiness, but one is missing `set_seed()` in SamHQ integration tests and the other is an xdist race on `captured_info.txt`. Not the same bug." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "FSDP2/PEFT optimizer-state bug and a cache-strategy typo are unrelated." + "reason": "Trainer optimizer-state corruption under FSDP2/PEFT is unrelated to a cache-strategy typo." + }, + { + "left": "issue:43064", + "right": "issue:44655", + "accept": false, + "reason": "Optimizer-state bug and `save_pretrained` pipeline persistence bug are different problems." }, { "left": "issue:43097", "right": "issue:43352", "accept": false, - "reason": "Missing tie_embeddings_and_encoder_decoder indication is unrelated to FlashAttention support errors." + "reason": "Missing `tie_embeddings_and_encoder_decoder` guidance in v5 and FlashAttention support errors are unrelated." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "State_dict ordering load failure and incorrect visual encoder keys during save_pretrained are different bugs." + "reason": "State-dict loading failure due to ordering and incorrect visual encoder keys on save are different save/load bugs." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Both concern Python/version compatibility, but they target different Python versions and failure modes." + "reason": "Both are Python-version compatibility complaints, but they are about different versions and different failure modes; not the same underlying issue." }, { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "UMT5Encoder missing weights is unrelated to the Qwen3.5 MoE from_pretrained regression." + "reason": "Different model families and distinct loading failures." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Tensor-parallel training bug and tensor-parallel OOM on model load are different problems." + "reason": "Tensor-parallel training bug and Qwen2-57B-A14B-Instruct OOM with TP are related only at a very high level; the concrete failures differ." }, { "left": "issue:42490", "right": "issue:44075", "accept": false, - "reason": "Local path trailing-slash import warning and unused SGD args are unrelated." + "reason": "Local path trailing-slash import warning and SGD args not being used are unrelated." }, { "left": "issue:43064", "right": "issue:44655", "accept": false, - "reason": "Optimizer state corruption in training is unrelated to save_pretrained for Pipeline objects." + "reason": "Different training/persistence issues with no shared concrete code path." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "PyTorch-missing ImportError and tokenizer-loading regression are different issues." + "reason": "PyTorch-not-found ImportError and tokenizer-loading regression are both import/load issues, but not the same bug." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "A DSA indexer/ReLU discussion is not the same as DeepseekV3MoE diverging from the remote implementation." + "reason": "DSA indexer ReLU discussion is unrelated to DeepseekV3MoE divergence from remote implementation." }, { "left": "issue:43867", "right": "issue:44484", "accept": false, - "reason": "A load error from sorted state_dict keys is unrelated to the 50GB max_shard_size question." + "reason": "State-dict loading failure is unrelated to the `max_shard_size` question." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "CLIPOutput attentions assignment and an AMP/CUDA crash in LwDetrImageLoss are different failures." + "reason": "CLIPOutput attentions regression and LwDetrImageLoss AMP/CUDA crash are separate failures." }, { "left": "issue:43618", "right": "issue:44977", "accept": false, - "reason": "CLIPOutput attentions regression is unrelated to Qwen3.5 generation issues with flash-attention." + "reason": "CLIPOutput attentions assignment and Qwen3.5 flash-attention generation are different bugs." }, { "left": "issue:43618", "right": "issue:45588", "accept": false, - "reason": "Both touch flash_attention code, but one is about CLIPOutput attentions and the other is an s_aux=None crash in sink-less models." + "reason": "CLIPOutput attentions regression is unrelated to `flash_attention.py` crashing on `s_aux=None`." }, { "left": "issue:44655", "right": "issue:44938", "accept": false, - "reason": "Saving Pipeline objects and Python 3.14 import/load failures are unrelated." + "reason": "Saving pipeline objects and Python 3.14 import failure are unrelated." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "A flaky SamHQ integration test is not the same bug as Phi-4-multimodal load failure." + "reason": "Seed-related test flakiness and Phi-4 multimodal loading failure are distinct." }, { "left": "issue:43901", "right": "issue:44938", "accept": false, - "reason": "Docs mismatch for return_all_scores is unrelated to Python 3.14 import failures." + "reason": "Pipeline docs mismatch and Python 3.14 import failure are unrelated." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Set_seed omissions in one test suite and a stale GraniteSpeech device override CI failure are different test bugs." + "reason": "SamHQ seeding flake and GraniteSpeech stale device override CI failure are different test issues." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "SamHQ test flakiness is unrelated to auto_mappings.py referencing removed Sam3LiteText configs." + "reason": "Test seeding/flakiness issue and removed `Sam3LiteText` config references in `auto_mappings.py` are unrelated." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "local_files_only network leakage and AddedToken keyword duplication are unrelated." + "reason": "Local-files-only network leak and `AddedToken()` keyword collision are completely different bugs." }, { - "left": "issue:44222", - "right": "issue:45125", - "accept": false, - "reason": "FP8 save_pretrained for MoE and missing _tp_plan for tensor parallelism are different subsystems and failure modes." + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Both are test-related, but one is xdist seeding/race flakiness and the other is a bug in `testing_utils.py`. Not the same issue." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 `save_pretrained` for MoE and missing `_tp_plan` for Qwen3.5 MoE are distinct tensor-parallel/save bugs." }, { "left": "issue:42548", "right": "issue:44351", "accept": false, - "reason": "Both are import-name errors, but they involve different exported symbols and likely different root causes." + "reason": "Both are top-level import errors, but for different missing exports (`PreTrainedModel` vs `HybridCache`)." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays grounded in the packet and the one accepted soft-edge pair (issue:41669 vs issue:44246) is conservative and well supported by the titles. The other verdicts also avoid overclaiming by treating only clear root-cause matches as non-duplicates. The canonical/best-issue choice is reasonable for this set." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -14617,12 +14635,12 @@ "issue:44261|issue:44360", "issue:44623|issue:45230" ], - "prepared_review_unit_hash": "eb146be0e324dbdb0b999d95831511b131c3852866bd3bf5024a1049e091777a", + "prepared_review_unit_hash": "996ad0f1265bb006ec33aff79d6ae36ab0006b46ccf84eafcbccd93069d2f5ae", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -14630,259 +14648,244 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 39, - "serialized_chars": 38248, - "estimated_input_tokens": 9562, - "estimated_eval_tokens": 19380 + "serialized_chars": 38250, + "estimated_input_tokens": 9563, + "estimated_eval_tokens": 19382 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a grab bag of unrelated Transformers issues; most look like distinct regressions, import failures, model-specific bugs, or docs problems. Only one soft edge looks plausibly like the same underlying breakage in the v5 `env` path.", - "confidence": 0.68, - "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster\u2019s recurring theme: upgrade-related import/loading breakages after moving from 4.57.x to v5.x.", + "summary": "This cluster is mostly about Transformers v5 upgrade regressions, especially import/load failures and related API breakages. None of the soft pairs are close enough to be treated as the same concrete bug, so I rejected all of them. The broadest representative issue is the one summarizing many import errors after upgrading to v5.5.0.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45278 best captures the shared theme of the cluster: a broad v5 upgrade regression causing multiple import failures. It is more representative than the narrowly scoped model- or feature-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 best fits as the umbrella issue because it is general enough to cover multiple nearby v5 regression reports without being as vague as the catch-all titles.", + "best_issue_reason": "issue:45278 is the most suitable cluster anchor because it is clear, general, and directly describes the kind of release-wide breakage that several other issues in the set also hint at.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "Different problems: one is an offline-loading network request bug, the other is a GLM-5 RoPE implementation discussion." + "reason": "Different problems: offline download behavior vs GLM-5 RoPE implementation." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "One concerns compiled Python file execution errors; the other is a post-upgrade data filtering slowdown." + "reason": "Unrelated bugs: compiled Python file errors vs a data filtering slowdown after upgrade." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3.5 MoE, but one is a missing tensor-parallel plan and the other is a from_pretrained load failure; not clearly the same bug." + "reason": "Both mention Qwen3.5 MoE, but one is tensor-parallel planning and the other is a from_pretrained load failure; not the same concrete bug." }, { "left": "issue:30990", "right": "issue:31795", "accept": false, - "reason": "Sentence-transformers loading hang and model.forward documentation confusion are unrelated." + "reason": "Sentence loading hang and documentation confusion are unrelated." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Tensor-parallel support and a missing config field in vision config are different code paths and symptoms." + "reason": "Different Qwen3.5 MoE issues: tensor parallelism vs missing config field serialization." }, { "left": "issue:30333", "right": "issue:33453", "accept": false, - "reason": "MLflow reporting on failed training jobs and tokenizer loading regression are unrelated." + "reason": "MLflow reporting on training failure is unrelated to tokenizer loading regression." }, { "left": "issue:44387", "right": "issue:44556", "accept": false, - "reason": "One is int4 quantization memory growth/OOM; the other is checkpoint reload compatibility after version upgrade." + "reason": "Int4 OOM from reserved memory and checkpoint reload failure are different bug paths." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "No-grad cache updates and import errors after a major upgrade are different bugs." + "reason": "Cache update no_grad behavior is unrelated to broad import errors after upgrade." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Tokenizer loading failure and checkpoint reload incompatibility are both loading-related but not the same concrete failure." + "reason": "Both involve loading, but one is tokenizer loading and the other is checkpoint reload compatibility; too broad to merge." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "FineGrainedFP8 accuracy drift and a missing image-patch method in a processor are distinct issues." + "reason": "FineGrainedFP8 accuracy and a missing image-patch helper are different issues." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Local-files-only API requests and FP8 save_pretrained for MoE are different behaviors." + "reason": "Local-files-only API calls are unrelated to FP8 save_pretrained behavior." }, { "left": "issue:43010", "right": "issue:44368", "accept": false, - "reason": "Cache update no-grad decoration and a tie_word_embeddings warning are unrelated." + "reason": "No-grad decoration for cache updates is unrelated to a tied-embeddings warning." }, { "left": "issue:43010", "right": "issue:45850", "accept": false, - "reason": "A specific autograd/cache bug is not the same as a generic 'too many models break' complaint." + "reason": "Cache update semantics are not the same as a general complaint that many models break." }, { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "Offline-loading API requests and the 'racoon' issue are unrelated." + "reason": "Offline API-request behavior and an unrelated generic issue title do not match." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Liger kernel application in model_init and a DSA ReLU omission are different subsystems." + "reason": "Liger Kernel application and a DSA indexer ReLU issue are unrelated." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Compiled Python file execution errors and PretrainedConfig type-checking breakage are unrelated." + "reason": "Compiled Python file errors and PretrainedConfig type-checking breakage are different regressions." }, { "left": "issue:30333", "right": "issue:36296", "accept": false, - "reason": "MLflow failure reporting and tensor-parallel training bugs do not share the same code path." + "reason": "MLflow reporting and tensor-parallel training bugs are unrelated." }, { "left": "issue:43827", "right": "issue:44961", "accept": false, - "reason": "Docs still referencing pipeline() and an unrelated 'racoon' issue are not duplicates." + "reason": "Docs still referencing pipeline() and a generic issue title are not the same bug." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Compiled-file errors and inability to save Pipeline objects are different failures." + "reason": "Running compiled Python files and saving Pipeline objects are separate failure modes." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "FP8 save_pretrained vs FP8 inference support for Qwen3-VL/Qwen3.5 MoE are related only at a broad feature level, not the same bug." + "reason": "FP8 save_pretrained for MoE and FP8 inference support are different code paths." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Both involve model loading after version changes, but they target different model families and likely different regressions." + "reason": "Different model-loading failures affecting different model families and versions; too broad to combine." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Tokenizer loading regression and Llama 3.2 Vision model loading failure are not the same issue." + "reason": "Tokenizer loading regression and Llama 3.2 vision model loading are distinct issues." }, { "left": "issue:42757", "right": "issue:43576", - "accept": true, - "reason": "These plausibly describe the same underlying v5 breakage in the `transformers env` path: an import/API removal around `is_offline_mode` that makes the command fail." + "accept": false, + "reason": "Missing huggingface_hub symbol and a broken transformers env command are not the same bug." }, { "left": "issue:33453", "right": "issue:36296", "accept": false, - "reason": "Tokenizer loading and tensor-parallel training are unrelated." + "reason": "Tokenizer loading regression and tensor parallel training bug are unrelated." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "The env-command failure and PretrainedConfig typing regression are different bugs." + "reason": "Broken env command and config type-checking regression are different failures." + }, + { + "left": "issue:43010", + "right": "issue:45656", + "accept": false, + "reason": "Cache update no_grad behavior is unrelated to optimizer stepping twice under deepspeed." }, { "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "The env command breakage and Qwen2.5-VL import failure are distinct model/import paths." + "reason": "Broken env command and Qwen2.5-VL import error are different API issues." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "Docs confusion around forward arguments and a missing import symbol are unrelated." + "reason": "Documentation confusion and an ImportError from configuration_utils are unrelated." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "AutoImageProcessor import failure and PreTrainedModel import failure are separate API regressions." + "reason": "Two different missing imports; same general area, but not the same concrete symbol or code-path problem." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "No meaningful relation between 'racoon' and Gemma-4 audio positional encoding." - }, - { - "left": "issue:43010", - "right": "issue:43976", - "accept": false, - "reason": "Cache update no-grad behavior and Python 3.9/3.10 compatibility are unrelated." - }, - { - "left": "issue:38617", - "right": "issue:43576", - "accept": false, - "reason": "Missing layer_type_validation import and env-command breakage are different import failures with different surfaces." - }, - { - "left": "issue:43576", - "right": "issue:44938", - "accept": false, - "reason": "The env command issue and Python 3.14 load failure are distinct compatibility problems." + "reason": "Generic issue title and Gemma-4 audio positional encoding bug do not match." }, { "left": "issue:44908", "right": "issue:44961", "accept": false, - "reason": "Scheduler kwargs handling and the 'racoon' issue are unrelated." + "reason": "Scheduler kwargs handling is unrelated to the generic racoon issue." }, { "left": "issue:36296", "right": "issue:43901", "accept": false, - "reason": "Tensor-parallel training and pipeline docs mentioning return_all_scores are unrelated." + "reason": "Tensor-parallel training and text classification docs behavior are unrelated." }, { "left": "issue:41628", "right": "issue:44355", "accept": false, - "reason": "AutoImageProcessor import failure and compiled-file execution errors are different failure modes." + "reason": "AutoImageProcessor import failure and compiled Python file errors are separate regressions." }, { "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 label reduction are unrelated." }, { "left": "issue:44261", "right": "issue:44360", "accept": false, - "reason": "MLA q_a_layernorm precision/config issue and a missing ReLU in DSA indexer are unrelated." + "reason": "MLA rms_norm_eps precision issue and DSA indexer ReLU omission are different bugs." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "Missing save_pretrained files and a generic 'Bug report' are not enough to establish a duplicate underlying bug." + "reason": "Processor.save_pretrained missing files and a placeholder bug report are not the same underlying issue." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "Grounded and conservative overall. The summary matches the packet well: most pairs are clearly unrelated, and the one accepted soft edge (issue:42757 vs issue:43576) is plausibly the same v5 env/import breakage. The choice of issue:45278 as the umbrella is supported by its broad title about many import errors after the 4.57.0 -> 5.5.0 upgrade." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -14989,12 +14992,12 @@ "issue:44360|issue:45468", "issue:44655|issue:44861" ], - "prepared_review_unit_hash": "b7f0786bd7c50702735e7e4b8bc7ce4c0a15fefea15a5c60662689b7c4c73656", + "prepared_review_unit_hash": "18a727ea631ef2dbfd8c3fd0e803d67f116edc5b6bb24d13cc6e46bf93c4c7ab", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -15011,252 +15014,255 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose mix of post-v5 regressions, import/lazy-loading problems, model-specific config errors, and a few docs/tests issues. None of the soft pairs look like the same concrete bug or change, so I would not merge any of them as duplicates.", - "confidence": 0.69, - "canonical_issue_reason": "issue:45278 is the best umbrella issue for the dominant theme here: broad import/load breakage after upgrading to v5. It is more representative than the very generic \"breaks too many models\" report and more central than the isolated model-specific bugs.", + "summary": "The cluster is still mostly heterogeneous. 44273 is the broad lazy-loading/import regression issue, but 45003 looks like a separate modeling_utils/sys.modules bug rather than a confirmed root cause, so the soft edge should be rejected.", + "confidence": 0.59, + "canonical_issue_reason": "44273 is the broad user-facing lazy-loading/import regression; 45003 appears to be a separate sys.modules access bug, not a proven root cause of the same issue.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 has the clearest, most general description of the recurring regression class in this cluster (many import errors after a v4.57 -> v5.5 upgrade), making it the most useful canonical reference despite the cluster being heterogeneous.", + "best_issue_reason": "44273 is the best representative issue because it captures the widest shared symptom class in this cluster, while the others are narrower, model-specific, or unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading/import mechanics vs a TensorFlow backend install error; different failure modes." + "reason": "Lazy loading/import machinery and a TensorFlow backend error are different failure modes." }, { "left": "issue:44273", "right": "issue:45003", "accept": false, - "reason": "Both touch import behavior, but one is lazy loading and the other is unsafe sys.modules access; not the same bug." + "reason": "They share broad import-related themes, but the evidence does not show the same underlying bug; 45003 should not be treated as the root cause of 44273." }, { "left": "issue:44393", "right": "issue:45753", "accept": false, - "reason": "Both are Qwen3-VL related, but one is 2D box output hallucination/error and the other is an infinite-loop case on a specific image." + "reason": "Different Qwen3-VL symptoms: 2D bbox output errors vs an infinite loop on a specific image." }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "Generic Qwen3-VL-Moe bug vs missing tensor-parallel plan for Qwen3_5Moe; different code paths." + "reason": "Different model families and different code paths: Qwen3-VL-Moe support vs Qwen3.5 MoE tensor-parallel metadata." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Model output bug vs global PretrainedConfig type-checking regression; unrelated." + "reason": "A multimodal Qwen error is not the same as a PretrainedConfig type-checking regression." }, { "left": "issue:44230", "right": "issue:46095", "accept": false, - "reason": "FP8 inference support request vs weight-conversion script question; not the same change or defect." + "reason": "FP8 support for Qwen3-VL/Qwen3.5 MoE is unrelated to a DeepSeek-v4 weight conversion-script request." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "Both are v5 upgrade regressions, but one is the env command and the other is checkpoint reload compatibility; too broad to be duplicates." + "reason": "Both are version-related, but they affect different paths: env command behavior vs checkpoint reload." }, { "left": "issue:44373", "right": "issue:44961", "accept": false, - "reason": "Docstring issue vs a content-free placeholder issue." + "reason": "A docstring issue about position_ids is unrelated to the placeholder bug report." }, { "left": "issue:41669", "right": "issue:43673", "accept": false, - "reason": "Import-star/lazy-loading performance regression vs missing GenerationMixin cache in chunked prefill; different problems." + "reason": "Import-star cleanup/performance and missing GenerationMixin cache during chunked prefill are different bugs." }, { "left": "issue:43525", "right": "issue:45850", "accept": false, - "reason": "Specific missing pad_token_id error vs a vague umbrella complaint about v5 breaking many models." + "reason": "A specific missing pad_token_id attribute is not the same as a broad 'many models break' report." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both concern loading older checkpoints in v5, but the concrete failure is different: generic reload compatibility vs InternVL2 remote-code/meta-init incompatibility." + "reason": "Both concern v5 compatibility, but one is checkpoint reload and the other is remote-code meta initialization." }, { "left": "issue:45362", "right": "issue:45850", "accept": false, - "reason": "Specific chat crash in one model vs broad v5 complaint; not the same underlying bug." + "reason": "A concrete Qwen3.5-35B chat crash is not the same as a generic broad breakage complaint." }, { "left": "issue:43572", "right": "issue:45850", "accept": false, - "reason": "Missing StableLmConfig field vs generic v5 breakage report; unrelated specifics." + "reason": "A StableLmConfig pad_token_id/pad_token_idx regression is a specific model-config bug, not the same as a broad umbrella issue." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "Integration-test seed flakiness vs inverse_sqrt scheduler kwargs handling; different subsystems." + "reason": "A flaky test missing set_seed() is unrelated to inverse_sqrt scheduler kwargs handling." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "AutoProcessor kwargs propagation bug vs scheduler kwargs bug; not the same code path." + "reason": "Passing kwargs to cached_file and scheduler timescale handling are separate code paths." }, { "left": "issue:44246", "right": "issue:44273", "accept": false, - "reason": "Import performance complaint vs lazy-loading malfunction; related theme, but not a concrete duplicate." + "reason": "Both involve import/lazy-loading themes, but the evidence only supports a loose similarity, not the same underlying bug." }, { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "Model-type warning vs Liger Kernel application with model_init; unrelated." + "reason": "sam3_tracker model-type warnings and Liger Kernel application under model_init are unrelated." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python-version compatibility issue vs a Qwen3-VL output bug; no overlap." + "reason": "Python version compatibility and Qwen3-VL bounding-box output are unrelated." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Flash Attention 2 support gap vs multimodal bounding-box error; different defects." + "reason": "Flash Attention 2.0 support gating for NemotronH is unrelated to a Qwen3-VL output error." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "Env command breakage vs scheduler-kwargs bug; both v5 issues, but not the same bug." + "reason": "Environment command breakage and inverse_sqrt scheduler kwargs are different regressions." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "Tokenizer save-pretrained class mismatch vs an unspecified bug report placeholder." + "reason": "Tokenizer save_pretrained metadata mismatch is unrelated to a generic bug report." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "position_ids buffer access error vs placeholder report; unrelated." + "reason": "Reading position_ids after buffer registration is unrelated to the generic placeholder issue." }, { "left": "issue:44908", "right": "issue:45071", "accept": false, - "reason": "Scheduler kwargs handling vs PretrainedConfig type-checking; different areas." + "reason": "Scheduler kwargs handling is not the same as PretrainedConfig type checking." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star/lazy-loading cost issue vs missing HybridCache import; similar import surface, but not the same bug." + "reason": "Import-star removal and a HybridCache import failure are different bugs." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Import performance refactor vs a small testing_utils bug; unrelated." + "reason": "Import-star cleanup and a bug in testing_utils.py are unrelated." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Chunked-prefill cache bug vs cache-strategies typo; one is functional, the other cosmetic." + "reason": "Missing generation cache and a cache-strategy typo are not the same bug." }, { "left": "issue:45278", "right": "issue:45850", "accept": false, - "reason": "Both are broad v5 breakage complaints, but they do not identify the same concrete failure mode." + "reason": "A concrete upgrade-to-5.5.0 import-error cluster is not the same as a vague broad breakage report." }, { "left": "issue:43874", "right": "issue:45850", "accept": false, - "reason": "Missing multimodal token helper in one processor vs generic v5 breakage; not the same issue." + "reason": "Missing get_number_of_image_patches in GLM46V is a specific multimodal API bug, not a general version-breakage complaint." }, { "left": "issue:44393", "right": "issue:45478", "accept": false, - "reason": "Qwen3-VL 2D box issue vs Qwen3.5-Moe from_pretrained error; different model families and code paths." + "reason": "Qwen3-VL hallucination/error in bbox output is different from Qwen3.5 MoE from_pretrained failures." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Optional post_init validation vs tied-weights-key list crash; unrelated implementation details." + "reason": "An optional post_init policy issue is unrelated to _get_tied_weight_keys crashing on a list." }, { "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Docs still referencing pipeline() vs inverse_sqrt scheduler kwargs; unrelated." + "reason": "Docs still referencing pipeline() after removals is unrelated to scheduler kwargs handling." }, { "left": "issue:41669", "right": "issue:44492", "accept": false, - "reason": "Import-star removal/performance concern vs cache-strategy typo; not the same bug." + "reason": "Import-star cleanup and a cache-strategy typo are unrelated." }, { "left": "issue:41669", "right": "issue:44908", "accept": false, - "reason": "Import/lazy-loading performance vs scheduler parameter plumbing; different subsystems." + "reason": "Import-star cleanup and scheduler kwargs are unrelated." }, { "left": "issue:31795", "right": "issue:33453", "accept": false, - "reason": "Doc confusion around model.forward arguments vs tokenizer loading regression; unrelated." + "reason": "Doc confusion about model.forward inputs is unrelated to tokenizer loading regression." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Liger Kernel not applied with model_init vs flash_attention_3 training degeneration; both training-related, but not the same bug." + "reason": "Liger Kernel not applied with model_init and flash_attention_3 degenerate training are different issues." }, { "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "TF32 API-settings warning vs cache-strategy typo; unrelated." + "reason": "TF32 behavior settings and cache-strategy typo are unrelated." }, { "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Broken env command vs lazy loading not working; both v5-related, but not the same concrete defect." + "reason": "Broken transformers env command and lazy loading malfunction are different user-facing regressions." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer ReLU discussion vs Gemma4 audio positional encoding bug; unrelated." + "reason": "A DSA indexer ReLU omission is unrelated to Gemma-4 audio positional encoding." }, { "left": "issue:44655", "right": "issue:44861", "accept": false, - "reason": "Pipeline save_pretrained issue vs tied-weight key crash; different serialization paths." + "reason": "Saving Pipeline objects and _get_tied_weight_keys crashing on list are different save-path issues." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "The summary is mostly cautious, but it overstates one soft edge: issue:44273 vs issue:45003. The packet only supports a loose thematic similarity (lazy loading / import internals) and the pair is explicitly marked deterministic_accept=false; calling 45003 a likely root cause of 44273 is stronger than the evidence. Since soft-edge verdicts need to stay conservative, this should be rejected." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -15351,12 +15357,12 @@ "issue:31515|issue:33453", "issue:43298|issue:44230" ], - "prepared_review_unit_hash": "83973ecc774fbaf59bcc0f9e9af39b1c9a53966f33bc92fb54fbcc6901efe392", + "prepared_review_unit_hash": "d618948ae678f21d767b314b3cce5605c28e3bd3403ca9549537e4a145205f54", "component_budget": { "node_count": 410, "item_count": 410, "soft_pair_count": 1534, - "serialized_chars": 791437, + "serialized_chars": 791439, "estimated_input_tokens": 197860, "estimated_eval_tokens": 395976 }, @@ -15373,228 +15379,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of Transformers v5/v5.x regression reports and model-specific bugs, but they do not appear to be duplicates of one another. No soft-edge pair looks safe to merge as the same underlying issue.", - "confidence": 0.84, - "canonical_issue_reason": "issue:45850 is the broadest representative of the cluster\u2019s theme: widespread breakage after the v5 transition. It best captures the general regression umbrella, even though it is vague.", + "summary": "This cluster is highly heterogeneous, but many items orbit Transformers v5 regression/compatibility complaints, especially tokenizer/model-loading breakages. No PRs are present. I chose the broadest regression report as the best representative, though most pairs are not true duplicates.", + "confidence": 0.46, + "canonical_issue_reason": "issue:45850 is the broadest umbrella report in the set ('Since 5.0 version it breaks too many models') and best matches the overall v5-regression theme, even though the cluster contains many unrelated specifics.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45850 is the best overall issue anchor because it generalizes the repeated \u201cv5 breaks models/loading\u201d pattern seen across the set and has the most umbrella-like scope.", + "best_issue_reason": "issue:45850 is the most suitable cluster representative because it is broad, regression-oriented, and overlaps thematically with several v5-related loading/tokenizer complaints.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "Both are model/processor-related failures, but one is a chat-template crash with tool calls and the other is a Qwen3.5 MoE from_pretrained error. Different code paths and symptoms." + "reason": "One is a generic placeholder bug report; the other is a specific Qwen3.5 MoE from_pretrained failure. Not enough evidence they are the same bug." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy loading failure is unrelated to inverse_sqrt scheduler kwargs being ignored." + "reason": "Lazy loading failure and inverse_sqrt scheduler kwargs handling are unrelated code paths." }, { "left": "issue:44492", "right": "issue:45850", "accept": false, - "reason": "A typo in cache strategy naming is not the same bug as broad v5 model breakage." + "reason": "A cache-strategy typo is not the same underlying problem as broad v5 model breakage." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are distinct model-specific bugs." + "reason": "Different model components and symptoms: Qwen3VL timestamp math vs GLM-5 RoPE implementation." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Timestamp calculation in a processor is unrelated to scheduler keyword handling." + "reason": "Timestamp calculation and scheduler kwargs handling are unrelated." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import-star cleanup/performance concerns are unrelated to a VRAM leak in dataloader threads." + "reason": "Import-star cleanup/performance issue is not the same as VRAM leakage in multithreaded inference." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Documentation confusion about input_ids/past_key_values is not the same as a model-loading regression." + "reason": "Documentation confusion about forward args is unrelated to a concrete model-loading regression." }, { "left": "issue:31795", "right": "issue:37428", "accept": false, - "reason": "Doc ambiguity and flash-attention import failure are different issues." + "reason": "Doc clarification and an ImportError in flash-attention utils are different issues." }, { "left": "issue:44315", "right": "issue:45092", "accept": false, - "reason": "Liger Kernel not applied with model_init and remote-code checkpoint/meta-init incompatibility are separate problems." + "reason": "Both mention initialization, but one is about Liger kernel application and the other about remote-code/meta-initialization incompatibility; different fixes." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "GLM-5 RoPE implementation and Gemma-4 audio positional encoding are different model-specific implementation bugs." + "reason": "Different model families and different implementation bugs." }, { "left": "issue:43704", "right": "issue:44485", "accept": false, - "reason": "VRAM leak under multithreaded dataloading is unrelated to RoPE implementation." + "reason": "VRAM leakage in threaded use and RoPE implementation are unrelated." }, { "left": "issue:43298", "right": "issue:46095", "accept": false, - "reason": "Audio-model continuous batching support and a DeepSeek weight-conversion question are not duplicates." + "reason": "Missing audio support in continuous batching and a weight-conversion-script question are not the same change." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "local_files_only still triggering API calls is unrelated to Qwen3VL timestamp computation." + "reason": "Local-files-only network leakage and Qwen3VL timestamp calculation are unrelated." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 settings guidance is unrelated to a VRAM leak bug." + "reason": "TF32 API guidance and VRAM leakage are different concerns." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Processor timestamp math and missing files from save_pretrained are different failure modes." + "reason": "Processor timestamp math and missing save_pretrained files are different processor bugs." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Timestamp calculation in a processor is unrelated to multithreaded VRAM leakage." + "reason": "Timestamp calculation and multithreaded VRAM leakage do not share the same code path." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split return values in a DSA indexer and Gemma-4 audio positional encoding are unrelated." + "reason": "DSA indexer torch.split behavior is unrelated to Gemma4 audio positional encoding." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch for bark-small and Gemma-4 audio positional encoding are different bugs." + "reason": "CPU offload device mismatch is a different failure mode from audio positional encoding." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "Missing rms_norm_eps in MLA q_a_layernorm is unrelated to Gemma-4 audio positional encoding." + "reason": "Missing rms_norm_eps in MLA q_a_layernorm is not the same underlying bug as Gemma4 audio encoding." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Both mention tokenizer behavior in v5, but one is a general behavior difference and the other is a specific wrong decoder type issue; not enough evidence they are the same bug." + "reason": "Both are tokenizer-related v5 complaints, but one is a broad behavior difference and the other is a decoder-type bug; too little evidence they share one concrete fix." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs handling and an unrelated unemployment placeholder issue are not duplicates." + "reason": "Scheduler kwargs handling and an unrelated non-technical issue are not duplicates." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "_get_tied_weight_keys AttributeError and apply_chat_template tool-call crash affect different code paths." + "reason": "Tied-weights key handling and chat-template crashes are different utility bugs." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Missing pad_token_id in Llama4Config is not the same as _get_tied_weight_keys crashing on a list." + "reason": "Missing pad_token_id in Llama4Config is unrelated to tied-weights key handling." }, { "left": "issue:43704", "right": "issue:44998", "accept": false, - "reason": "VRAM leak and an unrelated placeholder issue are not duplicates." + "reason": "VRAM leakage and an unrelated non-technical issue are not duplicates." }, { "left": "issue:41669", "right": "issue:44998", "accept": false, - "reason": "Import-star cleanup/perf work is unrelated to the placeholder issue." + "reason": "Import-star cleanup/performance and an unrelated non-technical issue are not duplicates." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig and tied-weight key extraction crash are unrelated." + "reason": "StableLmConfig pad_token_idx regression and tied-weights key crash are unrelated." }, { "left": "issue:43066", "right": "issue:44998", "accept": false, - "reason": "Tokenizer decoder-type regression and an unrelated placeholder issue are not duplicates." + "reason": "Tokenizer decoder-type regression and an unrelated non-technical issue are not duplicates." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "Slow from_pretrained checkpoint loading and an ImportError for layer_type_validation are different problems." + "reason": "Slow checkpoint loading and an ImportError from configuration_utils are different bugs." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "fp8 qwen3-vl/qwen3.5 support and processor.save_pretrained missing files are unrelated." + "reason": "FP8 support for Qwen3-VL/Qwen3.5 MoE is unrelated to processor.save_pretrained file omission." }, { "left": "issue:44360", "right": "issue:44998", "accept": false, - "reason": "DSA indexer ReLU discussion and the placeholder issue are not duplicates." + "reason": "Missing ReLU in DSA indexer and an unrelated non-technical issue are not duplicates." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLflow job-status reporting after failed training and slow checkpoint loading are unrelated." + "reason": "MLFlow reporting after failed training and slow from_pretrained checkpoint loading are unrelated." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "Slow from_pretrained checkpoint loading and a tensor-parallel training bug are different issues." + "reason": "Slow checkpoint loading and tensor-parallel training bug are different code paths." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Checkpoint loading slowness and documentation confusion in model.forward do not describe the same bug." + "reason": "From_pretrained performance regression and documentation confusion are not the same issue." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "FP8 support for qwen3-vl/qwen3.5 and a from_pretrained error on Qwen3.5 MoE are related only by model family, not the same concrete bug." - }, - { - "left": "issue:31515", - "right": "issue:33453", - "accept": false, - "reason": "Slow checkpoint loading and a tokenizer-loading regression are both loading-related but not the same failure." + "reason": "FP8 inference support for Qwen3-VL/Qwen3.5 MoE and a from_pretrained load error are related to the same model family but not the same concrete bug." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching for audio models and fp8 qwen3-vl/qwen3.5 support are different feature gaps." + "reason": "Continuous batching lacking audio support and Qwen3-VL/Qwen3.5 MoE FP8 support are distinct feature/bug reports." } ] }, @@ -15613,8 +15613,8 @@ "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-13244-3", "nodes": [ @@ -15684,8 +15684,8 @@ "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-26350-2", "nodes": [ @@ -15747,8 +15747,8 @@ "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-27088-5", "nodes": [ @@ -15879,8 +15879,8 @@ "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-36817-15", "nodes": [ @@ -16217,7 +16217,7 @@ "issue:43668|issue:44031", "issue:43668|issue:43813" ], - "prepared_review_unit_hash": "26bcf6253ddff442c315110dfd876b5b5b15ad9df724197f56aa4491b341c67b", + "prepared_review_unit_hash": "1bcb1b754d26fdcb73a611cdd038887ff763a7d5020b32f904653f31559e4d42", "component_budget": { "node_count": 7, "item_count": 7, @@ -16239,66 +16239,66 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These issues are only superficially similar (mostly model/config/tokenizer bugs) and do not describe the same underlying defect or change. I would not collapse any of the suggested pairs.", - "confidence": 0.96, - "canonical_issue_reason": "No clear canonical issue: the cluster spans unrelated topics (norm implementation, loss kwargs, config type hints, typo fixes, tokenizer warning behavior, and a model gating bug).", + "summary": "This set does not look like a true duplicate cluster: the issues span unrelated models, configs, tokenizer warnings, and a typo. I\u2019d anchor on #42936 only as the most substantive issue, not because it duplicates the others.", + "confidence": 0.89, + "canonical_issue_reason": "#42936 is the most developed and discussion-heavy report, with a concrete, reproducible bug description; the other issues are unrelated enough that no real canonical duplicate emerges.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43240 is the most concrete and broadly actionable runtime bug report: a specific code path where kwargs are not forwarded into loss computation.", + "best_issue_reason": "#42936 is the strongest standalone issue in the group: it is concrete, actionable, and has the most discussion. It is still not actually duplicate-similar to the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "Both are model/config questions, but one is about Llama4 normalization semantics and the other is a ModernBERT type hint; different code paths and fixes." + "reason": "Both mention normalization, but one is a model-design question and the other is a config type-hint issue for a different model; not the same bug." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Both are runtime behavior bugs, but one is Mask2Former ignore_value handling and the other is loss kwargs forwarding; unrelated implementations." + "reason": "Different failures: Mask2Former ignore_value usage vs. kwargs not being forwarded to a loss function. Similar loss-related area, but not the same code-path problem." }, { "left": "issue:42936", "right": "issue:46018", "accept": false, - "reason": "Different models and different defects: Mask2Former ignore_value usage vs DeepSeek-V4 expert gating." + "reason": "Mask2Former loss behavior and DeepSeek-V4 expert gating are unrelated model bugs." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "One is a model-specific ignore_value bug; the other is a global tokenizer regex warning regression. Not the same issue." + "reason": "A model-specific ignore_value bug is unrelated to tokenizer regex warning behavior." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "A model loss/label handling bug and a typo in a quantization attribute are unrelated." + "reason": "This is a typo in a quantization attribute name, not the same as a Mask2Former loss/ignore-value bug." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Different components and fix types: Mask2Former runtime bug vs ModernBERT config annotation typo." + "reason": "Different models and different defect types: unused ignore_value vs incorrect type hint." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "One is missing kwargs propagation into loss computation; the other is an incorrect type hint. No shared code defect." + "reason": "Loss-kwargs forwarding bug and ModernBERT config typing issue are unrelated." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "ModernBERT config typing and tokenizer warning regression are distinct changes in different subsystems." + "reason": "A type-hint bug in one config class is not the same as tokenizer regex warnings across models." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "A config type hint issue and a quantization-attribute typo are not the same underlying bug or change." + "reason": "Incorrect type hint and a typo in a quantization attribute are different, unrelated issues." } ] }, @@ -16317,8 +16317,8 @@ "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-41084-2", "nodes": [ @@ -16380,8 +16380,8 @@ "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-43598-2", "nodes": [ @@ -16443,8 +16443,8 @@ "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-43684-2", "nodes": [ @@ -16506,8 +16506,8 @@ "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-43835-4", "nodes": [ @@ -16609,8 +16609,8 @@ "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-44371-5", "nodes": [ @@ -16731,8 +16731,8 @@ "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-45145-2", "nodes": [ @@ -16862,7 +16862,7 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "3061d74a5bfbcf1ac5365aa8b2c10f409dbe7fa1743eebe1554b298cb0bc0b71", + "prepared_review_unit_hash": "2ba7602c785333c840314e61a346d9a26ff5cc50d3f7b6dc56df93d5f1188c1a", "component_budget": { "node_count": 2, "item_count": 2, @@ -16884,18 +16884,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are related to 4-bit inference, but they describe different bugs and different fix paths: Gemma4 uint8/NF4 audio-module TypeError vs. BitsAndBytesConfig skipping lm_head exclusion and triggering an AssertionError.", - "confidence": 0.96, - "canonical_issue_reason": "issue:45674 is the clearer representative of its own bug: it has a specific configuration trigger and a general 4-bit inference failure mode. issue:45672 is Gemma4/audio-module-specific and appears to be a separate problem.", + "summary": "These are two separate 4-bit inference bugs, not duplicates: one is a Gemma4 audio-module uint8/torch.finfo TypeError, the other is a BitsAndBytesConfig skip_modules regression that removes the default lm_head exclusion and triggers an AssertionError.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45674 is the better standalone issue to keep visible because its title and mechanism are more explicit; issue:45672 is a separate Gemma4-specific bug, not a duplicate.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "They share the same broad area (4-bit inference) but not the same underlying bug: one is a torch.finfo TypeError on uint8 weights in Gemma4 audio modules, the other is llm_int8_skip_modules clearing lm_head exclusion and causing an AssertionError. Different trigger, code path, and fix." + "reason": "Different failure modes and code paths: Gemma4 audio modules hit a uint8/torch.finfo TypeError, while BitsAndBytesConfig llm_int8_skip_modules changes lm_head exclusion handling and causes an AssertionError. Shared 4-bit context is too broad to treat as the same bug." } ] }, @@ -16914,8 +16914,8 @@ "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-45762-2", "nodes": [ @@ -16977,8 +16977,8 @@ "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-45764-2", "nodes": [ @@ -17040,8 +17040,8 @@ "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-45869-5", "nodes": [ @@ -17161,7 +17161,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-39895-16", + "cluster_id": "cluster-39895-14", "nodes": [ "pull_request:39895", "pull_request:41855", @@ -17176,9 +17176,7 @@ "pull_request:45213", "pull_request:45655", "pull_request:46048", - "pull_request:46065", - "pull_request:46067", - "pull_request:46091" + "pull_request:46067" ], "soft_pairs": [ "pull_request:39895|pull_request:43424", @@ -17187,251 +17185,189 @@ "pull_request:42668|pull_request:43823", "pull_request:43251|pull_request:46067", "pull_request:43251|pull_request:43823", - "pull_request:43251|pull_request:46065", "pull_request:43251|pull_request:44827", - "pull_request:43424|pull_request:46065", "pull_request:43424|pull_request:46067", - "pull_request:43424|pull_request:46091", "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", "pull_request:43823|pull_request:45655", "pull_request:43823|pull_request:43995", "pull_request:43823|pull_request:46067", - "pull_request:43995|pull_request:44827", "pull_request:43995|pull_request:45213", - "pull_request:44827|pull_request:46065", + "pull_request:43995|pull_request:44827", "pull_request:44827|pull_request:46067", - "pull_request:44827|pull_request:46091", "pull_request:44827|pull_request:45170", - "pull_request:45170|pull_request:46065", + "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:44827|pull_request:45655", - "pull_request:44827|pull_request:46048", - "pull_request:46065|pull_request:46067", - "pull_request:46065|pull_request:46091", - "pull_request:43823|pull_request:46091", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "fbca8782f98a5e01201dda1de68e65c7e7742477078765d95b141ea51ac8bfb8", + "prepared_review_unit_hash": "8d8b9254ca5de51faa645168084c5b0725f57e437c9adb98026097a7e9448f2c", "component_budget": { - "node_count": 16, - "item_count": 16, - "soft_pair_count": 30, - "serialized_chars": 45302, - "estimated_input_tokens": 11326, - "estimated_eval_tokens": 22908 + "node_count": 14, + "item_count": 14, + "soft_pair_count": 22, + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 }, "budget": { - "node_count": 16, - "item_count": 16, - "soft_pair_count": 30, - "serialized_chars": 45302, - "estimated_input_tokens": 11326, - "estimated_eval_tokens": 22908 + "node_count": 14, + "item_count": 14, + "soft_pair_count": 22, + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: new-model scaffolding, tokenizer/docs additions, export tests, conversion/loading fixes, and a single duplicated loss-function change. Only the two `fixed_cross_entropy` PRs appear to be the same underlying fix; the rest are broad-theme similarities, not true duplicates.", - "confidence": 0.96, + "summary": "This cluster is mostly a mix of unrelated PRs; the only true duplicate-style overlap is the pair of fixed_cross_entropy kwargs fixes (43251 and 43254), which target the same loss-function change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the best canonical representative of the duplicated loss fix: it has the explicit issue target, clearer bug framing, and more review/activity than 43254, which is a shorter near-identical variant.", + "canonical_pr_reason": "PR 43251 is the best canonical PR because it is the clearer, more explicit version of the shared cross-entropy fix: it targets issue #43240 and makes the concrete signature/body change in src/transformers/loss/loss_utils.py.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the strongest overall representative of the only true duplicate pair in the cluster and is the clearest concrete code change here.", + "best_pr_reason": "PR 43251 is the strongest representative of the actual duplicate pair: it directly addresses the same cross_entropy kwargs support bug, has the explicit issue link, and is narrowly scoped.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Unrelated PRs: VideoPrism model/docs addition vs Executorch dynamic-shape export test. Similarity is only that both are framework changes." + "reason": "VideoPrism model addition vs executorch dynamic-shape export test; unrelated changes." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Different fixes: Mistral tokenizer helper methods vs adding a new MobileLLM model. Shared tokenization/model codebase is too broad." + "reason": "Mistral tokenizer API addition vs new MobileLLM model implementation; different code paths." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer method addition vs executorch export test are not the same bug or change." + "reason": "Tokenizer method addition vs executorch export test; no shared underlying bug or change." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor auto-loading robustness changes are unrelated to adding a new model implementation." + "reason": "Processor robustness work across multimodal autos vs adding MobileLLM; not the same fix." }, { "left": "pull_request:43251", "right": "pull_request:46067", "accept": false, - "reason": "`cross_entropy` kwargs fix vs conversion loading/base_model_prefix handling are distinct codepaths and bugs." + "reason": "Cross-entropy kwargs support vs base_model_prefix conversion loading fixes; unrelated subsystems." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Loss-function kwargs fix is unrelated to MobileLLM model addition." - }, - { - "left": "pull_request:43251", - "right": "pull_request:46065", - "accept": false, - "reason": "Loss-function fix vs testing-utils device property fix are different issues with no shared concrete change." + "reason": "Loss-function kwargs fix vs adding MobileLLM; completely different artifacts." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Cross-entropy kwargs support is unrelated to Mistral4 test/model plumbing." - }, - { - "left": "pull_request:43424", - "right": "pull_request:46065", - "accept": false, - "reason": "Executorch dynamic-shape test and device-property test-utils fix are different concerns." + "reason": "Cross-entropy API fix vs Mistral4 test/model changes; not the same underlying issue." }, { "left": "pull_request:43424", "right": "pull_request:46067", "accept": false, - "reason": "Executorch export test vs conversion mapping/loading fix are unrelated." + "reason": "Executorch dynamic-shape test vs conversion loading prefix handling; unrelated." }, { "left": "pull_request:43424", - "right": "pull_request:46091", + "right": "pull_request:44827", "accept": false, - "reason": "Executorch export testing and tokenizer mapping updates are different changes." + "reason": "Executorch export test vs Mistral4 fixes; different model/code-path concerns." }, { "left": "pull_request:43424", - "right": "pull_request:44827", + "right": "pull_request:43823", "accept": false, - "reason": "Executorch export test does not match the Mistral4 test/model fix change-set." + "reason": "Executorch export test vs MobileLLM addition; no duplicate relationship." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New MobileLLM model addition is unrelated to the class-annotation/config resolution fix." + "reason": "New MobileLLM model vs config-resolution ordering fix; unrelated changes." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "Adding MobileLLM and refactoring Falcon output collection are different model integrations." + "reason": "MobileLLM model addition vs Falcon output-interface refactor; different model work." }, { "left": "pull_request:43823", "right": "pull_request:46067", "accept": false, - "reason": "MobileLLM model addition vs loading/conversion prefix fix are distinct changes." - }, - { - "left": "pull_request:43995", - "right": "pull_request:44827", - "accept": false, - "reason": "Falcon output-interface refactor is unrelated to Mistral4 test fixes." + "reason": "MobileLLM addition vs conversion-loading prefix fix; not the same bug/change." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon refactor vs model-creation skill PR are not the same underlying change." + "reason": "Falcon output-collection refactor vs model-creation skill/docs; only loosely connected by tooling, not same change." }, { - "left": "pull_request:44827", - "right": "pull_request:46065", + "left": "pull_request:43995", + "right": "pull_request:44827", "accept": false, - "reason": "Mistral4 test/model changes and testing-utils device detection fix are different bugs." + "reason": "Falcon refactor vs Mistral4 tests; unrelated model-specific work." }, { "left": "pull_request:44827", "right": "pull_request:46067", "accept": false, - "reason": "Mistral4 test/model changes do not match the conversion/loading prefix bug fix." - }, - { - "left": "pull_request:44827", - "right": "pull_request:46091", - "accept": false, - "reason": "Mistral4 test/model changes are unrelated to tokenizer mapping updates." + "reason": "Mistral4 test fixes vs conversion-loading prefix handling; different issues." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 fixes and `layrnorm`\u2192`layernorm` renaming are different model families and changes." - }, - { - "left": "pull_request:45170", - "right": "pull_request:46065", - "accept": false, - "reason": "Weight-renaming cleanup for CLIP-like models is unrelated to testing-utils device-property handling." - }, - { - "left": "pull_request:44827", - "right": "pull_request:45213", - "accept": false, - "reason": "Mistral4 test/model fixes are unrelated to the new-model skill PR." - }, - { - "left": "pull_request:43823", - "right": "pull_request:45213", - "accept": false, - "reason": "Model addition PR and a repository skill/documentation PR are not duplicates." - }, - { - "left": "pull_request:44827", - "right": "pull_request:45655", - "accept": false, - "reason": "Mistral4 fixes are unrelated to the `cls.config` resolution order bug." + "reason": "Mistral4 tests vs layrnorm->layernorm rename/conversion mapping; not the same concrete bug." }, { "left": "pull_request:44827", "right": "pull_request:46048", "accept": false, - "reason": "Mistral4 test/model fixes and TDT loss kernel integration are different features." + "reason": "Mistral4 test fixes vs TDT loss kernel integration; unrelated." }, { "left": "pull_request:44827", - "right": "pull_request:46067", + "right": "pull_request:45655", "accept": false, - "reason": "Different areas: Mistral4 test/model work vs conversion mapping/loading logic." + "reason": "Mistral4 tests vs config-resolution ordering fix; different code paths." }, { "left": "pull_request:44827", - "right": "pull_request:46091", - "accept": false, - "reason": "Mistral4 work and tokenizer backend mapping updates are unrelated." - }, - { - "left": "pull_request:46065", - "right": "pull_request:46067", + "right": "pull_request:45213", "accept": false, - "reason": "Testing-utils device detection and conversion-prefix loading fixes are distinct infrastructure bugs." + "reason": "Mistral4 tests vs model-creation skill; shared tooling references are too broad to be a duplicate." }, { - "left": "pull_request:46065", - "right": "pull_request:46091", + "left": "pull_request:43823", + "right": "pull_request:45213", "accept": false, - "reason": "Testing-utils device property fix is unrelated to tokenizer mapping changes." + "reason": "MobileLLM model addition vs model-creation skill; one is a model implementation, the other is a meta workflow/doc PR." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Same underlying bug fix in `fixed_cross_entropy`: both add supported kwargs/parameters to pass through to `nn.functional.cross_entropy` for the same loss-path behavior." + "reason": "Both PRs fix the same fixed_cross_entropy bug by adding supported kwargs to nn.functional.cross_entropy in src/transformers/loss/loss_utils.py; they could plausibly be merged into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The only accepted duplicate pair is the explicitly issue-linked `fixed_cross_entropy` pair, which is well supported by the shared file and matching kwargs passthrough change. The other rejections are appropriately cautious and do not overstate weak thematic similarity." + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted pair, 43251/43254, is well-supported by shared issue target 43240 and the same loss-utils change, and the other soft-edge rejections do not overstate similarity." }, "evaluator_used": true, "retried": false, @@ -17458,7 +17394,7 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "6d4d369d5afe4a6c3aa95fae32d0cff791020261a6a397495179e136b8cda45b", + "prepared_review_unit_hash": "c9b1a8a5b0418d288265bb5cd5adae09dd6e9139727e32d849b582d64af5f92e", "component_budget": { "node_count": 2, "item_count": 2, @@ -17480,18 +17416,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are about the VibeVoice model family, but they are not the same concrete change: #40546 implements the base VibeVoice support, while #42781 is a separate draft for VibeVoice Realtime with additional acoustic-tokenizer and generation changes. Shared files mostly reflect common model integration scaffolding, not a duplicate fix/change.", - "confidence": 0.82, + "summary": "These are related VibeVoice work, but they are not the same change: PR #40546 adds the core VibeVoice model and docs, while PR #42781 is a draft focused on a realtime variant plus acoustic tokenizer support. They overlap in shared package plumbing, but the underlying changes are different enough that they should not be merged as duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "#40546 is the broader, more established base VibeVoice implementation and is the best representative of this cluster; #42781 is a narrower follow-up/realtime variant and remains draft.", + "canonical_pr_reason": "PR #40546 is the better canonical VibeVoice PR because it is the non-draft, broader base implementation of VibeVoice with docs and auto-class wiring, and it has substantial review activity.", "best_issue_reason": null, - "best_pr_reason": "#40546 is the strongest standalone PR in the set because it introduces the core VibeVoice integration with docs, auto classes, modeling, processing, and conversion support. #42781 is related but not a duplicate and appears to extend the family rather than deliver the same change.", + "best_pr_reason": "PR #40546 is the best representative PR for this cluster since it covers the primary VibeVoice model integration and appears more mature than the draft realtime follow-up.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Do not merge as duplicates: #40546 adds the base VibeVoice model, while #42781 targets VibeVoice Realtime and includes extra acoustic-tokenizer/generation work. They overlap by subsystem, but they are distinct features and not the same code-path fix/change." + "reason": "Shared VibeVoice filenames mostly reflect common framework plumbing, but the scope differs: #40546 implements the base VibeVoice model, while #42781 adds a realtime-specific variant and acoustic tokenizer. They are not the same concrete change and would not plausibly be merged into one PR." } ] }, @@ -17521,7 +17457,7 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "f88250501340b4e3b4ce7094e5d7f4f1406dad5759742420ca297cf34075ecb9", + "prepared_review_unit_hash": "0f4db8cdf10389cd74f76d700276d27fca334e305ff9cc1584fe9562c0b5f90a", "component_budget": { "node_count": 2, "item_count": 2, @@ -17543,18 +17479,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are unrelated feature PRs: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. They do not look like the same underlying change.", + "summary": "These two pull requests are unrelated: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. They should not be deduplicated or clustered as the same change.", "confidence": 0.99, "canonical_issue_reason": null, - "canonical_pr_reason": "41224 is the more focused representative PR in this cluster; 40755 is a separate TimesFM feature change.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "41224 is the tighter, more self-contained PR to represent the cluster.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different models and different code paths: TimesFM covariate forecasting vs DINOv3 image-classification support. Not the same underlying change." + "reason": "Different models, different code paths, and different features: TimesFM forecasting with covariates vs DINOv3 image classification. They do not appear to fix the same underlying bug or implement the same change." } ] }, @@ -17649,12 +17585,12 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "29c1762d699376cf85062a61c38ba28acd831a4a3cd9ec059bcc37faca11e8b2", + "prepared_review_unit_hash": "8ad1d4ed5ef5efce201970aeeef7666c7bd3394e7a9f26b8f6ef3ece691c42b2", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10436, + "serialized_chars": 10435, "estimated_input_tokens": 2609, "estimated_eval_tokens": 5474 }, @@ -17662,7 +17598,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10436, + "serialized_chars": 10435, "estimated_input_tokens": 2609, "estimated_eval_tokens": 5474 }, @@ -17671,24 +17607,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three pull requests are separate new-model additions (DeepSeek OCR, Isaac, and Kimi2-6). They share the same Transformers scaffolding patterns, but they target different models and codepaths, so this is not a duplicate cluster.", + "summary": "These are three distinct model-addition PRs that share repository boilerplate but target different models and codepaths, so they should not be clustered as duplicates.", "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No true duplicate exists in this set. If a single representative PR is needed, 45630 is the broadest and most reviewed new-model addition, but it is still distinct from the others.", + "canonical_pr_reason": "No true canonical PR: DeepSeek OCR, Isaac, and Kimi2-6 are separate model additions with different implementation files, tests, and model-specific docs; the overlap is mostly shared registration/scaffolding.", "best_issue_reason": null, - "best_pr_reason": "45630 is the strongest representative for the cluster only in the sense that it is a substantial, well-scoped new model integration; however, it should not be treated as the same change as the other PRs.", + "best_pr_reason": "If one representative had to be picked, PR 45630 is the broadest and most actively reviewed of the three, but it is still not a duplicate of the others.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different models and different implementation targets: DeepSeek OCR vs Isaac. Shared framework files are just standard new-model plumbing, not the same underlying change." + "reason": "Different model additions (DeepSeek OCR vs Isaac). Shared auto/docs files are generic registration boilerplate, not the same underlying change or code-path fix." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "Different model families and different concrete integrations: Isaac vs Kimi2-6. The overlap is broad Transformers boilerplate, not one mergeable duplicate PR." + "reason": "Different model additions (Isaac vs Kimi2-6) with distinct model/config/processing implementations. Similar framework plumbing is insufficient to treat them as one PR." } ] }, @@ -17718,7 +17654,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "f34c4a6ea5ea4532b60237925aa2c443132eeba5ab4c94bc3661198bb12fa28d", + "prepared_review_unit_hash": "8b47e684eee483916e5331f58abc6b83fb36a67bac12ebd2c5a6caa694645f2c", "component_budget": { "node_count": 2, "item_count": 2, @@ -17740,18 +17676,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests are broad GitHub Actions maintenance bumps, but they target different concrete changes: PR 42975 is about Node 24 compatibility across many workflow actions, while PR 42976 is mainly upgrading docker/build-push-action and a few other workflow actions to newer versions. They overlap in the general area, but not the same underlying fix/change set.", - "confidence": 0.79, + "summary": "Two open PRs both modernize GitHub Actions workflows, but they target different actions and different compatibility upgrades, so they should not be treated as duplicates.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 42975 is the better canonical PR because it is the more explicit umbrella change and touches the larger set of workflow action upgrades tied to a clear compatibility goal.", + "canonical_pr_reason": "PR 42975 is the clearer representative because it has a specific, well-scoped goal: updating core workflow actions for Node 24 compatibility across multiple workflows.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the strongest representative of the cluster: it has the broader scope, clearer intent, and better matches the shared workflow-maintenance theme than the narrower docker-action bump in PR 42976.", + "best_pr_reason": "PR 42975 is the better canonical PR for this cluster since its title and diff describe a concrete compatibility-driven workflow update, while 42976 is a separate action-version bump focused on Docker build actions.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Do not merge: both are workflow-action upgrades, but one is a Node 24 compatibility sweep and the other is a docker/build-push-action version bump. They do not fix the same concrete bug and would not naturally collapse into one PR." + "reason": "Reject: both are GitHub Actions maintenance PRs, but they change different action families and do not appear to fix the same concrete problem. 42975 is about Node 24 compatibility for checkout/cache/upload-artifact; 42976 upgrades docker/build-push-action and related workflow versions. They could be separate PRs." } ] }, @@ -17788,70 +17724,70 @@ "pull_request:43448|pull_request:45490", "pull_request:43451|pull_request:45490" ], - "prepared_review_unit_hash": "1b74306bce56a92a327dd5c8711321ee4a8da684e4a1241172cdb8c39d446234", + "prepared_review_unit_hash": "f9f4c75b2f258575158e263e761d69c2f64710f80973951f16c632d07ae30a37", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 14074, - "estimated_input_tokens": 3519, - "estimated_eval_tokens": 7294 + "serialized_chars": 14204, + "estimated_input_tokens": 3551, + "estimated_eval_tokens": 7358 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 14074, - "estimated_input_tokens": 3519, - "estimated_eval_tokens": 7294 + "serialized_chars": 14204, + "estimated_input_tokens": 3551, + "estimated_eval_tokens": 7358 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All four PRs are separate model-addition efforts (ViT NEPA, Molmo, Molmo2, CTSM). They share the same repo-wide scaffolding patterns, but they do not fix the same bug or implement the same concrete change, so none should be merged as duplicates.", - "confidence": 0.96, + "summary": "These are four separate model-addition PRs (ViT NEPA, Molmo, Molmo2, CTSM). They share the usual Transformers scaffolding files, but they target different models, different configs/processors, and different implementation paths, so they are not duplicates of one another.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "Molmo2 is the broadest and most complete representative of the cluster (largest change surface and highest review activity), but it is not a duplicate of the others; the PRs are for different models.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "Molmo2 is the best single representative of this cluster because it is the most substantial model-integration PR and includes the most surrounding infrastructure (docs, auto mappings, processing, tests), making it the clearest exemplar of the pattern.", + "best_pr_reason": "PR 43451 is the broadest and most complete addition in this set (modeling, processing, image/video support, auto-mapping, docs, and many tests), 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": "ViT NEPA and Molmo are different model additions with different files, configs, and tests; shared auto/docs scaffolding is not enough to call them the same change." + "reason": "Different models and codepaths: ViT NEPA vs Molmo. Shared files like docs and auto registration are expected in separate model-addition PRs, not evidence of a duplicate fix." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "ViT NEPA and Molmo2 are distinct model integrations for different architectures and codepaths, not one concrete bugfix or reusable patch." + "reason": "Different models and codepaths: ViT NEPA vs Molmo2. They add separate architectures and support surfaces, so they cannot be merged as one duplicate PR." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "ViT NEPA and CTSM are unrelated model-addition PRs; they only overlap on standard registry/documentation plumbing." + "reason": "Different models and codepaths: ViT NEPA vs CTSM. The overlap is only in generic Transformers plumbing, not the same underlying change." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are separate model families/releases with different model, processor, and test implementations; they are not the same underlying change." + "reason": "Molmo and Molmo2 are related by name but are distinct model families with different files, processors, and tests. This is not the same concrete change." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo and CTSM target different model codepaths and add different APIs; the overlap is just common transformer boilerplate." + "reason": "Molmo vs CTSM are unrelated model additions. Shared auto/model-doc boilerplate is too generic to treat them as duplicates." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are different model integrations with no shared concrete bug or feature implementation beyond generic auto-registration/doc updates." + "reason": "Molmo2 and CTSM implement different models with different support components (including video processing for Molmo2). They do not fix or add the same thing." } ] }, @@ -17873,63 +17809,71 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-43291-3", + "cluster_id": "cluster-43291-4", "nodes": [ "pull_request:43291", "pull_request:43665", - "pull_request:44815" + "pull_request:44815", + "pull_request:46102" ], "soft_pairs": [ "pull_request:43291|pull_request:43665", "pull_request:43291|pull_request:44815", + "pull_request:43665|pull_request:46102", "pull_request:43665|pull_request:44815" ], - "prepared_review_unit_hash": "b1a3d641e8e77b93ba92656d2c5a7e9039c478ba593e582b608b28325fa1a749", + "prepared_review_unit_hash": "e81228c371ff23e30cd8b19a7b9c749bd448d24cca52bfdf6cb6995efebbde13", "component_budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 3, - "serialized_chars": 7093, - "estimated_input_tokens": 1774, - "estimated_eval_tokens": 3804 + "node_count": 4, + "item_count": 4, + "soft_pair_count": 4, + "serialized_chars": 9391, + "estimated_input_tokens": 2348, + "estimated_eval_tokens": 4952 }, "budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 3, - "serialized_chars": 7093, - "estimated_input_tokens": 1774, - "estimated_eval_tokens": 3804 + "node_count": 4, + "item_count": 4, + "soft_pair_count": 4, + "serialized_chars": 9391, + "estimated_input_tokens": 2348, + "estimated_eval_tokens": 4952 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are not duplicates of the same underlying change. One is a Whisper tokenization/test fix, one is a suspicious test-only PR in CLIP/ViT, and one is a broader FP8/dequantization fix across loading, quantization, and Mistral4 support. Shared similarity is only superficial.", - "confidence": 0.95, + "summary": "This cluster is heterogeneous: a Whisper test/tokenizer fix, a test-only placeholder patch for CLIP/ViT, an FP8 dequantization fix, and an FSDP tied-weights cleanup. They do not look like duplicates of one underlying bug or change.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44815 is the most substantial and coherent code change in the set, with real implementation updates across model loading and FP8 quantization; the others are narrower test-focused changes and do not align to the same bug or patch.", + "canonical_pr_reason": "pull_request:44815 is the strongest standalone change: it fixes a concrete FP8 dequantization/code-loading path and includes multiple implementation touchpoints plus tests.", "best_issue_reason": null, - "best_pr_reason": "PR #44815 is the best representative artifact because it contains the clearest non-test code path changes and the broadest implementation context, whereas #43291 is Whisper-specific and #43665 appears unrelated/synthetic test noise.", + "best_pr_reason": "pull_request:44815 is the most substantive and coherent fix in the set; the others are either test-only or address unrelated model/runtime paths.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "These touch different model families and different concerns: Whisper tokenization/timestamp decoding versus CLIP/ViT test edits. They do not appear to fix the same concrete bug or merge into one PR." + "reason": "Both are test-oriented, but they target different models and different changes; one updates Whisper tokenizer behavior, the other injects placeholder asserts into CLIP/ViT tests. Not the same underlying bug." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper decode/test changes are unrelated to FP8 dequantization, core model loading, and Mistral4 support. No shared code path or underlying defect is evident." + "reason": "Whisper decode/timestamp handling is unrelated to FP8 dequantization and model loading. Different subsystems and concrete fixes." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46102", + "accept": false, + "reason": "A dummy test patch for CLIP/ViT is unrelated to FSDP tied-weights cleanup. They do not fix the same code-path problem." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "The CLIP/ViT test-only changes are unrelated to the FP8/dequantization implementation work. They do not describe the same change and should not be merged as duplicates." + "reason": "The CLIP/ViT test-only assertions are not the same change as the FP8 dequant/loading fix; shared test activity is too broad to treat as a duplicate." } ] }, @@ -17951,55 +17895,89 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-43488-2", + "cluster_id": "cluster-43488-5", "nodes": [ "pull_request:43488", - "pull_request:43532" + "pull_request:43532", + "pull_request:45932", + "pull_request:46100", + "pull_request:46114" ], "soft_pairs": [ - "pull_request:43488|pull_request:43532" + "pull_request:43488|pull_request:46114", + "pull_request:43488|pull_request:43532", + "pull_request:43532|pull_request:46114", + "pull_request:45932|pull_request:46100", + "pull_request:46100|pull_request:46114" ], - "prepared_review_unit_hash": "ece7972da44f5ffecfa01cda37b58fb472316cc7a4f04dbc31437dd00f3a01ce", + "prepared_review_unit_hash": "3f540bc0dc3f077f65a36c7178c2a528aa381b020704a20e90971b85d5fecea0", "component_budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 4563, - "estimated_input_tokens": 1141, - "estimated_eval_tokens": 2538 + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 }, "budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 4563, - "estimated_input_tokens": 1141, - "estimated_eval_tokens": 2538 + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster contains two unrelated open PRs. #43488 is a repository-bot/check-format style change around setup/versioning and a ViT test, while #43532 changes CI/workflow and testing utility code. The only overlap is an incidental shared test file, so they should not be treated as duplicates.", - "confidence": 0.96, + "summary": "Most PRs are unrelated test/CI noise in ViT or repo-bot/Grafana experiments. The only true duplicate-looking pair is the two identical ViT test-migration PRs (46100 and 46114).", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43532 is the better canonical representative because it contains the more substantive code-path and workflow changes; #43488 is a narrow bot/check-style PR and is not representative of the cluster as a whole.", + "canonical_pr_reason": "PR 46100 is the best canonical representative because it is the earlier, cleaner instance of the exact ViT test-migration diff that is duplicated by 46114.", "best_issue_reason": null, - "best_pr_reason": "PR #43532 is the strongest standalone PR in the set: it makes a coherent set of workflow/testing updates across several files. #43488 is largely a synthetic repo-bot check and is less suitable as the cluster's representative PR.", + "best_pr_reason": "PR 46100 best represents the duplicate cluster: its diff matches 46114 exactly in the same ViT test file, while the other PRs are clearly different changes.", "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:46114", + "accept": false, + "reason": "Both are repo-test PRs touching ViT tests, but the actual changes are different: 43488 is a setup/version/dependency bot check, while 46114 is a failing-test migration snippet. Not the same underlying change." + }, { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "These PRs target different concerns: #43488 is a setup/version/dependency formatting check with a test tweak, while #43532 modifies CI workflows and testing utilities. The shared test file is incidental and they are not the same underlying change." + "reason": "Shared ViT test file is not enough; 43488 is a packaging/version check, whereas 43532 changes CI workflows and helpers. Different concrete problems." + }, + { + "left": "pull_request:43532", + "right": "pull_request:46114", + "accept": false, + "reason": "43532 is a workflow/CI plumbing change with multiple files, while 46114 is only a ViT test stub change. They do not fix the same bug or implement the same change." + }, + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "45932 is a broad observability/model config PR across many files; 46100 is a small ViT test migration. Same general test context, but not the same underlying change." + }, + { + "left": "pull_request:46100", + "right": "pull_request:46114", + "accept": true, + "reason": "These are effectively the same PR change: identical diff in the same ViT test file, adding the same failing test assertions. They could be merged as one PR." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: 46100 and 46114 have identical ViT test diffs and are a plausible duplicate pair, while the other pairings are reasonably rejected as different changes. The summary stays conservative and does not overstate evidence." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -18022,12 +18000,12 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "5bbd0bf4b60ef6819745b2b3045437c40372b1a8b9feb2bd90d607d6bdaf0adc", + "prepared_review_unit_hash": "bd01a73da5613cd1d14347800a847c0caa0f2ebd0b4005f9ec1b299eebd2a590", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6082, + "serialized_chars": 6081, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -18035,7 +18013,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6082, + "serialized_chars": 6081, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -18044,18 +18022,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs add different segmentation pipelines with overlapping scaffolding, but they implement distinct features and should not be deduplicated.", + "summary": "Two open PRs add separate pipeline features for different promptable segmentation tasks. They share scaffolding but target different APIs, docs, model integrations, and tests, so they are related siblings rather than duplicates.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No canonical PR: 43612 and 43613 implement different pipelines (concept vs visual segmentation) with different codepaths and model coverage.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "43613 is the broader, more integrated PR (more model adapters/tests) and is the better representative of this feature family, but it is still distinct from 43612.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different pipeline features: Promptable Concept Segmentation vs Promptable Visual Segmentation. They share registration/docs boilerplate, but the changed model/test files and task targets are not the same concrete code-path change." + "reason": "Shared pipeline/docs scaffolding is not enough; they expose different pipeline classes and work on different model stacks, so they are not the same underlying change and should not be merged as duplicates." } ] }, @@ -18085,40 +18063,40 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "64665a1bf929f59738ca6af85eea7b433cc49496f3e6124dbba11d36a201bc9f", + "prepared_review_unit_hash": "5ff24214fdb38904c51aa47b76c00348e76928a022bb481f6ad3abeb06fb4288", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4291, - "estimated_input_tokens": 1073, - "estimated_eval_tokens": 2402 + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4291, - "estimated_input_tokens": 1073, - "estimated_eval_tokens": 2402 + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate open PRs in `trainer.py`: one changes gradient-accumulation loss scaling, the other extracts signature-column filtering for customization. They share a file and broad trainer context but not the same underlying bug/change.", + "summary": "These are two unrelated trainer.py pull requests: one changes gradient-accumulation loss scaling behavior, the other extracts signature-column logic for overrideability. They share a file but not the same underlying bug or change.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: PR #43651 and PR #43663 address different trainer code paths and are not duplicates.", + "canonical_pr_reason": "No canonical PR: 43651 addresses loss scaling in training_step, while 43663 adds a separate override point for signature column filtering; they are distinct changes.", "best_issue_reason": null, - "best_pr_reason": "PR #43651 is the more concrete, behavior-changing fix in `training_step` and is the better representative of the cluster if one must be chosen.", + "best_pr_reason": "If a representative PR is needed, 43651 is the more concrete behavior fix, but it should not be treated as a duplicate of 43663.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Different features in the same file: #43651 adjusts gradient-accumulation loss normalization, while #43663 adds an overridable helper for signature-column filtering. They do not fix the same concrete bug and would not plausibly be merged as one PR." + "reason": "Different trainer extension points and different code paths: GA loss scaling vs. signature-column filtering. Sharing trainer.py is too broad to imply duplication." } ] }, @@ -18148,7 +18126,7 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "b3ab2816d0aec26356a0b212d5632f8ce2bb8c1c31fdeb80f584192e84ac211b", + "prepared_review_unit_hash": "405f758a3657c7cdafd1a46c2dc701f7d8437adcd3cc47c67460702516581f77", "component_budget": { "node_count": 2, "item_count": 2, @@ -18170,27 +18148,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs are related in `src/transformers/cli/serve.py` and both concern typing/runtime annotation handling, but the packet does not prove they fix the same concrete bug. #43842 is the narrower, explicitly targeted `TypeAdapter` NameError fix; #43656 is a broader annotations cleanup that may overlap but is not clearly the same issue.", - "confidence": 0.67, + "summary": "Two open PRs touch the same CLI module and both deal with import-time annotation/name-resolution problems, but they fix different concrete points in `serve.py` rather than the same exact bug.", + "confidence": 0.63, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43842 is the more specific and clearly scoped fix, directly targeting the `TypeAdapter` NameError path described in its title and diff.", + "canonical_pr_reason": "PR #43656 is the broader, earlier fix for the CLI NameError/annotation issue and has more discussion activity, so it is the better cluster representative.", "best_issue_reason": null, - "best_pr_reason": "PR #43842 is the best representative because it addresses a concrete, named runtime failure, while #43656 reads as a broader typing cleanup in the same file.", + "best_pr_reason": "PR #43842 is the most explicit and narrowly scoped fix: it directly targets the `TypeAdapter` NameError path and states the pydantic-absent case clearly.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": false, - "reason": "They touch the same file and both involve typing annotations, but the evidence does not show they fix the same underlying bug; #43656 is broader, while #43842 is a specific `TypeAdapter` NameError fix." + "reason": "Both modify `src/transformers/cli/serve.py`, but one fixes broader type-annotation/runtime issues with `__future__ annotations` while the other specifically handles `TypeAdapter` import/type-checking when pydantic is absent; they are related follow-ups, not the same concrete change." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The cluster is only weakly supported by the packet: both PRs touch `src/transformers/cli/serve.py` and are about typing/runtime annotation issues, but the evidence does not show they fix the exact same bug. PR #43656 appears to be a broader `from __future__ import annotations` / typing cleanup, while PR #43842 specifically moves `TypeAdapter` under `TYPE_CHECKING` and quotes a single annotation. The soft-edge verdict overstates the evidence by saying they fix the same underlying `TypeAdapter` NameError and can definitely be merged as one bugfix PR; the safer reading is that they are related but not proven to be the same issue." - }, - "evaluator_used": true, - "retried": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -18338,10 +18313,10 @@ "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:44071|pull_request:44086", "pull_request:44073|pull_request:44101", + "pull_request:44017|pull_request:44071", "pull_request:44059|pull_request:44071", "pull_request:44056|pull_request:44101", "pull_request:44073|pull_request:44161", @@ -18377,58 +18352,58 @@ "pull_request:44044|pull_request:44154", "pull_request:44024|pull_request:44076" ], - "prepared_review_unit_hash": "54e2d775bfe70ab9e82fe6f27859225129062a42bbeeda64f04a943ddc856d78", + "prepared_review_unit_hash": "ccea786088197cbc5b66c738a90be73b9b9e57d2dbfa1f3be50f612ebf494e19", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84507, - "estimated_input_tokens": 21127, - "estimated_eval_tokens": 42510 + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 }, "budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 72, - "serialized_chars": 76938, - "estimated_input_tokens": 19235, - "estimated_eval_tokens": 38726 + "serialized_chars": 76934, + "estimated_input_tokens": 19234, + "estimated_eval_tokens": 38724 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of model-specific refactors for the same standardized output-tracing initiative under issue #43979. The only true duplicate/revision threads are the GPT-Neo pair and the GPT-J/CodeGen pair; the rest are distinct per-model adaptations that share the same theme but not the same concrete code path.", - "confidence": 0.87, - "canonical_issue_reason": "All of these PRs point at the same umbrella tracker (#43979), but no separate issue artifact was included in the provided set, so there is no issue candidate to rank.", - "canonical_pr_reason": "PR 44722 is the best canonical representative of the actual duplicate subthread: it covers the GPT-J/CodeGen output-tracing refactor, overlaps directly with 44066 and 44085, and looks like the broadest/most complete version of that concrete change.", - "best_issue_reason": "No issue node is present in the input, so the shared issue #43979 can only be treated as the common umbrella target, not a ranked artifact.", - "best_pr_reason": "44722 is the strongest global representative in this cluster because it captures a concrete duplicated refactor, spans both GPT-J and CodeGen, and has direct overlap with the other GPT-J PRs rather than just sharing a generic theme.", + "summary": "This cluster is a broad standardized-output-tracing refactor across many model-specific PRs. Most pairs only share the migration pattern and are not the same underlying fix; the only clearly grounded duplicate-style linkage is the GPT-J/CodeGen follow-up around PR 44722.", + "confidence": 0.79, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44722 is the best representative because it is the clearest later GPT-J/CodeGen output-tracing refactor and the strongest grounded follow-up in the cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 44722 is the best single PR to anchor the cluster: it is the most concrete GPT-J/CodeGen standardized-output follow-up and the closest to the mergeable duplicate candidates.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families and different code paths; only the tracking issue overlaps." + "reason": "Same broad tracing theme, but FNet/CVT and GPT-J are different model paths." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Both are output-tracing refactors, but they touch unrelated model implementations." + "reason": "Different model families and code paths; shared refactor style is not enough." }, { "left": "pull_request:44066", "right": "pull_request:44085", - "accept": true, - "reason": "Same GPT-J output-tracing change on the same file; 44085 looks like a narrower revision of 44066." + "accept": false, + "reason": "Both are GPT-J-series refactors, but this looks like separate variant work rather than one mergeable bugfix." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "Different backbones and concrete implementations; same theme is not enough." + "reason": "Different vision backbones and files; same migration pattern only." }, { "left": "pull_request:44072", @@ -18440,247 +18415,223 @@ "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen versus EfficientNet; shared issue target only." + "reason": "GPT-J/CodeGen vs EfficientNet is only thematic overlap." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTa are unrelated model implementations." + "reason": "MobileNetV2 and DeBERTa v2 are unrelated implementations." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen refactor versus MGP-STR refactor; not the same concrete change." + "reason": "GPT-J/CodeGen and MGP-STR are different model code paths." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "Different models and different forward paths; only the standardized-output theme matches." + "reason": "GPT-J/CodeGen and MPT are different architectures." }, { "left": "pull_request:44018", "right": "pull_request:44068", - "accept": true, - "reason": "Same GPT-Neo model file/test set and the same output-tracing refactor; this looks like a revision of the same PR." + "accept": false, + "reason": "GPT-Neo pair is only soft similarity; not clearly the same mergeable bugfix." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen and GPT-Neo are different implementations." + "reason": "GPT-J/CodeGen vs GPT-Neo are separate model implementations." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "Different model families; the similarity is at the pattern level only." + "reason": "GPT-Neo and GPT-J/CodeGen are different code paths." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet output tracing is unrelated to GPT-J/CodeGen output tracing." + "reason": "ResNet and GPT-J/CodeGen are unrelated model paths." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPT-J/CodeGen are different code paths." + "reason": "MGP-STR and GPT-J/CodeGen are unrelated implementations." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPT-J are unrelated implementations." + "reason": "ResNet and GPT-J are different backbones." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo and ResNet are different model stacks." + "reason": "GPT-Neo and ResNet only share the migration theme." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet versus MPT; same infrastructure pattern but not the same change." + "reason": "ResNet and MPT are unrelated model paths." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are separate model families." + "reason": "GPT-Neo and GPT-J/CodeGen are not the same concrete fix." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo versus Flaubert/XLM; no shared code path." + "reason": "GPT-Neo and Flaubert/XLM are different models." }, { "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "MPT and GPT-J/CodeGen are unrelated implementations." + "reason": "MPT and GPT-J/CodeGen are unrelated code paths." }, { "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo and Flaubert are different code paths." + "reason": "GPT-Neo and Flaubert are unrelated implementations." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated models." + "reason": "GPT-Neo and MGP-STR are different model families." }, { "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo and GPT-J are different model families." + "reason": "GPT-Neo and GPT-J are separate model families." }, { "left": "pull_request:44068", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated implementations." + "reason": "GPT-Neo and MGP-STR are unrelated code paths." }, { "left": "pull_request:44071", - "right": "pull_request:44085", + "right": "pull_request:44086", "accept": false, - "reason": "MPT and GPT-J are not the same concrete refactor." + "reason": "MPT and MGP-STR are different implementations." }, { "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT-2 and MGP-STR are unrelated code paths." + "reason": "GPT-2 and MGP-STR are unrelated models." }, { "left": "pull_request:44085", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J and MGP-STR are different models; same issue target is not enough." + "reason": "GPT-J and MGP-STR are unrelated code paths." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT and Flaubert/XLM are different implementations." + "reason": "ViLT and Flaubert/XLM are different model families." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 and GPT-J/CodeGen are separate refactors." + "reason": "GPT-2 and GPT-J/CodeGen are not the same fix." }, { "left": "pull_request:44018", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are different code paths." + "reason": "GPT-Neo and MPT are different models." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer and GPT-J/CodeGen are unrelated model families." + "reason": "Segformer and GPT-J/CodeGen are unrelated code paths." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and LongT5 do not share a concrete implementation." + "reason": "Segformer and LongT5 are different model families." }, { "left": "pull_request:44017", "right": "pull_request:44073", "accept": false, - "reason": "SegFormer and VisualBert are different code paths." + "reason": "Segformer and VisualBert are unrelated implementations." }, { "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are unrelated implementations." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44086", - "accept": false, - "reason": "MPT and MGP-STR are different models." + "reason": "GPT-Neo and MPT are unrelated code paths." }, { "left": "pull_request:43997", "right": "pull_request:44073", "accept": false, - "reason": "RegNet and VisualBert are unrelated code paths." + "reason": "RegNet and VisualBert are different model families." }, { - "left": "pull_request:44073", - "right": "pull_request:44101", - "accept": false, - "reason": "VisualBert and Flaubert/XLM are different model implementations." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44071", - "accept": false, - "reason": "GPT-2 and MPT are unrelated refactors." - }, - { - "left": "pull_request:44056", - "right": "pull_request:44101", + "left": "pull_request:44071", + "right": "pull_request:44076", "accept": false, - "reason": "MPNet and Flaubert/XLM are different code paths." + "reason": "MPT and ImageGPT are unrelated architectures." }, { "left": "pull_request:44073", - "right": "pull_request:44161", + "right": "pull_request:44101", "accept": false, - "reason": "VisualBert and LongT5 are unrelated implementations." + "reason": "VisualBert and Flaubert/XLM are different implementations." }, { - "left": "pull_request:44059", - "right": "pull_request:44068", + "left": "pull_request:43997", + "right": "pull_request:44017", "accept": false, - "reason": "GPT-2 and GPT-Neo are different model families." + "reason": "RegNet and Segformer are different vision stacks." }, { - "left": "pull_request:44059", - "right": "pull_request:44018", + "left": "pull_request:43997", + "right": "pull_request:44056", "accept": false, - "reason": "GPT-2 and GPT-Neo are not the same concrete change." + "reason": "RegNet and MPNet are unrelated model paths." }, { - "left": "pull_request:44101", - "right": "pull_request:44161", + "left": "pull_request:44071", + "right": "pull_request:44129", "accept": false, - "reason": "Flaubert/XLM and LongT5 are unrelated code paths." + "reason": "MPT and SpeechT5 are different architectures." }, { "left": "pull_request:43997", - "right": "pull_request:44017", + "right": "pull_request:44018", "accept": false, - "reason": "RegNet and SegFormer are different implementations." + "reason": "RegNet and GPT-Neo are unrelated implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44056", - "accept": false, - "reason": "RegNet and MPNet are unrelated refactors." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44076", + "right": "pull_request:44161", "accept": false, - "reason": "MPT and ImageGPT are different model stacks." + "reason": "RegNet and LongT5 are different models." }, { "left": "pull_request:44074", @@ -18692,115 +18643,121 @@ "left": "pull_request:44056", "right": "pull_request:44161", "accept": false, - "reason": "MPNet and LongT5 are distinct model implementations." + "reason": "MPNet and LongT5 are different model implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44018", + "right": "pull_request:44071", "accept": false, - "reason": "RegNet and GPT-Neo are unrelated changes." + "reason": "RegNet and MPT are different architectures." }, { - "left": "pull_request:43997", - "right": "pull_request:44161", + "left": "pull_request:44010", + "right": "pull_request:44071", "accept": false, - "reason": "RegNet and LongT5 are unrelated implementations." + "reason": "SqueezeBert and MPT are unrelated models." }, { - "left": "pull_request:44074", - "right": "pull_request:44161", + "left": "pull_request:44010", + "right": "pull_request:44073", "accept": false, - "reason": "TextNet and LongT5 are different code paths." + "reason": "SqueezeBert and VisualBert are different model families." }, { - "left": "pull_request:44010", - "right": "pull_request:44071", + "left": "pull_request:44018", + "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and MPT are unrelated model implementations." + "reason": "GPT-Neo and SpeechT5 are unrelated code paths." }, { - "left": "pull_request:44010", - "right": "pull_request:44073", + "left": "pull_request:43997", + "right": "pull_request:44074", "accept": false, - "reason": "SqueezeBert and VisualBert are different code paths." + "reason": "RegNet and TextNet are different model families." }, { - "left": "pull_request:44018", + "left": "pull_request:43997", "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo and SpeechT5 are unrelated refactors." + "reason": "RegNet and SpeechT5 are unrelated implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are unrelated code paths." }, { "left": "pull_request:44010", "right": "pull_request:44018", "accept": false, - "reason": "SqueezeBert and GPT-Neo are different model families." + "reason": "SqueezeBert and GPT-Neo are unrelated implementations." }, { "left": "pull_request:43996", "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and LongT5 are unrelated implementations." + "reason": "FNet/CVT and LongT5 are different model paths." }, { "left": "pull_request:43996", "right": "pull_request:44017", "accept": false, - "reason": "CVT/FNet and SegFormer are distinct code paths." + "reason": "FNet/CVT and Segformer are unrelated models." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBert and DeBERTa are unrelated model implementations." + "reason": "SqueezeBert and DeBERTa v2 are different model families." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa and TextNet are different code paths." + "reason": "DeBERTa v2 and TextNet are unrelated models." }, { "left": "pull_request:44066", "right": "pull_request:44722", "accept": true, - "reason": "Same GPT-J/CodeGen refactor, overlapping files, and same concrete output-tracing change." + "reason": "Same GPT-J/CodeGen output-tracing refactor; 44722 looks like the grounded later follow-up on the same code path." }, { "left": "pull_request:44085", "right": "pull_request:44722", "accept": true, - "reason": "Same GPT-J output-tracing change; 44722 is the broader revision that subsumes it." + "reason": "Same GPT-J output-tracing change on the same file path; 44722 plausibly supersedes it." }, { "left": "pull_request:44085", "right": "pull_request:44129", "accept": false, - "reason": "GPT-J and SpeechT5 are unrelated implementations." + "reason": "GPT-J and SpeechT5 are unrelated model implementations." }, { "left": "pull_request:43996", "right": "pull_request:44007", "accept": false, - "reason": "CVT/FNet and RegNet/ResNet are separate model stacks." + "reason": "FNet/CVT and RegNet/ResNet are different model families." }, { "left": "pull_request:43996", "right": "pull_request:44129", "accept": false, - "reason": "CVT/FNet and SpeechT5 are unrelated code paths." + "reason": "FNet/CVT and SpeechT5 are unrelated code paths." }, { "left": "pull_request:44044", "right": "pull_request:44129", "accept": false, - "reason": "DeBERTa and SpeechT5 are unrelated changes." + "reason": "DeBERTa v2 and SpeechT5 are different models." }, { "left": "pull_request:44044", "right": "pull_request:44154", "accept": false, - "reason": "DeBERTa and VITS are different implementations." + "reason": "DeBERTa v2 and VITS are unrelated architectures." }, { "left": "pull_request:44024", @@ -18811,12 +18768,12 @@ ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative on the soft edges. The accepted pairs all look like mergeable revisions of the same concrete refactor within the same model/file area, and the rejected pairs are mostly cross-model or cross-path similarities only. I don\u2019t see evidence of overstatement." + "accept": false, + "feedback": "The cluster-level summary is mostly grounded, but the soft-edge verdicts are not conservative enough. In the packet, the GPT-Neo pair (44018, 44068) and the GPT-J pair (44066, 44085) are only supported by soft similarity; they are not deterministic accepts and do not clearly look like the same mergeable bugfix. Marking those pairs as true overstates the evidence. The accepted PR pairs that are clearly grounded are the GPT-J/CodeGen follow-ups involving 44722, but the analyst extends the duplicate call beyond what the packet supports." }, "evaluator_used": true, - "retried": false, - "accepted_nontrivial_soft_edge": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -18847,14 +18804,14 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "32ee689c87f6958580f7a8f4cc41687340b2b4450fa846181c27e490d2de4275", + "prepared_review_unit_hash": "942f15d66d4839bfcc115c8d896aa8d01e93fa1bcbb4a1372bc53a189fed7df4", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84507, - "estimated_input_tokens": 21127, - "estimated_eval_tokens": 42510 + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 }, "budget": { "node_count": 6, @@ -18869,42 +18826,42 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These PRs are all model-specific refactors around standardized output tracing/capture, but they touch different architectures and code paths. They are thematically similar rather than duplicates, so the cluster should not be merged across models.", - "confidence": 0.93, + "summary": "These PRs are all model-specific migrations to the new standardized output tracing/capture system, not duplicates of one another. They share a refactor theme, but each touches a different code path and model implementation.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44129 is the strongest representative because it has an explicit issue target (43979), a larger, more concrete SpeechT5 change, and directly frames the standardized output-capture refactor in the model that motivated it.", + "canonical_pr_reason": "PR #44056 is a clear exemplar of the shared migration pattern: it adds the capture_outputs plumbing and explicit _can_record_outputs mapping in a self-contained model implementation.", "best_issue_reason": null, - "best_pr_reason": "PR 44129 is the best global representative: it is the only one tied to an issue, appears to be a concrete end-to-end refactor in one model, and is more specific than the broader no-issue model migrations.", + "best_pr_reason": "PR #44056 is the strongest representative of the cluster because it most directly expresses the standardized output-tracing refactor, but it still targets MPNet specifically rather than duplicating the other PRs.", "soft_edge_verdicts": [ { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Same refactor theme, but different models (VisualBert vs TextNet) and different files/code paths; they are not the same underlying change." + "reason": "Both are output-tracing refactors, but they apply to different models (VisualBert vs TextNet) and different code paths. Same migration theme is not enough to treat them as one PR." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Both migrate output tracing, but MPNet and VisualBert are separate model implementations with distinct forward/capture logic, so this is not one duplicate PR." + "reason": "These are separate model migrations (MPNet vs VisualBert). They follow the same refactor pattern, but they are not the same underlying change and would not be merged as one PR." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "Different model code paths and model-specific output tracing implementations; too broad to treat as the same change." + "reason": "Different model implementations with model-specific output-capture wiring. Shared decorator usage does not make them duplicates." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model architectures; shared output-tracing refactor wording does not make them the same PR-worthy change." + "reason": "RegNet and SqueezeBert are distinct model code paths. Both migrate toward standardized outputs, but they fix different implementations and are not mergeable as a single concrete change." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and SpeechT5 are separate model refactors; SpeechT5 also has an explicit issue target, so these are not duplicates of the same concrete fix." + "reason": "SqueezeBert and SpeechT5 are unrelated model internals. Despite similar output-capture refactoring, they are separate changes with different affected layers and behavior." } ] }, @@ -18926,7 +18883,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-43998-13", + "cluster_id": "cluster-43998-14", "nodes": [ "pull_request:43998", "pull_request:43999", @@ -18940,7 +18897,8 @@ "pull_request:44027", "pull_request:44028", "pull_request:44029", - "pull_request:44030" + "pull_request:44030", + "pull_request:46106" ], "soft_pairs": [ "pull_request:43998|pull_request:44002", @@ -18982,276 +18940,311 @@ "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:44029|pull_request:44030", + "pull_request:44027|pull_request:46106", + "pull_request:44026|pull_request:46106", + "pull_request:44001|pull_request:46106", + "pull_request:44002|pull_request:46106", + "pull_request:43998|pull_request:46106" ], - "prepared_review_unit_hash": "ffcf7c6ab1b32f10a017b695a86d23c4b867fee860b7d0353928c87fc426ed5c", + "prepared_review_unit_hash": "30bb104ebea4a9a53c6f2cd55caa1684c607d23e165798c0bd9092691976fd72", "component_budget": { - "node_count": 13, - "item_count": 13, - "soft_pair_count": 40, - "serialized_chars": 40179, - "estimated_input_tokens": 10045, - "estimated_eval_tokens": 20346 + "node_count": 14, + "item_count": 14, + "soft_pair_count": 45, + "serialized_chars": 43774, + "estimated_input_tokens": 10944, + "estimated_eval_tokens": 22144 }, "budget": { - "node_count": 13, - "item_count": 13, - "soft_pair_count": 40, - "serialized_chars": 40179, - "estimated_input_tokens": 10045, - "estimated_eval_tokens": 20346 + "node_count": 14, + "item_count": 14, + "soft_pair_count": 45, + "serialized_chars": 43774, + "estimated_input_tokens": 10944, + "estimated_eval_tokens": 22144 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs share a common theme\u2014refactoring output tracing / tuple-return handling\u2014but they target different model families and code paths. They are related in style, not duplicates, so I would not merge any soft edge.", - "confidence": 0.91, + "summary": "These are mostly independent model-specific output-tracing refactors plus one unrelated docs toctree fix; they share a theme but not the same underlying bug or change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: the cluster contains independent model-specific output-tracing refactors across unrelated code paths, so none can safely absorb the others.", + "canonical_pr_reason": "No single canonical PR: each PR targets a different model/code path, so the cluster is not one duplicate change.", "best_issue_reason": null, - "best_pr_reason": "If one PR must represent the set, 44025 is the closest because it spans two sibling Depth Anything files and cleanly exemplifies the pattern, but it is still not a duplicate target for the rest.", + "best_pr_reason": "No PR is a true cluster representative; if forced, the broadest code change is 44003, but it still isn't the same fix as the others.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Different model implementations (timm_backbone vs upernet); same refactor theme, not the same bug/change." + "reason": "Different models (`timm_backbone` vs `upernet`); same refactor theme, not the same code path." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Different model code paths (timm_backbone vs vision_text_dual_encoder); only the output-tracing refactor pattern is shared." + "reason": "Separate model implementations (`timm_backbone` vs `vision_text_dual_encoder`)." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Unrelated models (timm_backbone vs univnet); not plausibly one merged fix." + "reason": "Different model families (`timm_backbone` vs `univnet`); not one merged fix." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Different architectures and files; both touch tracing helpers, but not the same underlying change." + "reason": "`timm_backbone` and `mamba`/`falcon_mamba` are unrelated model code paths." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "timm_backbone and rwkv are distinct model paths; this is thematic similarity only." + "reason": "Different architectures (`timm_backbone` vs `rwkv`); only the refactor pattern is shared." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "MobileNetV1 and UnivNet are unrelated model families; same refactor style, different changes." + "reason": "`mobilenet_v1` and `univnet` are unrelated models." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different model wrappers and code paths; not the same concrete fix." + "reason": "Separate model-specific changes (`mobilenet_v1` vs `upernet`)." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "MobileNetV1 vs Depth Anything are separate model implementations; not a duplicate PR." + "reason": "`mobilenet_v1` vs `depth_anything` are different code paths and tasks." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "Different model families; output-tracing cleanup alone is too broad to merge as one PR." + "reason": "Different model wrappers (`mobilenet_v1` vs `vision_encoder_decoder`)." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "MobileNetV1 and SpeechEncoderDecoder are unrelated code paths with only superficial similarity." + "reason": "Different model wrappers (`mobilenet_v1` vs `speech_encoder_decoder`)." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "Vision-text dual encoder and UnivNet are distinct models; no shared bug or change." + "reason": "Different models (`vision_text_dual_encoder` vs `univnet`); not the same bug." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Different model classes and fixes; same refactor theme only." + "reason": "Separate model code paths (`vision_text_dual_encoder` vs `upernet`)." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Separate model code paths (vision-text dual encoder vs mamba); not the same underlying issue." + "reason": "Different model families (`vision_text_dual_encoder` vs `mamba`)." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Different model families; the similar tracing refactor does not make these duplicates." + "reason": "`vision_text_dual_encoder` and `depth_anything` are distinct models with different fixes." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "UnivNet and UperNet are unrelated; they should remain separate PRs." + "reason": "`univnet` and `upernet` are unrelated model implementations." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "Different architectures and files; not the same concrete change." + "reason": "Different code paths (`univnet` vs `mamba`/`falcon_mamba`)." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "UnivNet vs Depth Anything are separate model-specific refactors." + "reason": "Separate model-specific refactors (`univnet` vs `depth_anything`)." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "UperNet and RWKV are unrelated model paths; only the refactor motif matches." + "reason": "Different model families (`upernet` vs `rwkv`)." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "Different subsystems and model implementations; not mergeable as one duplicate fix." + "reason": "`upernet` and `mamba` are different implementations; only the output-tracing theme matches." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "Falcon/Mamba output-tracing work and SuperPoint tracing work affect different code paths." + "reason": "Different models (`mamba` vs `superpoint`) and different forward logic." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "Mamba/Falcon-Mamba vs CodeGen are unrelated models; not the same bug/change." + "reason": "`mamba` and `codegen` touch unrelated code paths despite both refactoring outputs." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "timm_backbone and CodeGen are different model implementations; same broad refactor theme only." + "reason": "Different models (`timm_backbone` vs `codegen`)." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "UnivNet and CodeGen do not share a concrete code-path fix." + "reason": "Separate model implementations (`univnet` vs `codegen`)." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "UperNet and CodeGen are unrelated; not a single duplicate PR." + "reason": "`upernet` and `codegen` are unrelated model code paths." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "Different models and code paths; not plausibly one merged PR." + "reason": "Different models (`vision_text_dual_encoder` vs `codegen`)." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Depth Anything vs VisionEncoderDecoder are separate model families; similar refactor pattern is not enough." + "reason": "`depth_anything`/`prompt_depth_anything` vs `vision_encoder_decoder` are different model wrappers." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Depth Anything and SpeechEncoderDecoder are distinct code paths with different output semantics." + "reason": "Different model families (`depth_anything` vs `speech_encoder_decoder`)." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "Depth Anything and SuperPoint are unrelated model implementations." + "reason": "`depth_anything` and `superpoint` are separate vision models." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "Depth Anything and RWKV are different architectures; not the same change." + "reason": "Different model families (`depth_anything` vs `rwkv`)." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Depth Anything and DPR touch different model stacks; same refactor motif only." + "reason": "`depth_anything` and `dpr` are unrelated model code paths." }, { "left": "pull_request:44026", - "right": "pull_request:44027", + "right": "pull_request:44029", "accept": false, - "reason": "VisionEncoderDecoder and SpeechEncoderDecoder are sister wrappers, but these are still separate forward-path changes." + "reason": "`vision_encoder_decoder` and `rwkv` are unrelated models." }, { "left": "pull_request:44026", - "right": "pull_request:44028", + "right": "pull_request:44027", "accept": false, - "reason": "Different model families and output paths; not the same underlying fix." + "reason": "Different encoder-decoder wrappers (`vision_encoder_decoder` vs `speech_encoder_decoder`)." }, { "left": "pull_request:44026", - "right": "pull_request:44029", + "right": "pull_request:44028", "accept": false, - "reason": "VisionEncoderDecoder and RWKV are unrelated code paths." + "reason": "Different models (`vision_encoder_decoder` vs `superpoint`)." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "VisionEncoderDecoder and DPR are separate model implementations; not duplicate PRs." + "reason": "`vision_encoder_decoder` and `dpr` target different model code." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "SpeechEncoderDecoder and SuperPoint are unrelated model code paths." + "reason": "Different models (`speech_encoder_decoder` vs `superpoint`)." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "SpeechEncoderDecoder and RWKV are distinct architectures; not the same concrete change." + "reason": "Different model families (`speech_encoder_decoder` vs `rwkv`)." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "SpeechEncoderDecoder and DPR are separate model wrappers with different fixes." + "reason": "`speech_encoder_decoder` and `dpr` are separate model implementations." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "SuperPoint and RWKV are unrelated; only the output-tracing refactor theme overlaps." + "reason": "Different models (`superpoint` vs `rwkv`)." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "SuperPoint and DPR are different model families and code paths." + "reason": "`superpoint` and `dpr` are unrelated code paths." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "RWKV and DPR are unrelated model implementations; not one duplicate change." + "reason": "Different models (`rwkv` vs `dpr`) and different forward semantics." + }, + { + "left": "pull_request:44027", + "right": "pull_request:46106", + "accept": false, + "reason": "`speech_encoder_decoder` code change vs docs-only toctree edit; not the same fix." + }, + { + "left": "pull_request:44026", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs navigation change is unrelated to `vision_encoder_decoder` output tracing." + }, + { + "left": "pull_request:44001", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs-only toctree fix is unrelated to `univnet`." + }, + { + "left": "pull_request:44002", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs-only toctree fix is unrelated to `upernet`." + }, + { + "left": "pull_request:43998", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs-only toctree fix is unrelated to `timm_backbone`." } ] }, @@ -19284,7 +19277,7 @@ "pull_request:44662|pull_request:45075", "pull_request:45075|pull_request:45401" ], - "prepared_review_unit_hash": "f47efba7f8576885c917dd027ae984332f4ebdf58852f215036b8d61006d1b47", + "prepared_review_unit_hash": "5deb9c40b1e27830f8992722fb8802a53dc777220ecd20a6f8b2818654d720b6", "component_budget": { "node_count": 3, "item_count": 3, @@ -19306,30 +19299,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-implementation PRs for different architectures (PenguinVL, DeepSeek-OCR-2, Voxtral TTS). They share only generic Transformers boilerplate and registry/docs updates, not the same concrete bug or change.", - "confidence": 0.96, + "summary": "These three pull requests are unrelated model-addition implementations for different architectures (PenguinVL, DeepSeek-OCR-2, and Voxtral TTS). They share only generic Transformers boilerplate areas, so none should be treated as duplicates.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "No single PR is a duplicate of the others; they implement different models and touch different code paths. If one representative is needed, PR 45075 is the most substantial, but it is not canonical in a duplicate sense.", + "canonical_pr_reason": "No duplicate PR exists here; if a representative is needed, #45075 is the most complete/production-ready of the three because it includes model, image processor, converter, docs, and tests for a fully scoped model addition.", "best_issue_reason": null, - "best_pr_reason": "PR 45075 is the strongest representative of the cluster because it is the most complete, non-draft implementation with the broadest review activity. That said, it is not a better duplicate candidate than the others because the PRs are unrelated.", + "best_pr_reason": "#45075 is the strongest standalone PR to represent this cluster because it appears more complete than the draft TTS PR and is a full model implementation with conversion and test coverage. The other two PRs target entirely different models and should remain separate.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different model additions (PenguinVL vs Voxtral TTS); overlap is limited to shared Transformers scaffolding and docs/auto registry updates, not the same underlying change." + "reason": "Different model families and tasks: PenguinVL vs Voxtral TTS. Shared auto/docs scaffolding is generic and not evidence of the same bug or change." }, { "left": "pull_request:44662", "right": "pull_request:45075", "accept": false, - "reason": "These add different models (PenguinVL vs DeepSeek-OCR-2). Shared filenames are generic integration points, but the code paths, weights, and model implementations are unrelated." + "reason": "Distinct model additions for different architectures (PenguinVL vs DeepSeek-OCR-2). Overlap is only in standard Transformers registration files and docs, which is too broad to merge as duplicates." }, { "left": "pull_request:45075", "right": "pull_request:45401", "accept": false, - "reason": "DeepSeek-OCR-2 and Voxtral TTS are distinct model implementations with different functionality; they do not look mergeable as one PR or the same bug fix." + "reason": "OCR model implementation vs TTS model implementation are unrelated changes. They touch similar framework entry points, but they do not fix the same concrete code-path problem and should not be merged together." } ] }, @@ -19359,12 +19352,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "871f43949ded10a1ac586baee66f1ddaf1e313ac465f2da0d6e1a6030d6c44a5", + "prepared_review_unit_hash": "4a8eaf094cbb11deb704cd010c930b3548679762ca8547ec00d4f16feecff58d", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -19372,7 +19365,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -19381,18 +19374,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are related to module/kernel fusion, but they are not the same change: #44979 introduces a generic Module Fusion API, while #45363 wires n-to-1 kernel fusion through KernelConfig and model loading. They overlap conceptually yet target different code paths and could not be safely merged into one PR as duplicates.", - "confidence": 0.91, + "summary": "These are related only at a theme level: one PR adds a generic module-fusion API, while the other adds n-to-1 kernel fusion through KernelConfig and integration hooks. They are not the same underlying change and should not be merged as duplicates.", + "confidence": 0.9, "canonical_issue_reason": null, - "canonical_pr_reason": "#44979 is the more foundational, general-purpose API change; #45363 is a separate integration feature built around kernel config and model loading.", + "canonical_pr_reason": "PR #44979 is the most canonical anchor for the cluster because it introduces the reusable Module Fusion API itself; the other PR is a separate downstream kernel-fusion application built around KernelConfig.", "best_issue_reason": null, - "best_pr_reason": "#44979 is the cleaner representative for the cluster because it defines the core abstraction rather than a specific downstream use case.", + "best_pr_reason": "PR #44979 is the best representative PR here because it defines the core fusion abstraction and has focused tests, whereas PR #45363 is a broader, separate feature implementation on top of that idea.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Different concrete changes: generic module-fusion API vs. KernelConfig-driven n-to-1 kernel fusion integration. Shared theme is too broad to treat as the same underlying PR." + "reason": "Different changes: #44979 adds a new module-fusion API layer, while #45363 implements KernelConfig-driven n-to-1 kernel fusion across model loading/integration paths. Shared fusion terminology is not enough to treat them as the same fix/feature." } ] }, @@ -19422,7 +19415,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "aa9e661d9b1de3a4e2aafbded963009711bc0dd1713cc0e34cd5292ea972a9cd", + "prepared_review_unit_hash": "ab6d792972de836fa639a654a699f10af62ec7d582386bcd826a5c059b5e3873", "component_budget": { "node_count": 2, "item_count": 2, @@ -19444,18 +19437,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests address the same concrete AttributeError in `_patch_mistral_regex` by switching the pre-tokenizer access from `backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer` in the same code path. The second PR appears to be a fuller version of the same fix, including tests.", - "confidence": 0.98, + "summary": "Both PRs target the same Mistral regex patch bug on the same code path in `tokenization_utils_tokenizers.py`; the second PR is a slightly broader follow-up with tests, not a distinct fix.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45317 is the better canonical PR because it contains the same bug fix plus tests, making it the more complete and reviewable change set.", + "canonical_pr_reason": "PR 45317 is the better canonical PR because it addresses the same AttributeError fix as 45086 and adds test coverage, making it the more complete representation of the change.", "best_issue_reason": null, - "best_pr_reason": "PR 45317 is the best representative of the cluster: same fix as 45086, but with added test coverage and a clearer final form of the change.", + "best_pr_reason": "PR 45317 is the best PR overall since it fixes the same concrete bug as 45086 but is more complete and validated by tests.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same underlying AttributeError in `_patch_mistral_regex`, same file, same issue target, and same code-path fix; the second PR is effectively a superset with tests." + "reason": "Same underlying bug and same edited code path; 45317 appears to be an expanded version of 45086 with added tests, so they are mergeable into one PR." } ] }, @@ -19485,7 +19478,7 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "e933e92a0380631c46715b07a3fa98ded2e912f449bf37772f96cdf7d71f8fae", + "prepared_review_unit_hash": "8a8155fb74b461cb1a89636b667b781c56936eee7c6604de12c9f6df4b02e4f0", "component_budget": { "node_count": 2, "item_count": 2, @@ -19507,18 +19500,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two related NemotronH pull requests, but they address different scopes: one refactors inheritance/conversion mapping for NemotronH, while the other introduces/adjusts the NemotronH Dense and Sparse model split plus docs/tests. They overlap in the NemotronH code area, but are not the same underlying change.", - "confidence": 0.91, + "summary": "Two related NemotronH refactor PRs, but they are not duplicates: one is a narrower conversion/inheritance cleanup, the other is a broader model split and documentation/test update. The shared files indicate overlap in the same area, not the same concrete change.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:45569 is the broader, more complete implementation focused on the NemotronH family split and related docs/tests, so it is the better canonical PR for this cluster.", + "canonical_pr_reason": "PR 45569 is the broader, more complete NemotronH change: it introduces the dense/sparse model split, updates auto/model registrations, docs, and tests. PR 45115 is a narrower refactor focused on inheritance and conversion mapping, so 45569 is the better canonical representative for the cluster.", "best_issue_reason": null, - "best_pr_reason": "pull_request:45569 is the most comprehensive and user-facing change set, covering the actual model family structure and integration points rather than just the conversion/refactor groundwork.", + "best_pr_reason": "PR 45569 best represents the main architectural change because it covers the full model-family restructuring and user-facing integration, whereas 45115 is a smaller supporting refactor.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "They share NemotronH files, but 45115 is a narrow refactor/conversion-mapping change while 45569 adds the Dense/Sparse model split, docs, auto-mappings, and tests. These are related, not the same concrete change, and would not naturally merge into one PR." + "reason": "Both touch NemotronH internals, but they are not the same underlying fix. 45115 is a targeted refactor of conversion mapping and class inheritance, while 45569 adds the dense/sparse model variants plus docs, auto-mapping, and tests. They are related but not mergeable as one duplicate PR." } ] }, @@ -19548,12 +19541,12 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "35f43747ff0481f17c9674a01c5ee1e3d3d3cf90bf5e67de5b481f248cdca4fd", + "prepared_review_unit_hash": "b21b903d8d03230fc585efbd0d2d34a063c6b3497aa3f65e67aa2fbcbabe7bd1", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5397, + "serialized_chars": 5398, "estimated_input_tokens": 1350, "estimated_eval_tokens": 2956 }, @@ -19561,7 +19554,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5397, + "serialized_chars": 5398, "estimated_input_tokens": 1350, "estimated_eval_tokens": 2956 }, @@ -19570,18 +19563,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are closely related umbrella-feature PRs for heterogeneous model support, but they are not duplicates: one adds per-layer config plumbing, the other adds model/runtime support on top of it.", - "confidence": 0.87, + "summary": "Both PRs are about heterogeneous layer-wise configuration support, but one focuses on config plumbing while the other adds modeling/cache behavior on top of it. They are related, yet not the same concrete change and would not be safely merged as a single PR.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "No issue is present; as the representative PR, #45332 is the broader umbrella change and includes the runtime/modeling side in addition to heterogeneous config handling.", + "canonical_pr_reason": "PR #45332 is the broader umbrella change: it introduces heterogeneous model support across modeling/cache code, while PR #45333 is a narrower config-only foundation that appears to feed into it.", "best_issue_reason": null, - "best_pr_reason": "#45332 is the best cluster representative because it covers the larger end-to-end feature. #45333 is narrower, focused on per-layer configuration plumbing.", + "best_pr_reason": "PR #45332 is the best representative for the cluster because it spans the larger end-to-end heterogeneous-model feature, not just the configuration layer.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Related but not the same concrete change: #45333 adds heterogeneous config support, while #45332 adds heterogeneous modeling/runtime support and depends on the config work. They could be separate PRs and are not mergeable as one duplicate fix." + "reason": "Related feature area, but not the same fix/change: #45333 adds heterogeneous config support, while #45332 adds modeling and cache handling. They share files and concept, yet are separable PRs rather than one mergeable duplicate." } ] }, @@ -19611,12 +19604,12 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "644ae1fba2695055b662d19432d378fe9dcace5433a25e42bd291fa336d8e0ea", + "prepared_review_unit_hash": "c3b7f9a9cc8591ee478a8a646824841570296a966a261575a47504140b10583b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5514, + "serialized_chars": 5513, "estimated_input_tokens": 1379, "estimated_eval_tokens": 3014 }, @@ -19624,7 +19617,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5514, + "serialized_chars": 5513, "estimated_input_tokens": 1379, "estimated_eval_tokens": 3014 }, @@ -19633,18 +19626,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both items are open typing-related PRs, but they address different scopes: one is a broad repo-wide type-checking cleanup, the other is a more targeted typing-support update around modeling/utils integration. Shared filenames and theme alone aren\u2019t enough to call them duplicates.", - "confidence": 0.92, + "summary": "These two PRs are both typing-related, but they are not the same change: #45415 is a broad repo-wide type-checking sweep, while #45425 is a narrower typing/protocol update centered on modeling_utils/PEFT-related code. They should not be merged as duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45425 is the better representative because it is narrower and centered on specific typing support changes, whereas 45415 is a large sweep adding type checking across many transformer modules.", + "canonical_pr_reason": "No single canonical PR for a duplicate pair: #45415 is a broad type-checking sweep across many modules, while #45425 is a focused typing/protocol patch in a different area.", "best_issue_reason": null, - "best_pr_reason": "PR 45425 is the best single PR for this cluster since it reflects the more concrete, localized change; 45415 is too broad to serve as a canonical duplicate target.", + "best_pr_reason": "#45425 is the better representative PR for this cluster because it is more self-contained and specific, whereas #45415 is a much broader cleanup effort.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Reject: both are typing-focused, but 45415 is a broad type-checking pass across many files and config, while 45425 adds specific typing protocols/helpers for modeling_utils/PEFT-related code paths. They do not look like the same underlying change and would not plausibly merge into one PR as duplicates." + "reason": "Different underlying changes: #45415 is a broad typing sweep across many files, while #45425 adds typing aliases/protocols and related modeling_utils/PEFT updates. Same theme, not the same concrete fix." } ] }, @@ -19666,6 +19659,69 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45487-2", + "nodes": [ + "pull_request:45487", + "pull_request:46101" + ], + "soft_pairs": [ + "pull_request:45487|pull_request:46101" + ], + "prepared_review_unit_hash": "c456e2c7aec2dc7138b1427c0ebb47f9f214603b54c2d4167357185d08428a9a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5131, + "estimated_input_tokens": 1283, + "estimated_eval_tokens": 2822 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5131, + "estimated_input_tokens": 1283, + "estimated_eval_tokens": 2822 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are both device-placement/model-parallel fixes, but they target different model families and different failing code paths, so they are not duplicates. PR 45487 is the broader and more representative fix in this cluster.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45487 is the stronger canonical PR candidate because it addresses a concrete model-parallel device mismatch across multiple affected models/files and includes test coverage; the other PR is a separate Qwen2-Audio fix.", + "best_issue_reason": null, + "best_pr_reason": "PR 45487 is the best representative artifact for the cluster since it is broader, multi-model, and clearly fixes the underlying device mismatch bug with explicit code and test changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45487", + "right": "pull_request:46101", + "accept": false, + "reason": "Both are parallel/device fixes, but 45487 changes token_type_ids handling for AltCLIP/ChineseCLIP/etc. while 46101 fixes audio feature masking in Qwen2-Audio. Different models, different functions, and not plausibly one merged PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 23, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-45513-3", "nodes": [ "pull_request:45513", @@ -19676,7 +19732,7 @@ "pull_request:45513|pull_request:46084", "pull_request:46084|pull_request:46098" ], - "prepared_review_unit_hash": "2a0a150dc4048e54f292b0afc5435fe318cfe37fac96c5a8a60e80bd848c6aa1", + "prepared_review_unit_hash": "8287a8b5f71d5eb45b1b80536a1c9871ee84b88626b07f9e78a4d29fe6f6f220", "component_budget": { "node_count": 3, "item_count": 3, @@ -19698,30 +19754,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two PRs are closely related Mamba2 cache fixes for the same seq_len>1 use_cache bug; the Qwen3.5 PR is a separate model-specific change. PR #46098 is the best representative because it is the broader Mamba2 fix and explicitly targets the shared issue.", - "confidence": 0.86, + "summary": "The cluster splits into one unrelated PR ([Qwen3.5] GDN linear attention caching) and a Mamba2 cache bug pair. The Mamba2 PRs (46084 and 46098) target the same seq_len>1 use_cache incorrect-results issue; 46098 is the broader, more complete version. ", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46098 is the most complete representation of the Mamba2 bugfix: it targets the shared issue #46032, covers both CPU and GPU paths, and includes the accompanying docs/test updates.", + "canonical_pr_reason": "PR 46098 is the best representative because it states the full Mamba2 seq_len>1 use_cache bug, explicitly targets the tracking issue, and covers both CPU/GPU paths plus tests/docs.", "best_issue_reason": null, - "best_pr_reason": "PR #46098 is the best standalone PR for the cluster because it is the broadest, most explicit fix for the underlying Mamba2 cached-forward bug and subsumes the narrower #46084 change.", + "best_pr_reason": "PR 46098 is the strongest candidate to anchor the cluster: it is the most complete fix for the shared Mamba2 bug and is more self-contained than 46084.", "soft_edge_verdicts": [ { "left": "pull_request:45513", "right": "pull_request:46084", "accept": false, - "reason": "Both mention cached multi-token forward fixes, but they affect different models and code paths (Qwen3.5/OLMo hybrid linear attention vs Mamba2). Same pattern, not the same underlying bug." + "reason": "Different models and code paths: Qwen3.5/OLMO hybrid linear-attention caching versus Mamba2 use_cache handling. No evidence of the same underlying bug." }, { "left": "pull_request:46084", "right": "pull_request:46098", "accept": true, - "reason": "These are the same underlying Mamba2 use_cache seq_len>1 bug on the same files and issue target #46032; #46098 looks like a broader superset (CPU+GPU, docs/tests) that could plausibly be merged with #46084." + "reason": "Same underlying Mamba2 seq_len>1 cache correctness bug on the same model/file family; 46098 appears to be the broader superset fix, so these are plausibly mergeable into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster summary is grounded in the packet: 45513 is a separate Qwen3.5/OLMo hybrid change, while 46084 and 46098 both target the same Mamba2 use_cache seq_len>1 bug on shared files and the same issue (#46032). The mergeability judgment for 46084 <-> 46098 is conservative enough, since 46098 looks like a broader version of the same bugfix rather than a different change." + "feedback": "The summary is grounded in the packet: 45513 is clearly a different model/code path, and 46084/46098 both target the same Mamba2 seq_len>1 use_cache bug with shared issue target/files. The mergeability judgment for 46084 and 46098 is conservative and supported by the evidence. " }, "evaluator_used": true, "retried": false, @@ -19731,7 +19787,7 @@ }, { "label": "pull_request", - "component_index": 23, + "component_index": 24, "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, @@ -19748,7 +19804,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "e2c1424f415d79fcf39a0f20e36d39d72a225a0a57fcf0b193708699365f75da", + "prepared_review_unit_hash": "56cff4d840ed84fc50028e58a774d045b1da2e477aad75715346b71023bf24f3", "component_budget": { "node_count": 2, "item_count": 2, @@ -19770,18 +19826,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs both touch expert-parallel loading, but they fix different concrete failure modes: one is about EP + DeepSpeed ZeRO-3 / accelerate-launch serialization and model loading, while the other is about EP + FSDP2 experts being overwritten during distributed parameter assignment.", - "confidence": 0.93, + "summary": "These are related only at a very high level (both involve expert parallelism and model loading), but they fix different concrete bugs: one is about preserving/loading distributed_config for Accelerate + ZeRO-3, the other is about expert parameters being overwritten under FSDP2/rank-0 broadcast.", + "confidence": 0.95, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR 45662 is the broader and more representative EP bugfix in this cluster, with changes spanning loading, MoE, tensor-parallel, and trainer code paths.", + "best_pr_reason": "PR #45662 is the stronger canonical candidate for this cluster because it addresses a specific, well-scoped loading corruption bug across the relevant code path and has the broader set of coordinated changes; #45548 is a narrower serialization/flag-preservation fix for a different failure mode.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "They share the expert-parallel theme and one filename, but the underlying bugs are different: 45548 fixes distributed_config serialization plus a has_ep helper for EP + ZeRO-3 loading, while 45662 fixes rank-0 broadcast / DTensor wrapping during FSDP2 loading. These are not the same code-path problem and would not plausibly be merged as one PR." + "reason": "Both touch EP/loading, but they target different failure modes and code paths: #45548 preserves distributed_config and adds has_ep for Accelerate/ZeRO-3 loading, while #45662 fixes expert params being overwritten during FSDP2 post-shard handling. They would not plausibly merge as one PR." } ] }, @@ -19794,7 +19850,7 @@ }, { "label": "pull_request", - "component_index": 24, + "component_index": 25, "component_count": 27, "review_unit_index": 1, "review_unit_count": 1, @@ -19811,12 +19867,12 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "c91a51f9f1f2d75f9db553dc17ae6027982ba71bd4346bf5d36a548852d90bd4", + "prepared_review_unit_hash": "1b0ef316b0cca0fce3565772a8ba41f3f9cc00e3f7bec69cc3866c415abc979b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4589, + "serialized_chars": 4590, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -19824,7 +19880,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4589, + "serialized_chars": 4590, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -19833,81 +19889,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are closely related implementations of the same xdist-safe captured_info logging fix, with one PR acting as a broader superset of the other.", - "confidence": 0.87, + "summary": "Both PRs address the same xdist collision problem around captured_info artifacts in Transformers testing/debug logging. PR #45645 is the broader, more complete fix and subsumes the narrower code-only change in #45639.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45645 is the stronger canonical representative because it includes the core xdist-safe output-file change plus the CI workflow update needed to consume multiple captured_info files, making it the more complete fix.", + "canonical_pr_reason": "PR #45645 is the better canonical PR because it includes the core xdist-safe file change plus the CI/log-collection updates needed to preserve and read all captured_info shards.", "best_issue_reason": null, - "best_pr_reason": "PR #45645 is the best overall candidate: it addresses the same captured_info xdist collision bug as #45639 and also preserves downstream CI/debug-log behavior, so it is the most complete implementation in the cluster.", + "best_pr_reason": "PR #45645 is the strongest overall candidate: it fixes the same underlying captured_info collision bug and covers the operational CI path, while #45639 only changes the library-side helper.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Both PRs target the same captured_info xdist collision problem in testing_utils; #45645 is a superset that adds the workflow-side handling for the same fix, so they are close enough to treat as the same underlying change." - } - ] - }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, - "error_kind": null, - "error_message": null - }, - { - "label": "pull_request", - "component_index": 25, - "component_count": 27, - "review_unit_index": 1, - "review_unit_count": 1, - "status": "reviewed", - "reason": null, - "source": "cache", - "cache_hit": true, - "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-45683-2", - "nodes": [ - "pull_request:45683", - "pull_request:45734" - ], - "soft_pairs": [ - "pull_request:45683|pull_request:45734" - ], - "prepared_review_unit_hash": "eb709b85fdf0aaa42b264faf678be0edc1127137b5c90af86da076f2b207e68b", - "component_budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 4466, - "estimated_input_tokens": 1117, - "estimated_eval_tokens": 2490 - }, - "budget": { - "node_count": 2, - "item_count": 2, - "soft_pair_count": 1, - "serialized_chars": 4466, - "estimated_input_tokens": 1117, - "estimated_eval_tokens": 2490 - }, - "overflow_policy": "truncate_then_skip", - "trimmed": false, - "aggressively_trimmed": false, - "split": false, - "analyst_result": { - "summary": "These PRs are related but not duplicates: one excludes audio submodules from quantization, while the other makes auto-detected skip keys additive when users pass skip_modules. They touch the same quantizer path and share an issue target, but they fix different underlying bugs.", - "confidence": 0.92, - "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR for a duplicate cluster: the two PRs address separate behaviors in the same helper, not the same change.", - "best_issue_reason": null, - "best_pr_reason": "45734 is the better representative of the cluster because it fixes the more general control-flow bug in get_modules_to_not_convert and affects any user-supplied skip_modules case.", - "soft_edge_verdicts": [ - { - "left": "pull_request:45683", - "right": "pull_request:45734", - "accept": false, - "reason": "Shared files and issue target are not enough here: 45683 skips audio tower/embed_audio modules to avoid uint8 crashes, while 45734 changes skip_modules handling so default skips remain additive. Different bugs, different fixes." + "reason": "Both PRs target the same underlying xdist-safety bug for captured_info logging and touch the same concrete files; #45645 is an expanded superset of #45639 rather than a different change." } ] }, @@ -20003,40 +19996,40 @@ "soft_pairs": [ "pull_request:46069|pull_request:46073" ], - "prepared_review_unit_hash": "26fd8f39cdc1e8933ff15c2ef4aafe064eb780f2c19bff51f396ea023e7e0484", + "prepared_review_unit_hash": "d069e51ef43de099666188efc9294786b2ba4a6f440bddb1b14fcb9e5385bf71", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4330, - "estimated_input_tokens": 1083, - "estimated_eval_tokens": 2422 + "serialized_chars": 4394, + "estimated_input_tokens": 1099, + "estimated_eval_tokens": 2454 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4330, - "estimated_input_tokens": 1083, - "estimated_eval_tokens": 2422 + "serialized_chars": 4394, + "estimated_input_tokens": 1099, + "estimated_eval_tokens": 2454 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: both PRs add attn_implementation dispatch, but they target different model implementations (ShieldGemma2 vs CTRL) and patch different code paths/files.", - "confidence": 0.97, + "summary": "The two PRs are both about adding SDPA attention support, but they target different models (ShieldGemma2 vs CTRL) and different code paths, so they are not duplicates or mergeable into one change.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: the ShieldGemma2 PR and the CTRL PR are model-specific changes with distinct implementations and tests, not the same underlying fix.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "Neither PR is a clear global representative for the cluster because they address separate model backends rather than one shared bug or change.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:46069", "right": "pull_request:46073", "accept": false, - "reason": "Both PRs are about attn_implementation dispatch, but they apply to different models and different code paths; they are not the same change and would not plausibly merge into one PR." + "reason": "Different model implementations and test/docs updates; they share a broad attention-dispatch theme but not the same concrete bug or code path, so they should not be merged as one PR." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 719b1bfa7dd88ae10085c9d12acab095003208d2..8a093f54ff7b8c285a8e1ea3198995b431e0dfd8 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260520t060132z", + "analysis_id": "hybrid-model-20260520t120132z", "archived_artifacts": { - "hybrid": "snapshots/20260520T060132Z/analysis-runs/hybrid-model-20260520t060132z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260520T060132Z/analysis-runs/hybrid-model-20260520t060132z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-05-20T06:12:49Z", + "published_at": "2026-05-20T12:09:58Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260520T060132Z", + "snapshot_id": "20260520T120132Z", "variant": "hybrid" } diff --git a/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.json b/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..7e28c33b5b13dd1a9ecde8ea2b685360d796b07b --- /dev/null +++ b/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.json @@ -0,0 +1,1872 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260520T120132Z", + "generated_at": "2026-05-20T12:09:56Z", + "evidence_quality": "full", + "llm_enrichment": true, + "meta_bugs": [ + { + "cluster_id": "cluster-43979-11", + "summary": "Cluster of 1 issues and 10 PRs centered on issue #43979.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 43979, + "canonical_pr_number": 44007, + "issue_numbers": [ + 43979 + ], + "pr_numbers": [ + 43996, + 44007, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 43996, + "right_pr_number": 44007, + "code_similarity": 0.179, + "size_similarity": 0.576, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.429, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44013, + "code_similarity": 0.122, + "size_similarity": 0.318, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.392, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44044, + "code_similarity": 0.245, + "size_similarity": 0.864, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.479, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44066, + "code_similarity": 0.225, + "size_similarity": 0.818, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.408, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44072, + "code_similarity": 0.14, + "size_similarity": 0.303, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.528, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44085, + "code_similarity": 0.216, + "size_similarity": 0.783, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44129, + "code_similarity": 0.163, + "size_similarity": 0.643, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.229, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44154, + "code_similarity": 0.153, + "size_similarity": 0.535, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.31, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44722, + "code_similarity": 0.225, + "size_similarity": 0.848, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.368, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44013, + "code_similarity": 0.19, + "size_similarity": 0.553, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.531, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44044, + "code_similarity": 0.186, + "size_similarity": 0.667, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.354, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44066, + "code_similarity": 0.188, + "size_similarity": 0.704, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.315, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44072, + "code_similarity": 0.212, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.708, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44085, + "code_similarity": 0.195, + "size_similarity": 0.735, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.318, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44129, + "code_similarity": 0.103, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.191, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44154, + "code_similarity": 0.238, + "size_similarity": 0.93, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.344, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44722, + "code_similarity": 0.178, + "size_similarity": 0.679, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.28, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44044, + "code_similarity": 0.126, + "size_similarity": 0.368, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.351, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44066, + "code_similarity": 0.127, + "size_similarity": 0.389, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.325, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44072, + "code_similarity": 0.29, + "size_similarity": 0.952, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.667, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44085, + "code_similarity": 0.131, + "size_similarity": 0.406, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.329, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44129, + "code_similarity": 0.07, + "size_similarity": 0.205, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.192, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44154, + "code_similarity": 0.177, + "size_similarity": 0.594, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.389, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44722, + "code_similarity": 0.118, + "size_similarity": 0.375, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.287, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44066, + "code_similarity": 0.25, + "size_similarity": 0.947, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.404, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44072, + "code_similarity": 0.136, + "size_similarity": 0.351, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.442, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44085, + "code_similarity": 0.24, + "size_similarity": 0.906, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.394, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44129, + "code_similarity": 0.147, + "size_similarity": 0.555, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.243, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44154, + "code_similarity": 0.17, + "size_similarity": 0.62, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.306, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44722, + "code_similarity": 0.257, + "size_similarity": 0.982, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.402, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44072, + "code_similarity": 0.133, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.393, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44085, + "code_similarity": 0.763, + "size_similarity": 0.957, + "file_overlap": 0.5, + "area_overlap": 0.825, + "patch_similarity": 0.887, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ] + } + ] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44129, + "code_similarity": 0.145, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.263, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44154, + "code_similarity": 0.174, + "size_similarity": 0.654, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.286, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44722, + "code_similarity": 0.907, + "size_similarity": 0.964, + "file_overlap": 1.0, + "area_overlap": 0.808, + "patch_similarity": 0.874, + "shared_filenames": [ + "src/transformers/models/codegen/modeling_codegen.py", + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/codegen/modeling_codegen.py", + "left_ranges": [ + [ + 245, + 266 + ] + ], + "right_ranges": [ + [ + 228, + 234 + ] + ] + }, + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44085, + "code_similarity": 0.137, + "size_similarity": 0.387, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44129, + "code_similarity": 0.074, + "size_similarity": 0.195, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.231, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44154, + "code_similarity": 0.175, + "size_similarity": 0.566, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.414, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44722, + "code_similarity": 0.124, + "size_similarity": 0.357, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.347, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44129, + "code_similarity": 0.141, + "size_similarity": 0.503, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.272, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44154, + "code_similarity": 0.18, + "size_similarity": 0.684, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.289, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44722, + "code_similarity": 0.728, + "size_similarity": 0.923, + "file_overlap": 0.5, + "area_overlap": 0.791, + "patch_similarity": 0.78, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44154, + "code_similarity": 0.099, + "size_similarity": 0.344, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.199, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44722, + "code_similarity": 0.146, + "size_similarity": 0.545, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.247, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44154, + "right_pr_number": 44722, + "code_similarity": 0.164, + "size_similarity": 0.631, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.255, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, + { + "cluster_id": "cluster-41211-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41211.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41211, + "canonical_pr_number": 44339, + "issue_numbers": [ + 41211 + ], + "pr_numbers": [ + 41356, + 44339 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41356, + "right_pr_number": 44339, + "code_similarity": 0.155, + "size_similarity": 0.094, + "file_overlap": 0.273, + "area_overlap": 0.078, + "patch_similarity": 0.183, + "shared_filenames": [ + "docs/source/en/model_doc/deimv2.md", + "src/transformers/models/deimv2/__init__.py", + "src/transformers/models/deimv2/configuration_deimv2.py", + "src/transformers/models/deimv2/modeling_deimv2.py", + "tests/models/deimv2/__init__.py", + "tests/models/deimv2/test_modeling_deimv2.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/model_doc/deimv2.md", + "left_ranges": [ + [ + 1, + 132 + ] + ], + "right_ranges": [ + [ + 1, + 65 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/__init__.py", + "left_ranges": [ + [ + 1, + 15 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/configuration_deimv2.py", + "left_ranges": [ + [ + 1, + 74 + ] + ], + "right_ranges": [ + [ + 1, + 266 + ] + ] + }, + { + "filename": "tests/models/deimv2/test_modeling_deimv2.py", + "left_ranges": [ + [ + 1, + 15 + ] + ], + "right_ranges": [ + [ + 1, + 1734 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43240-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43240.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43240, + "canonical_pr_number": 43251, + "issue_numbers": [ + 43240 + ], + "pr_numbers": [ + 43251, + 43254 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43251, + "right_pr_number": 43254, + "code_similarity": 0.794, + "size_similarity": 0.64, + "file_overlap": 1.0, + "area_overlap": 0.667, + "patch_similarity": 0.882, + "shared_filenames": [ + "src/transformers/loss/loss_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/loss/loss_utils.py", + "left_ranges": [ + [ + 30, + 50 + ] + ], + "right_ranges": [ + [ + 30, + 43 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-41115-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41115.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41115, + "canonical_pr_number": 45613, + "issue_numbers": [ + 41115 + ], + "pr_numbers": [ + 41116, + 45613 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41116, + "right_pr_number": 45613, + "code_similarity": 0.805, + "size_similarity": 0.885, + "file_overlap": 0.75, + "area_overlap": 0.854, + "patch_similarity": 0.693, + "shared_filenames": [ + "docs/source/en/_toctree.yml", + "docs/source/en/model_doc/minicpm3.md", + "src/transformers/models/auto/modeling_auto.py", + "src/transformers/models/minicpm3/__init__.py", + "src/transformers/models/minicpm3/configuration_minicpm3.py", + "src/transformers/models/minicpm3/modeling_minicpm3.py", + "src/transformers/models/minicpm3/modular_minicpm3.py", + "tests/models/minicpm3/__init__.py", + "tests/models/minicpm3/test_modeling_minicpm3.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/_toctree.yml", + "left_ranges": [ + [ + 713, + 720 + ] + ], + "right_ranges": [ + [ + 713, + 720 + ] + ] + }, + { + "filename": "docs/source/en/model_doc/minicpm3.md", + "left_ranges": [ + [ + 1, + 67 + ] + ], + "right_ranges": [ + [ + 1, + 45 + ] + ] + }, + { + "filename": "src/transformers/models/auto/modeling_auto.py", + "left_ranges": [ + [ + 285, + 291 + ], + [ + 700, + 706 + ], + [ + 1298, + 1304 + ] + ], + "right_ranges": [ + [ + 699, + 705 + ], + [ + 1301, + 1307 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/__init__.py", + "left_ranges": [ + [ + 1, + 29 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/configuration_minicpm3.py", + "left_ranges": [ + [ + 1, + 141 + ] + ], + "right_ranges": [ + [ + 1, + 126 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 544 + ] + ], + "right_ranges": [ + [ + 1, + 522 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modular_minicpm3.py", + "left_ranges": [ + [ + 1, + 444 + ] + ], + "right_ranges": [ + [ + 1, + 342 + ] + ] + }, + { + "filename": "tests/models/minicpm3/test_modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 109 + ] + ], + "right_ranges": [ + [ + 1, + 136 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45647-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45647.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45647, + "canonical_pr_number": 45738, + "issue_numbers": [ + 45647 + ], + "pr_numbers": [ + 45737, + 45738 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 45737, + "right_pr_number": 45738, + "code_similarity": 0.526, + "size_similarity": 0.11, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.024, + "shared_filenames": [ + "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py", + "left_ranges": [ + [ + 491, + 497 + ] + ], + "right_ranges": [ + [ + 491, + 497 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46032-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46032.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 46032, + "canonical_pr_number": 46084, + "issue_numbers": [ + 46032 + ], + "pr_numbers": [ + 46084, + 46098 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46084, + "right_pr_number": 46098, + "code_similarity": 0.451, + "size_similarity": 0.531, + "file_overlap": 0.667, + "area_overlap": 0.272, + "patch_similarity": 0.33, + "shared_filenames": [ + "src/transformers/models/mamba2/modeling_mamba2.py", + "tests/models/mamba2/test_modeling_mamba2.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/mamba2/modeling_mamba2.py", + "left_ranges": [ + [ + 256, + 262 + ], + [ + 303, + 362 + ], + [ + 470, + 476 + ], + [ + 480, + 500 + ], + [ + 505, + 511 + ], + [ + 608, + 618 + ] + ], + "right_ranges": [ + [ + 255, + 373 + ], + [ + 481, + 495 + ] + ] + }, + { + "filename": "tests/models/mamba2/test_modeling_mamba2.py", + "left_ranges": [ + [ + 206, + 236 + ], + [ + 285, + 322 + ] + ], + "right_ranges": [ + [ + 476, + 542 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45561-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45561, + "canonical_pr_number": 45645, + "issue_numbers": [ + 45561 + ], + "pr_numbers": [ + 45639, + 45645 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45639, + "right_pr_number": 45645, + "code_similarity": 0.532, + "size_similarity": 0.619, + "file_overlap": 0.5, + "area_overlap": 0.543, + "patch_similarity": 0.451, + "shared_filenames": [ + "src/transformers/testing_utils.py", + "tests/utils/test_testing_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/testing_utils.py", + "left_ranges": [ + [ + 3529, + 3536 + ], + [ + 3753, + 3787 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 86 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46100-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 46100, + "issue_numbers": [], + "pr_numbers": [ + 46100, + 46114 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46100, + "right_pr_number": 46114, + "code_similarity": 1.0, + "size_similarity": 1.0, + "file_overlap": 1.0, + "area_overlap": 1.0, + "patch_similarity": 1.0, + "shared_filenames": [ + "tests/models/vit/test_modeling_vit.py" + ], + "shared_file_areas": [ + { + "filename": "tests/models/vit/test_modeling_vit.py", + "left_ranges": [ + [ + 216, + 226 + ] + ], + "right_ranges": [ + [ + 216, + 226 + ] + ] + } + ] + } + ] + } + ], + "duplicate_issues": [], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 45613, + "duplicate_pr_numbers": [ + 41116 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45561-3", + "canonical_pr_number": 45645, + "duplicate_pr_numbers": [ + 45639 + ], + "target_issue_number": 45561, + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45647-3", + "canonical_pr_number": 45738, + "duplicate_pr_numbers": [ + 45737 + ], + "target_issue_number": 45647, + "reason": "PRs in cluster-45647-3 are treated as duplicates because they converge on issue #45647 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46032-3", + "canonical_pr_number": 46084, + "duplicate_pr_numbers": [ + 46098 + ], + "target_issue_number": 46032, + "reason": "PRs in cluster-46032-3 are treated as duplicates because they converge on issue #46032 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46100-2", + "canonical_pr_number": 46100, + "duplicate_pr_numbers": [ + 46114 + ], + "target_issue_number": null, + "reason": "PRs in cluster-46100-2 are treated as duplicates because they share soft_similarity evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-45647-3", + "issue_number": 45647, + "reason": "Issue #45647 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 12 discussion comments plus 8 inbound references.", + "score": 162.0 + }, + "best_pr": { + "cluster_id": "cluster-43979-11", + "pr_number": 44007, + "reason": "PR #44007 is the strongest global PR candidate because it is open, belongs to a cluster with 11 artifacts, links to 1 issues, and carries 2 review/discussion events.", + "score": 238.886 + } +} diff --git a/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..3595b3b387dc200e8117a6aa66fa996d5cc3ecd7 --- /dev/null +++ b/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,20044 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260520T120132Z", + "generated_at": "2026-05-20T12:09:56Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43425", + "issue:43541", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45081", + "issue:45161", + "issue:45237", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:45874", + "issue:45901", + "issue:45902", + "issue:45920", + "issue:46015" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:45507", + "issue:44292|issue:45237", + "issue:30277|issue:35545", + "issue:35545|issue:42915", + "issue:35545|issue:41720", + "issue:45901|issue:46015", + "issue:44509|issue:44512", + "issue:35545|issue:36010", + "issue:44805|issue:45823", + "issue:43638|issue:44805", + "issue:43596|issue:44805", + "issue:44805|issue:45161", + "issue:44805|issue:44918", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43039|issue:45161", + "issue:43975|issue:44457", + "issue:44292|issue:44457", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:45859|issue:45902", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:39290|issue:45081", + "issue:45823|issue:45874", + "issue:43425|issue:44292", + "issue:43828|issue:45161", + "issue:34928|issue:41720", + "issue:44918|issue:45161", + "issue:43716|issue:45685", + "issue:43541|issue:43716", + "issue:43828|issue:44292", + "issue:43716|issue:44292", + "issue:45507|issue:45685", + "issue:44292|issue:44918", + "issue:43039|issue:44918", + "issue:41762|issue:43638", + "issue:43827|issue:44512", + "issue:43716|issue:44560", + "issue:43828|issue:45685", + "issue:44560|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:45237|issue:46015", + "issue:34928|issue:36010", + "issue:43039|issue:44805", + "issue:41762|issue:43596", + "issue:43901|issue:44512", + "issue:39290|issue:41720", + "issue:45237|issue:45901", + "issue:44841|issue:45823", + "issue:43039|issue:44560", + "issue:43975|issue:46015", + "issue:45812|issue:45920", + "issue:44863|issue:46015", + "issue:43866|issue:44863", + "issue:44292|issue:46015", + "issue:43854|issue:43866", + "issue:41093|issue:43638" + ], + "prepared_review_unit_hash": "de943234667c776ba187f80b6ec4683cc8c87f10bc1660e68809d4084b5458a4", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 66, + "serialized_chars": 47567, + "estimated_input_tokens": 11892, + "estimated_eval_tokens": 24040 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: most soft-similarity links connect unrelated model/runtime/doc issues. The clear duplicate subgroups are the AutoTokenizer token-ID regression reports and the ZeRO-3 Bert out-of-bounds reports; most other pairs should stay separate.", + "confidence": 0.69, + "canonical_issue_reason": "Issue 45920 is the broadest, most representative report in the only clearly duplicated subgroup here: it covers the AutoTokenizer wrong-token-ID regression across multiple models, not just one model family.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45920 is the best issue to keep as the umbrella report: open, clearly phrased, and broad enough to subsume the narrower Granite-only regression report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different failures: a missing config attribute in Qwen3.5 vs a chat/streaming inference failure. No same code path." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different models and hardware paths: GPT-OSS on AMD GPUs vs an MPS histogram op error in MoE code." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different bugs: Mistral-3 dtype mismatch vs GPT-OSS/AMD GPU support." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Both are dtype/device-related at a high level, but they are distinct models and distinct failure modes." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different model families and different invalid method/device issues; not the same bug." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Separate model/runtime failures with no clear shared code-path." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed ZeRO-3 collective mismatch is unrelated to ModernBERT ONNX export." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ONNX export failure vs Qwen3Moe FP8 loading failure; unrelated." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ModernBERT export and Qwen3 auto device mapping are different problems." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "Table QA crash vs outdated RAG docs/examples; not the same underlying issue." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are v5 docs cleanups, but they concern different removed items and different documentation locations." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "ONNX export error and GenerationMixin import error are unrelated." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask/index shape mismatch vs Gemma4 device_map forward device mismatch; different failures." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both mention indexing errors, but one is ZeRO-3 Bert out-of-bounds and the other is a mask-shape mismatch; not clearly one bug." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Same broad symptom class, but different models and failure details; too weak to treat as duplicates." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Mask-shape index error vs GPT-OSS MoE tensor-parallel issue are unrelated." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Different model/trainer failures with no shared concrete bug." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice preset bug is unrelated to ModernBERT export." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but one is collective mismatch and the other is an index-out-of-bounds during ZeRO-3 init." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 Bert out-of-bounds failure with nearly identical title and context; clearly the same underlying bug." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross_entropy routing and GPT-OSS tensor-parallel failure are different issues." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "DeepSeek detokenization regression and LoRA merge/save mismatch are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error and LoRA reload mismatch are distinct bugs." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "Both are shape/mismatch errors, but one is FSDP checkpointing and the other is a mask/indexing bug." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "DeepSpeed training errors, but one is a 2-D weight error and the other is a collective mismatch; not the same defect." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": false, + "reason": "Both involve Qwen3_5MoeTextRotaryEmbedding, but one is CPU-offload incompatibility and the other is uninitialized-memory after to_empty materialization." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "ZeRO-3 Bert index error vs GPT-OSS MoE tensor-parallel issue are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "Activation-checkpointing/FSDP tensor mismatch is unrelated to a CustomTrainer signature change." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "Different domains: ZeRO-3 Bert init error vs GPT-OSS TP failure." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Mistral-3 dtype mismatch and GraniteMoEHybrid invalid method are separate failures." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3 vLLM config attribute error and Mistral tokenizer regex patch crash are unrelated." + }, + { + "left": "issue:45823", + "right": "issue:45874", + "accept": false, + "reason": "Gemma4 device mismatch during forward vs expensive reverse embedding lookup are different issues." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "Torch 2.10 compatibility issue and Qwen-3-8B-NVFP4 runtime error are not the same bug." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "Qwen3.5 embedding unpacking in TRL SFT trainer is unrelated to GPT-OSS TP support." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": false, + "reason": "Both are ZeRO-3 out-of-bounds style errors, but they hit different models and usage paths; not clearly one underlying bug." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions in v5, but they reference different removed features and likely different pages." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Mistral-3 dtype mismatch vs Qwen3-vl video StopIteration are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross_entropy routing and video embedding StopIteration are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:46015", + "accept": false, + "reason": "Tokenizer detokenization regression vs outdated RAG examples are unrelated." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same underlying AutoTokenizer wrong-token-ID regression; 45920 is the broader umbrella report across multiple model families." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint loading failure is unrelated to outdated RAG docs." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Corrupted Ovis2 checkpoint and NemotronH checkpoint-loading implementation bug are different issues." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error is unrelated to docs/examples cleanup." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Unable to load GLM-4.7-Flash in tests vs corrupted Ovis2 checkpoint are different model-specific problems." + }, + { + "left": "issue:43425", + "right": "issue:43866", + "accept": false, + "reason": "Torch version incompatibility and checkpoint corruption are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch and MPS histogram op failure are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Video StopIteration in Qwen3-vl-embedding and MPS histogram op failure are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Different MoE/dtype failures with no shared code path." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch is unrelated to the CustomTrainer compute_loss signature change." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "GPT-OSS AMD inference issue and outdated RAG docs/examples are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP activation-checkpointing mismatch and GenerationMixin import error are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross_entropy issue and mask-shape indexing error are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Both are ZeRO-3 index-out-of-bounds reports, but the model contexts differ enough that this is not clearly one duplicate bug." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "TextClassificationPipeline docs mention is unrelated to the removed run command docs issue." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Gemma3 vLLM config attribute error and Qwen3 auto device mapping cuda assert are different issues." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "GPT-OSS AMD failure and table QA crash are unrelated." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Processor failure for Voxtral is unrelated to Gemma4 device mismatch." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross_entropy call and Qwen3-vl video StopIteration are unrelated." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same AutoTokenizer token-ID regression across models; the broader issue should absorb the Granite-specific report." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint loading and outdated RAG docs/examples are unrelated." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Corrupted checkpoint vs model-loading implementation bug are different failure modes." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Different models and different root causes; not the same issue." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both are index/mask-style runtime errors, but one is a mask-length mismatch and the other is a ZeRO-3 index-out-of-bounds bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The two accepted duplicate links are well supported by the titles/context: the ZeRO-3 Bert out-of-bounds pair and the AutoTokenizer wrong-token-ID regression pair. The rest of the soft-edge rejections are appropriately cautious and do not overclaim shared root cause." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 2, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36010", + "issue:39290", + "issue:40990", + "issue:41093", + "issue:41762", + "issue:42915", + "issue:43054", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43645", + "issue:43716", + "issue:43792", + "issue:43828", + "issue:43866", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44530", + "issue:44589", + "issue:44841", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45161", + "issue:45362", + "issue:45399", + "issue:45464", + "issue:45542", + "issue:45632", + "issue:45685", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45901", + "issue:45902" + ], + "soft_pairs": [ + "issue:39290|issue:43531", + "issue:43792|issue:45901", + "issue:43425|issue:45070", + "issue:43596|issue:45800", + "issue:43975|issue:45901", + "issue:43792|issue:45070", + "issue:45362|issue:45464", + "issue:44292|issue:45901", + "issue:39290|issue:45823", + "issue:43404|issue:43866", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:44841|issue:45084", + "issue:36010|issue:39290", + "issue:45800|issue:45823", + "issue:41093|issue:41762", + "issue:43296|issue:43366", + "issue:43572|issue:45070", + "issue:43541|issue:45161", + "issue:44387|issue:45005", + "issue:43828|issue:45464", + "issue:44291|issue:45070", + "issue:43366|issue:45464", + "issue:45020|issue:45800", + "issue:43531|issue:45070", + "issue:43541|issue:44589", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45542|issue:45800", + "issue:45759|issue:45902", + "issue:43716|issue:44918", + "issue:44918|issue:45685", + "issue:45632|issue:45698", + "issue:43404|issue:45005", + "issue:43645|issue:45399", + "issue:40990|issue:44960", + "issue:44530|issue:45005", + "issue:42915|issue:45399" + ], + "prepared_review_unit_hash": "01f4ef70e52ca3593d15cba457d49bbe9d22c6b24bdb4f84a98cbc4f10a4d90c", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 39, + "serialized_chars": 38525, + "estimated_input_tokens": 9632, + "estimated_eval_tokens": 19520 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a mixed cluster of Transformers/vLLM regression reports with a couple of near-duplicate subthreads; no PRs are present. The broadest issue here is the recent PretrainedConfig/pydantic regression, though most pairs are not true duplicates.", + "confidence": 0.56, + "canonical_issue_reason": "Issue 45070 is the broadest, most reusable regression report in the set: a core v5.4.0 PretrainedConfig breakage. It best serves as a canonical anchor for this mixed version-regression cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45070 has the clearest scope, strongest generality, and highest chance of covering adjacent version-breakage reports better than the narrower model-specific bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention sliding-window-related failures, but they hit different models and different failure modes; not the same concrete bug." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Completely different models and error paths; no shared underlying change." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch-version incompatibility and a pydantic/config regression are distinct problems." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "Different failures: DeepSpeed ZeRO-3 index error versus torch/transformers version incompatibility." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Different task/model bugs with no shared code path." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "No overlap beyond being Transformers bugs; the concrete failures 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; too different to merge." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "Different model-loading error versus task crash; not the same bug." + }, + { + "left": "issue:39290", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3 sliding-window attribute error and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Different models and distinct failures; tied-weights bug versus corrupted checkpoint." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 streaming/API failure is unrelated to the MPS histogram integer error." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Grouped_mm tracing failure and MPS integer histogram failure are different code paths." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Both are runtime dtype/tracing-style issues, but they affect different models and ops; not clearly one bug." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Processor loading and template compilation failures are different mechanisms." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "Importing GenerationMixin and a Gemma3 sliding-window attribute error are unrelated." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Version incompatibility and device mismatch are different regressions." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are index errors, but they occur in different contexts and likely different buggy code paths." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "PaddleOCR-VL loading failure and GGUF gpt-oss support are unrelated feature/compatibility issues." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Missing pad_token_idx and PretrainedConfig pydantic breakage are separate config regressions." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Torch-dynamo grouped_mm tracing and GPT-OSS TP-only behavior are different issues." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Int4 OOM regression and tied-weights translation-model behavior are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Different models and different symptoms; dtype mismatch under autocast is not the same as streaming inference failure." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Unexpected init_empty_weights argument and pydantic field breakage are separate API regressions." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "GGUF gpt-oss support and Qwen3.5 streaming inference are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Remote_code breakage and torch/transformers incompatibility are not the same concrete defect." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Sliding-window Qwen3-MoE failure and PretrainedConfig pydantic regression are different bugs." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Torch-dynamo grouped_mm tracing and Float8 storage lookup failure are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are index errors, but one is a mask shape mismatch and the other is a zero-size loading error; different fixes." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": true, + "reason": "Both describe SigLIP2 producing incorrect or nonsensical text embeddings/outputs for the same model family, so they likely share the same underlying implementation bug." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "TensorBoard-without-TF backend issue is unrelated to torch/transformers compatibility." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Both touch Qwen3.5-related configs, but one is from_config unwrapping and the other is embedding materialization; not the same bug." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Image-preprocessor dtype mismatch and SFT embedding unpacking failure are different model/training paths." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "No meaningful overlap beyond being runtime failures." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": true, + "reason": "Both report wrong custom/local module resolution when loading models, consistent with the same cache/path selection bug." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Tied weights in Mistral3 and tied-weights translation-model regressions are different model-specific bugs." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Custom-model notebook initialization and flash-attn fallback gating are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "Perplexity regression and GLM5 issue are unrelated model-specific reports." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and tied-weights regression are different subsystems." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "FineGrainedFP8 Qwen3Moe failure and flash-attn fallback blocking are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded in the packet overall: the summary is appropriately cautious, the broad canonical choice of issue:45070 is supported by the titles, and the two accepted soft-edge pairs look plausibly duplicate-level from the available evidence. No PR-pair mergeability concerns apply here because there are no PRs." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 3, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:39186", + "issue:39692", + "issue:41553", + "issue:41628", + "issue:41762", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43493", + "issue:43643", + "issue:43644", + "issue:43824", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44492", + "issue:44512", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:45005", + "issue:45020", + "issue:45507", + "issue:45520", + "issue:45663", + "issue:45685", + "issue:45701", + "issue:45800", + "issue:45823", + "issue:45859", + "issue:45902" + ], + "soft_pairs": [ + "issue:44360|issue:44512", + "issue:39186|issue:44918", + "issue:43873|issue:45005", + "issue:44863|issue:45005", + "issue:44492|issue:44512", + "issue:44918|issue:45507", + "issue:43975|issue:44292", + "issue:44292|issue:45507", + "issue:43950|issue:44534", + "issue:39692|issue:43054", + "issue:43644|issue:43950", + "issue:43828|issue:44928", + "issue:36331|issue:41762", + "issue:41628|issue:45800", + "issue:43866|issue:44403", + "issue:43950|issue:45902", + "issue:43872|issue:44589", + "issue:38175|issue:43054", + "issue:41553|issue:43883", + "issue:43054|issue:45701", + "issue:44910|issue:45859", + "issue:45663|issue:45902", + "issue:41553|issue:43881", + "issue:42947|issue:44387", + "issue:44403|issue:44451", + "issue:43054|issue:43493", + "issue:43643|issue:45020", + "issue:43975|issue:45520", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:41762|issue:42915", + "issue:45685|issue:45823", + "issue:43881|issue:43883", + "issue:43824|issue:43872" + ], + "prepared_review_unit_hash": "2c53fb31ab352fd5406e8fb26dc1c10745e22839f880366414e9ecd6d80c0399", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37546, + "estimated_input_tokens": 9387, + "estimated_eval_tokens": 19030 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous Transformers bug reports, with one clear duplicate subcluster around non-persistent buffer corruption in v5. Most soft pairs are only loosely related by subsystem or version, not the same underlying bug.", + "confidence": 0.82, + "canonical_issue_reason": "issue:43950 is the strongest representative of the recurring v5 regression about non-persistent buffers being corrupted; it is the clearest, most general statement of that bug and subsumes the closer follow-ups (43643/44534).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43950 has the broadest impact and best captures the repeated buffer-corruption regression theme seen in the cluster, making it the most suitable anchor issue.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "One is a DSA/ReLU discussion; the other is a docs note about a removed run command. Same docs area, not the same bug." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP weight-shape runtime error vs. Qwen3.5 embedding unpacking with TRL; different failing code paths." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior vs. tied-weights issues in translation models; related theme, but not the same defect." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading failure is model-specific and not the same as tied-weights handling in v5." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "A cache-strategy typo and a documentation note about a removed command are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Qwen3.5 embedding unpacking with TRL is unrelated to GraniteMoEHybrid calling an invalid method." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Both involve model/runtime behavior, but one is detokenization and the other is an NVFP4 execution error; not the same bug." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime failure and GraniteMoEHybrid invalid method call are different concrete failures." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": false, + "reason": "These are very close, but the wording suggests separate reports of the same regression rather than a clearly shared issue/PR target from the available data." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "SigLIP2 docs/example errors vs. text embedding quality regression; both about SigLIP2, but different problems." + }, + { + "left": "issue:43643", + "right": "issue:43950", + "accept": false, + "reason": "Both touch v5 configuration/buffer handling, but 43643 is remote-code config field loss while 43950 is buffer corruption." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Autocast dtype mismatch in Phi-tiny-MoE is not the same as Qwen3.5 position-id/SDPA issues." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Trainer compute_loss signature error and Gemma3 ZeRO-3 loading failure are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45800", + "accept": false, + "reason": "Importing AutoImageProcessor and torch/transformers version incompatibility are different issues." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Corrupted checkpoint vs. generic loading noise; not the same root cause." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "Both concern buffer/materialization state, but 45902 is Qwen3_5MoeTextRotaryEmbedding after to_empty(cuda), while 43950 is a broader v5 buffer corruption regression." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "bitsandbytes constructor-arg incompatibility and missing Float8 storage are separate backend errors." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Unexpected zero probabilities and worse text embeddings for SigLIP2 are different symptoms with different likely causes." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad AutoTokenizer error for Voxtral and Molmo tied-weights attribute error are unrelated." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "SigLIP2 text embedding quality and generic tokenization changes are not the same bug." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Both involve Qwen3.5 and offload/position embeddings, but one is flash-attn crash from packed-sequence misinterpretation and the other is CPU-offload incompatibility." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states bug and Qwen3_5Moe rotary-buffer initialization bug are different model-specific failures." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Voxtral tokenizer loading error and glm-4v-9b loading failure are distinct model-loading problems." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "LoRA gradient checkpointing ineffectiveness and int4 CUDA reserved-memory growth are different training regressions." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise versus inability to load a specific model; same area, not same bug." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "Both are SigLIP2-related, but one is embedding quality and the other is HF vs JAX implementation discrepancy." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Remote-code config field loss and broader breakage with recent transformers versions are not the same concrete failure." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Detokenization regression and Python 3.13 flash_attn import KeyError are unrelated." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Both mention model loading problems, but they are for different checkpoints and likely different root causes." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Gradient checkpointing/LoRA and Qwen3 MoE memory usage are separate training concerns." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading failure and Qwen3Moe FineGrainedFP8 failure are different model/backend paths." + }, + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "MPS histogram int error in MoE and Gemma4 PLE device mismatch under auto device map are unrelated." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "glm-4v-9b loading failure and Molmo missing all_tied_weights_keys are different model-specific bugs." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Qwen2.5-VL import failure and bitsandbytes _is_hf_initialized incompatibility are unrelated import/runtime issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 4, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43366", + "issue:43404", + "issue:43638", + "issue:43742", + "issue:43782", + "issue:43792", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43940", + "issue:43950", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44560", + "issue:44661", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45356", + "issue:45399", + "issue:45538", + "issue:45588", + "issue:45823", + "issue:45902", + "issue:45923" + ], + "soft_pairs": [ + "issue:42491|issue:44368", + "issue:45020|issue:45823", + "issue:44661|issue:45020", + "issue:44991|issue:45399", + "issue:43828|issue:45823", + "issue:43782|issue:44960", + "issue:45081|issue:45399", + "issue:43827|issue:44509", + "issue:45020|issue:45356", + "issue:43404|issue:43950", + "issue:43742|issue:45399", + "issue:42915|issue:45020", + "issue:43638|issue:45823", + "issue:43792|issue:44220", + "issue:43257|issue:43866", + "issue:43872|issue:44291", + "issue:42915|issue:43872", + "issue:43940|issue:45313", + "issue:41553|issue:44991", + "issue:43854|issue:44863", + "issue:36331|issue:41093", + "issue:42947|issue:45923", + "issue:44928|issue:45588", + "issue:43257|issue:45305", + "issue:43872|issue:45538", + "issue:41093|issue:44805", + "issue:44479|issue:44560", + "issue:41553|issue:44488", + "issue:44530|issue:44863", + "issue:24643|issue:30064", + "issue:43366|issue:43828", + "issue:45325|issue:45902" + ], + "prepared_review_unit_hash": "638f77f16c5f7f72ccd49b051a4be4af64de18db50dd4b876c493dbbd700171b", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37683, + "estimated_input_tokens": 9421, + "estimated_eval_tokens": 19098 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose collection of unrelated model-loading, tokenizer, and training regressions. I do not see any duplicate pairs that look like the same underlying bug/change, so all soft edges are rejected.", + "confidence": 0.44, + "canonical_issue_reason": "issue:45020 is the broadest, most central regression report in the set: it describes recent transformers versions breaking remote_code models, which is the closest thing here to a reusable umbrella issue.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best single issue to anchor the cluster because it is broad, recent, and likely to attract related loading/regression reports; however, the surrounding issues are too heterogeneous to treat as true duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Both mention Qwen3.x LoRA/tie-word-embeddings behavior, but one is an hf4.x vs hf5.x compatibility break and the other is a warning about missing tie_word_embeddings=False; not the same bug." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "One is a broad remote_code loading regression; the other is a Gemma4 PLE device_map/forward device mismatch. Different code paths and failure modes." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "add-new-model-like tokenizer mapping failure is a registry/helper bug, while 45020 is a generic remote_code loading regression." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading for EMBEDDIA/est-roberta is unrelated to the flash-attn fallback gating issue." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "torch.autocast dtype mismatch in Phi-tiny-MoE is a runtime precision issue, not a device-map placement bug." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only=True load error and GLM5 are different model-specific loading bugs." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Mistral tokenizer regex patch crash is a tokenizer backend attribute bug, not flash-attn fallback behavior." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": false, + "reason": "These are both docs about removed pipeline tasks, but they are distinct documentation issues, not one underlying bug." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "remote_code loading regressions and Kimi-K2.5 tokenizer codec/fix_mistral_regex problems are separate tokenizer/model-loading failures." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "lm_head tying in Mistral3 is a model weight-tying bug; non-persistent buffer corruption in from_pretrained is a different serialization regression." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "MobileLLM key error and flash-attn fallback gating are unrelated loading paths." + }, + { + "left": "issue:42915", + "right": "issue:45020", + "accept": false, + "reason": "Qwen3 MOE FineGrainedFP8Config failure is a model/config-specific precision bug, not the broad remote_code loading regression." + }, + { + "left": "issue:43638", + "right": "issue:45823", + "accept": false, + "reason": "DeepSpeed ZeRO-3 index error on non-pretrained Bert and Gemma4 PLE device mismatch are different runtime issues." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Whisper-large-v2 runtime failure and _torch_extract_fbank_features() are too vague to tie as the same underlying defect." + }, + { + "left": "issue:43257", + "right": "issue:43866", + "accept": false, + "reason": "Qwen3 MOE weights not converting under accelerate+deepspeed is unrelated to Ovis2 checkpoint corruption." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": false, + "reason": "bitsandbytes _is_hf_initialized incompatibility and init_empty_weights TypeError are the same broad symptom class but different triggering APIs and contexts; not safe to merge as one bug." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Qwen3 MOE FP8 loading and bitsandbytes Int8Params incompatibility differ in model, library, and failure mechanism." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Qwen3-Next ZeRO-3 all params missing and Qwen3.5 ZeRO-3 language_model loading are related to deepspeed, but they target different model families and failure sites." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Bad AutoTokenizer error for Voxtral and est-roberta tokenizer loading failures are separate model-specific tokenizer issues." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "GLM-4.7-Flash unit-test load failure and NemotronH checkpoint loading are different model implementations with different load paths." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Unexpected num_items_in_batch in CustomTrainer.compute_loss is a trainer API mismatch; mask shape IndexError is a data/tensor indexing bug." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Gradient checkpointing ineffective with PEFT LoRA and missing supports_gradient_checkpointing are related to checkpointing, but one is a behavioral bug and the other is a missing model capability flag." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "RLHF gradient explosion in Qwen3.5 due to 3D position_ids is unrelated to flash_attention.py crashing on s_aux=None." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed and GAS loss averaging with DeepSpeed are different training-time bugs." + }, + { + "left": "issue:43872", + "right": "issue:45538", + "accept": false, + "reason": "bitsandbytes Int8Params constructor incompatibility is unrelated to CLIPTokenizer model_max_length." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": false, + "reason": "These are both mask/tensor shape IndexErrors, but the reports do not indicate the same code path or model; too generic to merge." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "v5.3.0 video input regression across multiple Qwen variants and Qwen3-vl embedding StopIteration are likely separate video-processing failures." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Voxtral AutoTokenizer error and cjvt/sleng-bert loading failure are different tokenizer/model compatibility problems." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "PagedAttentionCache invalid group type on Qwen3.5 and NemotronH checkpoint loading do not share the same underlying bug." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "DeepSpeed weight 2-D training error and void segmentation map processing are entirely unrelated domains." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "GGUF gpt-oss support and Phi-tiny-MoE autocast dtype mismatch are unrelated feature/support and runtime precision issues." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Qwen2.5-VL rope_index temporal scaling and Qwen3_5MoeTextRotaryEmbedding uninitialized inv_freq after to_empty are different rotary/vision materialization bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 5, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43720", + "issue:43824", + "issue:43854", + "issue:43931", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44155", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:45030", + "issue:45276", + "issue:45313", + "issue:45357", + "issue:45584", + "issue:45663", + "issue:45823", + "issue:45834" + ], + "soft_pairs": [ + "issue:45663|issue:45823", + "issue:43366|issue:43531", + "issue:43950|issue:45834", + "issue:43550|issue:44910", + "issue:43366|issue:44589", + "issue:43645|issue:43646", + "issue:44560|issue:44918", + "issue:42915|issue:43957", + "issue:43824|issue:43957", + "issue:42617|issue:43366", + "issue:36010|issue:42915", + "issue:44661|issue:45823", + "issue:38175|issue:43994", + "issue:43632|issue:45834", + "issue:44560|issue:44805", + "issue:44155|issue:45584", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:44661|issue:44805", + "issue:44877|issue:45030", + "issue:43317|issue:44756", + "issue:38175|issue:39692", + "issue:42491|issue:43720", + "issue:44960|issue:44977", + "issue:45313|issue:45357", + "issue:44451|issue:44488", + "issue:41720|issue:42915", + "issue:43644|issue:44534", + "issue:42491|issue:45276", + "issue:43708|issue:44075", + "issue:42915|issue:44756", + "issue:43854|issue:43931" + ], + "prepared_review_unit_hash": "5cfc545504ecc305b1438ffa965bf448f3a8e4e0f0013ed9e44680b4256efffd", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37195, + "estimated_input_tokens": 9299, + "estimated_eval_tokens": 18854 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mixed set of Transformers v5.x regression reports. It contains a few clear duplicate pairs around non-persistent buffers, custom model initialization, SigLIP2 output corruption, and BERT model-loading, while the Zero-3 + BertModel initialization reports are related but not strong enough to call duplicates based on the available pairwise evidence.", + "confidence": 0.74, + "canonical_issue_reason": "Issue 43950 is the strongest canonical issue: it is a precise Transformers 5.x regression about silent corruption of non-persistent buffers, and it cleanly anchors the strongest duplicate pair (44534).", + "canonical_pr_reason": null, + "best_issue_reason": "43950 is the best representative issue for the cluster because it describes a concrete core regression rather than a broad symptom list, making it the most actionable and central report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different failures: FSDP2 KeyError in text attention vs device-mismatch during PLE forward." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Both mention Qwen3-related models, but one is GGUF/gpt-oss support and the other is a sliding_window bug in Qwen3-MoE." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Unrelated bugs: non-persistent buffer corruption vs Kosmos2.5 OCR long-input index error." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different code paths: Bamba torch.compile with SDPA vs Qwen3.5 flash-attention misreading 3D position_ids." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "GGUF architecture support and Float8 storage lookup are separate loading/runtime issues." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Very likely the same custom-model initialization regression; 43645 is the notebook-specific manifestation of the broader 43646 report." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Both involve Qwen models, but one is a video StopIteration bug and the other is TRL SFT embedding unpacking failure." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Different failures: FineGrainedFP8Config on Qwen3-MoE vs torch.device('meta') loading breakage." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "ImportError for Qwen2.5-VL export is not the same as meta-device model loading failures." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py runtime failure and GGUF gpt-oss support are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "ImportError for GenerationMixin and Qwen3-MoE FP8 issues are distinct bugs." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "Tokenizer-mapping failure when adding a model is unrelated to Gemma4 device mismatch during forward." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Same underlying SigLIP2 regression: corrupted/zero outputs for google/siglip2-base-patch16-224 in model inference." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "_is_hf_initialized flag regression and Kosmos2.5 long-OCR index error are different problems." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Both are index/error reports, but they concern different models and different failure mechanisms." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "AudioFlamingo3 batched leakage and Whisper empty-transcription failure are unrelated." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Zero-3 Bert initialization failure is not the same as tokenizer-mapping-name handling." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Deepspeed Zero-3 BertModel index error is unrelated to add-new-model-like tokenizer mapping logic." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping regression and tensor mask shape mismatch are different bugs." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Both are config-loading regressions, but the affected validation paths and models are not clearly the same bug." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "device_map/offload loading failure and mmap OOM on Strix Halo are not the same underlying issue." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "SigLIP2 zero probabilities and the doc-example model/processor mismatch plus quantization failure are related only by model family." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Qwen3-MoE LoRA compatibility and BitNet packed-weight unpacking are separate loading issues." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 and Qwen3.5 flash-attention generation failures are unrelated model-specific bugs." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "ZeRO-3 language_model weight loading and incorrect visual encoder save_pretrained keys are different regressions." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": true, + "reason": "These look like duplicate reports of the same BERT model-loading regression; 44488 explicitly says the current version also fails on another model." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Different Gemma4 failures: shared_kv_states KeyError vs device mismatch during forward." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Auto device-mapping cudaErrorAssert on A800 is not the same as FineGrainedFP8Config failure." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Clear duplicate regression: both report Transformers 5.x corrupting non-persistent buffers with junk data." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "Qwen3_MoE LoRA compatibility and Gemma4 resize_token_embeddings behavior are unrelated." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Checkpoint resume max_steps miscalculation is unrelated to SGD optimizer args being ignored." + }, + { + "left": "issue:43931", + "right": "issue:45357", + "accept": false, + "reason": "Weight-shape mismatch when loading Qwen3-VL and bad visual-encoder key saving are different issues." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary overstates evidence by calling the Zero-3 + BertModel init failures a set of true duplicates, but the listed soft-edge verdicts for those issues are all negative. Please keep that part conservative and describe them as related, not duplicates. The rest of the summary is mostly grounded, but this contradiction is enough to reject." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 6, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:42491", + "issue:42947", + "issue:43257", + "issue:43335", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43644", + "issue:43645", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44488", + "issue:44756", + "issue:44910", + "issue:44912", + "issue:44960", + "issue:44991", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:46050" + ], + "soft_pairs": [ + "issue:45488|issue:45812", + "issue:39692|issue:43994", + "issue:43257|issue:43854", + "issue:43856|issue:43873", + "issue:42491|issue:43931", + "issue:43644|issue:43645", + "issue:43856|issue:44387", + "issue:44910|issue:45588", + "issue:42491|issue:45292", + "issue:45362|issue:45406", + "issue:44488|issue:44991", + "issue:36010|issue:39404", + "issue:45276|issue:45335", + "issue:44164|issue:44291", + "issue:43404|issue:43454", + "issue:43425|issue:43957", + "issue:44402|issue:44488", + "issue:39692|issue:44960", + "issue:42947|issue:45704", + "issue:45663|issue:45759", + "issue:43335|issue:45663", + "issue:45663|issue:45859", + "issue:43873|issue:44387", + "issue:44155|issue:45823", + "issue:44291|issue:44756", + "issue:44960|issue:46050", + "issue:44292|issue:44912", + "issue:43854|issue:45216", + "issue:44387|issue:44756", + "issue:45276|issue:45292", + "issue:45684|issue:45698" + ], + "prepared_review_unit_hash": "d8ffb025d82003159022f7e3c8bab3fce68bf0944bb455829ebfa454f260d45a", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36586, + "estimated_input_tokens": 9147, + "estimated_eval_tokens": 18550 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These issues are broadly related to Transformers 5.x regressions, but they span different models and code paths (tokenizers, save/load, quantization, offload, FSDP, chat/serve, etc.). None of the soft pairs is a clear same-bug duplicate, so the cluster should not be collapsed aggressively.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 43644 is the strongest umbrella report: it describes a broad Transformers 5.0.0 regression with system-wide impact, making it the best anchor for this mostly regression-themed cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "43644 is the most general and representative issue in the set, with a broad failure mode and higher likelihood of subsuming adjacent regressions than the more model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they hit different model families and different mechanisms: LlamaTokenizer pre-tokenizer override vs Granite token ID generation." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is a docs/example error with quantization issues while the other is a runtime loading problem in tests; not the same bug." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Different models and paths: Qwen3 MoE weight conversion under accelerate/deepspeed vs GLM-4.7-Flash loading in unit tests." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "One is Qwen3 MoE training memory inefficiency; the other is offloading not working with quantization. Similar theme, different failure mode." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "Different problems: hf4.x LoRA compatibility with qwen3_moe vs a weight-shape mismatch when loading a Qwen3-VL model." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Both are Transformers 5 regressions, but one is junk non-persistent buffers and the other is Jupyter custom-model initialization behavior." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training inefficiency and the other is int4 quantization reserved-memory growth/OOM." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Different flash-attention failures: Qwen3.5 position_ids misinterpretation vs an AttributeError on s_aux=None in sink-less models." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "LoRA/qwen3_moe hf4.x compatibility and resize_token_embeddings/output_embeddings are unrelated bugs." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Different model/feature failures: Qwen3.5 chat crash vs Gemma4Processor missing _tokenizer in transformers serve." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading reports, but they concern different models and different observed failures." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "Completely different issues: an import error for GenerationMixin vs Whisper pipeline return_language behavior." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Similar API symptom, but the reports are model-specific (Gemma4 vs t5gemma) and do not clearly share the same root cause." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "One is save/from_pretrained extra_state handling; the other is init_empty_weights breaking due to an unexpected constructor argument." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both involve untied lm_head weights, but they target different model implementations and are not clearly the same underlying code-path bug." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Environment compatibility vs model loading with meta tensors; these are separate regressions." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Different tokenizer/model problems: vocab-size mismatch for bert-base-bg vs inability to load sleng-bert." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Unrelated models and symptoms: SigLIP2 docs/example issues vs GLM5 model loading." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with PEFT LoRA and T5 apex.FusedRMSNorm memory leak are different bugs." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Gemma-4 FSDP2 shared_kv_states KeyError vs Qwen 3.5/3.6 composite config unwrapping are distinct code paths." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation and Gemma-4 FSDP2 attention state rebuilding are unrelated." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Different model families and mechanisms: Gemma-4 FSDP2 attention state bug vs Qwen3_5MoeTextRotaryEmbedding CPU-offload incompatibility." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Both concern quantization/offloading and memory, but one is offloading behavior and the other is a reserved-memory regression." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "AudioFlamingo3 batched inference token leakage and Gemma4 PLE device mismatch are unrelated model bugs." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "init_empty_weights argument handling vs disabling mmap to avoid OOM are different failure modes." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "GLM5 loading and sinq quantized save/load breakage are separate issues." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Different quantization reports on different models: Qwen-3-8B-NVFP4 error vs git-oss-20b MXFP4 fallback to bf16." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Different model versions and operations: GLM-4.7-Flash loading in tests vs Qwen3.5 save_pretrained regression." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Both end in OOM, but one is a quantization-memory regression and the other is a Strix Halo mmap issue." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": false, + "reason": "The API symptom overlaps, but the model scope differs and there is no clear evidence they share the same root cause." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Both involve custom-model save/load, but one is read-only permission propagation and the other is loading the wrong custom module after save_pretrained." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 7, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36032", + "issue:39404", + "issue:41720", + "issue:42947", + "issue:43257", + "issue:43299", + "issue:43381", + "issue:43425", + "issue:43452", + "issue:43596", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44155", + "issue:44164", + "issue:44265", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44928", + "issue:45084", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45701", + "issue:45759", + "issue:45823" + ], + "soft_pairs": [ + "issue:43381|issue:44928", + "issue:42947|issue:43381", + "issue:44164|issue:44756", + "issue:44479|issue:44877", + "issue:41720|issue:44155", + "issue:45588|issue:45663", + "issue:43957|issue:44756", + "issue:43646|issue:43708", + "issue:45292|issue:45335", + "issue:43828|issue:43957", + "issue:44402|issue:44451", + "issue:43645|issue:44164", + "issue:43425|issue:43828", + "issue:43257|issue:44863", + "issue:44683|issue:45823", + "issue:44589|issue:44756", + "issue:43425|issue:45823", + "issue:43425|issue:43716", + "issue:39404|issue:45701", + "issue:43906|issue:44530", + "issue:43611|issue:43645", + "issue:44530|issue:45084", + "issue:43901|issue:44509", + "issue:44265|issue:44898", + "issue:39404|issue:44912", + "issue:44479|issue:44805", + "issue:43425|issue:44918", + "issue:44589|issue:45084", + "issue:43299|issue:43931", + "issue:43596|issue:44683", + "issue:43994|issue:44079", + "issue:44164|issue:45759", + "issue:36032|issue:43452" + ], + "prepared_review_unit_hash": "c84eeb287b688ca1f668aaced5bd48e696a426b50d6b5f9285d6457cd245cb0f", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37722, + "estimated_input_tokens": 9431, + "estimated_eval_tokens": 19118 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated issue reports. The only clear duplicate pair is the resize_token_embeddings regression affecting output/decoder embeddings; the Qwen3-VL and other loading/runtime pairs are similar in theme but not clearly the same concrete bug.", + "confidence": 0.69, + "canonical_issue_reason": "Issue 45292 is the broader, cleaner statement of the resize_token_embeddings bug; 45335 is a model-specific manifestation of the same missing update to downstream embeddings.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45292 is the best representative duplicate target because it describes the underlying API regression in general terms and is the most reusable canonical report for that bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Both mention training/runtime instability, but one is about gradient checkpointing in eval mode and the other is a Qwen3.5 RLHF NaN/fallback issue; different code paths and symptoms." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both involve gradient checkpointing, but one is about LoRA/PEFT not being effective and the other is about eval-mode restrictions; not the same bug." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "save/from_pretrained extra_state handling and disabling mmap for Strix Halo OOM are unrelated failure modes." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Different subsystems: a video-input regression for Qwen VL models versus a strict config-loading issue for granite_speech." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Qwen3 auto device mapping CUDA assert and AudioFlamingo3 track leakage are unrelated model-specific bugs." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Flash-attention sinkless-model crash and Gemma4 FSDP2 shared_kv_states KeyError are different code paths." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading failures and mmap OOM behavior are not the same underlying issue." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model init failures and Trainer resume/max_steps calculation are unrelated." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Both describe the same resize_token_embeddings regression: the resized embeddings are not propagated to downstream/output embeddings, with 45335 being the T5Gemma-specific case." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Autocast dtype mismatch and meta-device loading failures are distinct problems." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer vocab-size mismatch and a model-specific load failure are too different to merge as duplicates." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Jupyter custom-model initialization and extra_state save/load handling are different bugs." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "A Torch 2.10 compatibility report is broader and different from a specific dtype mismatch under autocast." + }, + { + "left": "issue:43257", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3 MOE weight conversion with accelerate/deepspeed and NemotronH checkpoint loading are different model/load paths." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device-mismatch during forward are unrelated." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Float8 storage deserialization and Strix Halo mmap OOM are not the same bug." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "Torch version incompatibility and Gemma4 device_map mismatch are separate issues." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "General Torch compatibility and Mistral-3 image-preprocessor dtype mismatch are different failures." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper pipeline return_language and tokenization changes are unrelated." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "The isolated reproduction of another issue and the Qwen3.5 PagedAttentionCache crash are not the same concrete bug." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "base_model_prefix loading breakage and notebook custom-model initialization are different failure surfaces." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention failure and non-template-node compile error are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs-related, but they concern different APIs/tasks and are not duplicates." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export with torch_compilable_check and Perceiver non-default-resolution failure are distinct." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper pipeline behavior and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Video-input regression for Qwen models and a mask-shape IndexError are different bugs." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch compatibility and Qwen3.5 embedding unpacking with TRL SFT trainer are not the same issue." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup failure and compile-time non-template-node error are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both involve Qwen3-VL loading, but one is a version-wide MoE loading regression and the other is a specific weight-shape mismatch; not enough to confirm the same bug." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Zero3 init IndexError with BertModel and compiled flex_attention failures are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "SigLIP2 bad outputs and ModelOutput key assignment bugs are different code paths." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "extra_state save/load and AutoModelForCausalLM.from_config text_config unwrapping are unrelated." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5Tokenizer add_special_tokens conflict and gguf_file load path failures are different issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about the soft-edge pairs. The only clearly supported duplicate is 45292/45335, and the reasoning matches the issue titles. No PR-mergeability concerns apply here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 8, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:32090", + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42915", + "issue:43262", + "issue:43317", + "issue:43377", + "issue:43421", + "issue:43452", + "issue:43482", + "issue:43541", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43856", + "issue:43883", + "issue:44164", + "issue:44206", + "issue:44220", + "issue:44451", + "issue:44464", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44661", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45507", + "issue:45584", + "issue:45859", + "issue:45987" + ], + "soft_pairs": [ + "issue:39404|issue:44671", + "issue:43856|issue:44756", + "issue:44991|issue:45081", + "issue:44671|issue:44991", + "issue:44737|issue:45859", + "issue:32090|issue:39404", + "issue:44164|issue:44479", + "issue:44661|issue:44683", + "issue:44488|issue:45081", + "issue:43582|issue:43632", + "issue:43611|issue:43644", + "issue:40990|issue:43421", + "issue:43716|issue:45084", + "issue:44206|issue:44220", + "issue:44488|issue:44987", + "issue:43643|issue:44530", + "issue:36032|issue:42915", + "issue:34567|issue:40444", + "issue:44811|issue:45584", + "issue:43883|issue:44488", + "issue:44451|issue:44987", + "issue:43541|issue:45084", + "issue:43452|issue:43482", + "issue:44464|issue:45987", + "issue:42915|issue:44530", + "issue:43644|issue:43646", + "issue:43317|issue:43482", + "issue:44849|issue:45325", + "issue:36032|issue:43645", + "issue:45084|issue:45507", + "issue:43262|issue:43377" + ], + "prepared_review_unit_hash": "7ccfce1e8308c630c09e85b4477a6e6dd10330864cccd7056869ba2c97ffa629", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36947, + "estimated_input_tokens": 9237, + "estimated_eval_tokens": 18730 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These issues are mostly unrelated model- or feature-specific regressions. I did not find any clear PRs, and nearly all soft pairs are too broad or touch different code paths to be treated as duplicates. The only plausible near-duplicate is the GGUF loading regression pair.", + "confidence": 0.44, + "canonical_issue_reason": "Issue 43644 is the broadest concrete regression in the set and is centered on core Transformers v5 initialization/state handling, making it the most reasonable anchor issue among these disparate reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43644 is the best representative because it describes a clear, upstream v5 regression with a core runtime symptom rather than a model-specific edge case.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline language return vs CamemBERT MLM predictions are different models and different failures." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "MoE training memory usage and Strix Halo mmap OOM are unrelated causes and code paths." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-load regressions, but they involve different models and different tokenizer fix paths." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "Incorrect masked-LM predictions is not the same bug as tokenizer loading failure." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "Both mention device/offload behavior, but XLNet positional encoding and Qwen3.5 MoE rotary embedding are distinct implementations." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast TypeError and Whisper return_language failure are unrelated." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "Save/from_pretrained extra_state handling and video input regression affect different subsystems and symptoms." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Tokenizer mapping change failure and compiled flex_attention failure are unrelated issues." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both are loading problems, but they concern different models and different tokenizer-loading logic." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError is unrelated to the _is_hf_initialized flag regression." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "base_model_prefix loading breakage and non-persistent buffer junk are separate v5 regressions." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "High perplexity on a model and tokenizer post-processor update behavior are not the same bug." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Image preprocessor dtype mismatch and template compilation error are unrelated." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Both are audio-feature related, but the titles point to different failing functions and no shared concrete bug is evident." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Different model-loading failures; same broad theme is not enough to merge." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "AutoConfig trust_remote_code field loss and PagedAttentionCache linear_attention crash are unrelated." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer special-token conflict and Qwen3Moe FP8 failure are different code paths." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "TrainerState token counting and multi-image iterable dataset finetuning are unrelated." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Whisper batch_decode skip_special_tokens handling and empty-transcription generation after align_special_tokens are related only at a high level, not the same concrete bug." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Molmo tied-weights attribute error and BERT model loading failure are unrelated." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Different model-loading regressions with different affected models and likely different root causes." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Mixtral torch-dynamo grouped_mm tracing failure is unrelated to the template-node compile error." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": true, + "reason": "Both reports are about GGUF loading breaking in transformers v5, and the titles suggest the same gguf_file/from_pretrained code path." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked generation with compiled forward and StaticCache seq-length return type are different bugs." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Qwen3Moe FP8 training failure and Qwen3.5 PagedAttentionCache crash are separate issues." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 init regressions, but one is about buffer initialization and the other about custom model initialization; the concrete bug is not the same." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "gpu+cpu offload dequantized-model loading and GGUF loading failure are not the same specific bug." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3.5 hidden-states output bug and Qwen2.5-VL RoPE position-id scaling are unrelated." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer special-token conflict and Jupyter custom-model initialization are different failure modes." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Template compilation error and GraniteMoEHybrid invalid method call are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio processor sample-rate default and MIMI encoder padding-mask behavior are different bugs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is conservative and grounded in the packet. The issue pair verdicts generally match the titles/scope, and the summary correctly avoids overclaiming by treating the GGUF pair as only a plausible near-duplicate rather than a definite match. No PR pair evidence is present in the packet, so the lack of PR conclusions is appropriate." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 9, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36032", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43089", + "issue:43296", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43540", + "issue:43611", + "issue:43632", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43792", + "issue:43819", + "issue:43883", + "issue:43994", + "issue:44062", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45084", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45542", + "issue:45987" + ], + "soft_pairs": [ + "issue:43089|issue:45987", + "issue:43883|issue:44991", + "issue:43404|issue:44466", + "issue:44987|issue:44991", + "issue:43994|issue:44610", + "issue:43646|issue:44534", + "issue:43322|issue:43540", + "issue:44403|issue:45081", + "issue:43377|issue:43819", + "issue:42915|issue:45084", + "issue:44265|issue:44610", + "issue:43645|issue:44534", + "issue:44220|issue:44683", + "issue:44451|issue:45081", + "issue:44987|issue:45081", + "issue:45216|issue:45357", + "issue:43404|issue:45127", + "issue:42915|issue:44291", + "issue:43632|issue:44756", + "issue:36032|issue:43742", + "issue:44062|issue:44987", + "issue:42915|issue:44589", + "issue:43296|issue:43531", + "issue:43531|issue:44877", + "issue:43425|issue:44403", + "issue:43611|issue:44534", + "issue:44479|issue:45325", + "issue:36032|issue:41720", + "issue:42175|issue:45542", + "issue:43792|issue:44589", + "issue:40990|issue:44811", + "issue:44683|issue:45020", + "issue:44403|issue:44991" + ], + "prepared_review_unit_hash": "574dfc8082600ec93f5f9db9b5d228bbd78ddeb261a0411167f8339b86835b2d", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37507, + "estimated_input_tokens": 9377, + "estimated_eval_tokens": 19010 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of Transformers v5-era regression reports; only the Qwen3.5 save_pretrained pair looks like the same underlying bug. The rest are superficially similar but target different models, code paths, or failure modes.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43646 is the best canonical issue: it is a concrete, broadly relevant Transformers 5.0.0 regression about custom model initialization, and its wording matches several nearby init/loading breakage reports better than the more model-specific tickets.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is also the best standalone issue to keep because it is specific enough to be actionable while still serving as the cleanest anchor for closely related v5 initialization regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Different symptoms and code paths: generation performance overhead vs a StaticCache return-type contract bug." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Unrelated failures: missing all_tied_weights_keys on Molmo vs tokenizer loading for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Both mention tied weights, but one is a model-specific untied lm_head bug and the other is save_pretrained serialization varying by device." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Different affected models and likely different regressions; both are loading failures but not the same bug." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Different problem classes: nonsensical model outputs vs an image processor size mismatch." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 regressions, but one is custom model initialization and the other is non-persistent buffer corruption; not the same underlying bug." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Different multimodal failures affecting different model families and different stages of loading/processing." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Generic loading noise issue vs a specific Mistral tokenizer regex crash." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Different audio-model correctness bugs: batching/padding-mask mismatch vs DAC latent conversion mismatch." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "FP8 config failure on Qwen3Moe is unrelated to the template-node compilation error." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "torch.export/exportability bug vs processor output-size mismatch; different layers and failure modes." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Notebook-specific custom model init regression is not the same as junk-filled non-persistent buffers." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio feature extraction bug vs compiled flex_attention failure on newer torch." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different models and different tokenizer/config loading failures." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Different regressions: model loading for one checkpoint vs a Mistral tokenizer regex patch crash." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Same underlying Qwen3.5 save_pretrained regression; 45357 looks like a narrower follow-up about incorrect visual encoder keys." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both involve weights, but one is a missing tie in a specific model and the other is LoRA merge collapse after vocabulary extension." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Both are model-loading problems, but FP8 config failure is unrelated to the init_empty_weights/_is_hf_initialized argument bug." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "Different root causes: transformers v5 flag handling vs mmap-related OOM avoidance on Strix Halo." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Tokenizer constructor conflict vs a KeyError when loading a different model." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "Generic AddedToken constructor issue is not the same as the physical-intelligence/fast loading failure." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Both mention fp8/floating-point storage, but one is a config compatibility issue and the other is a missing storage type error." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "Different model families and different failure modes: PaddleOCR-VL load error vs Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Sliding-window regression in Qwen3-MoE is unrelated to strict config blocking granite_speech loading." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch version incompatibility is not the same as the transformer-loading noise report." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Base-model-prefix loading breakage and junk non-persistent buffers are distinct v5 regressions." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Both touch Qwen2.5-VL/Qwen3-VL media position handling, but one is a broad video-input regression and the other is a specific still-image temporal position_ids scaling bug." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer load-time AttributeError and Qwen3 auto device mapping cudaErrorAssert are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "These are both backend/dependency complaints, but one is about missing TensorFlow in a torch install and the other is tensorboard installing without TensorFlow." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Different issues: Whisper runtime failure vs missing Float8 storage type." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity quality issue is unrelated to Whisper batch_decode skipping special-token handling." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Compiled flex_attention on new torch and remote_code loading breakages are different bugs." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Unnecessary loading noise is not the same as tokenizer load failure for EMBEDDIA/est-roberta." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analysis stays grounded in the packet and the soft-edge calls are conservative. The selected canonical issue, issue:43646, is a reasonable generic anchor for the v5 custom-init regressions, and the only accepted pair (issue:45216 / issue:45357) is well supported by the titles. I don\u2019t see evidence of overstatement." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 10, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:33666", + "issue:40444", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43381", + "issue:43421", + "issue:43454", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43749", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44462", + "issue:44466", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44829", + "issue:44898", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45663", + "issue:45812", + "issue:45834", + "issue:45923" + ], + "soft_pairs": [ + "issue:43421|issue:44568", + "issue:43632|issue:44534", + "issue:41553|issue:43582", + "issue:43818|issue:45357", + "issue:33666|issue:40444", + "issue:44079|issue:45663", + "issue:44462|issue:45812", + "issue:43482|issue:45538", + "issue:43317|issue:43856", + "issue:44829|issue:45479", + "issue:43381|issue:45923", + "issue:40444|issue:45325", + "issue:44291|issue:44589", + "issue:43482|issue:43525", + "issue:43377|issue:45491", + "issue:41720|issue:45081", + "issue:42175|issue:43296", + "issue:43531|issue:44291", + "issue:44265|issue:45072", + "issue:43994|issue:44898", + "issue:43632|issue:45072", + "issue:43611|issue:43646", + "issue:43792|issue:43957", + "issue:44079|issue:44534", + "issue:45325|issue:45834", + "issue:43454|issue:44466", + "issue:42915|issue:43482", + "issue:40444|issue:45381", + "issue:45325|issue:45381", + "issue:43644|issue:44079", + "issue:43749|issue:44079" + ], + "prepared_review_unit_hash": "aab414a1ce923cd7436cebde1783c959009b49ee4c223e762bfb66fe19ca4567", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37278, + "estimated_input_tokens": 9320, + "estimated_eval_tokens": 18896 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly distinct bugs/regressions across different model families and v5 migration paths. A few are superficially similar (especially model-loading/init regressions), but none are clearly the same concrete issue based on titles alone, so I would not merge any soft pairs.", + "confidence": 0.39, + "canonical_issue_reason": "Issue 43611 is the clearest representative of the only plausible duplicate family in this set: a Transformers v5 model-loading regression. Its title is specific enough to anchor triage, while the nearby init/loading reports are broader or refer to different failure modes.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43611 is the best single issue to keep as the representative anchor because it names a concrete v5 loading breakage and is more specific than the neighboring generic initialization reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Both involve tokenizer/special-token behavior, but one is about runtime post-processor updates and the other about add_special_tokens BOS/EOS insertion for a specific tokenizer; different bugs and code paths." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one concerns `_is_hf_initialized` during model init and the other non-persistent buffer contents; distinct underlying failures." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Unrelated problems: bad Voxtral error message versus an Apple Silicon TypeError in caching allocator warmup." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Different model families and failures: Video-LLaVA weight/temporal-attention issue versus Qwen3.5 save_pretrained visual-key regression." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both are Qwen2-VL training issues, but one is multi-GPU training and the other is IterableDataset/multi-image finetuning failure; not the same bug." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment is unrelated to Gemma4 FSDP2 shared_kv_states rebuilding." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Both touch AutoTokenizer, but one ignores repository tokenizer.json while the other returns wrong token IDs for Granite models; different symptoms and likely different fixes." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Qwen2.5-GGUF loading failure and CLIPTokenizer model_max_length are unrelated." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Device-map/offload loading failure is not the same as Qwen3 MoE memory inefficiency during training." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Both are degenerate-training complaints, but one is FA3-specific classification behavior and the other is a single-label/num_labels=1 loss bug across models." + }, + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "Gradient checkpointing in eval mode is different from a missing supports_gradient_checkpointing flag on a remote-code model." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen2.5-VL, but one is finetuning with multiple images per prompt and the other is still-image temporal position scaling; not the same code-path." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Unexpected `_is_hf_initialized` argument during init-empty-weights loading is unrelated to missing Float8 storage registration." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Qwen2.5-GGUF loading and missing `pad_token_id` on Llama4Config are different issues." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "MIMI encoder padding-mask batching bug and Gemma3 mixed-length sliding-window NaNs are different model/code paths." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Both are Qwen-related failures, but auto device mapping cudaErrorAssert and Mistral tokenizer regex patch crashes are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "TensorFlow-backend packaging omission is unrelated to PaddleOCR-VL loading in vLLM." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and init_empty_weights `_is_hf_initialized` argument mismatch are separate regressions." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export failures from torch_compilable_check are unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Siglip2 nonsensical outputs with AutoModel/pipeline and Perceiver non-default resolution failure are different vision-model bugs." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "`_is_hf_initialized` flag regression and bfloat16 dtype mismatch are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 model-init/loading regressions, but one is specifically about `base_model_prefix` during loading and the other about custom model initialization; too broad to treat as the same bug." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Whisper-large-v2 failing to run and meta-device loading breakage are different issues." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer junk are separate regression areas." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Qwen2.5-VL position-id scaling bug and Kosmos2.5 OCR long-input index error are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "AyaVision lm_head tying bug and inconsistent lm_head.weight serialization are not the same problem." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and Qwen2.5-GGUF loading failure involve different models and mechanisms." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "IterableDataset multi-image finetuning failure is not the same as video vision_position_ids being wrong." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL positional issues, but one is still-image rope scaling and the other is video input vision_position_ids; different scenarios and likely different fixes." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Non-persistent buffer junk and ModelOutput key assignment are unrelated regressions." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP_CPU_RAM_EFFICIENT_LOADING and ModelOutput key assignment do not share the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 11, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42947", + "issue:43257", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43701", + "issue:43720", + "issue:43746", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44008", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44387", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44568", + "issue:44610", + "issue:44617", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45325", + "issue:45357", + "issue:45584", + "issue:45701", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45920", + "issue:45923" + ], + "soft_pairs": [ + "issue:43296|issue:45800", + "issue:44479|issue:45823", + "issue:35707|issue:40444", + "issue:44291|issue:44877", + "issue:44387|issue:44617", + "issue:43606|issue:44265", + "issue:43881|issue:44488", + "issue:44568|issue:45920", + "issue:43701|issue:45325", + "issue:44186|issue:44265", + "issue:43606|issue:43819", + "issue:41720|issue:45823", + "issue:43454|issue:43550", + "issue:43262|issue:43746", + "issue:43377|issue:44610", + "issue:43317|issue:43873", + "issue:39692|issue:40444", + "issue:43454|issue:45072", + "issue:43572|issue:45759", + "issue:43377|issue:44898", + "issue:43257|issue:45923", + "issue:40990|issue:45584", + "issue:40990|issue:42947", + "issue:44466|issue:45357", + "issue:43454|issue:43720", + "issue:43526|issue:43883", + "issue:43873|issue:45701", + "issue:43296|issue:44877", + "issue:43881|issue:44991", + "issue:41720|issue:44977", + "issue:44008|issue:45823" + ], + "prepared_review_unit_hash": "e2886579934085600a8e169eb855384733b5cf985964ca668484253d74992523", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37323, + "estimated_input_tokens": 9331, + "estimated_eval_tokens": 18918 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is highly heterogeneous: most pairs only share a broad subsystem (tokenizers, offloading, multimodal loading, quantization) but not the same concrete bug. I would not merge any of the soft-edge pairs as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": "issue:43873 is the broadest, most central open report in the set, with relatively high discussion/inbound activity and a title that can plausibly anchor nearby offloading/quantization failures.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best overall issue anchor for this cluster because it is open, broadly scoped, and has the strongest signal of ongoing impact.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different failure classes: model/processor loading in vLLM vs a torch/transformers version incompatibility." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "One is a video-input regression in Qwen VL models; the other is a Gemma4 PLE device-mismatch forward bug." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to finetuning Qwen2.5-VL on multi-image iterable data." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Unexpected _is_hf_initialized handling is a constructor/init bug, while granite_speech loading is a strict-config issue." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but one is an int4 CUDA reserved-memory regression and the other is Sam3Video memory usage." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU offload device mismatch for bark-small is unrelated to torch.export failures around torch_compilable_check." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "These are different models and different loading problems; they do not point to the same bug." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "Tokenizer special-token insertion for one model is not the same concrete regression as wrong token IDs across multiple models." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint key mismatch and Qwen2.5-VL RoPE index scaling are unrelated code paths." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer padding/truncation crashes are unrelated to export-time torch_compilable_check failures." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Device-mismatch on offload and DAC from_latents/STE mismatch are different bugs in different models." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both are device-related, but Qwen3 cudaErrorAssert and Gemma4 PLE device mismatch are different failures." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "lm_head weight tying/garbage generation is unrelated to Bamba torch.compile SDPA behavior." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and GraniteSpeech adapter checkpoint loading are separate issues." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "Missing padding-mask support in MIMI encoding is unrelated to OmDet processor/model resolution mismatch." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both involve offload/quantization, but one is a load failure on gpu+cpu offload and the other is a broader quantization offloading complaint." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "SigLIP2 docs errors and multi-image Qwen2.5-VL finetuning failure are different problems." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Weight tying serialization and bfloat16 dtype mismatches are not the same bug." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "A missing pad_token_idx config field is unrelated to unwrapping text_config for composite multimodal configs." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Padding-mask batching bug in MIMI and Perceiver non-default-resolution failure are different model behaviors." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "MoE weight conversion with accelerate/deepspeed is unrelated to a missing supports_gradient_checkpointing flag." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on gpt-oss-20b and Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity regression and gradient-checkpointing ineffectiveness are different issues." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "lm_head serialization inconsistency is not the same as wrong visual-encoder keys in Qwen3.5 save_pretrained." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Weight tying causing bad generation is unrelated to packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "BEiT label reduction bug and missing all_tied_weights_keys attribute are unrelated." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Quantization/offloading failures do not match a tokenizer-version-change regression." + }, + { + "left": "issue:43296", + "right": "issue:44877", + "accept": false, + "reason": "PaddleOCR-VL load failure and strict granite_speech config loading are separate model/config issues." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Model loading failure and tokenizer loading failure for a different model are not the same bug." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3 auto device-map cuda assert and Qwen3.5 flash-attention generation failure are different concrete code paths." + }, + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3n forward variable collision and Gemma4 PLE device mismatch are unrelated failures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 12, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:36010", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43295", + "issue:43316", + "issue:43317", + "issue:43322", + "issue:43404", + "issue:43450", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43708", + "issue:43720", + "issue:43749", + "issue:43873", + "issue:43881", + "issue:44186", + "issue:44387", + "issue:44423", + "issue:44479", + "issue:44534", + "issue:44610", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44918", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45704", + "issue:45812", + "issue:46050" + ], + "soft_pairs": [ + "issue:43404|issue:45357", + "issue:34634|issue:35707", + "issue:42617|issue:42915", + "issue:35707|issue:41720", + "issue:43577|issue:43708", + "issue:45632|issue:45684", + "issue:44186|issue:45072", + "issue:44423|issue:45406", + "issue:44877|issue:44991", + "issue:43550|issue:43720", + "issue:36331|issue:42915", + "issue:43450|issue:45701", + "issue:43749|issue:44534", + "issue:36331|issue:42617", + "issue:44610|issue:44898", + "issue:43873|issue:44756", + "issue:44186|issue:44610", + "issue:43322|issue:46050", + "issue:44387|issue:45704", + "issue:44479|issue:44918", + "issue:43550|issue:43606", + "issue:42915|issue:43317", + "issue:43316|issue:44913", + "issue:41720|issue:43322", + "issue:43295|issue:45701", + "issue:40990|issue:45701", + "issue:41720|issue:45161", + "issue:45081|issue:45812", + "issue:41720|issue:42617", + "issue:43531|issue:43572", + "issue:43526|issue:43881", + "issue:36010|issue:42617" + ], + "prepared_review_unit_hash": "5e4bfdb7db60ddd9add9ed20d154201a755b2f56bfb9416b7061e5fadd364808", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37291, + "estimated_input_tokens": 9323, + "estimated_eval_tokens": 18902 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These issues are broadly heterogeneous; the soft links look like loose topical similarity rather than true duplicates, so no pair should be merged. If one issue must serve as the anchor, issue 41720 is the most central runtime/device-mapping failure among the candidates, but the cluster does not form a single duplicate family.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 41720 is the most connected item in the set and is a concrete runtime/device-mapping failure, but the overall cluster is too mixed to treat it as a true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 41720 is the best anchor only because it is relatively central and has a clear, reproducible failure mode; however, it should not be interpreted as a true duplicate representative of the whole set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Different bugs: tied lm_head weights in Mistral3 vs incorrect visual encoder keys in Qwen3.5 save_pretrained." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "Unrelated models and failures: BarkProcessor voice_preset vs progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Different failure modes: inability to run 3d_parallel.py vs Qwen3Moe with FineGrainedFP8Config." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Both are generation/runtime issues, but one is inputs_embeds/past_key_values and the other is Qwen3 auto device mapping on A800." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "Dtype propagation bug in Blip2 is unrelated to Trainer resume_from_checkpoint max_steps calculation." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache path collision for trust_remote_code is unrelated to read-only permissions propagation in save_pretrained." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer crash on NER inputs is a different code path from dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Both involve serve/multimodal crashes, but one is a str.to error in continuous batching and the other is a missing _tokenizer attribute in Gemma4Processor." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Strict config loading for granite_speech is unrelated to tokenizer loading regression for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Both touch model loading/compilation, but one is torch.compile+SDPA on Bamba and the other is packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Trainer compute_loss signature mismatch is unrelated to Qwen3Moe FineGrainedFP8Config failures." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video processor batching shape bug is unrelated to a broad tokenization behavior change." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP CPU RAM efficient loading is a loading/offload issue, while the other is non-persistent buffers being filled with junk." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Trainer API mismatch and 3d_parallel.py execution failure are different subsystems and bugs." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Processor output size mismatch for OmDet-Turbo is unrelated to Perceiver interpolation failure." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Quantization offloading problems are only loosely related; one is general offloading with quantization, the other is mmap OOM on Strix Halo." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer crash and processor resolution mismatch are different model components and failure paths." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Segmentation fault loading Llava Next models is unrelated to sinq quantized save/load breakage." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Int4 CUDA reserved-memory OOM is a quantization memory issue, while the other is a T5 RMSNorm memory leak through apex." + }, + { + "left": "issue:44479", + "right": "issue:44918", + "accept": false, + "reason": "Video input regression across Qwen VL variants is unrelated to unpacking Qwen3.5 input embeddings with TRL SFTTrainer." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Both are load/runtime failures, but one is torch.compile+SDPA on Bamba and the other is CPU offload device mismatch on Bark-small." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to dequantized model loading with gpu+cpu offload." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy is unrelated to GPTNeoX rotary_pct not persisting on reload." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Different symptoms and subsystems: auto device mapping cudaErrorAssert vs segmentation fault while loading Llava Next." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor.tokenizer regression in v4.57.5 is unrelated to a tokenization behavior change report." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity anomaly on gpt-oss-20b is not the same underlying bug as generic version-dependent tokenization changes." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Both mention Qwen3 MoE, but one is auto device mapping on A800 and the other is only-TP behavior; not the same bug." + }, + { + "left": "issue:45081", + "right": "issue:45812", + "accept": false, + "reason": "Mistral tokenizer regex patch crash and Granite tokenizer ID regression are distinct tokenizer bugs." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Qwen3 auto device mapping failure and inability to run 3d_parallel.py do not share the same concrete code path." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Sliding_window behavior in Qwen3-MoE is unrelated to missing pad_token_idx in StableLmConfig." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to glm-4v-9b loading failure." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "GenerationMixin import error and 3d_parallel.py execution failure are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 13, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:29159", + "issue:30064", + "issue:33666", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36032", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:43322", + "issue:43329", + "issue:43388", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43873", + "issue:43927", + "issue:44186", + "issue:44361", + "issue:44368", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44671", + "issue:44743", + "issue:44898", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45059", + "issue:45072", + "issue:45357", + "issue:45406", + "issue:45464", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45812" + ], + "soft_pairs": [ + "issue:43540|issue:43782", + "issue:44671|issue:44743", + "issue:43421|issue:45812", + "issue:44186|issue:44361", + "issue:45406|issue:45464", + "issue:29159|issue:36032", + "issue:30064|issue:34634", + "issue:44492|issue:44509", + "issue:43421|issue:45701", + "issue:43606|issue:43720", + "issue:43550|issue:43653", + "issue:42915|issue:43322", + "issue:43388|issue:45059", + "issue:43606|issue:43653", + "issue:44514|issue:44545", + "issue:43531|issue:45759", + "issue:34567|issue:35707", + "issue:43421|issue:43927", + "issue:44368|issue:45357", + "issue:44451|issue:44991", + "issue:42490|issue:43329", + "issue:41720|issue:43746", + "issue:44898|issue:45072", + "issue:43873|issue:45704", + "issue:44987|issue:45042", + "issue:43531|issue:44589", + "issue:33666|issue:35707" + ], + "prepared_review_unit_hash": "f118d158912b48c1e995609c7aafabbd2cee241aa3c13d9f194b43db65c92eb7", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "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": "This cluster is mostly a collection of separate tokenizer/model/processor bugs. The strongest anchor remains issue:45812 as a broad cross-model tokenizer regression. Several pairs are superficially similar, but the packet usually only shows shared themes or titles; there is not enough body-level evidence to confidently merge them as duplicates. The Qwen2_5_VLProcessor pair is especially close in title, but still needs the underlying failure details to be shown before calling it a duplicate.", + "confidence": 0.79, + "canonical_issue_reason": "issue:45812 is the broadest and most representative standalone issue here: it describes a cross-model AutoTokenizer mis-tokenization regression with clear user impact and a likely shared root cause.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 is the best issue anchor because it is concrete, high-impact, and general enough to represent the cluster\u2019s recurring tokenizer-regression theme.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Both are Qwen-related, but one is video-input processing and the other is from_pretrained weight_only loading. Same family, different bug." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "Unrelated code paths: CamemBERT MLM prediction regression vs Qwen3.5 recurrent-state reset in modular generation." + }, + { + "left": "issue:43421", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer-related, but one is runtime post-processor syncing after special-token edits and the other is wrong token IDs for Granite models. Not the same concrete bug." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizer failures for different models and inputs; no shared concrete failure path is evidenced here." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Both involve serving/inference, but one is a Gemma4Processor attribute bug and the other is a streaming chat/completions failure. Too different to merge." + }, + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Both mention tokenizers, but one is slow/fast consistency and the other is a T5 tokenizer load-time attribute conflict. Separate issues." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Different processor implementations and different failures: void segmentation maps vs Bark voice_preset handling." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs-ish, but one is a cache-strategy typo and the other is stale pipeline documentation. Not duplicate content." + }, + { + "left": "issue:43421", + "right": "issue:45701", + "accept": false, + "reason": "Tokenization is the broad theme, but one is special-token post-processor syncing and the other is a version-related tokenization change. Not the same bug." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both involve accelerate/offload/loading, but the concrete failures differ: CPU offload device mismatch vs packed-weight unpacking during loading." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Completely different subsystems: torch.compile/SDPA runtime failure vs tokenizer special-token registration and decode output." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Different model-loading/runtime problems with no shared code path: FineGrainedFP8Config failure vs Llava Next segmentation fault." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Different bug classes: distributed metrics collation vs SAM3 text/box behavior." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "Different root causes and code paths: offload device mismatch vs tokenizer special-token registration." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": false, + "reason": "The titles are nearly identical, but the packet does not show matching body-level failure details beyond the shared symptom and cross-reference signals. That is not enough evidence to merge them as the same crash." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Both are Qwen3/Qwen3-MoE related, but one is sliding_window behavior and the other is config unwrapping for composite multimodal models." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token-count tracking vs progressive generation with inputs_embeds/past_key_values are unrelated." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Different tokenizer state issues: runtime special-token updates not refreshing post-processing vs save/load losing custom token IDs." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "One is a warning about tied embeddings during LoRA fine-tuning; the other is incorrect visual encoder keys saved for Qwen3.5. Different failure mode." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Both are version-sensitive tokenizer loads, but they affect different model families and likely different compatibility problems." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Local-path import warning and multimodal token counting bug are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Different runtime problems: auto device mapping cuda assert vs PEFT adapter loading for GraniteSpeech." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Different model classes and failure paths: image-size interpolation issue vs bfloat16 dtype mismatch." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior and T5 apex RMSNorm memory leak are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Both are version regressions, but one is model loading and the other is image-processor torchvision dependency handling. Not the same bug." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Sliding-window behavior for Qwen3-MoE is unrelated to the Float8 storage deserialization error." + }, + { + "left": "issue:33666", + "right": "issue:35707", + "accept": false, + "reason": "Multi-GPU training and progressive generation are different workflows and bugs." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The overall grouping is mostly grounded, but one soft-edge verdict is too strong for the packet: the 44514/44545 note claims they have the \"same crash,\" which is not evidenced here beyond matching titles and cross-references. The summary also leans a bit too hard on certainty with \"only\" true duplicates. Please keep the duplicate call conservative unless the packet shows identical body-level failure details." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 14, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:39692", + "issue:41720", + "issue:42886", + "issue:43381", + "issue:43421", + "issue:43525", + "issue:43531", + "issue:43540", + "issue:43596", + "issue:43618", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43819", + "issue:43844", + "issue:43873", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44077", + "issue:44186", + "issue:44220", + "issue:44451", + "issue:44610", + "issue:44625", + "issue:44661", + "issue:44671", + "issue:44843", + "issue:44918", + "issue:44936", + "issue:45081", + "issue:45183", + "issue:45356", + "issue:45381", + "issue:45479", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45859", + "issue:45907", + "issue:45923", + "issue:46050" + ], + "soft_pairs": [ + "issue:45759|issue:45859", + "issue:44062|issue:44451", + "issue:39692|issue:44610", + "issue:45704|issue:45923", + "issue:45588|issue:45759", + "issue:45663|issue:45923", + "issue:44918|issue:44936", + "issue:42886|issue:44843", + "issue:43421|issue:43653", + "issue:43381|issue:43844", + "issue:43421|issue:44186", + "issue:43525|issue:43531", + "issue:45704|issue:45859", + "issue:34634|issue:41720", + "issue:43782|issue:45381", + "issue:45081|issue:45356", + "issue:39692|issue:43873", + "issue:45356|issue:45907", + "issue:44625|issue:45479", + "issue:43994|issue:44671", + "issue:43596|issue:44220", + "issue:44220|issue:44661", + "issue:43720|issue:43819", + "issue:30064|issue:36010", + "issue:43540|issue:44008", + "issue:43746|issue:46050", + "issue:43618|issue:43697", + "issue:43653|issue:43720", + "issue:45183|issue:45479", + "issue:44077|issue:44625" + ], + "prepared_review_unit_hash": "7a089531b5a5e5426d83f7f05de0f62a35740aa4365e2e34f52dcfe78269af27", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37049, + "estimated_input_tokens": 9263, + "estimated_eval_tokens": 18782 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly a set of unrelated Transformers bugs/regressions across tokenizers, multimodal models, quantization/offload, and training code. I did not find any soft pairs that look like the same underlying defect.", + "confidence": 0.44, + "canonical_issue_reason": "issue:43873 is the strongest standalone issue in this set: it is active, has multiple references/discussion, and describes a concrete high-impact runtime regression around quantization/offloading.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is also the best issue to keep as the representative bug report because it is broad, reproducible, and directly actionable for maintainers.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Different bugs: one is text_config unwrapping for composite Qwen multimodal configs, the other is CPU-offload incompatibility in Qwen3_5MoeTextRotaryEmbedding." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "Both are tokenizer-loading failures, but the root causes differ: AddedToken keyword duplication vs a model-specific load failure for ScandiBERT." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Different models and different failures: SigLIP2 example/model mismatch plus quantization issues vs OmDet-Turbo processor size mismatch." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "Unrelated bugs: a T5 apex RMSNorm memory leak vs a missing supports_gradient_checkpointing flag on a trust_remote_code model." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Different code paths: flash_attention s_aux=None crash vs Qwen composite config unwrapping." + }, + { + "left": "issue:45663", + "right": "issue:45923", + "accept": false, + "reason": "Gemma-4 FSDP2 KeyError in attention state handling is unrelated to a gradient-checkpointing flag omission." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Both involve trainers, but the failures are distinct: Qwen3.5 embedding unpacking in SFT vs evaluate() breaking after train()." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": false, + "reason": "Both are offline-mode related, but one is cache loading failure and the other is an unconditional model_info() call in _patch_mistral_regex; not the same concrete bug." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both touch special tokens, but one is runtime post-processor updates and the other is BigBird mask-token registration/decoding." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Eval-mode gradient checkpointing restriction and ZeRO-3 gradient explosion are different problems with different triggers." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "TokenizersBackend post-processor updates and LayoutLMv2 NER padding/truncation crashes are unrelated failures." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Different model/config bugs: missing pad_token_id on Llama4Config vs sliding_window behavior in Qwen3-MoE." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "Different root causes: apex FusedRMSNorm memory leak in T5 vs CPU-offload incompatibility in Qwen3_5MoeTextRotaryEmbedding." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset failure is unrelated to Qwen3 auto device-mapping cuda asserts." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen multimodal issues, but one is from_pretrained with weight_only=True and the other is incorrect video vision_position_ids; not the same bug." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both mention fix_mistral_regex, but one is a crash in _patch_mistral_regex while loading a tokenizer and the other is a Kimi-K2.5 codec/warning regression." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "SigLIP2 example/model mismatch and quantization failure is unrelated to the general offloading/quantization bug in issue:43873." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regression in Kimi-K2.5 is unrelated to list_repo_files failing to forward revision in CI." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 and single-label classification zero-loss behavior are different model/config bugs." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different models and different regressions: SigLIP2 nonsense outputs vs CamemBERT masked-LM prediction errors." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Deepspeed ZeRO-3 initialization index error in BertModel is unrelated to _torch_extract_fbank_features()." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio feature extraction and tokenizer-mapping-name handling are separate subsystems and failures." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "Packed-weight loading in BitNet and DAC.from_latents/STE mismatch are unrelated implementation bugs." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Image segmentation-map processing and a GenerationMixin import error do not share the same root cause." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Qwen3OmniMoe video-input processing failure is unrelated to Gemma3n's audio_mel_mask variable collision." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "Local checkpoint loading with PEFT adapters and sinq quantized save/load breakage are different code-path problems." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIPOutput attentions not being assigned and RTDetrV2 output changes in v5 are distinct regressions." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "BigBird special-token decode failure and BitNet packed-weight loading are unrelated bugs." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Wrong/misleading input validation errors are not the same as the zero-loss classification bug for num_labels=1." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "patchtsmixer post_init policy and Qwen3.5 num_labels propagation are unrelated issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 15, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:30064", + "issue:36331", + "issue:39692", + "issue:42915", + "issue:43295", + "issue:43335", + "issue:43388", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43577", + "issue:43643", + "issue:43653", + "issue:43716", + "issue:43746", + "issue:43756", + "issue:43782", + "issue:43818", + "issue:43819", + "issue:43825", + "issue:43873", + "issue:43950", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44448", + "issue:44485", + "issue:44488", + "issue:44509", + "issue:44561", + "issue:44671", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45381", + "issue:45584", + "issue:45834", + "issue:45987", + "issue:46032" + ], + "soft_pairs": [ + "issue:43388|issue:46032", + "issue:42915|issue:43746", + "issue:43335|issue:43756", + "issue:43716|issue:44008", + "issue:43295|issue:45834", + "issue:43550|issue:45072", + "issue:43550|issue:43819", + "issue:43746|issue:43873", + "issue:43950|issue:45987", + "issue:44186|issue:44898", + "issue:44062|issue:44488", + "issue:44843|issue:45081", + "issue:43653|issue:43819", + "issue:43531|issue:43643", + "issue:43643|issue:44877", + "issue:43540|issue:45381", + "issue:43577|issue:43818", + "issue:43782|issue:44977", + "issue:44448|issue:44671", + "issue:43421|issue:44811", + "issue:30064|issue:36331", + "issue:44561|issue:45020", + "issue:44821|issue:45084", + "issue:39692|issue:45834", + "issue:44485|issue:44509", + "issue:45584|issue:45834", + "issue:43825|issue:43994" + ], + "prepared_review_unit_hash": "dcc36a166ffc30ba8ffce8c037835a67b6c6fabf44c9ddf4e56909b5b382e330", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36161, + "estimated_input_tokens": 9041, + "estimated_eval_tokens": 18338 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a true duplicate cluster: the items span many unrelated regressions across tokenizers, vision, quantization, model loading, and training APIs. The soft pairs are mostly thematic similarities, not the same concrete bug or change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the broadest, most cluster-central report (remote_code regressions in recent transformers versions), so it is the best representative issue, but it does not actually cover the rest of the heterogeneous set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Different bugs: gather_for_metrics label truncation vs Mamba2 use_cache corruption." + }, + { + "left": "issue:42915", + "right": "issue:43746", + "accept": false, + "reason": "Different model families and failures: Qwen3Moe FP8 config vs GraniteSpeech PEFT checkpoint loading." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "Different config/model behaviors: sparse encoder layer creation vs Smollm3 RoPE layer drop." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Both are dtype/vision-adjacent, but one is image preprocessor mismatch and the other is Gemma3n audio mask name collision." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Different regressions: processor.tokenizer/image forwarding vs long OCR index error." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Different code paths: torch.compile + SDPA bug vs bfloat16 inference dtype mismatch." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Different failure modes: SDPA/compile issue vs DAC.from_latents vs forward mismatch." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Different root causes: PEFT local checkpoint loading vs quantization offloading behavior." + }, + { + "left": "issue:43950", + "right": "issue:45987", + "accept": false, + "reason": "Different API contracts: non-persistent buffer corruption vs StaticCache.get_seq_length return type." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "Different models and symptoms: LayoutLMv2 tokenizer crashes vs Perceiver image classification with non-default resolution." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "Different tokenizer issues: AddedToken special arg collision vs failure to load a specific model." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Different offline/regression bugs: unconditional model_info call vs Mistral regex patch crashing on tokenizer internals." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Different tokenizer/model areas: BigBirdTokenizer special-token registration vs DAC latent/forward mismatch." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Both involve loading/configuration, but one is sliding_window behavior in Qwen3-MoE and the other is missing fields from trust_remote_code AutoConfig." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Related to config loading, but not the same bug: missing fields in AutoConfig return vs strict granite_speech config rejection." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Different multimodal bugs: video input processing error in Qwen3OmniMoe vs incorrect vision_position_ids in Qwen2.5-VL." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Different model components: BLIP2 dtype loading issue vs Video-LLaVA missing temporal attention/shared weights." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Different generation failures in different models: Qwen3VL weight_only load error vs Qwen3.5 flash-attention generation issue." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Both are output-quality regressions, but Pegasus v4/v5 differences and CamemBERT masked LM predictions are separate model-specific bugs." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Related to tokenization post-processing, but one is runtime special-token updates and the other is Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Different domains and APIs: image processor void segmentation maps vs CustomTrainer compute_loss signature mismatch." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "Both mention remote_code/trust_remote_code regressions, but they are different concrete breakages in different code paths." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "Different loader/template problems: AutoImageProcessor URL loading vs template compilation error." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "Different problems: SigLIP2 doc/example issues and quantization failure vs Kosmos2.5 long OCR input index error." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "Different topic scope: GLM-5 RoPE implementation discussion vs docs still mentioning removed pipeline tasks." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Different multimodal failures: Whisper empty transcription after alignment vs Kosmos2.5 OCR input length indexing." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "Different issue types: pipeline translation-task messaging vs SigLIP2 model/pipeline nonsensical results." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 16, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36010", + "issue:42175", + "issue:43262", + "issue:43322", + "issue:43329", + "issue:43404", + "issue:43452", + "issue:43526", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43650", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43782", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:44060", + "issue:44186", + "issue:44368", + "issue:44442", + "issue:44479", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45406", + "issue:45459", + "issue:45538", + "issue:45600", + "issue:45685", + "issue:45701", + "issue:45803", + "issue:45834" + ], + "soft_pairs": [ + "issue:43322|issue:45685", + "issue:44479|issue:45834", + "issue:43606|issue:45072", + "issue:43825|issue:43827", + "issue:44186|issue:44442", + "issue:34634|issue:36010", + "issue:43577|issue:43582", + "issue:44805|issue:44936", + "issue:28218|issue:29159", + "issue:43452|issue:44991", + "issue:44610|issue:45072", + "issue:42175|issue:45081", + "issue:43720|issue:45072", + "issue:45081|issue:45406", + "issue:43404|issue:43782", + "issue:43582|issue:44857", + "issue:43650|issue:43906", + "issue:44671|issue:45701", + "issue:44060|issue:44368", + "issue:43262|issue:43819", + "issue:43526|issue:43697", + "issue:45084|issue:45538", + "issue:43742|issue:45020", + "issue:43653|issue:44912", + "issue:43329|issue:45600", + "issue:45459|issue:45803" + ], + "prepared_review_unit_hash": "da577da15239fb0ad6dbd0e62c03f3603704370e8d62ef302a82696a36f26caa", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35743, + "estimated_input_tokens": 8936, + "estimated_eval_tokens": 18128 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft-edge pair looks like a true duplicate/one-change match; most are only loosely related by subsystem or symptom. The cluster is heterogeneous, so there is no PR to choose and only a broad issue can serve as a weak anchor.", + "confidence": 0.73, + "canonical_issue_reason": "Issue 45701 is the broadest, most central report in this set (tokenization changes across versions) and has the strongest linkage/activity, so it works best as an umbrella issue if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "45701 is the most representative issue-level anchor because it describes a general regression class rather than a narrow model-specific failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Different failures: Llava Next loading causes a segfault, while mps histogram_mps is an Apple Silicon Int type error in a different code path." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Both are multimodal regressions, but one is a video-input regression in several Qwen VL models and the other is a long-OCR index error in Kosmos2.5." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Both mention loading/inference mismatches, but the concrete bugs differ: CPU offload device placement vs bf16 dtype mismatches for different models." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "One is a bad pipeline() error message, the other is stale docs referencing removed pipeline APIs; related theme, not the same bug/change." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer problems for different models: LayoutLMv2 NER/batching crash versus FastSpeech2ConformerTokenizer load failure." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "BarkProcessor voice_preset failure and GenerationMixin import breakage are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "One is an incorrect float32 dtype on load, the other is a TypeError in caching_allocator_warmup on Apple Silicon." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "A mask/index shape mismatch during model work is unrelated to trainer.evaluate() failing after train()." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": false, + "reason": "Both concern tokenizer behavior, but the reports are not clearly the same underlying bug; one is added-token spacing, the other is slow/fast inconsistency." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "Both involve loading tokenizers, but one is gguf_file handling and the other is a specific tokenizer failing on transformers>=5." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Completely different bugs: OmDet-Turbo input sizing versus dtype mismatch warnings." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "TensorFlow backend packaging and a Mistral tokenizer regex crash are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is a different failure mode than bf16 dtype mismatches." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Both are loader crashes, but they affect different components/models and different missing attributes/code paths." + }, + { + "left": "issue:43404", + "right": "issue:43782", + "accept": false, + "reason": "Both are weight/loading issues, but Mistral3 tied weights and Qwen3VL weight_only are distinct model-specific problems." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "AppleSilicon warmup TypeError and CUDA float16 AMP crashes are unrelated." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "A data request and a reproduction of another issue are not duplicate bug reports." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "CamemBERT masked-LM prediction regression and a generic tokenization-version-change report are not the same bug." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Related only by tied-weights warnings; the underlying Qwen3-Next warning and the Qwen3.5 LoRA warning are different model/config issues." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and DAC from_latents mismatch are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug and RTDetrV2 output differences are different model failures." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "TypeError compiling template nodes and CLIPTokenizer model_max_length are unrelated tokenizer issues." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "Loading a specific model key error and general remote_code breakage are not the same concrete issue." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Undefined helper names in a video-token branch and stale auto_mappings entries for removed configs are different maintenance bugs." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Both involve overly broad exception handling, but they affect different functions and error types; not the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 17, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:29159", + "issue:34567", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:42490", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43454", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43606", + "issue:43638", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43854", + "issue:43927", + "issue:44291", + "issue:44479", + "issue:44554", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44811", + "issue:44849", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:45042", + "issue:45356", + "issue:45357", + "issue:45488", + "issue:45701", + "issue:45759", + "issue:45859" + ], + "soft_pairs": [ + "issue:43638|issue:44936", + "issue:43644|issue:43749", + "issue:44589|issue:45042", + "issue:43746|issue:44811", + "issue:38175|issue:44610", + "issue:39692|issue:44912", + "issue:43819|issue:43927", + "issue:36010|issue:43645", + "issue:43854|issue:44977", + "issue:43550|issue:44912", + "issue:34634|issue:36331", + "issue:43526|issue:44554", + "issue:43720|issue:44610", + "issue:43454|issue:43577", + "issue:43749|issue:44291", + "issue:43531|issue:45042", + "issue:36010|issue:36331", + "issue:43257|issue:43329", + "issue:43262|issue:44811", + "issue:44568|issue:45356", + "issue:42490|issue:43257", + "issue:43643|issue:45759", + "issue:43329|issue:43450", + "issue:43329|issue:44479", + "issue:34567|issue:40990", + "issue:44568|issue:45488", + "issue:43606|issue:45859", + "issue:44849|issue:45701", + "issue:29159|issue:34567", + "issue:43697|issue:45357" + ], + "prepared_review_unit_hash": "d76d0ddfe96a9414c94a1be20d3628c84934d25112362862d4e3a465af49592d", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37063, + "estimated_input_tokens": 9266, + "estimated_eval_tokens": 18788 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: it mixes tokenizer regressions, trainer-state bugs, audio/video processor issues, model-loading failures, quantization problems, and unrelated runtime errors. I do not see any pair that is clearly the same underlying bug/change, so no soft edge should be merged.", + "confidence": 0.93, + "canonical_issue_reason": "No single issue is a safe canonical duplicate for the whole cluster; the broadest tokenization-regression report is issue 45701, but the cluster spans many unrelated subsystems, so it should not be treated as one duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45701 is the broadest and most cross-cutting report in the set, but it is still only suitable as a weak thematic anchor, not a true duplicate canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Different trainer failures: one is deepspeed zero3 with a non-pretrained Bert model, the other is evaluate() breaking after train()." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Both concern loading/state handling, but one is about non-persistent buffers in v5.0.0 and the other is FSDP CPU RAM efficient loading; different bugs and code paths." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated areas: a Float8 storage import error versus PIL image processors incorrectly requiring torchvision." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Different subsystems and symptoms: local PEFT checkpoint loading versus Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Both are vision-related, but one is zero probabilities in SigLIP2 and the other is a processor/model input-size mismatch in OmDet-Turbo." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "One is a SigLIP2 docs/example issue with model/processor mismatch and quantization failure; the other is MXFP4 loading fallback for gpt-oss-20b." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "Different model bugs: DAC forward/from_latents mismatch versus DiaConfig losing custom token IDs after save/load." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "Both involve transformers v5 breakage, but one is a GenerationMixin import error and the other is a custom-model initialization issue in Jupyter notebooks." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "Different models and failure modes: GLM-4.7-Flash loading in tests versus Qwen3.5 flash-attention generation problems." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "Unrelated model/runtime issues: Bamba-9B-v2 torch.compile SDPA failure versus gpt-oss-20b MXFP4 quantization fallback." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "Different components: BarkProcessor voice_preset versus a CustomTrainer.compute_loss signature error." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "No overlap in bug type: BeitImageProcessorFast label reduction versus an MPS attention correctness issue." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Different failure modes and code paths: BitNet packed-weight loading versus OmDet-Turbo image-size mismatch." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "Both are model-loading/configuration related, but one is lm_head tying for AyaVision and the other is dtype propagation for Blip2; not the same bug." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Both relate to loading, but FSDP RAM-efficient loading and init_empty_weights/_is_hf_initialized are separate regressions." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Completely different domains: Qwen3-MoE sliding window behavior versus PIL backend image processors requiring torchvision." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "ImportError for GenerationMixin and a CustomTrainer loss-signature TypeError are unrelated." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Different areas: accelerate+deepspeed weight conversion for Qwen3 MOE versus undefined functions in multimodal video token counting." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Both are audio processor issues, but one is apply_chat_template sampling-rate defaulting and the other is batch_decode skip_special_tokens being ignored." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different mechanisms (special tokens insertion vs codec/warning behavior)." + }, + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "A trailing-slash path warning is not the same bug as missing MOE weight conversion during accelerate+deepspeed loading." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Both involve config handling, but one is missing fields with trust_remote_code and the other is unwrap logic for composite multimodal text_config." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Both touch video processors, but one is a token-counting bug with undefined helpers and the other is batched-shape output from the processor." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Related only at a very high level; the video-token helper bug is not clearly the same as the broader v5.3.0 video-input regression." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token counting and high perplexity on WikiText are unrelated issues." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but they affect different tokenizer implementations and distinct failure mechanisms." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "Both mention CPU offload, but one is a Bark device mismatch and the other is a Qwen3_5Moe rotary-embedding compatibility issue; not mergeable as one fix." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Different symptoms: Qwen3.5 output_hidden_states regression versus a broad tokenization change across versions." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency and TrainerState token counting are unrelated." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "Different model regressions: RTDetrV2 output differences versus Qwen3.5 visual-encoder key saving issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 18, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:24643", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43404", + "issue:43452", + "issue:43493", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43646", + "issue:43696", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43906", + "issue:43927", + "issue:43931", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44560", + "issue:44737", + "issue:44821", + "issue:44898", + "issue:44912", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45137", + "issue:45381", + "issue:45405", + "issue:45538", + "issue:45588", + "issue:45823", + "issue:45907", + "issue:45987" + ], + "soft_pairs": [ + "issue:43257|issue:43931", + "issue:44291|issue:45042", + "issue:43452|issue:43582", + "issue:42915|issue:43742", + "issue:41720|issue:43540", + "issue:33357|issue:34567", + "issue:42915|issue:44821", + "issue:43927|issue:44912", + "issue:44987|issue:45538", + "issue:42915|issue:45538", + "issue:44479|issue:45381", + "issue:44987|issue:45907", + "issue:43540|issue:44479", + "issue:43317|issue:45538", + "issue:43540|issue:44560", + "issue:38175|issue:44898", + "issue:43577|issue:45072", + "issue:43404|issue:45987", + "issue:45137|issue:45405", + "issue:24643|issue:29159", + "issue:44737|issue:45588", + "issue:43742|issue:44291", + "issue:43257|issue:43756", + "issue:44062|issue:44991", + "issue:43540|issue:45823", + "issue:43257|issue:43696", + "issue:43906|issue:45381", + "issue:34567|issue:36331", + "issue:44062|issue:45081", + "issue:43493|issue:43994", + "issue:43646|issue:43749", + "issue:36010|issue:42175" + ], + "prepared_review_unit_hash": "59341eb4ec80f21001fbc787035abcac0d4a65ae029a8d4f7a704686697ad1f8", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37345, + "estimated_input_tokens": 9337, + "estimated_eval_tokens": 18930 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous, with two plausible duplicate families standing out: SigLIP2 output-correctness reports and Qwen video-input regression reports. Everything else appears to be different bugs touching different code paths.", + "confidence": 0.64, + "canonical_issue_reason": "Issue 44479 is the broadest concrete regression report and has the clearest duplicate-like companion (45381); it is the best single anchor if one issue must be chosen from this mixed set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44479 is the most representative actionable issue here because it describes a concrete, versioned regression with multiple affected models and a clear fix surface.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Different Qwen3-related loading failures: MOE weight conversion under accelerate+deepspeed vs Qwen3-VL shape mismatch." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated bugs: init_empty_weights argument handling vs PIL image processors incorrectly requiring torchvision." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "Tokenizer/model gguf_file loading breakage and Apple Silicon allocator warmup TypeError are different code paths." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different failures on different models: FineGrainedFP8Config on Qwen3Moe vs a key error loading MobileLLM-125M." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Both are Qwen-family failures, but one is a device-mapping assert on A800 and the other is a video-input ValueError in Qwen3OmniMoe; not the same bug." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "Completely different symptoms and subsystems: MacOS bus error on CLIP vs TrainerState token counter not updating." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3Moe FP8 loading and AutoImageProcessor URL loading are unrelated." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Config save/load custom token IDs and MXFP4 quantization fallback are different issues." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Tokenizer loading regression for a specific model and CLIPTokenizer max_length bug are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Different model families and failures; no shared code path." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Both describe the same Qwen video-input regression in 5.3.0; 45381 is a narrower report of the wrong vision_position_ids path covered by 44479." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "One is an isolated reproduction of issue 38071; the other is a Qwen2.5-VL video-position bug. Not the same underlying bug." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Both involve video inputs, but the models and reported failures differ (Qwen3OmniMoe ValueError vs broader Qwen2.5/Qwen3 video regression)." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Device-map/offload loading failure and CLIPTokenizer model_max_length bug are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different models and failure modes: SigLIP2 zero probabilities vs Perceiver resolution/interpolation failure." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "BLIP2 dtype propagation and SwitchTransformers/TimmWrapperModel bfloat16 mismatch are different bugs." + }, + { + "left": "issue:43404", + "right": "issue:45987", + "accept": false, + "reason": "Mistral3 lm_head tying bug and StaticCache seq-length contract issue are separate code paths." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "DeepSpeed training weight-shape runtime error and tokenizer slow/fast inconsistency are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU tensor creation bug and flash-attention sink handling crash are unrelated." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Model loading key error vs init_empty_weights argument mismatch; different underlying problems." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MoE deepspeed weight-conversion bug and Smollm3 RoPE-layer drop are different issues." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "AddedToken constructor regression and tokenizer loading failure for EMBIDDIA/est-roberta are not the same bug." + }, + { + "left": "issue:43540", + "right": "issue:45823", + "accept": false, + "reason": "Different model families and failure modes: Qwen3OmniMoe video ValueError vs Gemma4 PLE device mismatch." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Qwen3Moe MOE conversion failure and GPT-oss OOM are unrelated." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "Reproduction of an unrelated upstream issue vs a Qwen2.5-VL video-position bug." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token-count update bug and custom compute_loss signature breakage are different training issues." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "AddedToken keyword conflict and Mistral tokenizer regex patch crash are different tokenizer bugs." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "Both point to the same SigLIP2 correctness problem on the same model family; one is a JAX-vs-HF discrepancy, the other reports clearly wrong outputs." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Transformers 5.0 custom initialization breakage and FSDP CPU RAM efficient loading are distinct issues." + }, + { + "left": "issue:36010", + "right": "issue:42175", + "accept": false, + "reason": "GenerationMixin import regression and missing TensorFlow backend in a torch install are unrelated packaging/import issues." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque error and MIN_PEFT_VERSION bump are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Model-loading regression in transformers>=5.1.0 and list_repo_files revision forwarding are different issues." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Same general area only superficially; the concrete failures and affected models differ." + }, + { + "left": "issue:43742", + "right": "issue:45538", + "accept": false, + "reason": "Loading key error for MobileLLM-125M and CLIPTokenizer max_length bug are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The two accepted duplicate families (Qwen video-input regression and SigLIP2 correctness reports) are plausibly same-bug pairs from the titles alone, and the remaining rejections are framed cautiously rather than overclaimed. No obvious mergeability or evidence overstatement issues stand out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 19, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28282", + "issue:29127", + "issue:34689", + "issue:41762", + "issue:42175", + "issue:43295", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43582", + "issue:43643", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43824", + "issue:44291", + "issue:44484", + "issue:44488", + "issue:44514", + "issue:44589", + "issue:44625", + "issue:44704", + "issue:44756", + "issue:44792", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44936", + "issue:44938", + "issue:44964", + "issue:45042", + "issue:45081", + "issue:45216", + "issue:45290", + "issue:45325", + "issue:45356", + "issue:45357", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45600", + "issue:45920" + ], + "soft_pairs": [ + "issue:43749|issue:44756", + "issue:42175|issue:43824", + "issue:44291|issue:44821", + "issue:45357|issue:45920", + "issue:44792|issue:44964", + "issue:43452|issue:44488", + "issue:43643|issue:45375", + "issue:45042|issue:45081", + "issue:28282|issue:34689", + "issue:43742|issue:44589", + "issue:43526|issue:45479", + "issue:43756|issue:45325", + "issue:45405|issue:45600", + "issue:43531|issue:44821", + "issue:43697|issue:45216", + "issue:28282|issue:29127", + "issue:43720|issue:44898", + "issue:44514|issue:45381", + "issue:43582|issue:43782", + "issue:41762|issue:44936", + "issue:44484|issue:44704", + "issue:42175|issue:44938", + "issue:44625|issue:44877", + "issue:43452|issue:43577", + "issue:43452|issue:44843", + "issue:44514|issue:45290", + "issue:43295|issue:45356" + ], + "prepared_review_unit_hash": "2d7cd37159edac5068f34adda1fe1045ee6d358868c02ddc392224a66cf9a2dc", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36229, + "estimated_input_tokens": 9058, + "estimated_eval_tokens": 18372 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly distinct Transformers bug reports clustered by broad theme (loading, tokenization, config, and vision/multimodal regressions). No soft-edge pair looks like the same concrete bug/change, so I\u2019d keep them separate. If forced to pick an umbrella representative, the broadest recent regression is the AutoTokenizer token-id issue.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45920 is the broadest issue in the set: it describes a v5 regression affecting AutoTokenizer across multiple model families, which makes it the best umbrella representative for the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the strongest global issue candidate because it is recent, regression-focused, and spans several affected models rather than a single narrow edge case.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Both are loading/memory-related, but one is FSDP CPU RAM efficient loading and the other is mmap/OOM on Strix Halo; different failure modes and code paths." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "TensorFlow backend packaging and a missing Qwen2.5-VL class import are unrelated issues; no shared underlying bug is evident." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "`init_empty_weights` with an unexpected argument and AutoImageProcessor URL loading are different APIs and different failures." + }, + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Qwen3.5 save_pretrained key mapping is a checkpoint-writing bug, while 45920 is tokenizer output regression; not the same code path." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Both involve multimodal model loading, but they affect different models and likely different loaders; too broad to merge." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading failures, but 43452 is about gguf_file handling in from_pretrained while 44488 is a specific model load failure with no concrete overlap shown." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both mention missing config fields, but one is a trust_remote_code/AutoConfig issue and the other is a strict-config field-drop regression in a specific Qwen config." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Image processor torchvision dependency and Mistral regex patch crash are unrelated subsystems and unrelated bugs." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "Generic import-error-from-missing-dependency and a Llama 3.2 Vision load regression are not the same underlying issue." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "A key error loading MobileLLM-125M and a Float8 storage lookup failure are unrelated loading errors." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "One is an image processor label-reduction bug; the other is a sequence-classification loss/num_labels bug. Same word 'labels' but different functionality." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both involve RoPE/position ids in vision models, but they target different model families and distinct computations; too speculative to unify." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "A dependency version bump and stale auto_mappings references are unrelated maintenance/CI issues." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and AutoImageProcessor URL loading are unrelated bugs." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Different regressions in different model flows: RTDetrV2 output drift vs Qwen3.5 save_pretrained checkpoint corruption." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "BitNet packed-weight loading and Perceiver interpolation failure are different model-specific bugs." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL, but one is chat template batching/padding and the other is video position-id scaling; not the same concrete bug." + }, + { + "left": "issue:43582", + "right": "issue:43782", + "accept": false, + "reason": "Apple Silicon caching allocator warmup TypeError and Qwen3VL weight_only loading error are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "DeepSpeed ZeRO-3 model loading IndexError and trainer.evaluate() after train() are different training/runtime failures." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "A save_pretrained shard-size question and missing kwargs in AutoProcessor/cached_file are not the same issue." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Backend packaging for torch install and Python 3.14 load failure are unrelated environment/setup reports." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Both touch config handling, but one is num_labels propagation in Qwen3.5 and the other is strict config blocking granite_speech loading." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "gguf_file loading breakage and BLIP2 dtype propagation are different model-loading problems with no shared code-path evidence." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "gguf_file handling and Mistral regex/offline model_info behavior are separate tokenizer-loading concerns." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both are apply_chat_template crashes, but one is batched padding=False and the other is assistant tool-call/no-content handling; likely separate edge cases, so too risky to merge." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "A regression around processor.tokenizer/images passing and a Kimi-K2.5 codec/mistral-regex regression are unrelated tokenizer issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 20, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:29159", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:34567", + "issue:35141", + "issue:38175", + "issue:41720", + "issue:43122", + "issue:43295", + "issue:43504", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43643", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43819", + "issue:43906", + "issue:43927", + "issue:44361", + "issue:44442", + "issue:44464", + "issue:44466", + "issue:44493", + "issue:44561", + "issue:44568", + "issue:44610", + "issue:44743", + "issue:44811", + "issue:44843", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45005", + "issue:45440", + "issue:45491", + "issue:45588", + "issue:45823", + "issue:46002", + "issue:46050" + ], + "soft_pairs": [ + "issue:44466|issue:45005", + "issue:24643|issue:29127", + "issue:43819|issue:44811", + "issue:44936|issue:45823", + "issue:30064|issue:33666", + "issue:38175|issue:43295", + "issue:43653|issue:44361", + "issue:44361|issue:44442", + "issue:43653|issue:44568", + "issue:43577|issue:43697", + "issue:44843|issue:44991", + "issue:44464|issue:44743", + "issue:43643|issue:44561", + "issue:41720|issue:43906", + "issue:44493|issue:46050", + "issue:29159|issue:33357", + "issue:45491|issue:46002", + "issue:43531|issue:43742", + "issue:43653|issue:43927", + "issue:44464|issue:44977", + "issue:43550|issue:43927", + "issue:43688|issue:45440", + "issue:43577|issue:45588", + "issue:43550|issue:44361", + "issue:32090|issue:34567", + "issue:43504|issue:43720", + "issue:43504|issue:43550", + "issue:43504|issue:44610", + "issue:44466|issue:46050", + "issue:28282|issue:35141", + "issue:43122|issue:43906" + ], + "prepared_review_unit_hash": "777a0e146fe27a4fbd8831ee590331fc21c4cf6032d8e84b13aef3408257f08d", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37100, + "estimated_input_tokens": 9275, + "estimated_eval_tokens": 18806 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is highly heterogeneous; the soft-similarity pairs mostly share only broad subsystem keywords (tokenizers, generation, loading, MoE, etc.) rather than the same concrete bug. I would not merge any of the proposed pairs as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 45005 is the clearest representative of the closest theme in this set: a concrete v5 tied-weights regression with a broader scope that can subsume related tied-weight reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44466 is the most concrete, reproducible report in the tied-weights group and has the clearest symptom description for triage.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Related to tied weights in v5, but one is a serialization/device-dependent symptom and the other is a broader translation-model report; not the same concrete bug." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "Completely different areas: DeepSpeed training/runtime error vs LayoutLMv3 box-validation messaging." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "Different models and failures: DAC latent conversion mismatch vs Whisper batch_decode skipping special tokens." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Unrelated Trainer lifecycle failure vs Gemma4 device-map mismatch during forward." + }, + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Different domains: segmentation-map processing vs Qwen2-VL multi-GPU training." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Unrelated symptom classes: zero probabilities in SigLIP2 vs processor/tokenizer regression in v5.7.5." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Both are tokenizer bugs, but they affect different tokenizers and different failure modes." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer implementations and errors; no shared code-path bug is evident." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both mention special tokens, but one is BigBird mask-token registration and the other is mDeBERTa BOS/EOS insertion; not the same bug." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Different problems: Blip2 dtype loading vs RTDetrV2 output drift across v5." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but the root causes and affected models differ." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Both involve generation/caching, but one is compiled chunked generation and the other is recurrent state reset in a specific modular model." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both touch trust_remote_code, but missing config fields and removal of is_torch_fx_available are different failures." + }, + { + "left": "issue:41720", + "right": "issue:43906", + "accept": false, + "reason": "The second item is only an isolated reproduction reference, and the underlying bug is not clearly the same as the Qwen3 device-mapping crash." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Different serialization/loading issues with different symptoms: unexpected position-id keys vs quantized-model save/load breakage." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer inconsistency vs a MacOS bus error in a CLIP model; not the same bug." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Both mention NaNs, but they arise in different models and different parts of the pipeline." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Unrelated: Qwen3-MoE sliding-window behavior vs MobileLLM loading key error." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "Both are token-related regressions, but one is a tokenizer special-token registration issue and the other is config save/load losing custom IDs." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Both concern generation, but the concrete failures differ: compiled chunked generation vs flash-attention behavior in Qwen3.5." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "Torch.compile/SDPA bug is unrelated to DiaConfig save/load token-ID loss." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Both involve MoE models, but one is auxiliary-loss normalization and the other is a broader remote-vs-native divergence report." + }, + { + "left": "issue:43577", + "right": "issue:45588", + "accept": false, + "reason": "Different code paths: Blip2 dtype loading vs flash_attention integration crashing on sink-less models." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Completely unrelated failures: torch.compile with SDPA vs tokenizer AttributeError." + }, + { + "left": "issue:32090", + "right": "issue:34567", + "accept": false, + "reason": "Trainer broadcast TypeError and TrainerState token counter not updating are separate issues." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Both are load-time bugs, but they affect different models and different mechanisms (legacy field vs packed-weight unpacking)." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Different model families and failure modes; no shared concrete bug." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Beit semantic-segmentation preset loading and OmDet-Turbo processor size mismatch are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "Both involve save/load or serialization, but the specific tied-weight/device regression is not the same as sinq quantization breakage." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "ImportError for missing PyTorch and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:43906", + "accept": false, + "reason": "The second item is a reproduction of another issue, but there is no evidence it is the same tokenization regression as 4.57.3-to-5.0 drift." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 21, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28282", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43262", + "issue:43295", + "issue:43334", + "issue:43452", + "issue:43493", + "issue:43525", + "issue:43540", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43688", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43825", + "issue:43927", + "issue:44361", + "issue:44451", + "issue:44464", + "issue:44479", + "issue:44488", + "issue:44554", + "issue:44568", + "issue:44625", + "issue:44821", + "issue:44843", + "issue:44849", + "issue:44913", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45357", + "issue:45405", + "issue:45704", + "issue:45803", + "issue:45812", + "issue:45823", + "issue:45987", + "issue:46050" + ], + "soft_pairs": [ + "issue:43742|issue:44821", + "issue:32090|issue:33666", + "issue:43452|issue:44451", + "issue:41720|issue:44464", + "issue:45823|issue:46050", + "issue:33666|issue:36010", + "issue:43493|issue:43825", + "issue:44849|issue:45357", + "issue:43606|issue:43927", + "issue:43525|issue:43572", + "issue:43720|issue:43927", + "issue:43756|issue:44913", + "issue:43262|issue:43295", + "issue:28282|issue:33357", + "issue:43688|issue:43749", + "issue:44568|issue:45812", + "issue:28282|issue:32090", + "issue:43720|issue:44361", + "issue:33666|issue:36331", + "issue:39692|issue:43493", + "issue:43577|issue:44554", + "issue:44964|issue:45405", + "issue:43540|issue:44977", + "issue:44479|issue:44625", + "issue:43688|issue:45704", + "issue:45803|issue:45987", + "issue:43334|issue:43572", + "issue:43756|issue:45083", + "issue:44488|issue:44843" + ], + "prepared_review_unit_hash": "230d7471aae9932714e99e75633a036b9557e4c41865278922953cb2e2f37f57", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36557, + "estimated_input_tokens": 9140, + "estimated_eval_tokens": 18536 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a heterogeneous mix of unrelated bug reports across loading, tokenizers, configs, multimodal models, training, and docs; none of the soft pairs look like the same underlying issue, so there is no clear canonical duplicate in the cluster.", + "confidence": 0.95, + "canonical_issue_reason": "No single issue is a credible canonical representative: the reports span many distinct subsystems and failure modes rather than one shared bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a strong global best candidate; any one would be an arbitrary pick because the cluster is not a duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "Different loading paths and failures: MobileLLM key error vs AutoImageProcessor URL loading." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Both involve training, but one is a `_gpu_broadcast_one` NoneType error and the other is Qwen2-VL multi-GPU training; not the same bug." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "Both are model-loading related, but one is `gguf_file` handling in auto classes and the other is a specific model that cannot load; too different." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Unrelated failures: device-mapping CUDA assert vs inconsistent chunked generation with compiled forward." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Different issues: Gemma4 device mismatch on `device_map=\"auto\"` vs quantized model save/load breakage." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Multi-GPU training support vs an import error for `GenerationMixin`; no shared underlying bug." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy vs pipeline error messaging for translation tasks; unrelated." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Both are Qwen3.5 regressions, but one is `output_hidden_states` behavior and the other is incorrect save_pretrained visual keys." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "Device mismatch during CPU offload vs DiaConfig token ID persistence; different code paths." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing padding-related config fields after v5, but they affect different model configs and are not clearly the same bug." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "BitNet packed-weight loading failure vs DiaConfig save/load token IDs; unrelated." + }, + { + "left": "issue:43756", + "right": "issue:44913", + "accept": false, + "reason": "Both involve RoPE/rotary settings, but one is a Smollm3 architecture bug and the other is GPTNeoX config reload behavior." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio chat-template sampling-rate default vs processor/tokenizer access regression; different functionality." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "PyTorch-missing ImportError vs MacOS bus error on a CLIP model; not the same issue." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Auxiliary-loss normalization bug vs FSDP CPU RAM efficient loading; different training internals." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different model families and concrete tokenization bugs." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "ImportError for missing PyTorch vs Trainer broadcast NoneType error; unrelated." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet accelerated loading bug vs MLukeTokenizer attribute error; no shared code-path." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training support vs custom trainer `compute_loss` signature change; distinct trainer issues." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "SigLIP2 docs/example problems vs a model implementation discrepancy; same model family but not the same bug." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Blip2 dtype propagation bug vs MPS attention correctness issue; unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model load failure vs dependency version bump; no common underlying defect." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3OmniMoe video-processing error vs Qwen3.5 flash-attention generation bug; different failure modes." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression across Qwen variants vs `num_labels` propagation into Qwen3.5 text config; unrelated." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Aux-loss normalization bug vs T5 apex RMSNorm memory leak; different model families and symptoms." + }, + { + "left": "issue:45803", + "right": "issue:45987", + "accept": false, + "reason": "Fuyu tensor conversion exception swallowing vs StaticCache return-type contract; unrelated." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-VL missing `pad_token_id` vs StableLmConfig missing `pad_token_idx`; similar shape, but different models and fields." + }, + { + "left": "issue:43756", + "right": "issue:45083", + "accept": false, + "reason": "Smollm3 RoPE-layer issue vs a helper length-calculation bug in qwen3_omni_moe; not the same bug." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Model loading failure for a specific BERT model vs offline-mode breakage from an unconditional hub call; different root causes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 22, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28218", + "issue:29942", + "issue:30333", + "issue:33357", + "issue:33666", + "issue:34567", + "issue:39692", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43493", + "issue:43504", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43824", + "issue:43844", + "issue:43927", + "issue:43994", + "issue:44112", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44464", + "issue:44466", + "issue:44493", + "issue:44521", + "issue:44589", + "issue:44625", + "issue:44734", + "issue:44792", + "issue:44964", + "issue:45200", + "issue:45229", + "issue:45290", + "issue:45405", + "issue:45520", + "issue:45584", + "issue:45588", + "issue:45701", + "issue:45759", + "issue:45823", + "issue:46050" + ], + "soft_pairs": [ + "issue:45200|issue:45229", + "issue:43493|issue:43844", + "issue:43334|issue:45759", + "issue:43824|issue:44589", + "issue:39692|issue:43295", + "issue:43257|issue:43299", + "issue:43646|issue:44466", + "issue:45229|issue:45823", + "issue:43572|issue:44291", + "issue:43278|issue:43844", + "issue:43504|issue:43606", + "issue:44589|issue:45520", + "issue:43688|issue:45701", + "issue:33357|issue:33666", + "issue:43927|issue:44361", + "issue:43504|issue:43653", + "issue:44423|issue:44734", + "issue:43606|issue:44112", + "issue:43575|issue:45405", + "issue:44792|issue:45405", + "issue:44625|issue:45759", + "issue:43994|issue:44493", + "issue:45588|issue:46050", + "issue:29942|issue:30333", + "issue:44464|issue:45584", + "issue:44521|issue:45290", + "issue:43844|issue:44964", + "issue:28218|issue:34567" + ], + "prepared_review_unit_hash": "a0221b611560d3ecb53fa8a51871e7a168fbf77f58489374c9126923a67ea54c", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36181, + "estimated_input_tokens": 9046, + "estimated_eval_tokens": 18348 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Loose similarity cluster of mostly unrelated Transformers issues; the soft pairs are thematic lookalikes, not true duplicates or mergeable PR equivalents.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:45759 is the most concrete, reproducible loading/config regression in the set and is a reasonable representative of the broader model-loading subset, though the overall cluster is not duplicate-tight.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Same family (Gemma 4) but different failures: missing mm_token_type_ids default vs multi-GPU CUDA OOM." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "Both relate to model correctness, but one is SigLIP2 implementation mismatch and the other is DeepSpeed/ZeRO gradient blow-up." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Both touch Qwen multimodal loading, but one is a missing pad_token_id attribute and the other is AutoModel.from_config not unwrapping text_config." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "ImportError for a missing class vs a storage-type lookup failure; different code paths and symptoms." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "Docs/example errors and a regression in processor/tokenizer API access are related only at a very high level." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both mention Qwen3 MoE, but one is accelerate+deepspeed conversion and the other is a dev-version loading regression for Qwen3VL." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Custom model initialization breakage is not the same bug as device-dependent lm_head.weight serialization." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Both are Gemma 4-related, but one is inference OOM and the other is a device mismatch in forward with device_map=auto." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Both involve init/load-time compatibility, but one is a missing pad_token_idx in StableLmConfig and the other is an unexpected _is_hf_initialized argument." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Different bugs: dtype drift between train/eval vs gradient normalization under HfDeepSpeedConfig." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy-field load failure in Beit semantic segmentation is unrelated to CPU offload device mismatch in bark-small." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Both are type/compatibility problems, but one is missing Float8 storage handling and the other is a Python 3.13 flash_attn import key error." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Aux-loss normalization and version-dependent tokenization are unrelated bugs." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "Both are multimodal/model-scale issues, but MacOS bus error on CLIP and Qwen2-VL multi-GPU training are different failure modes." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "Both mention tokenizer/config state, but DiaConfig token IDs and MLukeTokenizer AttributeError are distinct problems." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Loading a legacy BEiT field and BigBirdTokenizer special-token registration are not the same underlying bug." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Serving crashes with multimodal .to() handling are different from KV-cache continuation indexing bugs." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Both are CI/device-related, but one is a real offload mismatch and the other is a stale GraniteSpeech test." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "Tensor-parallel OOM is unrelated to a version pin bump for PEFT." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "A failing janus generate-images test is not the same issue as the unreleased PEFT version constraint." + }, + { + "left": "issue:44625", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen multimodal config propagation, but num_labels plumbing and text_config unwrapping are different API bugs." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 bad outputs and unexpected position-id keys are separate correctness problems." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "Flash-attention sinkless-model AttributeError and sinq save/load quantization breakage are unrelated." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "CI Flash Attention 2 failures and MLflow reporting after training failure are different subsystems and symptoms." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Chunked generation with compiled forward and Whisper empty-transcription failures have no concrete shared code path." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both are chat-template edge cases, but all-zero assistant masks and tool-call content crashes are distinct failures." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "DeepSpeed gradient growth and Phi-4 multimodal load failure are unrelated." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer spacing regression and TrainerState token counter not updating are unrelated trainer/tokenizer bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 23, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29159", + "issue:32090", + "issue:33357", + "issue:37428", + "issue:38175", + "issue:42490", + "issue:42915", + "issue:43232", + "issue:43278", + "issue:43295", + "issue:43334", + "issue:43450", + "issue:43493", + "issue:43525", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43582", + "issue:43688", + "issue:43696", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43957", + "issue:43975", + "issue:44315", + "issue:44466", + "issue:44554", + "issue:44610", + "issue:44849", + "issue:44898", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45084", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45325", + "issue:45356", + "issue:45412", + "issue:45446", + "issue:45520", + "issue:45823", + "issue:46050" + ], + "soft_pairs": [ + "issue:44554|issue:45446", + "issue:43688|issue:43873", + "issue:45084|issue:45520", + "issue:29159|issue:32090", + "issue:43540|issue:45200", + "issue:43582|issue:44977", + "issue:15354|issue:38175", + "issue:43232|issue:44849", + "issue:45200|issue:45823", + "issue:45072|issue:45198", + "issue:43873|issue:46050", + "issue:43278|issue:43701", + "issue:44315|issue:44849", + "issue:28218|issue:33357", + "issue:43295|issue:44610", + "issue:44554|issue:45356", + "issue:44898|issue:45412", + "issue:43572|issue:44987", + "issue:43493|issue:43701", + "issue:43572|issue:43957", + "issue:44466|issue:45127", + "issue:45200|issue:45325", + "issue:37428|issue:43957", + "issue:43295|issue:43450", + "issue:43550|issue:43844", + "issue:43334|issue:43525", + "issue:43295|issue:43873", + "issue:43824|issue:43975", + "issue:38175|issue:45412", + "issue:42490|issue:43696", + "issue:42915|issue:43824", + "issue:37428|issue:43824", + "issue:38175|issue:43493" + ], + "prepared_review_unit_hash": "4173defa13453ab17c1c3689fcd065a399600c22be3a186a555e326d91d57bc9", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37688, + "estimated_input_tokens": 9422, + "estimated_eval_tokens": 19100 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Loose similarity cluster of unrelated issue reports across tokenization, generation, quantization/offloading, multimodal processors, and import/runtime regressions. No soft pair looks like a safe duplicate/one-PR merge candidate.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the cluster is a broad topical bucket, not a duplicate set, so any representative would be misleading.", + "canonical_pr_reason": null, + "best_issue_reason": "If one issue must anchor the cluster, #43873 is the broadest and most active item in the closest quantization/offloading subgroup, but it is still not a true duplicate representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "Different bugs: MPS attention correctness vs a flex_attention import/version check. Same general area only, not the same code-path." + }, + { + "left": "issue:43688", + "right": "issue:43873", + "accept": false, + "reason": "Aux-loss normalization in specific MoE models is unrelated to offloading/quantization behavior. Different failure mode and fix." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "One is a compile/template-node error, the other a flash_attn import KeyError on Python 3.13. Unrelated runtime paths." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency and Trainer GPU broadcast NoneType error are separate subsystems and symptoms." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Both are multimodal-related, but one is a Qwen3OmniMoe video-processing error and the other is Gemma 4 text-only token-type defaults. Not the same bug." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation issues are unrelated." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "JIT scripting support for generator expressions vs SigLIP2 zero probabilities are different issues entirely." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Both involve generation/model outputs, but one is a sync_gpus kwargs update bug and the other is output_hidden_states handling in Qwen3.5. Different code paths." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different concrete defects: missing mm_token_type_ids defaults vs device placement mismatch with device_map=\"auto\"." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "dtype mismatch in inference is not the same as Wav2Vec2 save/tokenization failure. No shared concrete bug." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Quantization offloading failure and sinq save/load breakage are both quantization-adjacent, but they target different mechanisms and code paths." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype drift during eval and checkpoint key mismatch are unrelated training-state bugs." + }, + { + "left": "issue:44315", + "right": "issue:44849", + "accept": false, + "reason": "Liger Kernel application via model_init and Qwen3.5 hidden-state output behavior are separate issues." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer spacing after added tokens and a MacOS bus error loading CLIP are not the same defect." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Processor/tokenizer API regression and OmDet-Turbo image-size mismatch both affect processors, but the concrete bugs differ." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness and Kimi-K2.5 tokenizer codec regression are unrelated." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver interpolation failure and RT-DETR memory retention are different model-specific bugs." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig and a loading failure for a different model are separate regressions." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Both are model-loading regressions, but one is a missing config attribute and the other is a meta-device loading breakage. Not the same bug." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Serialization of tied lm_head weights and LoRA vocabulary-merging collapse are distinct weight-handling problems." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Different multimodal token/index bugs in different models; no shared concrete code-path." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "Flash-attention import failure and meta-device loading breakage are different import/loading regressions." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor.tokenizer regression and batched video-shape bug are separate processor issues." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "torch.compile/SDPA failure and ZeRO-3 gradient growth are unrelated training/runtime bugs." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Both mention missing pad_token_id on config objects, but they affect different model families and are not enough evidence for the same underlying fix." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor tokenizer regression and quantization offloading behavior are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "Import error for Qwen2.5-VL and DeepSeek detokenization regression are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 zero-probability outputs and RT-DETR memory release issues are completely different bugs." + }, + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Path-with-slash import warning and GPT-oss OOM are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and Qwen2.5-VL import error are different model-load issues." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but they involve different symbols and different failure modes." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Both concern SigLIP2, but one reports zero probabilities while the other is a broader HF-vs-JAX discrepancy; not enough to treat as the same concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 24, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:37428", + "issue:39692", + "issue:42490", + "issue:43262", + "issue:43278", + "issue:43335", + "issue:43388", + "issue:43425", + "issue:43450", + "issue:43493", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43856", + "issue:43901", + "issue:43931", + "issue:44112", + "issue:44206", + "issue:44351", + "issue:44360", + "issue:44442", + "issue:44448", + "issue:44493", + "issue:44514", + "issue:44521", + "issue:44561", + "issue:44610", + "issue:44743", + "issue:44792", + "issue:44857", + "issue:45083", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45245", + "issue:45290", + "issue:45381", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45823" + ], + "soft_pairs": [ + "issue:44514|issue:44521", + "issue:43653|issue:44112", + "issue:44442|issue:45198", + "issue:43278|issue:43856", + "issue:43450|issue:43575", + "issue:43827|issue:43901", + "issue:42490|issue:43493", + "issue:43262|issue:43450", + "issue:43819|issue:44112", + "issue:43819|issue:45083", + "issue:44610|issue:44857", + "issue:44206|issue:45584", + "issue:45127|issue:45245", + "issue:37428|issue:43425", + "issue:44792|issue:45563", + "issue:43335|issue:43550", + "issue:43606|issue:45083", + "issue:45290|issue:45381", + "issue:44857|issue:45593", + "issue:44448|issue:44743", + "issue:45200|issue:45588", + "issue:43278|issue:43825", + "issue:37428|issue:44351", + "issue:44206|issue:45823", + "issue:43611|issue:43931", + "issue:43335|issue:43653", + "issue:44360|issue:44561", + "issue:39692|issue:43526", + "issue:43550|issue:44112", + "issue:43388|issue:43526", + "issue:43819|issue:44493" + ], + "prepared_review_unit_hash": "a67bcd94fa0ca045b36cda4ce8fbf1f5866beeaa5538fa9619f3527aea79330c", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37069, + "estimated_input_tokens": 9268, + "estimated_eval_tokens": 18792 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is not a true duplicate cluster: the items span many unrelated bugs (chat templates, tokenizers, model loading, vision/audio processors, training/runtime regressions). None of the soft pairs look like the same underlying change or failure mode.", + "confidence": 0.95, + "canonical_issue_reason": "issue:44521 is the broadest and most active representative of the apply_chat_template / multimodal processor subgroup, but it does not actually subsume the other items as duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 is the best single representative issue because it has the most central chat-template/multimodal scope and the strongest discussion activity among the items, even though the cluster is not a duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Same general area (apply_chat_template) but different concrete failures: batched-padding crash vs assistant_mask all zeros." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "Completely unrelated: BigBirdTokenizer decode/special-token bug vs GraniteSpeech CI device-override test." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Both involve tokenizers, but the failures are model-specific and unrelated (FastSpeech2Conformer loading vs Wav2Vec2 save/tokenization)." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Different subsystems and symptoms: dtype drift in evaluation vs Qwen3 MoE memory usage." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video processor shape bug and tensor-parallel OOM are unrelated issues." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs-related, but one is about pipeline removals and the other about return_all_scores; not the same change." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "Local-path import warning and SigLIP2 HF-vs-JAX discrepancy are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate bug vs video batched-shape bug; different code paths." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC latent/STE mismatch is unrelated to a GraniteSpeech CI device-override failure." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "Different audio/model internals: DAC latent behavior vs qwen3_omni_moe feature-length helper." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Both are vision-related, but one is processor input sizing and the other is AMP/CUDA loss crashing." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Audio feature-extractor crash and Whisper empty-transcription failure are separate bugs." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge collapse and category-count runtime limit are unrelated." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "An import error for a flash-attention helper is not the same as a broad Torch 2.10 incompatibility report." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "A failing Janus image-generation test and a stale generate() warning are different problems." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "SwitchTransformers config bug and Bamba torch.compile/SDPA failure are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "CPU offload device mismatch does not match the qwen3_omni_moe helper-length issue." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Both touch multimodal/chat-template areas, but one is tool-call tokenization crash and the other is video vision_position_ids mismatch." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "AMP/CUDA crash in LwDetrImageLoss is not the same as missing auxiliary losses in D-FINE." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Pegasus v4/v5 output drift and Qwen3_5 recurrent-state reset are unrelated." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 mm_token_type_ids defaults and flash_attention s_aux=None crash are different code paths." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype regression and pipeline error-message docs issue are unrelated." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import-related, but they concern different missing symbols and different breakages." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "LasrFeatureExtractor center-arg crash and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "General base_model_prefix loading regression and a specific Qwen3-VL weight-shape mismatch are distinct failures." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Sparse-layer config bug and BigBirdTokenizer special-token bug are unrelated." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer ReLU discussion and removal of is_torch_fx_available are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "SigLIP2 docs example errors and BeitImageProcessorFast reduce_labels bug are different issues." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba and GraniteSpeech CI device-overrides are unrelated." + }, + { + "left": "issue:43388", + "right": "issue:44493", + "accept": false, + "reason": "Last-batch gather_for_metrics label trimming and unexpected position-id keys are not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 25, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28218", + "issue:29127", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:38175", + "issue:39692", + "issue:40444", + "issue:42490", + "issue:43335", + "issue:43412", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43701", + "issue:43720", + "issue:43854", + "issue:43927", + "issue:44112", + "issue:44292", + "issue:44448", + "issue:44462", + "issue:44493", + "issue:44561", + "issue:44568", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45083", + "issue:45198", + "issue:45229", + "issue:45310", + "issue:45372", + "issue:45459", + "issue:45478", + "issue:45491", + "issue:45520", + "issue:45563", + "issue:45588", + "issue:45698", + "issue:45712", + "issue:45803", + "issue:45859" + ], + "soft_pairs": [ + "issue:44855|issue:45083", + "issue:43335|issue:43504", + "issue:45310|issue:45478", + "issue:43335|issue:45083", + "issue:45588|issue:45859", + "issue:29942|issue:32090", + "issue:39692|issue:44493", + "issue:43720|issue:44112", + "issue:43335|issue:45072", + "issue:44448|issue:44779", + "issue:43335|issue:44112", + "issue:38175|issue:44493", + "issue:44292|issue:45520", + "issue:28218|issue:35141", + "issue:44977|issue:45229", + "issue:42490|issue:43854", + "issue:43606|issue:44857", + "issue:44561|issue:44779", + "issue:44462|issue:44561", + "issue:44857|issue:45072", + "issue:44448|issue:44977", + "issue:43701|issue:45083", + "issue:43550|issue:45083", + "issue:34689|issue:35141", + "issue:43927|issue:45698", + "issue:29127|issue:32090", + "issue:40444|issue:43526", + "issue:44568|issue:45491", + "issue:45372|issue:45588", + "issue:43412|issue:45588", + "issue:44977|issue:45563", + "issue:45198|issue:45803", + "issue:45459|issue:45712" + ], + "prepared_review_unit_hash": "739570aa18f35f07765d534acd5c0fada620074f66a84e6fcfef69abcb905a70", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37239, + "estimated_input_tokens": 9310, + "estimated_eval_tokens": 18876 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly noisy: the titles span many unrelated Transformers bugs. Only one clear duplicate pair stands out: the two Qwen3.5 MoE from_pretrained issues (45310/45478), which describe the same failure mode.", + "confidence": 0.92, + "canonical_issue_reason": "issue:45310 is the best representative because it is the earlier, more detailed report of the same Qwen3.5 MoE from_pretrained failure that 45478 repeats.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the strongest global issue candidate: it is a real duplicate target, specific enough to anchor the cluster, and has the fuller bug context.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Different problems: Python 3.13 import/torch.jit parsing vs a Qwen3 Omni helper output-length behavior issue." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "Unrelated models and code paths: SwitchTransformers config creation vs BEiT preset loading with a legacy field." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Same underlying bug: both report Qwen3.5 MoE from_pretrained failures on transformers>=5.4.0, with the later issue repeating the same error." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "Both are model/config bugs, but they affect different components and failure modes." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Both touch flash-attention/CPU-offload-adjacent areas, but one is an s_aux=None AttributeError and the other is a rotary embedding incompatibility." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Different failures: Flash Attention 2 test breakage vs Trainer/_gpu_broadcast_one NoneType type error." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "One is a SigLIP2 documentation example issue; the other is a broad position-id regression affecting many models." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "Different CI failures and root causes: unpacking packed weights during accelerate loading vs a stale device override test." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "Both are dtype-related in different models, but they are distinct bugs in separate code paths." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus v4/v5 output drift is not the same bug as Deepseek tokenizer regressions in v5." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "SwitchTransformers config creation and GraniteSpeech CI device override are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Different symptom classes: zero probabilities in SigLIP2 vs unexpected position-id keys across many models." + }, + { + "left": "issue:44292", + "right": "issue:45520", + "accept": false, + "reason": "One is a Qwen-3-8B-NVFP4 runtime problem; the other is a Python 3.13 flash_attn import_utils KeyError." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization after resize when tie_word_embedding is false." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Both are generation/inference problems, but they involve different models and distinct failure modes (flash-attention vs multi-GPU OOM)." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path import warning and GLM-4.7-Flash unit-test load failure are not the same issue." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch in Bark is unrelated to float16 AMP crashes in LwDetrImageLoss." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "A v5 API removal breaking trust_remote_code models is a different regression than Deepseek tokenizer behavior changes." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json and the is_torch_fx_available removal are separate loader/API issues." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "Both are numeric-precision problems, but they affect different models and different code paths." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus output differences and Qwen3.5 flash-attention generation failures are unrelated." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint key mismatch and Qwen3 Omni helper behavior are distinct issues." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Bamba SDPA torch.compile failure is unrelated to the Qwen3 Omni feature-length helper bug." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Llama 3.2 Vision model loading regression and output embedding reinitialization are separate bugs." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "DiaConfig save/load token-ID loss and wrong custom module loading after save_pretrained are different serialization bugs." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and Trainer broadcast type errors are unrelated." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Qwen2.5-VL iterable-dataset finetuning failure and BEiT reduce_labels returning the wrong shape/value are different issues." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "mDeBERTa special-token insertion regression is not the same as Gemma3 NaN embeddings with mixed-length batching." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Gemma 4 processor loading ImportError and flash_attention s_aux=None crash are different failure points." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export problems for gemma3n are unrelated to the flash_attention AttributeError on sink-less models." + }, + { + "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 bug." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save_pretrained/tokenization failure and a bare except swallowing errors in FuyuBatchFeature are unrelated." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer protobuf error masking and leftover dummy classes in dummy_pt_objects.py are different maintenance/failure issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the cluster is broad and the identical Qwen3.5 MoE from_pretrained titles (45310/45478) are the only clear duplicate pair. The soft-edge verdicts are generally conservative and do not overstate similarity beyond what the titles support." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 26, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:25251", + "issue:29127", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:34567", + "issue:35141", + "issue:36010", + "issue:37428", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43381", + "issue:43575", + "issue:43653", + "issue:43976", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44448", + "issue:44451", + "issue:44462", + "issue:44466", + "issue:44479", + "issue:44610", + "issue:44734", + "issue:44743", + "issue:44857", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45520", + "issue:45563", + "issue:45588", + "issue:45823" + ], + "soft_pairs": [ + "issue:36010|issue:37428", + "issue:29942|issue:35141", + "issue:42490|issue:43575", + "issue:44857|issue:45588", + "issue:43065|issue:45412", + "issue:45245|issue:45563", + "issue:45372|issue:45459", + "issue:44448|issue:44466", + "issue:45588|issue:45823", + "issue:43653|issue:44743", + "issue:43278|issue:43381", + "issue:43976|issue:45440", + "issue:43994|issue:45412", + "issue:25251|issue:32090", + "issue:44186|issue:45198", + "issue:43066|issue:44462", + "issue:44206|issue:44479", + "issue:44451|issue:45520", + "issue:42915|issue:42994", + "issue:44871|issue:44977", + "issue:44361|issue:45198", + "issue:43065|issue:43994", + "issue:44265|issue:45198", + "issue:33357|issue:39692", + "issue:29942|issue:30064", + "issue:44734|issue:45406", + "issue:29127|issue:34567", + "issue:44610|issue:45412" + ], + "prepared_review_unit_hash": "7db875d41f119fe783f590566776c35c55256c0df840cd2b5a5491fa58350443", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36128, + "estimated_input_tokens": 9032, + "estimated_eval_tokens": 18320 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a heterogeneous set of Transformers bug reports, not a tight duplicate cluster. The soft pairs mostly share only broad subsystem similarity (tokenizers, serving, vision models, flash-attn, quantization) but point to different code paths and failure modes.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 44462 is the clearest standalone regression report here: it names a specific tokenizer-loading problem, has a concrete symptom, and is more actionable than the other broadly similar tokenizer issues.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44462 is the strongest representative issue in the set because it is specific, reproducible, and directly tied to user-facing loading behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36010", + "right": "issue:37428", + "accept": false, + "reason": "Both are import errors, but they involve different symbols, modules, and root causes." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention test failures and embedding reinitialization are unrelated bugs." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "A local-path import warning and a tensor-parallel OOM are completely different issues." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "Both are runtime crashes, but in different components and code paths." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d initialization and RT-DETR memory retention are unrelated model behaviors." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-count overflow and a stale generate warning are not the same bug." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Processor loading import failure and protobuf-error masking are different tokenizer-loading problems, not one defect." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Different output behavior and serialization inconsistency are separate regression classes." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "Flash-attention crash on s_aux=None and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "Tokenizer special-token registration and recurrent-state reset are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Eval-time dtype change and gradient checkpointing in eval are distinct issues." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python version compatibility and DeepSeek implementation divergence are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 inference mismatch and RT-DETR memory release are different model bugs." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline output nesting and Trainer broadcast NoneType errors are unrelated." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "Both affect tokenization, but the failing models and failure modes differ." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer decoder typing and ignoring tokenizer.json are related broadly, but not the same underlying bug." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Unsupported image-processor arg and video-input regression are separate regressions." + }, + { + "left": "issue:44451", + "right": "issue:45520", + "accept": false, + "reason": "Model loading failure and a flash_attn import KeyError are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Quantization config failure and quantized saving failure are different code paths." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "EOS-token config inconsistency and flash-attention generation issues are unrelated." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizers and different crash points; not the same bug." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export with torch_compilable_check and Wav2Vec2 save/tokenization failure are unrelated." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "MacOS bus error for a community CLIP model and SigLIP2 documentation errors are not the same issue." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention test failures and void segmentation map processing are unrelated." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Serving KV-cache indexing and Gemma4Processor missing _tokenizer are different serve-time failures." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and num_input_tokens_seen updates are unrelated." + }, + { + "left": "issue:44610", + "right": "issue:45412", + "accept": false, + "reason": "Processor size mismatch and memory release after deletion are unrelated issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 27, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:29942", + "issue:32090", + "issue:33357", + "issue:36010", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43329", + "issue:43334", + "issue:43381", + "issue:43421", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43653", + "issue:43701", + "issue:43723", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43873", + "issue:43931", + "issue:44062", + "issue:44186", + "issue:44206", + "issue:44279", + "issue:44387", + "issue:44442", + "issue:44554", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45103", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45440", + "issue:45491", + "issue:45593", + "issue:45834", + "issue:45859", + "issue:45907" + ], + "soft_pairs": [ + "issue:43421|issue:45491", + "issue:10000|issue:15354", + "issue:43756|issue:45440", + "issue:43873|issue:45229", + "issue:45372|issue:45406", + "issue:42490|issue:43873", + "issue:43278|issue:43575", + "issue:43723|issue:44554", + "issue:43653|issue:44442", + "issue:41720|issue:45229", + "issue:43381|issue:43701", + "issue:43526|issue:45834", + "issue:45103|issue:45372", + "issue:43761|issue:45859", + "issue:45440|issue:45593", + "issue:43278|issue:43329", + "issue:29942|issue:36010", + "issue:32090|issue:33357", + "issue:43334|issue:43531", + "issue:43526|issue:44186", + "issue:44977|issue:45245", + "issue:43824|issue:44279", + "issue:43526|issue:44857", + "issue:42490|issue:43381", + "issue:44442|issue:45072", + "issue:43262|issue:44206", + "issue:42915|issue:43931", + "issue:44062|issue:45907", + "issue:43575|issue:43701", + "issue:43575|issue:44387" + ], + "prepared_review_unit_hash": "59c1bf8d6822926173238040ba1540263557dcb0af998241187d1aea8195b679", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36177, + "estimated_input_tokens": 9045, + "estimated_eval_tokens": 18346 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the issues span unrelated subsystems (tokenizers, model loading, quantization/offloading, multimodal processors, training, and inference). The title similarity is mostly superficial, and none of the soft-edge pairs looks like a true duplicate.", + "confidence": 0.9, + "canonical_issue_reason": "No clear canonical issue: the items do not describe one underlying bug cluster, so there is no strong duplicate target to collapse onto.", + "canonical_pr_reason": null, + "best_issue_reason": "If a representative issue must be chosen, issue 43873 is the broadest runtime bug report in the set (offloading/quantization) and has the most generic scope, but it is still not a duplicate of the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Different subsystems and failure modes: tokenizer post-processing update bug vs Gemma3 NaN embeddings from sliding-window batching." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "Unrelated models and errors: DistilBERT German issue vs TorchScript generator-expression export failure for ViT." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Both concern model implementation mismatches, but they affect different architectures and different bugs (RoPE layer dropout vs DeepSeekV3MoE divergence)." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Both mention memory/offloading, but one is a quantization offloading problem and the other is a Gemma4 multi-GPU OOM; not the same bug." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both are Gemma4-related, but one is an import dependency break and the other is a missing processor attribute in serve; distinct code paths." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Local-path import warning and quantization offloading are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Mixed BF16/FP32 dtype behavior during eval is unrelated to tensor-parallel OOM on a specific Qwen2 load." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "Tokenizer loading in v5 and an MPS attention correctness issue are different components and failure modes." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "Both involve tokenizers, but one is a BigBird special-token registration bug and the other is AutoTokenizer failing for a specific tokenizer class." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Different models and runtime failures: Qwen3 device-mapping cuda assert vs Gemma4 OOM." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Gradient checkpointing in eval mode and resume-from-checkpoint key mismatch are separate training-control bugs." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "ImageProcessor label reduction bug and Kosmos2.5 OCR index error are unrelated." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "auto_docstring annotation handling is unrelated to the Gemma4 processor import failure." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIPVision hidden_states regression and Qwen3_5Moe CPU-offload incompatibility are unrelated." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "DeepSeekV3MoE divergence and D-FINE auxiliary-loss behavior are different model bugs." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Dtype drift in eval and undefined multimodal-token helpers are unrelated." + }, + { + "left": "issue:29942", + "right": "issue:36010", + "accept": false, + "reason": "Flash Attention 2 test failures and a GenerationMixin import error are unrelated." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer broadcast NoneType error and MacOS CLIP bus error are different failure classes." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Qwen3-VL pad_token_id load failure and Qwen3-MoE sliding_window issue are distinct model bugs." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "Both are tokenizer/processor-adjacent, but one is label reduction in BeitImageProcessorFast and the other is LayoutLMv2Tokenizer padding/truncation crashes." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug and category-count RuntimeError are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "A specific missing export/import and a generic dependency issue are not the same underlying bug." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "Image label reduction and float16 AMP loss crash are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "Path-with-slash import warning and gradient checkpointing in eval mode are unrelated." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "Different areas and bugs: tokenizer loading vs bfloat16 dtype mismatches in inference." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio chat-template sampling-rate default and unsupported feature-extractor argument are different audio-processing issues." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3Moe FP8 configuration failure and Qwen3-VL weight-shape mismatch are unrelated model-specific load errors." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken special-argument conflict and list_repo_files revision forwarding are unrelated API bugs." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Tensor-parallel OOM and resume-from-checkpoint key mismatch are different problems." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both involve memory, but one is model-load OOM for a specific Qwen2 config and the other is a broader int4 reserved-memory regression; not the same concrete code path." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 28, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28218", + "issue:29127", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:38175", + "issue:41628", + "issue:42907", + "issue:43316", + "issue:43381", + "issue:43441", + "issue:43493", + "issue:43525", + "issue:43550", + "issue:43611", + "issue:43643", + "issue:43646", + "issue:43720", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44568", + "issue:44743", + "issue:44811", + "issue:44898", + "issue:44913", + "issue:44964", + "issue:44977", + "issue:45216", + "issue:45290", + "issue:45372", + "issue:45440", + "issue:45479", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45698", + "issue:45759", + "issue:45803" + ], + "soft_pairs": [ + "issue:43441|issue:45588", + "issue:43646|issue:44077", + "issue:44743|issue:44811", + "issue:29127|issue:35141", + "issue:44977|issue:45584", + "issue:44464|issue:45491", + "issue:33357|issue:38175", + "issue:44964|issue:45440", + "issue:43316|issue:43525", + "issue:44479|issue:45290", + "issue:44442|issue:45803", + "issue:28218|issue:29127", + "issue:43761|issue:44079", + "issue:44442|issue:44898", + "issue:42907|issue:45698", + "issue:33290|issue:34567", + "issue:34689|issue:43611", + "issue:43381|issue:43873", + "issue:43931|issue:45216", + "issue:43493|issue:43827", + "issue:43761|issue:44188", + "issue:43316|issue:45759", + "issue:43550|issue:44743", + "issue:43720|issue:44442", + "issue:41628|issue:45372", + "issue:43643|issue:44913", + "issue:44568|issue:45479" + ], + "prepared_review_unit_hash": "3d2ccb14fe79a623f0e76df4c605d3ac9c2dcbb8b7ee84fdcc95eebc84842125", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36057, + "estimated_input_tokens": 9015, + "estimated_eval_tokens": 18286 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set does not look like a true duplicate cluster: the items span many unrelated bugs across tokenization, loading, config, attention, generation, and docs. All soft-edge pairs are only topic-level similar and should be rejected.", + "confidence": 0.83, + "canonical_issue_reason": "Issue 43873 is the most active open issue in the set and the closest thing to a broad anchor, but the cluster is too heterogeneous for a real canonical duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the best representative by discussion and inbound references, with a broad runtime-impact bug report; however, it should be treated as a standalone issue, not a duplicate hub.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both involve flash attention, but they are different failures: model compatibility/regression vs an AttributeError in flash_attention.py for sink-less models." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Both mention init/post_init behavior, but one is a v5 custom model initialization regression and the other is a patchtsmixer-specific post_init constraint." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Different code paths: recurrent state reset in modular_qwen3_5.py vs Whisper processor batch_decode skipping special tokens." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "Unrelated bugs: LayoutLMv3 error-message clarity vs token embedding resize reinitializing output embeddings." + }, + { + "left": "issue:44977", + "right": "issue:45584", + "accept": false, + "reason": "Different model paths and symptoms: Qwen3.5 flash-attention generation bug vs Whisper empty transcription after align_special_tokens." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Both are generation/attention-related, but one is compiled forward chunked generation inconsistency and the other is Gemma3 NaN embeddings with mixed-length batches." + }, + { + "left": "issue:33357", + "right": "issue:38175", + "accept": false, + "reason": "Different models and failure modes: MacOS bus error on CLIP loading vs unexpected zero probabilities on SigLIP2." + }, + { + "left": "issue:44964", + "right": "issue:45440", + "accept": false, + "reason": "Different loading/implementation problems for different model families; not the same underlying bug." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config/API issues, but they concern different fields and model families (`Gemma3TextConfig` vs `Llama4Config.pad_token_id`)." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Different regressions: video-input handling for Qwen VL models vs chat-template crashes on assistant tool-call messages." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer loading failure and a bare-except bug in Fuyu tensor conversion are unrelated code paths." + }, + { + "left": "issue:28218", + "right": "issue:29127", + "accept": false, + "reason": "Tokenizer spacing behavior vs a LayoutLMv3 error-message issue; no shared underlying bug." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "Both involve outputs, but one is CLIPVision hidden_states regression and the other is ModelOutput key assignment semantics." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading and Perceiver non-default resolution failure are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Both are save/load related, but one is dequantized checkpoint saving and the other is incorrect custom-module resolution after save_pretrained." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "An optimizer OOM in deepspeed is not the same bug as TrainerState token counting not updating." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both are loading regressions, but they affect different mechanisms and model classes; not mergeable as one fix." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing in eval mode and quantization offloading are separate subsystems with different failure modes." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch on load vs a Qwen3.5 save_pretrained regression are different bugs." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "A model discrepancy between HF and JAX is unrelated to docs still referencing pipeline() after v5 removals." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "Different attention/output issues: missing hidden_states vs divergent kernels from allow_is_bidirectional_skip branching under torch.compile." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Both touch config handling, but one is an API discrepancy in Gemma3TextConfig and the other is unwrapping text_config for composite Qwen multimodal configs." + }, + { + "left": "issue:43550", + "right": "issue:44743", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure and Qwen recurrent-state reset are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "Accelerate loading of packed BitNet weights is unrelated to AutoTokenizer failing to load a tokenizer class." + }, + { + "left": "issue:41628", + "right": "issue:45372", + "accept": false, + "reason": "Both are import-related, but they involve different missing symbols and different processor/loading paths." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "trust_remote_code missing fields and GPTNeoX rotary_pct reset are distinct config/reload bugs." + }, + { + "left": "issue:44568", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer special-token insertion and sequence-classification zero-loss behavior are different problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 29, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:25251", + "issue:28282", + "issue:29127", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:36010", + "issue:36331", + "issue:41950", + "issue:42617", + "issue:43065", + "issue:43232", + "issue:43295", + "issue:43316", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43452", + "issue:43475", + "issue:43540", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43701", + "issue:43716", + "issue:43742", + "issue:43761", + "issue:43825", + "issue:43873", + "issue:44077", + "issue:44162", + "issue:44190", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44617", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44877", + "issue:45127", + "issue:45137", + "issue:45200", + "issue:45584", + "issue:45656", + "issue:46050" + ], + "soft_pairs": [ + "issue:32090|issue:33290", + "issue:34689|issue:36010", + "issue:43873|issue:44871", + "issue:44190|issue:44496", + "issue:44077|issue:44479", + "issue:44855|issue:45137", + "issue:29127|issue:33357", + "issue:33290|issue:34689", + "issue:43316|issue:44877", + "issue:43540|issue:44514", + "issue:33290|issue:45656", + "issue:43316|issue:43716", + "issue:43404|issue:43441", + "issue:43295|issue:43825", + "issue:44162|issue:44792", + "issue:25251|issue:28282", + "issue:41950|issue:43475", + "issue:43550|issue:43701", + "issue:43606|issue:43701", + "issue:44871|issue:45584", + "issue:43065|issue:44617", + "issue:43295|issue:44871", + "issue:45127|issue:45200", + "issue:42617|issue:43575", + "issue:33290|issue:35141", + "issue:43232|issue:43388", + "issue:33290|issue:36331", + "issue:43761|issue:44479", + "issue:43575|issue:43742", + "issue:43452|issue:44496", + "issue:44871|issue:46050" + ], + "prepared_review_unit_hash": "8f64af31adf721bb076a5e665e6b5aa33cb4f41d92d4f0410b54c85d9e496008", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37024, + "estimated_input_tokens": 9256, + "estimated_eval_tokens": 18768 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: the proposed soft pairs mostly connect unrelated issues across model loading, training, pipeline, and quantization/offload regressions. I would not merge any of the soft edges into duplicate clusters.", + "confidence": 0.86, + "canonical_issue_reason": "No single issue cleanly represents the whole set because the cluster is not a true duplicate group; if forced to pick an anchor, #43873 is the broadest active issue, but it does not subsume the others.", + "canonical_pr_reason": null, + "best_issue_reason": "#43873 is the best issue anchor because it is open, actively discussed, and broad enough to serve as a representative bug report, even though it is not a duplicate of the other issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Different bugs: `_gpu_broadcast_one` NoneType trainer failure vs DeepSpeed+Adafactor OOM." + }, + { + "left": "issue:34689", + "right": "issue:36010", + "accept": false, + "reason": "Both are import/model-loading errors, but the failures are unrelated: Llama 3.2 vision loading vs missing `GenerationMixin` import." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offload behavior is not the same as Gemma-3 eos token configuration mismatch." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in a no-trainer script is unrelated to an unrecognized model/config loading error." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "`post_init` allowance for patchtsmixer is unrelated to the v5 video-input regression in Qwen models." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 TorchScript parsing failure is unrelated to DeepSpeed ZeRO3 deque/pop errors." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 box-content error and MacOS CLIP bus error are different model/runtime problems." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Adafactor/DeepSpeed OOM does not match a specific model-loading regression for Llama 3.2 vision." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy is unrelated to strict config blocking granite_speech loading." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Qwen3OmniMoe video-input ValueError is not the same as Qwen2_5_VL batched chat-template crash." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Both involve training/optimizer behavior, but one is OOM in DeepSpeed Adafactor and the other is double optimizer stepping." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Config API mismatch and image preprocessor dtype mismatch are different issues in different code paths." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Mistral3 lm_head tying bug is unrelated to Ministral-3 FlashAttention failure." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression is unrelated to the pipeline translation-task error message regression." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 breakage and janus image-generation test failure do not describe the same bug." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k nesting bug is unrelated to PyTorch missing ImportError for AutoModel." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification pipeline processor selection is unrelated to Sam3VisionEncoderOutput missing `fpn_position_embeddings`." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile+SDPA failure in Bamba is unrelated to a resume_from_checkpoint key mismatch." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch is not the same as checkpoint key mismatch." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma eos-token config mismatch is unrelated to Whisper empty-transcription generation after align_special_tokens." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder is a model-implementation concern, not the same as Sam3Video CUDA OOM." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor/tokenizer regression and Gemma eos-token list-vs-single-value mismatch are different model-interface bugs." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA merge collapse with extended vocab is not the same as Gemma 4 requiring `mm_token_type_ids` to default to zeros." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "Running 3d_parallel.py and OOM loading Qwen2-57B are separate failures with different causes." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "DeepSpeed Adafactor OOM is unrelated to embedding reinitialization in `post_init`." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs sync issue is unrelated to `gather_for_metrics` dropping tuple label elements." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "Adafactor OOM under DeepSpeed is unrelated to `compute_loss` signature incompatibility." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression is unrelated to Qwen video-input regressions." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "OOM loading a 57B model via tensor parallelism is unrelated to a key error loading MobileLLM-125M." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "gguf_file handling for tokenizer/model loading is unrelated to the generic unrecognized-model config error." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "Gemma eos-token configuration mismatch is unrelated to quantized model save/load breakage after sinq." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 30, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:41720", + "issue:43065", + "issue:43335", + "issue:43504", + "issue:43540", + "issue:43550", + "issue:43644", + "issue:43646", + "issue:43673", + "issue:43761", + "issue:43827", + "issue:43906", + "issue:43931", + "issue:43976", + "issue:43994", + "issue:44060", + "issue:44077", + "issue:44190", + "issue:44242", + "issue:44462", + "issue:44479", + "issue:44521", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44737", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44938", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45245", + "issue:45290", + "issue:45357", + "issue:45381", + "issue:45542", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:45803", + "issue:45907" + ], + "soft_pairs": [ + "issue:43646|issue:44913", + "issue:43504|issue:45698", + "issue:43673|issue:44743", + "issue:43761|issue:44242", + "issue:44077|issue:44534", + "issue:44479|issue:44734", + "issue:43065|issue:44610", + "issue:44617|issue:45584", + "issue:44792|issue:44977", + "issue:43827|issue:43994", + "issue:43906|issue:45290", + "issue:43931|issue:44060", + "issue:44190|issue:44964", + "issue:44534|issue:44913", + "issue:41720|issue:45245", + "issue:44855|issue:45588", + "issue:44792|issue:45600", + "issue:44737|issue:45803", + "issue:45127|issue:45357", + "issue:44521|issue:45381", + "issue:43335|issue:44855", + "issue:44462|issue:45907", + "issue:43976|issue:44855", + "issue:43550|issue:44190", + "issue:44938|issue:45542", + "issue:43540|issue:44734", + "issue:43644|issue:44913", + "issue:44077|issue:44877" + ], + "prepared_review_unit_hash": "f7220f2be3f0a7e769922bb8aa5f1f3dbc13d0360e17ec89abfa682ff2329268", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35542, + "estimated_input_tokens": 8886, + "estimated_eval_tokens": 18028 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "One clear duplicate pair stands out: the two reports about non-persistent buffers being filled with junk in Transformers v5. The rest are distinct bugs affecting different models, APIs, or code paths.", + "confidence": 0.89, + "canonical_issue_reason": "issue:43644 is the earlier, more specific report of the non-persistent-buffer regression and has the stronger provenance; issue:44534 reads like a later rewording of the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43644 best represents the underlying bug cluster because it is the original, version-specific report and captures the concrete failure mode clearly.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Both involve model/config behavior, but one is custom model initialization and the other is GPTNeoX config reload semantics; different bugs." + }, + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Legacy-field preset loading and wrong custom-module loading after save_pretrained are different serialization/loading problems." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both mention cache, but one is missing generation cache in chunked prefill and the other is recurrent-state reset logic; not the same code path." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "Unrelated issues: CLIP hidden_states regression versus MoE load-balancing loss omission." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "Optional post_init validation for patchtsmixer is unrelated to the v5 non-persistent buffer junk regression." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Different subsystems and failures: multimodal video regression versus KV-cache continuation tensor indexing." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Sam3PixelDecoder dummy layer and OmDet-Turbo processor size mismatch are unrelated model-specific bugs." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "CUDA OOM in Sam3Video and Whisper empty-transcription failure are unrelated runtime bugs." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation bug do not describe the same defect." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Documentation still referencing pipeline() is not the same as SigLIP2 producing bad outputs through AutoModel/pipeline." + }, + { + "left": "issue:43906", + "right": "issue:45290", + "accept": false, + "reason": "43906 is only an isolated reproduction of an older issue; there is not enough evidence here to say it is the same concrete crash as the tool-call apply_chat_template bug." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Weight-shape mismatch during model loading and an incorrect tied-weights warning are different model-definition issues." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading in a training example and Phi-4 multimodal model loading are unrelated." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer initialization regression is unrelated to GPTNeoX rotary_pct not persisting on reload." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "CUDA assert on A800 with auto device mapping is unrelated to the category-count limit error." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 IndentationError in DebertaV2Model is unrelated to flash-attention crashing on sink-less models." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "A Janus test failure and stale auto_mappings entries for removed Sam3LiteText configs are different problems." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet CPU tensor creation bug and a bare-except swallowing errors in FuyuBatchFeature are unrelated." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "LoRA merge collapse with extended vocab is a different failure from Qwen3.5 saving incorrect visual-encoder keys." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "All-zero assistant masks in chat templating are different from wrong vision_position_ids in qwen2.5-vl video inputs." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "Sparse-layer creation in SwitchTransformers and a DebertaV2 import parse error are unrelated." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer file selection and list_repo_files revision forwarding are separate repository API issues." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Python version support regression and Python 3.13 import parse failure are not the same bug." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba is unrelated to local dataset loading in the training script." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 import/load failure and tensorboard-without-TensorFlow backend error are different environment/setup issues." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "patchtsmixer post_init policy and strict config blocking granite_speech loading are not the same code-path problem." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk in v5 and GPTNeoX rotary_pct resetting on reload are unrelated regression classes." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "These touch initialization, but they target different behaviors and cannot plausibly be merged as one fix." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Same underlying Transformers v5 regression: non-persistent buffers are being initialized with junk; 44534 is a later restatement of the same symptom." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 31, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:35141", + "issue:36246", + "issue:38175", + "issue:38617", + "issue:39692", + "issue:41628", + "issue:41950", + "issue:42490", + "issue:42907", + "issue:43065", + "issue:43295", + "issue:43335", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43479", + "issue:43643", + "issue:43650", + "issue:43688", + "issue:43854", + "issue:44016", + "issue:44112", + "issue:44186", + "issue:44190", + "issue:44220", + "issue:44246", + "issue:44464", + "issue:44466", + "issue:44496", + "issue:44561", + "issue:44683", + "issue:44734", + "issue:44913", + "issue:44933", + "issue:44964", + "issue:45003", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45362", + "issue:45375", + "issue:45446", + "issue:45563", + "issue:45584", + "issue:45600", + "issue:45632", + "issue:45803" + ], + "soft_pairs": [ + "issue:35141|issue:42490", + "issue:44112|issue:45600", + "issue:43335|issue:44190", + "issue:43650|issue:44220", + "issue:39692|issue:44190", + "issue:44496|issue:45803", + "issue:45276|issue:45584", + "issue:36246|issue:41628", + "issue:43479|issue:45375", + "issue:45245|issue:45584", + "issue:43295|issue:45276", + "issue:44464|issue:45563", + "issue:44186|issue:44734", + "issue:38175|issue:43065", + "issue:44683|issue:45446", + "issue:44466|issue:44913", + "issue:44933|issue:45362", + "issue:43404|issue:45216", + "issue:44246|issue:45003", + "issue:44734|issue:45632", + "issue:44561|issue:45632", + "issue:36246|issue:38617", + "issue:45563|issue:45584", + "issue:43643|issue:44496", + "issue:44964|issue:45600", + "issue:44112|issue:44964", + "issue:41950|issue:44016", + "issue:42490|issue:45632", + "issue:43854|issue:44964", + "issue:42907|issue:43441", + "issue:43388|issue:43688" + ], + "prepared_review_unit_hash": "0e5aded3bb19f96d684a911a4c7bedfa738247bf64cbba9744f2887558566440", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36557, + "estimated_input_tokens": 9140, + "estimated_eval_tokens": 18536 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: it mixes unrelated import errors, model/config regressions, pipeline bugs, and docs/test issues. The soft edges are driven by superficial lexical similarity (for example, many mentions of imports or model loading), but they do not describe the same underlying bug or change, so none should be merged.", + "confidence": 0.89, + "canonical_issue_reason": "If forced to pick a representative issue, issue 41628 is the broadest user-facing regression in the set: a straightforward import failure on a public API. It is closer to the recurring import/API-failure theme than the more specialized model or pipeline bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 41628 is the clearest, most general, and most likely to attract duplicate reports among this mixed set. That said, the cluster is not a true duplicate group, so this is only a representative choice.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "One is a token-embedding resize/tied-weights bug; the other is a local-model path trailing-slash import warning. Different code paths and failure modes." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "CI stale-device-override test failure vs. auto_mappings referencing removed Sam3LiteText configs. Both are CI-related but clearly different breakages." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction bug vs. local dataset loading in an example script. No shared underlying defect." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "Placeholder-style issue title vs. audio feature extraction bug. No substantive overlap." + }, + { + "left": "issue:39692", + "right": "issue:44190", + "accept": false, + "reason": "SigLIP2 docs example errors vs. local dataset loading in an image-classification script. Different problems, different surfaces." + }, + { + "left": "issue:44496", + "right": "issue:45803", + "accept": false, + "reason": "Unrecognized model/config error vs. bare-except swallowing errors in Fuyu batch conversion. Different failure mechanisms." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 resize_token_embeddings behavior vs. Whisper empty transcription failure. Both involve generation/model behavior, but not the same bug." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Different import targets and likely different missing exports: Qwen2_5_VLImageProcessor vs. AutoImageProcessor." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Phi4 multimodal default config initialization vs. Qwen3_5MoeVisionConfig missing a field. Both are config issues, but not the same one." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Category-count runtime error is unrelated to Whisper's empty-transcription generation failure." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Processor/tokenizer regression in v4.57.5 vs. Gemma4 embedding resize bug. Different API surfaces and symptoms." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Chunked generation inconsistency with compiled forward vs. stale warning in paged generate. Different issues, one correctness and one warning behavior." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Tokenizer padding/truncation crash vs. serve KV-cache continuation indexing bug. Both are runtime bugs, but on different components." + }, + { + "left": "issue:38175", + "right": "issue:43065", + "accept": false, + "reason": "SigLIP2 zero-probabilities model behavior vs. dummy Conv2d in Sam3PixelDecoder. Unrelated model internals." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Compiled flex_attention failure on torch>=2.9 vs. incorrect torch version gate for AuxRequest import. Related to torch checks only at a very broad level." + }, + { + "left": "issue:44466", + "right": "issue:44913", + "accept": false, + "reason": "v5 serialization of lm_head.weight vs. GPTNeoX rotary_pct reload regression. Both are config/serialization related, but separate bugs." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Nonexistent import from image_utils vs. Qwen3.5-35B chat crash. Import surface vs. model chat runtime failure." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Mistral3 lm_head tying bug vs. Qwen3.5 save_pretrained regression. Different models and different persistence/weight-tying problems." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import taking long sometimes vs. unsafe sys.modules access in modeling_utils. Both touch import-time behavior, but one is performance and the other is unsafe access." + }, + { + "left": "issue:44734", + "right": "issue:45632", + "accept": false, + "reason": "Serve KV cache indexing crash vs. trust_remote_code cache-path collision. Different subsystems and failure modes." + }, + { + "left": "issue:44561", + "right": "issue:45632", + "accept": false, + "reason": "Removal of is_torch_fx_available breaking trust_remote_code models vs. cache-path collision for local models. Both affect remote/local model loading, but not the same bug." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Two separate import errors for different symbols in different modules. Same broad area, but not the same underlying missing export." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "Paged generate warning mismatch vs. Whisper empty-transcription failure. Warning bug and functional generation bug are different." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "trust_remote_code AutoConfig missing fields vs. unrecognized model/config.json error. Both involve config loading, but one loses fields after loading and the other cannot resolve the model." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Phi-4 multimodal loading failure vs. Sam3LiteText auto_mappings CI breakage. Different models and different breakpoints." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "CI stale device override test vs. model loading failure for Phi-4 multimodal. No shared code path." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Video-classification pipeline using image processors vs. syntax error in a notebook. Completely unrelated." + }, + { + "left": "issue:42490", + "right": "issue:45632", + "accept": false, + "reason": "Trailing-slash warning on local model paths vs. cache-path collision between local models. Both concern path handling, but one is warning behavior and the other is cache key collision." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Unit-test loading failure for GLM-4.7-Flash vs. Phi-4 multimodal loading failure. Different models and likely different root causes." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Saving dequantized Ministral/Devstrals vs. FlashAttention failure in Ministral-3. Same family name, but different operations and bug class." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric gathering drops tuple labels in the last batch vs. auxiliary-loss normalization in OLMoE/GPT Oss. Both training-related, but not the same defect." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 32, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:32090", + "issue:35141", + "issue:38175", + "issue:41950", + "issue:42175", + "issue:43295", + "issue:43299", + "issue:43335", + "issue:43381", + "issue:43479", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43673", + "issue:43824", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44220", + "issue:44242", + "issue:44279", + "issue:44336", + "issue:44351", + "issue:44368", + "issue:44496", + "issue:44617", + "issue:44683", + "issue:44779", + "issue:44933", + "issue:45020", + "issue:45042", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45335", + "issue:45357", + "issue:45405", + "issue:45542", + "issue:45584", + "issue:45593", + "issue:45600", + "issue:45759", + "issue:45812", + "issue:45823" + ], + "soft_pairs": [ + "issue:43673|issue:44336", + "issue:44016|issue:44220", + "issue:43295|issue:45335", + "issue:43531|issue:45600", + "issue:42175|issue:44279", + "issue:43526|issue:43901", + "issue:44496|issue:45042", + "issue:43299|issue:45216", + "issue:43335|issue:45593", + "issue:44075|issue:44368", + "issue:43381|issue:45593", + "issue:45335|issue:45584", + "issue:43901|issue:44496", + "issue:43299|issue:45357", + "issue:44496|issue:45759", + "issue:43976|issue:45405", + "issue:41950|issue:45020", + "issue:44242|issue:45200", + "issue:44351|issue:45278", + "issue:43479|issue:43525", + "issue:43824|issue:44933", + "issue:44779|issue:45812", + "issue:44683|issue:45542", + "issue:44016|issue:44062", + "issue:41950|issue:45823", + "issue:43577|issue:45216", + "issue:38175|issue:44617", + "issue:32090|issue:35141", + "issue:43299|issue:43854" + ], + "prepared_review_unit_hash": "db466fb854fbb375907b1b1bb95cd01f71de377210a5e24e9b417df113179603", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36382, + "estimated_input_tokens": 9096, + "estimated_eval_tokens": 18448 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No duplicate cluster here: the issues span unrelated models, pipelines, configs, tokenizer behavior, packaging, and docs/runtime regressions. I would not merge any of the soft-edge pairs.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Different subsystems and failure modes: GenerationMixin chunked-prefill cache handling vs ANSI formatting in loading_report." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error is unrelated to audio feature extraction in _torch_extract_fbank_features()." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Both are regressions, but one is about processor/tokenizer input handling and the other about resizing token embeddings in t5gemma." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and stale Sam3LiteText auto_mappings references are unrelated bugs." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "Both mention dependencies, but one is a pip extras/backend omission and the other is a vague transformers dependency issue with no shared concrete code path." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to a documentation mismatch about return_all_scores." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Model config resolution failure and PIL image processor torchvision dependency checks are different issues." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3VL MoE loading regression and Qwen3.5 save_pretrained checkpoint corruption affect different code paths." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction and D-FINE auxiliary-loss logic are unrelated model-specific bugs." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "SGD optimizer arguments not being used is unrelated to a tie_word_embeddings warning during Qwen fine-tuning." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary losses are separate behaviors with no shared root cause." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "t5gemma embedding resize propagation and Whisper empty-transcription handling after align_special_tokens are different bugs." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Docs wording for TextClassificationPipeline is unrelated to a model-loading config error." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Both concern Qwen multimodal models, but one is a loading regression and the other is incorrect save_pretrained visual encoder keys." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Unrecognized model config and failure to unwrap text_config in composite Qwen configs are different loader failures." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python version incompatibility is unrelated to the PEFT minimum-version bump." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video-classification processor lookup and remote_code breakage are unrelated subsystems." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "MoE load-balancing loss computation and Gemma4 mm_token_type_ids defaults are different model-training issues." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "A missing HybridCache export is too specific to merge with a broad 'many import errors' report." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Phi4MultimodalConfig default config initialization and Llama4Config missing pad_token_id are separate config bugs." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "These are both import errors, but they concern different symbols/modules and do not share the same underlying defect." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Tokenizer regression for Deepseek and wrong token IDs for Granite are model-specific and not the same bug." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and TensorFlow/tensorboard installation behavior are unrelated." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error is unrelated to AddedToken receiving duplicate special arguments." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Video-classification pipeline processor lookup and Gemma4 device_map forward placement are different issues." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Blip2 dtype propagation on load and Qwen3.5 save_pretrained key corruption are unrelated regressions." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 and CUDA OOM in Sam3Video are different failure classes." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer _gpu_broadcast_one TypeError and token embedding reinitialization in post_init are unrelated bugs." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "A single missing HybridCache import is not the same underlying problem as a broad post-upgrade import breakage report." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 33, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:39692", + "issue:41950", + "issue:42994", + "issue:43065", + "issue:43122", + "issue:43295", + "issue:43317", + "issue:43441", + "issue:43450", + "issue:43650", + "issue:43653", + "issue:43723", + "issue:43784", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44291", + "issue:44297", + "issue:44361", + "issue:44462", + "issue:44466", + "issue:44521", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:44898", + "issue:44987", + "issue:45081", + "issue:45276", + "issue:45290", + "issue:45325", + "issue:45397", + "issue:45405", + "issue:45542", + "issue:45593", + "issue:45600", + "issue:45712", + "issue:45823", + "issue:45834", + "issue:45907" + ], + "soft_pairs": [ + "issue:39692|issue:43901", + "issue:44279|issue:45405", + "issue:43295|issue:44779", + "issue:44462|issue:45081", + "issue:41950|issue:43650", + "issue:43976|issue:44987", + "issue:44297|issue:44462", + "issue:43976|issue:44279", + "issue:45081|issue:45907", + "issue:44521|issue:45325", + "issue:43450|issue:44279", + "issue:43441|issue:45834", + "issue:45600|issue:45712", + "issue:43065|issue:44898", + "issue:42994|issue:43317", + "issue:44261|issue:44263", + "issue:44466|issue:45276", + "issue:43441|issue:44683", + "issue:44623|issue:45397", + "issue:44623|issue:44749", + "issue:43650|issue:45290", + "issue:43653|issue:45593", + "issue:43723|issue:44162", + "issue:44016|issue:45823", + "issue:44361|issue:45593", + "issue:44671|issue:44779", + "issue:44987|issue:45542", + "issue:43901|issue:43994", + "issue:43784|issue:45542", + "issue:43122|issue:43650", + "issue:44016|issue:44291" + ], + "prepared_review_unit_hash": "28984978ee6370ac7f88c7f86ea0905e1706c71a380fe07b6311e49f34c2e687", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36421, + "estimated_input_tokens": 9106, + "estimated_eval_tokens": 18468 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft-edge pair looks like a true duplicate/change-equivalent match. The cluster has some tokenizer/processor/loading regressions, but the paired artifacts still point to different concrete failure modes, models, or code paths.", + "confidence": 0.95, + "canonical_issue_reason": "issue:43723 is the best cluster anchor: it is a clear v5 tokenizer-loading regression and the broadest representative of the tokenizer-related near-matches.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43723 is the most actionable and general issue in this set; it captures the shared tokenizer-loading regression theme better than the more model-specific or unrelated reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both involve docs/examples around pipelines, but one reports SigLIP2 example errors and quantization failure while the other is a TextClassificationPipeline docs mismatch; different bugs." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "A generic dependency problem is not the same as bumping MIN_PEFT_VERSION; they describe different package/version issues." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "One is a processor/tokenizer regression in custom model code, the other is Deepseek tokenization output differences; related area, different underlying bug." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-loading related, but one ignores repository tokenizer.json and the other crashes in the Mistral regex patch path; not the same failure." + }, + { + "left": "issue:41950", + "right": "issue:43650", + "accept": false, + "reason": "Video-classification processor lookup and an unrelated placeholder issue ('ADD THE DATA') are not the same bug." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Python-version compatibility and a model-loading failure for physical-intelligence/fast are different issues." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Both concern tokenizer serialization/loading, but one is tokenizer_class metadata mismatch on save and the other is AutoTokenizer ignoring tokenizer.json; distinct code paths." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "A Python support regression is not the same as a vague dependency issue." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regex patch crash and list_repo_files missing revision forwarding are unrelated." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Multimodal assistant-mask generation and Qwen2.5-VL rope-index scaling are different multimodal bugs." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "Video-processor batch shape bug and a dependency issue do not match." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "FlashAttention on Ministral-3 and Kosmos2.5 OCR index errors are unrelated model-specific failures." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are repo/CI hygiene issues, but one is stale auto_mappings references and the other is dummy class cleanup; not the same problem." + }, + { + "left": "issue:43065", + "right": "issue:44898", + "accept": false, + "reason": "Both are vision-model issues, but one is a dummy Conv2d in Sam3PixelDecoder and the other is a Perceiver resize/classification failure; different code paths." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized model saving and dequantized model loading with device_map=auto are different quantization/offload failures." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "One is a missing config epsilon in MLA norm, the other is torch.split return handling in GlmMoeDsaIndexer; unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Inconsistent lm_head serialization and resize_token_embeddings not propagating are different model-state bugs." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "FlashAttention failure on Ministral-3 and compiled flex_attention failing on torch>=2.9 are separate attention-stack issues." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Processor save_pretrained missing files and gemma-4 zero3 from_pretrained are different save/load paths." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Processor serialization missing files and a Chinese report about slower filtering in v5.3.0 are not the same bug." + }, + { + "left": "issue:43650", + "right": "issue:45290", + "accept": false, + "reason": "'ADD THE DATA' is unrelated to apply_chat_template crashing on tool-call assistant messages." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "AutoTokenizer.from_pretrained loading in v5 and ESM2 breaking are different model/loading problems." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax error and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "MLukeTokenizer task-time AttributeError and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "Both are wrong-output regressions, but CamemBERT masked LM predictions and Deepseek tokenizer results are different code paths/models." + }, + { + "left": "issue:44987", + "right": "issue:45542", + "accept": false, + "reason": "Model-loading failure and a TensorBoard/TensorFlow backend dependency issue are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Documentation wording for return_all_scores and SigLIP2 nonsensical inference results are different issues." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "A NameError when importing sentence-transformers and a TensorBoard backend dependency problem are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer-output drift and 'ADD THE DATA' are not the same bug." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "A notebook syntax error and an init_empty_weights TypeError are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 34, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:36246", + "issue:41628", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:43262", + "issue:43299", + "issue:43352", + "issue:43450", + "issue:43525", + "issue:43643", + "issue:43673", + "issue:43761", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:44162", + "issue:44188", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44351", + "issue:44373", + "issue:44485", + "issue:44496", + "issue:44521", + "issue:44561", + "issue:44589", + "issue:44617", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:44964", + "issue:45003", + "issue:45081", + "issue:45200", + "issue:45250", + "issue:45292", + "issue:45362", + "issue:45397", + "issue:45593" + ], + "soft_pairs": [ + "issue:44188|issue:44336", + "issue:44297|issue:45397", + "issue:43450|issue:44617", + "issue:43262|issue:44521", + "issue:43673|issue:43761", + "issue:43525|issue:43881", + "issue:43525|issue:43643", + "issue:44589|issue:45362", + "issue:41628|issue:43976", + "issue:43901|issue:44246", + "issue:44336|issue:44373", + "issue:42491|issue:43299", + "issue:44242|issue:45593", + "issue:16998|issue:22355", + "issue:44295|issue:45397", + "issue:44246|issue:44496", + "issue:41950|issue:44220", + "issue:44336|issue:44485", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:43352|issue:45250", + "issue:43824|issue:43976", + "issue:43976|issue:44964", + "issue:43673|issue:44561", + "issue:44246|issue:44871", + "issue:44242|issue:44857", + "issue:45081|issue:45362", + "issue:44295|issue:44749", + "issue:44162|issue:45200", + "issue:44779|issue:45292", + "issue:43901|issue:45003" + ], + "prepared_review_unit_hash": "dc21e7da78ada6acec5a1f3988168cf37393633ad51420830b649dee179a1d62", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36764, + "estimated_input_tokens": 9191, + "estimated_eval_tokens": 18638 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated Transformers bugs, regressions, and import/config issues. None of the proposed soft edges look like the same underlying bug or change, so all are rejected. The strongest standalone issue in this set is 44521.", + "confidence": 0.66, + "canonical_issue_reason": "Issue 44521 is the clearest anchor: it is an open, concrete bug in apply_chat_template for multimodal inputs, with clear repro details and multiple inbound references.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the best overall issue candidate because it is specific, current, and well-scoped; the rest are mostly isolated import errors, version regressions, or unrelated subsystem complaints.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "Different subsystems and failure modes: attention-kernel divergence under torch.compile vs ANSI codes emitted by loading_report." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer config metadata mismatch is unrelated to gemma-4 zero3 from_pretrained loading failure." + }, + { + "left": "issue:43450", + "right": "issue:44617", + "accept": false, + "reason": "Video processor batched-shape bug and Sam3Video CUDA OOM are different problems, not one code path." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Both involve apply_chat_template, but one is audio sampling-rate defaulting and the other is multimodal assistant_mask generation; not the same bug." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache missing in chunked_prefill and CLIPVisionModel hidden_states regression are separate model/runtime issues." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Llama4Config pad_token_id missing and glm-4v loading failed are model-specific loading issues with different causes." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Both mention config fields, but one is a missing pad_token_id attribute and the other is trust_remote_code dropping fields; not the same defect." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage import/type error is unrelated to Qwen3.5 chat crash." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "Cannot import AutoImageProcessor and Python-version compatibility are different import/runtime problems." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "A docs mismatch about return_all_scores is unrelated to intermittent import slowness." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "ANSI-code leakage in loading_report and a wrong docstring for position_ids are unrelated." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen models, but one is LoRA compatibility across hf versions and the other is Qwen3VL Moe loading on v5; different bugs." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Load-balancing loss omission and D-FINE auxiliary-loss omission are different model-training issues." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "Model_max_length question for DeBERTa-V3 and missing transformers.onnx are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "Position_ids buffer access error and gemma-4 zero3 loading failure are not the same code-path problem." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import slowdown and unrecognized-model config error are different failure classes." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline selecting image processors and _torch_extract_fbank_features are different pipeline/audio paths." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI-code formatting leakage and GLM-5 RoPE implementation discussion are unrelated." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Missing Qwen2_5_VLImageProcessor and missing is_offline_mode are both import errors, but for different symbols and modules." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Missing Qwen2_5_VLImageProcessor and missing HybridCache are distinct export/import problems." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "Nemotron-specific Flash Attention 2 support error is not the same as a generic Flash Attention 2 topic." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Missing Qwen2_5_VLForConditionalGeneration and Python-version support failure are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44964", + "accept": false, + "reason": "Python-version incompatibility and Phi-4 multimodal loading failure do not share the same underlying bug." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Chunked_prefill cache regression and removal of is_torch_fx_available are separate v5 breakages." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import slowness and Gemma-3 eos_token_id inconsistency are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Load-balancing loss not being added and float16 AMP CUDA crash in LwDetrImageLoss are different bugs." + }, + { + "left": "issue:45081", + "right": "issue:45362", + "accept": false, + "reason": "Mistral regex patch crash and Qwen3.5 chat crash are unrelated model-specific failures." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "Reading position_ids after buffer registration and Chinese slowdown after upgrading transformers are not the same issue." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaulting are separate model problems." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "Deepseek tokenizer correctness regression and resize_token_embeddings not affecting output embeddings are different code paths." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Docstring mismatch and unsafe sys.modules access in modeling_utils are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 35, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36246", + "issue:42886", + "issue:43065", + "issue:43232", + "issue:43257", + "issue:43262", + "issue:43408", + "issue:43475", + "issue:43526", + "issue:43540", + "issue:43596", + "issue:43653", + "issue:43673", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44246", + "issue:44336", + "issue:44368", + "issue:44466", + "issue:44496", + "issue:44734", + "issue:44877", + "issue:44933", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45292", + "issue:45335", + "issue:45375", + "issue:45397", + "issue:45459", + "issue:45561", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45656", + "issue:45701", + "issue:45759", + "issue:45987", + "issue:46082" + ], + "soft_pairs": [ + "issue:45987|issue:46082", + "issue:36246|issue:43596", + "issue:44877|issue:45375", + "issue:45292|issue:45584", + "issue:45561|issue:45563", + "issue:43526|issue:45003", + "issue:45292|issue:45701", + "issue:43232|issue:44336", + "issue:44987|issue:45459", + "issue:44496|issue:45003", + "issue:43065|issue:43408", + "issue:45375|issue:45759", + "issue:43262|issue:44246", + "issue:43257|issue:45137", + "issue:44995|issue:45200", + "issue:43475|issue:43540", + "issue:43653|issue:44162", + "issue:44995|issue:45588", + "issue:45397|issue:45656", + "issue:44466|issue:45335", + "issue:44368|issue:45563", + "issue:44368|issue:44977", + "issue:42886|issue:45229", + "issue:45003|issue:45561", + "issue:43673|issue:44734", + "issue:45137|issue:45600", + "issue:44060|issue:44077", + "issue:43931|issue:44368", + "issue:44933|issue:45459" + ], + "prepared_review_unit_hash": "4139387bc5b2dce812fadf4a2be39656d607aeb503fa5fefed6737e54e484b90", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35939, + "estimated_input_tokens": 8985, + "estimated_eval_tokens": 18226 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most candidate pairs are clearly unrelated. The only soft duplicate-like pair is the resize_token_embeddings bug in the generic issue and the T5Gemma-specific manifestation.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45292 is the broader, model-agnostic report of resize_token_embeddings not updating output embeddings, while issue:45335 is a narrower T5Gemma-specific instance of the same behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45292 is the best representative issue because it states the underlying bug in the most general form and can subsume the model-specific follow-up report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "Different problems: StaticCache seq length contract vs LlamaConfig head_dim validation." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "ImportError for Qwen2.5-VL image processor is unrelated to DeepSpeed ZeRO3/BertModel indexing." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both mention config handling, but they are different model configs and different failure modes; not the same concrete bug." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "resize_token_embeddings/output embeddings bug is unrelated to Whisper empty transcription after align_special_tokens." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "Testing-utils file race and paged generate stale warning are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to sys.modules access in modeling_utils." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "Embedding resize bug is unrelated to tokenization changing across versions." + }, + { + "left": "issue:43232", + "right": "issue:44336", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to ANSI output formatting in loading_report." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Model loading failure and protobuf-related tokenizer error masking are different issues." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model config and unsafe sys.modules access do not describe the same bug." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "SAM3 dummy Conv2d and SAM3 tracker-vs-video warning are related only by subsystem, not by the same bug." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both involve composite Qwen configs, but one is missing a field under strict parsing and the other is text_config unwrapping in from_config." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio sampling-rate default bug is unrelated to import latency." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Both involve DeepSpeed, but one is MOE weight conversion during loading and the other is a deque IndexError in ZeRO3." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Stale indexer cache crash and mm_token_type_ids defaults are different issues." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "SAM3 vision encoder attribute error and Qwen3OmniMoe video-processing ValueError are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token registration is unrelated to ESM2 breakage." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Second-forward stale cache crash is unrelated to flash_attention s_aux=None AttributeError." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 ZeRO3 from_pretrained bug and duplicated optimizer.step under deepspeed are not the same failure." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both mention embedding-related behavior, but serialization of tied weights and resize_token_embeddings not updating decoder embeddings are different codepaths." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "Qwen3.5 tied-weights warning is unrelated to a stale num_return_sequences warning." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Warning about tie_word_embeddings is not the same as flash-attention generation failure." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline cache tokenizer loading and Gemma4 CUDA OOM are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "sys.modules access bug and captured_info.txt race are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44734", + "accept": false, + "reason": "Chunked-prefill cache missing and serve KV-indexing crash are different generation bugs." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "ZeRO3 deque error is unrelated to stale Sam3LiteText references in auto_mappings.py." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Qwen3-Next tied-weights warning and patchtsmixer post_init policy are unrelated." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch is unrelated to a tied-weights warning in Qwen3.5 fine-tuning." + }, + { + "left": "issue:44933", + "right": "issue:45459", + "accept": false, + "reason": "Removed image_utils import and protobuf decode-error masking are unrelated import issues." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "These describe the same underlying resize_token_embeddings bug: the generic report says output embeddings are not updated, and the T5Gemma report is a model-specific manifestation on decoder.embed_tokens." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 36, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42371", + "issue:42673", + "issue:42831", + "issue:42994", + "issue:43232", + "issue:43452", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43704", + "issue:43824", + "issue:43827", + "issue:44038", + "issue:44162", + "issue:44222", + "issue:44297", + "issue:44351", + "issue:44361", + "issue:44496", + "issue:44623", + "issue:44661", + "issue:44749", + "issue:44869", + "issue:44933", + "issue:44964", + "issue:45003", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45478", + "issue:45561", + "issue:45656", + "issue:45712" + ], + "soft_pairs": [ + "issue:44869|issue:45356", + "issue:44297|issue:44623", + "issue:44361|issue:44749", + "issue:43646|issue:44162", + "issue:43452|issue:45003", + "issue:43673|issue:43827", + "issue:44623|issue:45656", + "issue:44749|issue:45656", + "issue:44749|issue:45310", + "issue:42994|issue:43482", + "issue:43232|issue:43673", + "issue:42371|issue:45092", + "issue:28282|issue:30990", + "issue:43475|issue:45381", + "issue:45230|issue:45397", + "issue:43475|issue:43643", + "issue:44038|issue:44623", + "issue:41628|issue:44933", + "issue:43475|issue:44661", + "issue:45081|issue:45103", + "issue:42994|issue:43525", + "issue:45375|issue:45478", + "issue:45310|issue:45375", + "issue:41628|issue:44351", + "issue:42673|issue:43704", + "issue:42831|issue:44222", + "issue:44222|issue:45310", + "issue:42994|issue:43824", + "issue:36010|issue:41628", + "issue:42175|issue:45712", + "issue:44038|issue:45310", + "issue:44496|issue:45561", + "issue:44964|issue:45137" + ], + "prepared_review_unit_hash": "e23beb15d72be912bf9473f0e30e7552b85786d9fba9a7b2b56625f8aa5015a0", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37422, + "estimated_input_tokens": 9356, + "estimated_eval_tokens": 18968 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous Transformers regression reports (imports, model loading, generation, quantization, and save/load issues). Only one soft pair looks like a true duplicate; the rest are related only at a broad subsystem/theme level and should stay separate.", + "confidence": 0.35, + "canonical_issue_reason": "Issue 42673 is the cleanest canonical issue because it is a concrete, well-scoped bug report and has an exact duplicate companion (43704) with the same title and failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43646 is the best cluster representative overall because it most clearly summarizes the broader theme of Transformers v5 breaking existing model initialization paths, which matches the dominant pattern across the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer-related regressions, but one is a Whisper timestamp decode crash and the other is a Kimi tokenizer codec/regex regression; different code paths and fixes." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Both involve save_pretrained, but one is tokenizer metadata mismatch and the other is processor file omission; not the same underlying bug." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "These are unrelated: one is an MLukeTokenizer AttributeError during tasks, the other is a broad performance regression after upgrading Transformers." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Both are Transformers regressions, but one is custom model initialization and the other is ESM2 breakage; too different to merge." + }, + { + "left": "issue:43452", + "right": "issue:45003", + "accept": false, + "reason": "Both touch model-loading internals, but one is gguf/AutoModel loading and the other is unsafe sys.modules access in modeling_utils; different failures." + }, + { + "left": "issue:43673", + "right": "issue:43827", + "accept": false, + "reason": "One is a generation cache bug in chunked prefill, the other is a docs issue about pipeline() references; not the same change." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "save_pretrained missing files and optimizer.step being called twice are unrelated bugs." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "A data-filtering slowdown and a double optimizer step under deepspeed are different issues." + }, + { + "left": "issue:44749", + "right": "issue:45310", + "accept": false, + "reason": "Different symptoms and code paths: performance regression in data filtering vs Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are separate serialization/loading problems." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Both are generation-related, but one is _update_model_kwargs_for_generation after sync_gpus and the other is missing GenerationMixin cache in v5.0.0; not the same bug." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 behavior settings and old InternVL2 meta-initialization incompatibility are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "AutoModel missing PyTorch and Sentence Transformers hanging on load are not the same underlying problem." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Different multimodal model bugs: missing fpn_position_embeddings vs wrong vision_position_ids." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Generic bug report vs gemma-4 ZeRO3 from_pretrained issue; no shared failure mode." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "A missing attribute on a SAM3 output object is unrelated to trust_remote_code causing missing fields in AutoConfig." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe v5 regression and processor.save_pretrained missing files are different bugs." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Both are import errors, but they concern different symbols and likely different export paths." + }, + { + "left": "issue:43475", + "right": "issue:44661", + "accept": false, + "reason": "SAM3 output attribute failure and add-new-model-like/tokenizer mapping failure are unrelated." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "These are unrelated attribute errors in different utilities; they do not share the same code path." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized save failure and Llama4Config missing pad_token_id are separate model/config bugs." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5 MoE-related, but one is a missing config field silently dropped by strict validation and the other is a from_pretrained error; not the same concrete bug." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "A from_pretrained loading error and a config-field dropping issue are related only by model family, not by the same fix." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import failures, but they are for different names and likely separate API/export regressions." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact same bug report and title about Qwen3ForCausalLM leaking VRAM across multiple dataloader threads; clearly a duplicate." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "FP8 accuracy degradation and FP8 save_pretrained failure are different issues." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen3.5 MoE problems: saving FP8 checkpoints vs loading from_pretrained." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving and a missing Qwen2_5_VLForConditionalGeneration import are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors after a version change, but they involve different symbols and different affected paths." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Packaging/backend dependency omission and leftover dummy class cleanup are separate maintenance issues." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Both mention Qwen MoE and v5, but the reported failures are different model variants and different loading errors." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "An unrecognized model/config error and a pytest-xdist race condition are unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "A model loading failure and a DeepSpeed ZeRO3 deque IndexError are unrelated runtime bugs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The canonical duplicate callout for 42673/43704 is supported by the packet, and the summary appropriately keeps the rest of the pairs separate as broad-theme similarities rather than 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": 17, + "review_unit_index": 37, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:42175", + "issue:42222", + "issue:42898", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43673", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43901", + "issue:43976", + "issue:44060", + "issue:44079", + "issue:44188", + "issue:44190", + "issue:44222", + "issue:44242", + "issue:44295", + "issue:44297", + "issue:44373", + "issue:44448", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44779", + "issue:44792", + "issue:44933", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45137", + "issue:45200", + "issue:45310", + "issue:45356", + "issue:45405", + "issue:45561", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45812", + "issue:46082" + ], + "soft_pairs": [ + "issue:43377|issue:44242", + "issue:45561|issue:45600", + "issue:42222|issue:45200", + "issue:43901|issue:45561", + "issue:43232|issue:44792", + "issue:42898|issue:44568", + "issue:30990|issue:35141", + "issue:43232|issue:44188", + "issue:44623|issue:45310", + "issue:44188|issue:44373", + "issue:43784|issue:43824", + "issue:42898|issue:45812", + "issue:43278|issue:45137", + "issue:44734|issue:45103", + "issue:30333|issue:33290", + "issue:43874|issue:45092", + "issue:43723|issue:44568", + "issue:42898|issue:45356", + "issue:42898|issue:44779", + "issue:44060|issue:45125", + "issue:45405|issue:45712", + "issue:43673|issue:44188", + "issue:42898|issue:44448", + "issue:44295|issue:45656", + "issue:44933|issue:45712", + "issue:42175|issue:44933", + "issue:43479|issue:44079", + "issue:44222|issue:44297", + "issue:43976|issue:44933", + "issue:44190|issue:46082" + ], + "prepared_review_unit_hash": "155917dd5282e710ea2f0f364a36a5d8bea880bc806081c1d70acfcab74d67ff", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36773, + "estimated_input_tokens": 9194, + "estimated_eval_tokens": 18644 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a noisy mix of unrelated issues, with a loose concentration around tokenizer/v5 regressions. None of the proposed soft edges look like true duplicates of the same underlying bug/change, so all should be rejected.", + "confidence": 0.9, + "canonical_issue_reason": "No single duplicate set is evident across the whole cluster. If one issue must represent the strongest canonical candidate, issue:45812 is the clearest broad, concrete tokenizer regression report with a silent v4\u2192v5 impact and actionable scope.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 stands out as the best representative issue because it describes a widespread, user-visible regression ('wrong token IDs' across Granite models) with a clear version boundary and concrete failure mode.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Different bugs: MIMI batched-vs-single encoder output mismatch vs MoE load-balancing loss gating." + }, + { + "left": "issue:45561", + "right": "issue:45600", + "accept": false, + "reason": "Unrelated: xdist file race in testing utils vs stale auto_mappings references to removed configs." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Different areas: broken VitPose model loading vs Gemma 4 text-only mm_token_type_ids defaulting." + }, + { + "left": "issue:43901", + "right": "issue:45561", + "accept": false, + "reason": "Docs/API mismatch for TextClassificationPipeline vs test utility race; not the same bug." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation kwargs/cache issue vs Janus image-generation test failure; no shared code-path evidence." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Both tokenizer-related, but one is clean_up_tokenization_spaces behavior and the other is add_special_tokens/BOS-EOS handling; too different to merge." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Sentence Transformers loading hang is unrelated to token embedding reinitialization on resize." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Generation cache update issue vs attention kernel divergence under torch.compile; different subsystems and failure modes." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Processor.save_pretrained missing files is unrelated to Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "Runtime kernel divergence is unrelated to a docstring typo for position_ids." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Importing sentence-transformers with latest transformers vs missing Qwen2.5-VL export are separate import-time problems." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different behaviors and model families; not the same underlying defect." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift during evaluation is unrelated to DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Serving KV-cache tensor indexing crash vs auto-docstring annotation parsing crash; unrelated code paths." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLFlow reporting on failed training jobs is unrelated to Adafactor OOM with DeepSpeed." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Missing image-patch method in GLM46V processor is unrelated to InternVL2 meta-initialization incompatibility." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer loading failure vs tokenizer add_special_tokens behavior regression; not the same concrete bug." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Different tokenizer regressions in v5 with different symptoms and model-specific handling." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both concern tokenizer correctness, but the reported failures and affected models differ enough to avoid merging." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Incorrect tied-weights warning vs missing _tp_plan for tensor parallelism are distinct Qwen3-related issues." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "PEFT version pin issue is unrelated to leftover dummy classes leaking into dir(transformers)." + }, + { + "left": "issue:43673", + "right": "issue:44188", + "accept": false, + "reason": "Chunked prefill cache missing in GenerationMixin is unrelated to torch.compile attention-kernel branching." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Different output between v4 and v5 for Pegasus is not shown to be the same root cause as clean_up_tokenization_spaces behavior change." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Position_ids buffer read error and double optimizer.step under DeepSpeed are different runtime bugs." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Missing import from image_utils and leftover dummy classes are separate repo hygiene/import issues." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Packaging backend selection for pip extras is unrelated to an import from image_utils." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to image_utils import issues." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "FP8 MoE save_pretrained behavior and tokenizer_class metadata mismatch are unrelated save/load bugs." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Local dataset loading in an example script is unrelated to LlamaConfig head_dim validation." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 38, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28282", + "issue:30333", + "issue:30990", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:38175", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42907", + "issue:43116", + "issue:43232", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43761", + "issue:43873", + "issue:44190", + "issue:44206", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44623", + "issue:44829", + "issue:44869", + "issue:44871", + "issue:44945", + "issue:44977", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45656", + "issue:45834", + "issue:46002", + "issue:46050", + "issue:46082" + ], + "soft_pairs": [ + "issue:43116|issue:44190", + "issue:44297|issue:45656", + "issue:42831|issue:44623", + "issue:44222|issue:44623", + "issue:43479|issue:44869", + "issue:44206|issue:45103", + "issue:41628|issue:42757", + "issue:30333|issue:36331", + "issue:45071|issue:46082", + "issue:30333|issue:34689", + "issue:44297|issue:45310", + "issue:44222|issue:44484", + "issue:44263|issue:45092", + "issue:28282|issue:30333", + "issue:43723|issue:44977", + "issue:43232|issue:45834", + "issue:30333|issue:35141", + "issue:43761|issue:46082", + "issue:43873|issue:44945", + "issue:44373|issue:44485", + "issue:44483|issue:45406", + "issue:44263|issue:44360", + "issue:44295|issue:44623", + "issue:44514|issue:45103", + "issue:44261|issue:45092", + "issue:43479|issue:43644", + "issue:45092|issue:45357", + "issue:43723|issue:44871", + "issue:30990|issue:38175", + "issue:44829|issue:46002", + "issue:42907|issue:46050" + ], + "prepared_review_unit_hash": "ff92a12ede27969af87fb7c95a4f06fa5683418a5c78edbc0dea9ceb0631a953", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36836, + "estimated_input_tokens": 9209, + "estimated_eval_tokens": 18674 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are distinct Transformers bugs across imports, training, save/load, generation, and model-specific regressions. None of the soft edges look like the same underlying fix, so they should stay separate.", + "confidence": 0.92, + "canonical_issue_reason": "No single canonical issue: these are mostly unrelated bugs spanning imports, training loops, save/load, generation, and model-specific regressions, so none cleanly absorbs the others.", + "canonical_pr_reason": null, + "best_issue_reason": "If one issue must anchor the cluster, issue 43723 (tokenizer loading in v5) is the most general user-facing regression, but the cluster is too mixed for it to be a true duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Both involve example scripts, but one is empty multi-label classification output and the other is local dataset loading failure in a different script and code path." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch and a deepspeed optimizer double-step bug are unrelated problems." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FP8 accuracy loss and missing files from processor.save_pretrained affect different features and failure modes." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "Both mention save_pretrained, but one is FP8 MoE checkpoint saving and the other is processor file export; not the same bug." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Phi4 multimodal config initialization and Whisper timestamp decode are separate model-specific issues." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "LasrFeatureExtractor center-arg crash and auto_docstring AttributeError with future annotations are unrelated subsystems." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different dependency problems." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "MLflow failure reporting and a Trainer compute_loss signature break are different training-stack bugs." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "PretrainedConfig type-checking regression and LlamaConfig head_dim validation are distinct config issues." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "Failed training-job reporting in MLflow and a model-loading regression for Llama 3.2 Vision are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer_class metadata mismatch on save and a Qwen3.5 from_pretrained error are different save/load failures." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 MoE save_pretrained bug and the 50GB max_shard_size default are not the same change or bug." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "A torch.split return-value issue in GlmMoeDsaIndexer is unrelated to InternVL2 meta-initialization incompatibility." + }, + { + "left": "issue:28282", + "right": "issue:30333", + "accept": false, + "reason": "Missing PyTorch import and MLflow reporting on failed training jobs are completely different issues." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading regression and Qwen3.5 flash-attention generation failure are different code paths." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation state update after sync_gpus and a Kosmos2.5 OCR index error are unrelated runtime bugs." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "MLflow job reporting and embedding resizing/post_init reinitialization are unrelated concerns." + }, + { + "left": "issue:43761", + "right": "issue:46082", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression and LlamaConfig head_dim validation are different problems." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization offloading behavior and pipeline-parallel LLM output corruption are separate infra bugs." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A docstring mistake for position_ids and GLM-5 RoPE implementation are not the same underlying issue." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat-completions request rejection and Gemma4Processor missing _tokenizer crash are different server failures." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Torch.split return values and a missing ReLU in the DSA indexer describe different code defects." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Reading position_ids after buffer registration and processor.save_pretrained missing files are unrelated bugs." + }, + { + "left": "issue:44514", + "right": "issue:45103", + "accept": false, + "reason": "Batched apply_chat_template crashing and auto_docstring failing with future annotations are different features." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "Missing rms_norm_eps precision issue in MLA q_a_layernorm and InternVL2 meta-init incompatibility are not the same bug." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Phi4MultimodalConfig None-handling and non-persistent buffer junk in v5 are distinct regressions." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "InternVL2 meta-init compatibility and Qwen3.5 visual-encoder key saving are different save/load issues." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Tokenizer loading failure and Gemma3 eos_token_id inconsistency affect different parts of model initialization." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Sentence Transformers loading hangs and SigLIP2 zero-probability outputs are unrelated." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Flash-attention training degeneration and NaNs in classification heads after checkout/init are different bugs." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Dequantized model save failure and quantized model save/load breakage are related by theme but not the same concrete bug or fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 39, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43122", + "issue:43232", + "issue:43479", + "issue:43576", + "issue:43646", + "issue:43723", + "issue:43756", + "issue:43784", + "issue:43824", + "issue:43937", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44373", + "issue:44442", + "issue:44464", + "issue:44514", + "issue:44568", + "issue:44704", + "issue:44779", + "issue:44913", + "issue:44945", + "issue:45278", + "issue:45290", + "issue:45310", + "issue:45593", + "issue:45600", + "issue:45753", + "issue:46002", + "issue:46082" + ], + "soft_pairs": [ + "issue:41628|issue:43824", + "issue:42831|issue:45310", + "issue:43479|issue:44913", + "issue:44263|issue:44373", + "issue:22355|issue:28282", + "issue:42831|issue:44038", + "issue:39401|issue:44568", + "issue:42757|issue:45278", + "issue:44945|issue:45753", + "issue:42886|issue:42994", + "issue:43576|issue:43784", + "issue:42886|issue:43066", + "issue:44464|issue:46082", + "issue:42371|issue:44263", + "issue:44117|issue:44704", + "issue:43232|issue:45593", + "issue:43646|issue:46002", + "issue:30990|issue:33357", + "issue:44704|issue:45290", + "issue:44295|issue:44297", + "issue:43756|issue:44060", + "issue:42907|issue:44945", + "issue:44261|issue:44373", + "issue:41628|issue:45278", + "issue:43122|issue:44514", + "issue:43723|issue:44442", + "issue:43723|issue:44779", + "issue:43784|issue:45600", + "issue:42907|issue:43937" + ], + "prepared_review_unit_hash": "c26c650a2ca37cad21f5048c1606fffa44b7b1fdacc99002257a292245e9d46f", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35314, + "estimated_input_tokens": 8829, + "estimated_eval_tokens": 17914 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose pile of Transformers issues with superficial title overlap around imports, tokenizers, and v5 regressions. The soft pairs are not true duplicates; they usually point to different symbols, different model families, or different code paths. If a single issue has to anchor the cluster, the broad v5 import-regression report is the best representative.", + "confidence": 0.88, + "canonical_issue_reason": "issue:45278 is the broadest umbrella here: it captures the general \u2018many import errors after upgrade to v5.5.0\u2019 theme that overlaps several import-related reports more than any other single issue.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the most useful representative for triage because it summarizes the common failure mode seen in the import-related subset, even though the cluster is not actually duplicate-tight.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different subsystems (AutoImageProcessor vs Qwen2.5-VL model export/import)." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "One is a FineGrainedFP8 accuracy concern; the other is a Qwen3.5 MoE from_pretrained error. Different bug classes." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Both involve config initialization/reload, but they affect different models and fields (multimodal vision/audio defaults vs GPTNeoX rotary_pct)." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "Torch.split return-value behavior is unrelated to a docstring mistake for position_ids." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Different root causes: missing transformers.onnx module vs PyTorch not installed for AutoModel." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "Both mention model-specific problems, but the concrete failures differ: FP8 accuracy vs Qwen3-VL-Moe behavior in v5." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer offset_mapping correctness and BOS/EOS insertion are different tokenizer bugs on different models." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "One is a specific huggingface_hub symbol import failure; the other is a broad report about many import errors after a version upgrade." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline-parallelism output corruption is unrelated to a Qwen3.5 infinite-loop-on-image report." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline cache loading for tokenizers and quantized model saving are different failure modes." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Broken `transformers env` CLI and a `sentence-transformers` import NameError are separate code paths." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline tokenizer cache loading is unrelated to a wrong decoder type in Transformers v5." + }, + { + "left": "issue:44464", + "right": "issue:46082", + "accept": false, + "reason": "Chunked generation with compiled forward and LlamaConfig head_dim validation are different issues." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 configuration guidance has nothing to do with torch.split behavior." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "A tokenizer mapping assumption bug and an AutoProcessor cached_file kwargs bug affect different call paths." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs after sync_gpus is unrelated to D-FINE auxiliary-loss computation." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Custom model init breakage in v5 and NaNs after checkout/init are different underlying defects." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Sentence Transformers hanging on load and a MacOS bus error on CLIP are not the same bug." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "Missing kwargs to cached_file and chat-template crashes with tool-call messages are distinct processor bugs." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Reading a registered position_ids buffer and saving tokenizer_class metadata are unrelated." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping and Qwen3-Next tied-weights warnings are different model/config problems." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Quantized model saving failures do not match pipeline-parallelism output errors." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "A precision error from missing rms_norm_eps is unrelated to a wrong docstring." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Both are import-error themed, but 41628 is a specific missing class while 45278 is a broad upgrade regression report." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Different tokenization between versions and a batched chat-template crash are separate bugs." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "Both mention AutoTokenizer loading, but they concern different tokenizer classes and likely different failure mechanisms." + }, + { + "left": "issue:43723", + "right": "issue:44779", + "accept": false, + "reason": "A tokenizer loading failure and incorrect Deepseek tokenization results are not the same underlying bug." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "The sentence-transformers import NameError and removed Sam3LiteText auto-mapping references are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Dequantized-model saving and invalid GenerationConfig are different problem areas." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 40, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:22355", + "issue:28218", + "issue:29127", + "issue:29159", + "issue:30990", + "issue:36296", + "issue:39401", + "issue:42371", + "issue:42757", + "issue:42907", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43452", + "issue:43475", + "issue:43493", + "issue:43611", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44222", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44360", + "issue:44485", + "issue:44488", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44779", + "issue:44829", + "issue:44964", + "issue:44977", + "issue:45200", + "issue:45216", + "issue:45245", + "issue:45446", + "issue:45478", + "issue:45753", + "issue:46002" + ], + "soft_pairs": [ + "issue:43408|issue:43475", + "issue:44360|issue:44485", + "issue:43232|issue:45245", + "issue:43122|issue:44521", + "issue:44625|issue:44829", + "issue:43116|issue:45200", + "issue:42907|issue:44964", + "issue:44038|issue:44297", + "issue:29127|issue:30990", + "issue:43122|issue:44779", + "issue:44038|issue:44295", + "issue:36296|issue:43116", + "issue:42907|issue:43452", + "issue:43824|issue:44351", + "issue:39401|issue:44488", + "issue:44038|issue:44222", + "issue:45216|issue:45478", + "issue:39401|issue:44521", + "issue:42757|issue:44351", + "issue:43784|issue:45446", + "issue:22355|issue:28218", + "issue:43122|issue:44977", + "issue:22355|issue:29159", + "issue:43992|issue:44704", + "issue:43066|issue:44779", + "issue:43937|issue:44485", + "issue:43611|issue:46002", + "issue:42371|issue:43874", + "issue:43976|issue:45446", + "issue:43493|issue:45753" + ], + "prepared_review_unit_hash": "c886bd4081207718304126cda98168e597a7625a7913a29eee66d3e01c87147a", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35680, + "estimated_input_tokens": 8920, + "estimated_eval_tokens": 18096 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is not a duplicate set: it mixes unrelated tokenizer, multimodal, generation, import, and version-check bugs. The soft pairs are mostly superficial topic overlaps rather than the same underlying issue.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Same SAM 3 Video area, but one is a model-type warning and the other is a missing attribute on encoder output; different code paths and fixes." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "Completely different topics: DSA indexer/ReLU discussion versus GLM-5 RoPE implementation." + }, + { + "left": "issue:43232", + "right": "issue:45245", + "accept": false, + "reason": "Generation kwargs handling after sync_gpus is unrelated to the 2^24 categories runtime limit." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Both involve tokenization APIs, but one is a version-to-version tokenization regression and the other is assistant mask handling in apply_chat_template for multimodal inputs." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and flash_attention_3 degenerate training are different bugs in different subsystems." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Multi-label classification script behavior is unrelated to Gemma 4 mm_token_type_ids defaults." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Saving dequantized Ministrals/Devstrals and loading Phi-4 multimodal instruct are distinct model-loading/saving failures." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe loading/FP8 issues do not match a tokenizer_class mismatch in tokenizer_config.json." + }, + { + "left": "issue:29127", + "right": "issue:30990", + "accept": false, + "reason": "LayoutLMv3 error clarity and Sentence Transformers stuck loading are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different models and behaviors; not the same concrete bug." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Qwen3-VL-Moe save/load issues are unrelated to the position_ids buffer read error." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor parallel training and the multi-label classification example are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Dequantized model save failures and gguf_file loading failures are different code paths and likely different fixes." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different dependency/export problems." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Wrong Qwen3 offset_mapping and failure to load cjvt/sleng-bert are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Same broad model family, but one is a Qwen3-VL-Moe issue and the other is FP8 save_pretrained for moe; not the same bug." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5 and transformers 5.4+, but one is save_pretrained regression and the other is from_pretrained loading error; different failures." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Offset mapping correctness and multimodal assistant mask generation are different tokenizer-related issues." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Missing is_offline_mode from huggingface_hub and missing HybridCache from transformers are separate import/dependency issues." + }, + { + "left": "issue:43784", + "right": "issue:45446", + "accept": false, + "reason": "Sentence-transformers import failure due to nn and a flex_attention version check bug are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "No module named transformers.onnx and an extra space after an added token are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:44977", + "accept": false, + "reason": "Tokenization regression across versions and Qwen3.5 flash-attention generation problems are different code paths." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "Missing transformers.onnx module and slow/fast tokenizer inconsistency are not the same issue." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "UMT5Encoder missing embed_tokens.weight and AutoProcessor kwargs forwarding are unrelated." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Wrong tokenizer decoder type in v5 and Deepseek tokenization errors in v5 are both tokenizer-related but not the same concrete bug." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "GLM-5 GenerationConfig validation and GLM-5 RoPE implementation are different problems." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "base_model_prefix loading breaks and NaNs in classification heads are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API settings warning and missing get_number_of_image_patches are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python version support and AuxRequest import version checks are different version-compatibility issues." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and Qwen3_5 infinite-loop on a specific image are distinct multimodal model bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 41, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:33453", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42673", + "issue:42831", + "issue:43064", + "issue:43116", + "issue:43381", + "issue:43421", + "issue:43441", + "issue:43450", + "issue:43493", + "issue:43673", + "issue:43825", + "issue:43827", + "issue:43937", + "issue:43986", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44380", + "issue:44410", + "issue:44466", + "issue:44625", + "issue:44779", + "issue:44829", + "issue:44936", + "issue:44945", + "issue:44977", + "issue:45072", + "issue:45081", + "issue:45083", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45405", + "issue:45446", + "issue:45478", + "issue:45563", + "issue:45588", + "issue:45753", + "issue:46002", + "issue:46077" + ], + "soft_pairs": [ + "issue:42222|issue:43441", + "issue:42831|issue:45753", + "issue:44075|issue:44936", + "issue:45357|issue:45478", + "issue:44038|issue:45478", + "issue:43116|issue:43825", + "issue:45083|issue:45446", + "issue:43116|issue:43493", + "issue:43937|issue:45563", + "issue:38617|issue:45278", + "issue:43064|issue:43673", + "issue:44380|issue:46077", + "issue:33453|issue:39401", + "issue:44117|issue:45588", + "issue:44625|issue:46002", + "issue:43827|issue:45446", + "issue:39401|issue:44779", + "issue:22355|issue:29127", + "issue:43381|issue:44945", + "issue:43937|issue:44977", + "issue:44466|issue:46002", + "issue:45405|issue:45446", + "issue:43450|issue:43986", + "issue:44117|issue:45081", + "issue:42673|issue:45072", + "issue:38617|issue:41628", + "issue:44410|issue:45310", + "issue:44829|issue:45216", + "issue:39401|issue:43421" + ], + "prepared_review_unit_hash": "95e2bb9c68b78ae1bb9efc2a1ca4b8378c6c967d8655cdd11fdccabebff41f4e", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36249, + "estimated_input_tokens": 9063, + "estimated_eval_tokens": 18382 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a collection of unrelated issue reports across different models and subsystems. I found no soft-edge pair that is clearly the same underlying bug; all candidate links should be kept separate. The strongest likely duplicate among the listed issues is the repeated Qwen3.5 Moe from_pretrained report.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45310 is the clearest canonical issue because it is an early, focused report of the Qwen3.5 Moe from_pretrained failure and matches a later same-title duplicate (issue:45478).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the best representative issue in this set because it is specific, likely the original report for the repeated Qwen3.5 Moe from_pretrained regression, and is more canonical than the later duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "Different failures: VitPose model breakage vs Ministral-3 FlashAttention crash." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "Accuracy regression in FineGrainedFP8 is unrelated to a Qwen3_5 image-generation infinite loop." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "SGD argument handling and trainer.evaluate() after train() are different code paths and bugs." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5, but one is save_pretrained visual-key serialization and the other is a from_pretrained error; not the same concrete bug." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both mention Qwen3 variants, but the model and failure mode differ (Qwen3-VL-Moe vs Qwen3.5 Moe from_pretrained)." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Multi-label classification example output is unrelated to a v5 pipeline translation-task error message." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "qwen3_omni_moe feature-length helper behavior is unrelated to a flex_attention PyTorch version check." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "run_classification.py empty multi-label results and SigLIP2 discrepancy are unrelated bugs." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "GenerationConfig validation failure and a stale paged-generate warning are not the same issue." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "A specific missing import and a broad upgrade-related import-error cluster are too different to merge." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption and chunked_prefill cache missing are separate training vs generation problems." + }, + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "GPT2 attention scaling being ignored and Gemma4 causal-mask construction are different attention bugs." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Tokenizer loading regression and Qwen3 offset_mapping correctness are distinct tokenizer issues." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Tokenizer mapping None handling and flash_attention s_aux=None crash are unrelated." + }, + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and NaNs in classification heads are different model/config problems." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs still referencing pipeline() and a flex_attention import version check are unrelated." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer bugs, but they concern different models and different incorrect behavior." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "Missing transformers.onnx module and a layoutlmv3 error-message clarity issue are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:44945", + "accept": false, + "reason": "Gradient checkpointing in eval mode and incorrect output under pipeline parallelism are separate behaviors." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "Invalid GenerationConfig and Qwen3.5 flash-attention generation failure are not the same bug." + }, + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "lm_head.weight serialization inconsistency and classification-head NaNs are different failure modes." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video processor shape bug and AutoProcessor crash without torchvision are related only by video processing, not the same defect." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer mapping fallback logic and Mistral regex patch crash are unrelated code paths." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM threads and bfloat16 dtype mismatches are distinct issues." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and likely different fixes." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Missing qkv/gate projections in qwen3next is a model-definition issue, not a from_pretrained error." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 degenerate training and incorrect save_pretrained checkpoint contents are unrelated." + }, + { + "left": "issue:39401", + "right": "issue:43421", + "accept": false, + "reason": "Tokenizer offset mapping and runtime special-token post-processor updates are different tokenizer behaviors." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 42, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:38617", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42831", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43335", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43577", + "issue:43742", + "issue:43874", + "issue:43906", + "issue:43931", + "issue:43937", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44291", + "issue:44368", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44496", + "issue:44829", + "issue:44964", + "issue:44977", + "issue:44987", + "issue:45071", + "issue:45278", + "issue:45310", + "issue:45459", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45656" + ], + "soft_pairs": [ + "issue:44291|issue:45071", + "issue:44484|issue:45310", + "issue:43408|issue:43742", + "issue:42222|issue:45542", + "issue:43937|issue:44496", + "issue:42175|issue:42222", + "issue:44964|issue:45278", + "issue:44483|issue:45464", + "issue:42371|issue:44261", + "issue:44075|issue:45656", + "issue:43066|issue:43122", + "issue:43012|issue:43577", + "issue:43874|issue:44261", + "issue:44410|issue:45478", + "issue:44484|issue:45478", + "issue:38617|issue:44220", + "issue:44263|issue:44485", + "issue:43097|issue:44493", + "issue:41628|issue:45071", + "issue:43012|issue:43408", + "issue:42222|issue:45459", + "issue:45071|issue:45542", + "issue:43441|issue:44829", + "issue:42371|issue:42831", + "issue:42673|issue:44977", + "issue:42673|issue:43906", + "issue:43931|issue:44038", + "issue:43335|issue:45071", + "issue:43097|issue:43452", + "issue:43874|issue:44263", + "issue:43408|issue:44368", + "issue:44117|issue:44987" + ], + "prepared_review_unit_hash": "73e14938b39263c89ebe51ce982f8df037c6e0a402ee06177817c40ee9090984", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36591, + "estimated_input_tokens": 9148, + "estimated_eval_tokens": 18552 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is largely heterogeneous: many separate Transformers v5-era bug reports, but not the same underlying defect. The soft pairs are mostly only thematically similar (imports, tokenizer, model loading, backend regressions) rather than true duplicates.", + "confidence": 0.93, + "canonical_issue_reason": "issue:45278 is the broadest and most representative cluster member: it captures multiple import failures after a Transformers upgrade, which aligns with several other version-regression reports here.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best standalone issue in this set because it is broad, user-facing, and plausibly serves as an umbrella for related upgrade/import breakages.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Different bugs: init_empty_weights/_is_hf_initialized TypeError vs PreTrainedConfig type-checking regression." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "Different failures: max_shard_size behavior vs Qwen3.5-Moe from_pretrained error." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "Different model-loading warnings/errors: SAM3 model-type mismatch vs MobileLLM key error." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "One is a VitPose model breakage; the other is a TensorFlow/tensorboard backend installation issue." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Unrelated: GenerationConfig validation error vs unrecognized model/config.json error." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "Different subsystems: backend dependency packaging vs VitPose model breakage." + }, + { + "left": "issue:44964", + "right": "issue:45278", + "accept": false, + "reason": "Different concrete problems: Phi-4 multimodal load failure vs general import errors after upgrade." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Different API surfaces: chat/completions request rejection vs streaming inference failure on Qwen3.5-0.8B." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 API warning/behavior vs MLA q_a_layernorm precision mismatch." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "Different optimizer bugs: SGD args ignored vs optimizer.step called twice with deepspeed." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Related tokenizer/version issues, but one is wrong decoder type and the other is changed tokenization behavior." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Different dtype issues: PyTorch warning during compile vs BLIP2 components staying float32." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Different multimodal/audio bugs: missing image-patch method vs MLA layernorm epsilon precision issue." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Different model architectures and failures: qwen3next missing projections vs Qwen3.5-Moe from_pretrained error." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "Different problems: save_pretrained sharding limit question vs Qwen3.5-Moe loading failure." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "Different import/API issues: missing layer_type_validation vs _torch_extract_fbank_features failure." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "Different GLM-related bugs: torch.split return values vs RoPE implementation discussion." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Different version-regression complaints: removed tie_embeddings_and_encoder_decoder vs unexpected position-id keys." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "Different regressions: missing AutoImageProcessor import vs PreTrainedConfig type-checking breakage." + }, + { + "left": "issue:43012", + "right": "issue:43408", + "accept": false, + "reason": "Different issues: compile warning for bfloat16 vs model-type mismatch warning." + }, + { + "left": "issue:42222", + "right": "issue:45459", + "accept": false, + "reason": "Different causes: VitPose model breakage vs tokenizer error handling when protobuf is absent." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Type-checking regression is unrelated to backend installation/error handling." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Different code paths: Ministral FlashAttention failure vs sequence-classification training degeneracy with flash_attention_3." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 configuration warning is unrelated to FineGrainedFP8 accuracy regression." + }, + { + "left": "issue:42673", + "right": "issue:44977", + "accept": false, + "reason": "Both concern Qwen3.5 but different manifestations: VRAM leak in dataloader threads vs bad generation with flash-attention." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "One is a VRAM leak; the other is an isolated reproduction of a different earlier issue." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Different Qwen3-VL failures: weight-shape mismatch vs a separate transformers 5.0 bug report." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "Config constructor bug in SwitchTransformers is unrelated to PreTrainedConfig type-checking changes." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "Removed tie_embeddings API issue is unrelated to gguf_file loading failures." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Different GLM46V vs GLM MoE bugs, with distinct failing functions and symptoms." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "Different warning types: SAM3 model-type mismatch vs tie_word_embeddings warning during Qwen3.5 fine-tuning." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Different tokenizer bugs: TOKENIZER_MAPPING_NAMES None handling vs a model-loading failure in transformers>=5.1.0." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 43, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:38617", + "issue:42617", + "issue:42673", + "issue:42757", + "issue:42886", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43381", + "issue:43502", + "issue:43550", + "issue:43825", + "issue:43873", + "issue:43906", + "issue:43931", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44315", + "issue:44351", + "issue:44368", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44655", + "issue:44704", + "issue:44945", + "issue:44964", + "issue:45103", + "issue:45200", + "issue:45290", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45561", + "issue:45588", + "issue:45753", + "issue:45859", + "issue:46050" + ], + "soft_pairs": [ + "issue:43976|issue:44945", + "issue:43986|issue:44484", + "issue:38617|issue:42757", + "issue:43064|issue:44315", + "issue:43097|issue:46050", + "issue:44410|issue:45375", + "issue:44062|issue:44117", + "issue:38617|issue:44351", + "issue:43097|issue:45588", + "issue:42673|issue:45200", + "issue:44038|issue:44410", + "issue:44704|issue:45103", + "issue:43931|issue:44410", + "issue:43502|issue:43992", + "issue:44261|issue:44485", + "issue:44964|issue:45341", + "issue:45341|issue:45859", + "issue:43097|issue:43873", + "issue:43986|issue:44117", + "issue:43066|issue:43906", + "issue:42617|issue:45341", + "issue:43986|issue:45290", + "issue:43986|issue:45372", + "issue:45341|issue:45561", + "issue:43986|issue:45341", + "issue:43825|issue:44655", + "issue:45341|issue:45588", + "issue:43992|issue:44484", + "issue:43208|issue:43550", + "issue:43012|issue:43550", + "issue:43012|issue:44368", + "issue:42886|issue:43381", + "issue:42907|issue:45753", + "issue:42907|issue:44655", + "issue:44483|issue:44514" + ], + "prepared_review_unit_hash": "7c7d875a084d8d3f121b5fffd2abe1a17aac54b3270f28cbd68b66fb0cbb018a", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38017, + "estimated_input_tokens": 9505, + "estimated_eval_tokens": 19266 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated Transformers bug reports that only share broad topical similarity (imports, tokenizers, model loading, offload/quantization, training, and API behavior). None of the soft pairs look like the same underlying bug or change, so I reject all duplicate links.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical duplicate emerges: the issues span distinct failure modes and subsystems, so there isn't one issue that cleanly represents the whole set as a duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the most active open report with the broadest concrete impact (quantization/offloading), so it is the best representative if one issue must be chosen; even so, it is not a duplicate of the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Different problems: a Transformers/Python compatibility issue versus incorrect pipeline-parallel LLM output." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Different bug classes: optional torchvision-dependent processor crash versus a question about save_pretrained shard size." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different missing symbols from different modules (`layer_type_validation` vs `is_offline_mode`)." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Different training issues: FSDP2/PEFT optimizer state corruption versus Liger Kernel not applied with `model_init`." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Different serialization/config topics: removed tie-embeddings API warning versus broken quantized save/load after `sinq`." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Different model/config bugs: missing QKV/gate projections in qwen3next versus a Qwen3_5MoeVisionConfig field being dropped." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "Different tokenizer failures: `AddedToken(special=...)` argument conflict versus `TOKENIZER_MAPPING_NAMES` being `None` unexpectedly." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports from different packages (`layer_type_validation` in configuration utils vs `HybridCache` from transformers)." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Unrelated: tie-embeddings deprecation/removal warning versus an `integrations/flash_attention.py` crash on `s_aux=None`." + }, + { + "left": "issue:42673", + "right": "issue:45200", + "accept": false, + "reason": "Different areas: VRAM leak with multiple dataloader threads versus Gemma 4 text-only fine-tuning needing `mm_token_type_ids` defaulting." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Different Qwen3-family bugs in different model variants: Qwen3-VL-Moe loading versus qwen3next missing projections." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "Different layers: `AutoProcessor.from_pretrained` kwarg forwarding versus an `auto_docstring` crash under future annotations." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Different load-time failures: Qwen3-VL-30B-A3B weight shape mismatch versus missing qwen3next projections." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Different code paths: offline/local_files_only network leakage versus UMT5Encoder missing `embed_tokens.weight` on load." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Different numerical/model issues: missing `rms_norm_eps` precision sensitivity versus GLM-5 RoPE implementation concerns." + }, + { + "left": "issue:44964", + "right": "issue:45341", + "accept": false, + "reason": "Different failures: Phi-4 multimodal model loading versus a small testing_utils bug." + }, + { + "left": "issue:45341", + "right": "issue:45859", + "accept": false, + "reason": "Different domains: testing utility bug versus Qwen3_5 rotary embedding incompatibility with CPU offload." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Related broadly to v5 model behavior, but one is an API/config removal warning and the other is a quantization/offloading bug." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Different issues: missing torchvision for video processor loading versus tokenizer mapping returning `None`." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Different tokenizer/model concerns: wrong decoder type in v5 versus an isolated reproduction of another issue." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "Different problems: failure running `3d_parallel.py` versus a minor testing utility bug." + }, + { + "left": "issue:43986", + "right": "issue:45290", + "accept": false, + "reason": "Different processor/template issues: missing torchvision crash versus batched chat-template handling with tool calls." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Different missing-dependency/import problems: torchvision absence versus `ReasoningEffort` import failure in mistral_common." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "Different test infra bugs: a small testing_utils issue versus an xdist race on `captured_info.txt`." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "Different scope: video-model processor dependency crash versus a testing_utils bug." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Different API regressions: pipeline task support/error messaging versus inability to save Pipeline objects." + }, + { + "left": "issue:45341", + "right": "issue:45588", + "accept": false, + "reason": "Different layers: testing utility bug versus flash-attention handling of `s_aux=None`." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Different concerns: missing checkpoint weight on UMT5 load versus save_pretrained shard-size default behavior." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "Different model/back-end bugs: xLSTM training blockers versus Bamba SDPA/torch.compile failure." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "Different symptoms: a bfloat16 compile precision warning versus an actual model compile failure with SDPA." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Both mention warnings, but one is a PyTorch precision warning and the other is a model config warning about `tie_word_embeddings`." + }, + { + "left": "issue:42886", + "right": "issue:43381", + "accept": false, + "reason": "Different runtime constraints: offline cache loading bug versus gradient checkpointing being unavailable in eval mode." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Different model behaviors: save/load of dequantized models versus an infinite loop on a specific image." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Different serialization paths: dequantized model save failure versus saving Pipeline objects." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Different request-processing bugs: v5 chat/completions schema rejection versus batched `apply_chat_template` crashing on tool-call assistant messages." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 44, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:36010", + "issue:36683", + "issue:38617", + "issue:42490", + "issue:42757", + "issue:42898", + "issue:42994", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43502", + "issue:43525", + "issue:43526", + "issue:43611", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43867", + "issue:43881", + "issue:43901", + "issue:43906", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44222", + "issue:44230", + "issue:44242", + "issue:44263", + "issue:44380", + "issue:44483", + "issue:44485", + "issue:44625", + "issue:44655", + "issue:44683", + "issue:45216", + "issue:45276", + "issue:45290", + "issue:45381", + "issue:45440", + "issue:45478", + "issue:45753", + "issue:46077", + "issue:46082" + ], + "soft_pairs": [ + "issue:44222|issue:45478", + "issue:43208|issue:43526", + "issue:43012|issue:44683", + "issue:42490|issue:43012", + "issue:42757|issue:43824", + "issue:38617|issue:43502", + "issue:44038|issue:44230", + "issue:44483|issue:45290", + "issue:43618|issue:43761", + "issue:43064|issue:44380", + "issue:44483|issue:45381", + "issue:43299|issue:43867", + "issue:43901|issue:44655", + "issue:36683|issue:43475", + "issue:42994|issue:44655", + "issue:45276|issue:46077", + "issue:43352|issue:43937", + "issue:36010|issue:42757", + "issue:43867|issue:45216", + "issue:42898|issue:43066", + "issue:43756|issue:44485", + "issue:43611|issue:44655", + "issue:44625|issue:46077", + "issue:46077|issue:46082", + "issue:36683|issue:43525", + "issue:43618|issue:45216", + "issue:43906|issue:45753", + "issue:44263|issue:45440", + "issue:36683|issue:44242", + "issue:43299|issue:45753", + "issue:36683|issue:43881", + "issue:43901|issue:46077", + "issue:43352|issue:43976" + ], + "prepared_review_unit_hash": "b6951fa258bf00cb15c996e11c29e01dbf7fc36809e0ac6d15b4d23c19623ecf", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37178, + "estimated_input_tokens": 9295, + "estimated_eval_tokens": 18846 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Loose similarity cluster of many unrelated Transformers regressions: import failures, model load/save bugs, config propagation errors, attention/back-end issues, and a few model-specific vision/MoE problems. None of the soft pairs look like true duplicates; they mostly share only broad version/subsystem context.", + "confidence": 0.96, + "canonical_issue_reason": "If one issue must anchor the cluster, issue 43611 is the broadest representative of the v5 loading-regression theme; several others are narrower model-specific load/save failures.", + "canonical_pr_reason": null, + "best_issue_reason": "Best overall representative of the dominant regression pattern in this set, but still only a partial umbrella because the cluster also spans configs, attention masks, tokenizer/template behavior, and backend warnings.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5/MoE and save/load APIs, but one is FP8 save_pretrained while the other is a from_pretrained loading error; different concrete bugs." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "Completely different models and code paths: xLSTM training bugs vs BeitImageProcessorFast label reduction." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "Both touch compile/backend warnings, but one is a bfloat16 precision warning and the other is a flex_attention failure on torch>=2.9." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "Import warning on local-path handling is unrelated to the PyTorch bfloat16 precision warning." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but they concern different symbols and different causes." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "One is an import failure, the other is network requests despite local_files_only=True; not the same bug." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Both are Qwen/MoE-related, but one is a loading regression and the other is FP8 inference support." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Chat/completions request rejection is unrelated to a chat-template crash on tool-call assistant messages." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Same model family, but different regressions: missing attentions vs hidden_states=None." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "FSDP2/PEFT optimizer-state corruption is unrelated to GPT-2 attention scaling under SDPA/FlashAttention." + }, + { + "left": "issue:44483", + "right": "issue:45381", + "accept": false, + "reason": "Different areas entirely: chat API request validation vs Qwen2.5-VL video position IDs." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Both are loading failures, but one is Qwen3-VL-MoE specific and the other is a state_dict ordering issue." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Documentation mismatch about return_all_scores is not the same as failing to save Pipeline objects." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Both are attribute errors, but they involve different models and missing fields." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Both mention saving, but quantized model saving and Pipeline.save_pretrained are different failure modes." + }, + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Same Gemma4 family, but resize_token_embeddings and bidirectional-attention masking are separate code paths." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Different issues: FlashAttention 2 support check vs invalid GenerationConfig for GLM-5." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Different missing imports from different modules; no shared underlying defect." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "Both are model loading/saving problems, but one is state_dict sorting and the other is Qwen3.5 checkpoint correctness in save_pretrained." + }, + { + "left": "issue:42898", + "right": "issue:43066", + "accept": false, + "reason": "Tokenizer behavior change and wrong decoder type are distinct tokenizer regressions." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both reference RoPE, but they are different models and different implementation bugs." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Both are v5 regressions, but one is about loading with base_model_prefix and the other is Pipeline.save_pretrained." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "Num-label propagation and Gemma4 attention masking are unrelated configuration bugs." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Different config/attention bugs in different models; not the same underlying issue." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Both are missing config attributes, but they are separate model-specific regressions." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIP output attentions regression is unrelated to Qwen3.5 save_pretrained checkpoint corruption." + }, + { + "left": "issue:43906", + "right": "issue:45753", + "accept": false, + "reason": "An isolated reproduction of an earlier issue is not the same bug as the Qwen3_5 infinite-loop image case." + }, + { + "left": "issue:44263", + "right": "issue:45440", + "accept": false, + "reason": "Different MoE internals and different models; these are not the same concrete code-path problem." + }, + { + "left": "issue:36683", + "right": "issue:44242", + "accept": false, + "reason": "Missing Gemma3Config vocab_size is unrelated to the load-balancing-loss omission." + }, + { + "left": "issue:43299", + "right": "issue:45753", + "accept": false, + "reason": "Both involve Qwen vision models, but loading failure and infinite-loop generation are distinct defects." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "FlashAttention support for Nemotron is unrelated to the Python version compatibility issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 45, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:28282", + "issue:33453", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42490", + "issue:42548", + "issue:42890", + "issue:43064", + "issue:43097", + "issue:43352", + "issue:43441", + "issue:43502", + "issue:43575", + "issue:43618", + "issue:43867", + "issue:43901", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44117", + "issue:44222", + "issue:44246", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44380", + "issue:44484", + "issue:44492", + "issue:44655", + "issue:44855", + "issue:44857", + "issue:44938", + "issue:44964", + "issue:44977", + "issue:45125", + "issue:45310", + "issue:45341", + "issue:45357", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45479", + "issue:45561", + "issue:45588", + "issue:45600" + ], + "soft_pairs": [ + "issue:43992|issue:45310", + "issue:43441|issue:44380", + "issue:44380|issue:45479", + "issue:43352|issue:44380", + "issue:43867|issue:45375", + "issue:43352|issue:45478", + "issue:41669|issue:44246", + "issue:44355|issue:44855", + "issue:43618|issue:44380", + "issue:36683|issue:44117", + "issue:42890|issue:45561", + "issue:43064|issue:44492", + "issue:43097|issue:43352", + "issue:43867|issue:45357", + "issue:43976|issue:44938", + "issue:43992|issue:45478", + "issue:36296|issue:43575", + "issue:42490|issue:44075", + "issue:43064|issue:44655", + "issue:28282|issue:33453", + "issue:44360|issue:45440", + "issue:43867|issue:44484", + "issue:43618|issue:44857", + "issue:43618|issue:44977", + "issue:43618|issue:45588", + "issue:44655|issue:44938", + "issue:42890|issue:44964", + "issue:43901|issue:44938", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:43502|issue:44062", + "issue:42890|issue:45341", + "issue:44222|issue:45125", + "issue:42548|issue:44351" + ], + "prepared_review_unit_hash": "f8cb859e7eaf4a7dfe3d42a9789409bf3c7b1f34817348abe2e2014a43a435a3", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37457, + "estimated_input_tokens": 9365, + "estimated_eval_tokens": 18986 + }, + "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 (imports, model loading, FlashAttention, tensor-parallel, tests). The only strong duplicate-looking pair is the import-time slowdown reports around `transformers` import performance.", + "confidence": 0.68, + "canonical_issue_reason": "issue:41669 is the best canonical issue because it is the broadest, most clearly scoped performance regression in the set and has the closest duplicate-style neighbor (`issue:44246`).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41669 is the strongest global issue candidate: clear title, concrete symptom, and it represents the only clearly related duplicate cluster in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Both are loading errors, but for different models and causes (`UMT5Encoder` missing weights vs Qwen3.5 MoE `from_pretrained` failure). Not the same bug." + }, + { + "left": "issue:43441", + "right": "issue:44380", + "accept": false, + "reason": "Both mention FlashAttention/backend behavior, but one is a model-support error for Ministral-3 and the other is GPT2 attention scaling being ignored. Different code paths." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "GPT2 backend scaling bug is unrelated to the sequence-classification zero-loss bug. Same broad modeling area, not the same issue." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Nemotron FlashAttention 2.0 support error is a model-specific capability issue; GPT2 attention scaling under SDPA/FA is a different backend bug." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "One is a state-dict loading/order failure, the other is a config field being dropped by `@strict`. Different underlying failures." + }, + { + "left": "issue:43352", + "right": "issue:45478", + "accept": false, + "reason": "Different models and symptoms: Nemotron FA2 unsupported vs Qwen3.5 MoE `from_pretrained` error." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "Both report the same underlying regression: `transformers` import becoming slow, with the latter describing the same import-time slowdown symptom in broader terms." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Both fail while importing/using Python files, but one is about compiled Python files and the other is a PyTorch JIT parsing/indentation issue on Python 3.13. Different root causes." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIPOutput attentions assignment regression is unrelated to GPT2 attention scaling under SDPA/FlashAttention." + }, + { + "left": "issue:36683", + "right": "issue:44117", + "accept": false, + "reason": "Gemma3Config missing `vocab_size` and `TOKENIZER_MAPPING_NAMES` returning `None` are distinct `from_pretrained` failures." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Both are test flakiness, but one is missing `set_seed()` in SamHQ integration tests and the other is an xdist race on `captured_info.txt`. Not the same bug." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Trainer optimizer-state corruption under FSDP2/PEFT is unrelated to a cache-strategy typo." + }, + { + "left": "issue:43064", + "right": "issue:44655", + "accept": false, + "reason": "Optimizer-state bug and `save_pretrained` pipeline persistence bug are different problems." + }, + { + "left": "issue:43097", + "right": "issue:43352", + "accept": false, + "reason": "Missing `tie_embeddings_and_encoder_decoder` guidance in v5 and FlashAttention support errors are unrelated." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "State-dict loading failure due to ordering and incorrect visual encoder keys on save are different save/load bugs." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version compatibility complaints, but they are about different versions and different failure modes; not the same underlying issue." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Different model families and distinct loading failures." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug and Qwen2-57B-A14B-Instruct OOM with TP are related only at a very high level; the concrete failures differ." + }, + { + "left": "issue:42490", + "right": "issue:44075", + "accept": false, + "reason": "Local path trailing-slash import warning and SGD args not being used are unrelated." + }, + { + "left": "issue:43064", + "right": "issue:44655", + "accept": false, + "reason": "Different training/persistence issues with no shared concrete code path." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "PyTorch-not-found ImportError and tokenizer-loading regression are both import/load issues, but not the same bug." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "DSA indexer ReLU discussion is unrelated to DeepseekV3MoE divergence from remote implementation." + }, + { + "left": "issue:43867", + "right": "issue:44484", + "accept": false, + "reason": "State-dict loading failure is unrelated to the `max_shard_size` question." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "CLIPOutput attentions regression and LwDetrImageLoss AMP/CUDA crash are separate failures." + }, + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "CLIPOutput attentions assignment and Qwen3.5 flash-attention generation are different bugs." + }, + { + "left": "issue:43618", + "right": "issue:45588", + "accept": false, + "reason": "CLIPOutput attentions regression is unrelated to `flash_attention.py` crashing on `s_aux=None`." + }, + { + "left": "issue:44655", + "right": "issue:44938", + "accept": false, + "reason": "Saving pipeline objects and Python 3.14 import failure are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Seed-related test flakiness and Phi-4 multimodal loading failure are distinct." + }, + { + "left": "issue:43901", + "right": "issue:44938", + "accept": false, + "reason": "Pipeline docs mismatch and Python 3.14 import failure are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "SamHQ seeding flake and GraniteSpeech stale device override CI failure are different test issues." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Test seeding/flakiness issue and removed `Sam3LiteText` config references in `auto_mappings.py` are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "Local-files-only network leak and `AddedToken()` keyword collision are completely different bugs." + }, + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Both are test-related, but one is xdist seeding/race flakiness and the other is a bug in `testing_utils.py`. Not the same issue." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 `save_pretrained` for MoE and missing `_tp_plan` for Qwen3.5 MoE are distinct tensor-parallel/save bugs." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Both are top-level import errors, but for different missing exports (`PreTrainedModel` vs `HybridCache`)." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays grounded in the packet and the one accepted soft-edge pair (issue:41669 vs issue:44246) is conservative and well supported by the titles. The other verdicts also avoid overclaiming by treating only clear root-cause matches as non-duplicates. The canonical/best-issue choice is reasonable for this set." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 46, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "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:43299", + "issue:43502", + "issue:43526", + "issue:43576", + "issue:43723", + "issue:43824", + "issue:43827", + "issue:43874", + "issue:43901", + "issue:43976", + "issue:44222", + "issue:44230", + "issue:44261", + "issue:44315", + "issue:44355", + "issue:44360", + "issue:44368", + "issue:44387", + "issue:44485", + "issue:44556", + "issue:44623", + "issue:44655", + "issue:44749", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:45071", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45375", + "issue:45468", + "issue:45478", + "issue:45656", + "issue:45850" + ], + "soft_pairs": [ + "issue:43502|issue:44485", + "issue:44355|issue:44749", + "issue:45125|issue:45478", + "issue:30990|issue:31795", + "issue:45125|issue:45375", + "issue:30333|issue:33453", + "issue:44387|issue:44556", + "issue:43010|issue:45278", + "issue:43723|issue:44556", + "issue:42831|issue:43874", + "issue:43502|issue:44222", + "issue:43010|issue:44368", + "issue:43010|issue:45850", + "issue:43502|issue:44961", + "issue:44315|issue:44360", + "issue:44355|issue:45071", + "issue:30333|issue:36296", + "issue:43827|issue:44961", + "issue:44355|issue:44655", + "issue:44222|issue:44230", + "issue:43299|issue:44556", + "issue:33453|issue:34689", + "issue:42757|issue:43576", + "issue:33453|issue:36296", + "issue:43576|issue:45071", + "issue:43010|issue:45656", + "issue:43576|issue:43824", + "issue:31795|issue:38617", + "issue:41628|issue:42548", + "issue:44961|issue:45468", + "issue:43010|issue:43976", + "issue:38617|issue:43576", + "issue:43576|issue:44938", + "issue:44908|issue:44961", + "issue:36296|issue:43901", + "issue:41628|issue:44355", + "issue:36296|issue:43526", + "issue:44261|issue:44360", + "issue:44623|issue:45230" + ], + "prepared_review_unit_hash": "996ad0f1265bb006ec33aff79d6ae36ab0006b46ccf84eafcbccd93069d2f5ae", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 39, + "serialized_chars": 38250, + "estimated_input_tokens": 9563, + "estimated_eval_tokens": 19382 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly about Transformers v5 upgrade regressions, especially import/load failures and related API breakages. None of the soft pairs are close enough to be treated as the same concrete bug, so I rejected all of them. The broadest representative issue is the one summarizing many import errors after upgrading to v5.5.0.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45278 best captures the shared theme of the cluster: a broad v5 upgrade regression causing multiple import failures. It is more representative than the narrowly scoped model- or feature-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the most suitable cluster anchor because it is clear, general, and directly describes the kind of release-wide breakage that several other issues in the set also hint at.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Different problems: offline download behavior vs GLM-5 RoPE implementation." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated bugs: compiled Python file errors vs a data filtering slowdown after upgrade." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Both mention Qwen3.5 MoE, but one is tensor-parallel planning and the other is a from_pretrained load failure; not the same concrete bug." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Sentence loading hang and documentation confusion are unrelated." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen3.5 MoE issues: tensor parallelism vs missing config field serialization." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLflow reporting on training failure is unrelated to tokenizer loading regression." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Int4 OOM from reserved memory and checkpoint reload failure are different bug paths." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Cache update no_grad behavior is unrelated to broad import errors after upgrade." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Both involve loading, but one is tokenizer loading and the other is checkpoint reload compatibility; too broad to merge." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FineGrainedFP8 accuracy and a missing image-patch helper are different issues." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Local-files-only API calls are unrelated to FP8 save_pretrained behavior." + }, + { + "left": "issue:43010", + "right": "issue:44368", + "accept": false, + "reason": "No-grad decoration for cache updates is unrelated to a tied-embeddings warning." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "Cache update semantics are not the same as a general complaint that many models break." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "Offline API-request behavior and an unrelated generic issue title do not match." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel application and a DSA indexer ReLU issue are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled Python file errors and PretrainedConfig type-checking breakage are different regressions." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLflow reporting and tensor-parallel training bugs are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs still referencing pipeline() and a generic issue title are not the same bug." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Running compiled Python files and saving Pipeline objects are separate failure modes." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "FP8 save_pretrained for MoE and FP8 inference support are different code paths." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Different model-loading failures affecting different model families and versions; too broad to combine." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer loading regression and Llama 3.2 vision model loading are distinct issues." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Missing huggingface_hub symbol and a broken transformers env command are not the same bug." + }, + { + "left": "issue:33453", + "right": "issue:36296", + "accept": false, + "reason": "Tokenizer loading regression and tensor parallel training bug are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Broken env command and config type-checking regression are different failures." + }, + { + "left": "issue:43010", + "right": "issue:45656", + "accept": false, + "reason": "Cache update no_grad behavior is unrelated to optimizer stepping twice under deepspeed." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "Broken env command and Qwen2.5-VL import error are different API issues." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation confusion and an ImportError from configuration_utils are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Two different missing imports; same general area, but not the same concrete symbol or code-path problem." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "Generic issue title and Gemma-4 audio positional encoding bug do not match." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs handling is unrelated to the generic racoon issue." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training and text classification docs behavior are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44355", + "accept": false, + "reason": "AutoImageProcessor import failure and compiled Python file errors are separate regressions." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor parallel training and BeitImageProcessorFast label reduction are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44360", + "accept": false, + "reason": "MLA rms_norm_eps precision issue and DSA indexer ReLU omission are different bugs." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Processor.save_pretrained missing files and a placeholder bug report are not the same underlying issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 47, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:41669", + "issue:42371", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43827", + "issue:43874", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44230", + "issue:44246", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44373", + "issue:44393", + "issue:44492", + "issue:44556", + "issue:44655", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44961", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45341", + "issue:45362", + "issue:45468", + "issue:45478", + "issue:45542", + "issue:45753", + "issue:45850", + "issue:46095" + ], + "soft_pairs": [ + "issue:44273|issue:45542", + "issue:44273|issue:45003", + "issue:44393|issue:45753", + "issue:44038|issue:45125", + "issue:44393|issue:45071", + "issue:44230|issue:46095", + "issue:43576|issue:44556", + "issue:44373|issue:44961", + "issue:41669|issue:43673", + "issue:43525|issue:45850", + "issue:44556|issue:45092", + "issue:45362|issue:45850", + "issue:43572|issue:45850", + "issue:42890|issue:44908", + "issue:44704|issue:44908", + "issue:44246|issue:44273", + "issue:43408|issue:44315", + "issue:43976|issue:44393", + "issue:43352|issue:44393", + "issue:43576|issue:44908", + "issue:44297|issue:45230", + "issue:44295|issue:45230", + "issue:44908|issue:45071", + "issue:41669|issue:44351", + "issue:41669|issue:45341", + "issue:43673|issue:44492", + "issue:45278|issue:45850", + "issue:43874|issue:45850", + "issue:44393|issue:45478", + "issue:44077|issue:44861", + "issue:43827|issue:44908", + "issue:41669|issue:44492", + "issue:41669|issue:44908", + "issue:31795|issue:33453", + "issue:44315|issue:44829", + "issue:42371|issue:44492", + "issue:43576|issue:44273", + "issue:44360|issue:45468", + "issue:44655|issue:44861" + ], + "prepared_review_unit_hash": "18a727ea631ef2dbfd8c3fd0e803d67f116edc5b6bb24d13cc6e46bf93c4c7ab", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 39, + "serialized_chars": 38768, + "estimated_input_tokens": 9692, + "estimated_eval_tokens": 19640 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is still mostly heterogeneous. 44273 is the broad lazy-loading/import regression issue, but 45003 looks like a separate modeling_utils/sys.modules bug rather than a confirmed root cause, so the soft edge should be rejected.", + "confidence": 0.59, + "canonical_issue_reason": "44273 is the broad user-facing lazy-loading/import regression; 45003 appears to be a separate sys.modules access bug, not a proven root cause of the same issue.", + "canonical_pr_reason": null, + "best_issue_reason": "44273 is the best representative issue because it captures the widest shared symptom class in this cluster, while the others are narrower, model-specific, or unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading/import machinery and a TensorFlow backend error are different failure modes." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "They share broad import-related themes, but the evidence does not show the same underlying bug; 45003 should not be treated as the root cause of 44273." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Different Qwen3-VL symptoms: 2D bbox output errors vs an infinite loop on a specific image." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Different model families and different code paths: Qwen3-VL-Moe support vs Qwen3.5 MoE tensor-parallel metadata." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "A multimodal Qwen error is not the same as a PretrainedConfig type-checking regression." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 support for Qwen3-VL/Qwen3.5 MoE is unrelated to a DeepSeek-v4 weight conversion-script request." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Both are version-related, but they affect different paths: env command behavior vs checkpoint reload." + }, + { + "left": "issue:44373", + "right": "issue:44961", + "accept": false, + "reason": "A docstring issue about position_ids is unrelated to the placeholder bug report." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star cleanup/performance and missing GenerationMixin cache during chunked prefill are different bugs." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "A specific missing pad_token_id attribute is not the same as a broad 'many models break' report." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both concern v5 compatibility, but one is checkpoint reload and the other is remote-code meta initialization." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "A concrete Qwen3.5-35B chat crash is not the same as a generic broad breakage complaint." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "A StableLmConfig pad_token_id/pad_token_idx regression is a specific model-config bug, not the same as a broad umbrella issue." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "A flaky test missing set_seed() is unrelated to inverse_sqrt scheduler kwargs handling." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "Passing kwargs to cached_file and scheduler timescale handling are separate code paths." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Both involve import/lazy-loading themes, but the evidence only supports a loose similarity, not the same underlying bug." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "sam3_tracker model-type warnings and Liger Kernel application under model_init are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python version compatibility and Qwen3-VL bounding-box output are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention 2.0 support gating for NemotronH is unrelated to a Qwen3-VL output error." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "Environment command breakage and inverse_sqrt scheduler kwargs are different regressions." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch is unrelated to a generic bug report." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "Reading position_ids after buffer registration is unrelated to the generic placeholder issue." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs handling is not the same as PretrainedConfig type checking." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star removal and a HybridCache import failure are different bugs." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Import-star cleanup and a bug in testing_utils.py are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Missing generation cache and a cache-strategy typo are not the same bug." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "A concrete upgrade-to-5.5.0 import-error cluster is not the same as a vague broad breakage report." + }, + { + "left": "issue:43874", + "right": "issue:45850", + "accept": false, + "reason": "Missing get_number_of_image_patches in GLM46V is a specific multimodal API bug, not a general version-breakage complaint." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL hallucination/error in bbox output is different from Qwen3.5 MoE from_pretrained failures." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "An optional post_init policy issue is unrelated to _get_tied_weight_keys crashing on a list." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing pipeline() after removals is unrelated to scheduler kwargs handling." + }, + { + "left": "issue:41669", + "right": "issue:44492", + "accept": false, + "reason": "Import-star cleanup and a cache-strategy typo are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Import-star cleanup and scheduler kwargs are unrelated." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Doc confusion about model.forward inputs is unrelated to tokenizer loading regression." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel not applied with model_init and flash_attention_3 degenerate training are different issues." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 behavior settings and cache-strategy typo are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken transformers env command and lazy loading malfunction are different user-facing regressions." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "A DSA indexer ReLU omission is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Saving Pipeline objects and _get_tied_weight_keys crashing on list are different save-path issues." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary is mostly cautious, but it overstates one soft edge: issue:44273 vs issue:45003. The packet only supports a loose thematic similarity (lazy loading / import internals) and the pair is explicitly marked deterministic_accept=false; calling 45003 a likely root cause of 44273 is stronger than the evidence. Since soft-edge verdicts need to stay conservative, this should be rejected." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 48, + "review_unit_count": 48, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-410", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:37428", + "issue:38617", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:43066", + "issue:43298", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43606", + "issue:43704", + "issue:44230", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44315", + "issue:44360", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44861", + "issue:44908", + "issue:44998", + "issue:45092", + "issue:45230", + "issue:45290", + "issue:45310", + "issue:45468", + "issue:45850", + "issue:46095" + ], + "soft_pairs": [ + "issue:45230|issue:45310", + "issue:44273|issue:44908", + "issue:44492|issue:45850", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:31795|issue:37428", + "issue:44315|issue:45092", + "issue:44485|issue:45468", + "issue:43704|issue:44485", + "issue:43298|issue:46095", + "issue:43502|issue:43519", + "issue:42371|issue:43704", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:44261|issue:45468", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44861|issue:45290", + "issue:43525|issue:44861", + "issue:43704|issue:44998", + "issue:41669|issue:44998", + "issue:43572|issue:44861", + "issue:43066|issue:44998", + "issue:31515|issue:38617", + "issue:44230|issue:44623", + "issue:44360|issue:44998", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:31515|issue:31795", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:43298|issue:44230" + ], + "prepared_review_unit_hash": "d618948ae678f21d767b314b3cce5605c28e3bd3403ca9549537e4a145205f54", + "component_budget": { + "node_count": 410, + "item_count": 410, + "soft_pair_count": 1534, + "serialized_chars": 791439, + "estimated_input_tokens": 197860, + "estimated_eval_tokens": 395976 + }, + "budget": { + "node_count": 38, + "item_count": 38, + "soft_pair_count": 36, + "serialized_chars": 32173, + "estimated_input_tokens": 8044, + "estimated_eval_tokens": 16344 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous, but many items orbit Transformers v5 regression/compatibility complaints, especially tokenizer/model-loading breakages. No PRs are present. I chose the broadest regression report as the best representative, though most pairs are not true duplicates.", + "confidence": 0.46, + "canonical_issue_reason": "issue:45850 is the broadest umbrella report in the set ('Since 5.0 version it breaks too many models') and best matches the overall v5-regression theme, even though the cluster contains many unrelated specifics.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45850 is the most suitable cluster representative because it is broad, regression-oriented, and overlaps thematically with several v5-related loading/tokenizer complaints.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "One is a generic placeholder bug report; the other is a specific Qwen3.5 MoE from_pretrained failure. Not enough evidence they are the same bug." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading failure and inverse_sqrt scheduler kwargs handling are unrelated code paths." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "A cache-strategy typo is not the same underlying problem as broad v5 model breakage." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Different model components and symptoms: Qwen3VL timestamp math vs GLM-5 RoPE implementation." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Timestamp calculation and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star cleanup/performance issue is not the same as VRAM leakage in multithreaded inference." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation confusion about forward args is unrelated to a concrete model-loading regression." + }, + { + "left": "issue:31795", + "right": "issue:37428", + "accept": false, + "reason": "Doc clarification and an ImportError in flash-attention utils are different issues." + }, + { + "left": "issue:44315", + "right": "issue:45092", + "accept": false, + "reason": "Both mention initialization, but one is about Liger kernel application and the other about remote-code/meta-initialization incompatibility; different fixes." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "Different model families and different implementation bugs." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leakage in threaded use and RoPE implementation are unrelated." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Missing audio support in continuous batching and a weight-conversion-script question are not the same change." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "Local-files-only network leakage and Qwen3VL timestamp calculation are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 API guidance and VRAM leakage are different concerns." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp math and missing save_pretrained files are different processor bugs." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Timestamp calculation and multithreaded VRAM leakage do not share the same code path." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer torch.split behavior is unrelated to Gemma4 audio positional encoding." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch is a different failure mode from audio positional encoding." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "Missing rms_norm_eps in MLA q_a_layernorm is not the same underlying bug as Gemma4 audio encoding." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Both are tokenizer-related v5 complaints, but one is a broad behavior difference and the other is a decoder-type bug; too little evidence they share one concrete fix." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs handling and an unrelated non-technical issue are not duplicates." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weights key handling and chat-template crashes are different utility bugs." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config is unrelated to tied-weights key handling." + }, + { + "left": "issue:43704", + "right": "issue:44998", + "accept": false, + "reason": "VRAM leakage and an unrelated non-technical issue are not duplicates." + }, + { + "left": "issue:41669", + "right": "issue:44998", + "accept": false, + "reason": "Import-star cleanup/performance and an unrelated non-technical issue are not duplicates." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression and tied-weights key crash are unrelated." + }, + { + "left": "issue:43066", + "right": "issue:44998", + "accept": false, + "reason": "Tokenizer decoder-type regression and an unrelated non-technical issue are not duplicates." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow checkpoint loading and an ImportError from configuration_utils are different bugs." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 support for Qwen3-VL/Qwen3.5 MoE is unrelated to processor.save_pretrained file omission." + }, + { + "left": "issue:44360", + "right": "issue:44998", + "accept": false, + "reason": "Missing ReLU in DSA indexer and an unrelated non-technical issue are not duplicates." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLFlow reporting after failed training and slow from_pretrained checkpoint loading are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Slow checkpoint loading and tensor-parallel training bug are different code paths." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "From_pretrained performance regression and documentation confusion are not the same issue." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 inference support for Qwen3-VL/Qwen3.5 MoE and a from_pretrained load error are related to the same model family but not the same concrete bug." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Continuous batching lacking audio support and Qwen3-VL/Qwen3.5 MoE FP8 support are distinct feature/bug reports." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "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": "7f6a67aa1ca6adbc0edf86a1ab96017269c143cd65659e7cfa5288f136be94bb", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2478, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2478, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Only the two model-loading issues are loosely related at the subsystem level; the Tapas tokenization issue is unrelated. The available titles do not support true duplicate status for either soft pair.", + "confidence": 0.81, + "canonical_issue_reason": "issue:16618 is the broadest and most central of the remaining issues ('Can't load the model for bert-base-uncased'), so it is the best candidate to represent the model-loading side of the cluster, even though it is not a confirmed duplicate of the auth-related issue.", + "canonical_pr_reason": null, + "best_issue_reason": "As an issue representative, issue:16618 is the most general user-visible failure and most likely to anchor discussion, but the cluster still appears to contain at least one unrelated report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Different products/problems: Tapas tokenization vs. model loading for bert-base-uncased. They do not look like the same underlying bug." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Both mention model download/load failures, but one is specifically a private-model login/error-message issue while the other is a generic inability to load a public model; not enough evidence that they are the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 3, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-26350-2", + "nodes": [ + "issue:26350", + "issue:33260" + ], + "soft_pairs": [ + "issue:26350|issue:33260" + ], + "prepared_review_unit_hash": "22b69d2b1b2df1131c18a278e5b5dc2ca321781a2185f11dcdd9a93c8d37f6d0", + "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": "The two issues are about different requested features: Flash Attention 2 support for more architectures vs GGUF support for more architectures. They are not duplicates and should not be clustered as the same underlying request.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different feature requests and code paths; one targets Flash Attention 2 architecture support, the other targets GGUF support. They are not the same underlying bug/change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-27088-5", + "nodes": [ + "issue:27088", + "issue:28936", + "issue:43834", + "issue:44365", + "issue:44370" + ], + "soft_pairs": [ + "issue:27088|issue:44370", + "issue:27088|issue:44365", + "issue:27088|issue:43834", + "issue:27088|issue:28936", + "issue:28936|issue:43834", + "issue:28936|issue:44365", + "issue:28936|issue:44370", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:44365|issue:44370" + ], + "prepared_review_unit_hash": "0d6e10c8deeaef48aa9d40d70e5b739aa3a9c4bdfb11ecd1d681c0916c1b623e", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 10, + "serialized_chars": 6443, + "estimated_input_tokens": 1611, + "estimated_eval_tokens": 3478 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 10, + "serialized_chars": 6443, + "estimated_input_tokens": 1611, + "estimated_eval_tokens": 3478 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster is not one true duplicate set. Three issues are near-identical generic template requests for translating docs to , and those look like duplicates of each other. The Turkish and Spanish issues are separate language-specific requests and should not be merged with the generic template group or with each other.", + "confidence": 0.92, + "canonical_issue_reason": "issue:43834 is the earliest of the near-identical generic template translation issues and is the best representative of that duplicate subgroup.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43834 is the strongest overall issue representative for the duplicate subgroup because it is the earliest and effectively identical to issue:44365 and issue:44370; the Turkish and Spanish issues are distinct requests.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Different target languages; this is Turkish vs a generic placeholder translation request, so they are not the same underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "Different target languages; Turkish does not duplicate the generic placeholder request." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Different target languages and different concrete request; do not collapse Turkish into the generic placeholder issue." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "Turkish and Spanish are separate translation requests, not the same issue." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "Different target languages; Spanish is not a duplicate of the generic placeholder request." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "Different target languages; Spanish is distinct from the generic placeholder translation issue." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Different target languages; Spanish and the generic placeholder request are not the same underlying issue." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": true, + "reason": "These look like the same generic translation issue with nearly identical wording and no distinguishing language-specific details." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": true, + "reason": "Near-identical generic template issue; same request, just a duplicate opening." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": true, + "reason": "Effectively identical placeholder translation issues; same underlying request." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the Turkish and Spanish issues are clearly language-specific by title, while 43834/44365/44370 are near-identical generic template issues. The duplicate claims among the generic-template trio are conservative enough given the titles and high similarity scores. No overreach stands out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 5, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-36817-15", + "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" + ], + "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:43837|issue:45322", + "issue:43837|issue:43909", + "issue:43909|issue:44205", + "issue:43671|issue:43909", + "issue:44121|issue:45322", + "issue:43837|issue:44121", + "issue:43837|issue:45027", + "issue:37015|issue:45027", + "issue:43837|issue:45177" + ], + "prepared_review_unit_hash": "d5b9a4e2686f1431c8705b2af41063fd7131afb3d49e186bf961ff8cae5bf9c5", + "component_budget": { + "node_count": 15, + "item_count": 15, + "soft_pair_count": 35, + "serialized_chars": 20703, + "estimated_input_tokens": 5176, + "estimated_eval_tokens": 10608 + }, + "budget": { + "node_count": 15, + "item_count": 15, + "soft_pair_count": 35, + "serialized_chars": 20703, + "estimated_input_tokens": 5176, + "estimated_eval_tokens": 10608 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a loose theme of model-add/support requests, not true duplicates. The items span unrelated model families and modalities (BERT variants, detection, audio TTS/ASR, OCR, sparse transformers), so the shared wording is superficial.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:43671 is the most central representative of the repeated 'add model support' pattern, but the cluster is too heterogeneous for it to serve as a true duplicate canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Both are model-add requests, but EuroBert and DEIMv2 are unrelated models and likely different code paths." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert config support and Qwen3-TTS support are different model families and different functionality." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert and RF-DETR are unrelated models; this is only a broad 'add model' similarity." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are different detection models with no evidence of the same underlying fix." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR and LFM2.5 Audio 1.5B are different tasks and code paths." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "A detection model request and a sparse-transformer request are not the same underlying change." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR and Qwen3-ASR are unrelated models; same broad support theme only." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "RF-DETR and Qwen3-TTS are different model families and different implementation work." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are unrelated model additions with different behavior and integration points." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT and the weight-sparse transformer request are distinct architectures, not one concrete bug/change." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT and an audio model request are clearly different artifacts." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are unrelated model requests; no shared underlying defect." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are both BERT-like, but they are different model-specific integrations." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are different model requests." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 and the sparse-transformer request are distinct model/architecture additions." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 and ModernVBERT are different model-specific changes, not one mergeable PR/issue." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 and Voxtral TTS are different model families and modalities." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated requests; only the generic support wording overlaps." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT and Qwen3-ASR are different models and tasks." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are unrelated model support requests." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT and 'BERT-like models with RoPE' are related only at a high level, not the same concrete change." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "A broad RoPE-related BERT proposal and Voxtral TTS support are different changes." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Qwen3-TTS and Qwen3-ASR are sibling modalities, but they are separate model integrations and not one underlying fix." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-TTS and Voxtral TTS are different speech models and would not plausibly merge into one PR." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and code paths." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "TTS support and a sparse-transformer model request are unrelated changes." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "Qwen3-ASR and EUPE are different model families and tasks." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio are both audio-related, but they are distinct model requests." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio and SAM3-LiteText are unrelated model support requests." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio are different audio models with different implementation details." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Sparse-transformer support and EUPE are not the same underlying model change." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR and the sparse-transformer request are unrelated model integrations." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral TTS are different speech model requests." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral TTS are unrelated models and code paths." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-ASR and DeepSeek-OCR2 are different modalities and should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-37934-7", + "nodes": [ + "issue:37934", + "issue:42936", + "issue:43240", + "issue:43668", + "issue:43813", + "issue:44031", + "issue:46018" + ], + "soft_pairs": [ + "issue:37934|issue:43668", + "issue:42936|issue:43240", + "issue:42936|issue:46018", + "issue:42936|issue:44031", + "issue:42936|issue:43813", + "issue:42936|issue:43668", + "issue:43240|issue:43668", + "issue:43668|issue:44031", + "issue:43668|issue:43813" + ], + "prepared_review_unit_hash": "1bcb1b754d26fdcb73a611cdd038887ff763a7d5020b32f904653f31559e4d42", + "component_budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 9, + "serialized_chars": 7185, + "estimated_input_tokens": 1797, + "estimated_eval_tokens": 3850 + }, + "budget": { + "node_count": 7, + "item_count": 7, + "soft_pair_count": 9, + "serialized_chars": 7185, + "estimated_input_tokens": 1797, + "estimated_eval_tokens": 3850 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This set does not look like a true duplicate cluster: the issues span unrelated models, configs, tokenizer warnings, and a typo. I\u2019d anchor on #42936 only as the most substantive issue, not because it duplicates the others.", + "confidence": 0.89, + "canonical_issue_reason": "#42936 is the most developed and discussion-heavy report, with a concrete, reproducible bug description; the other issues are unrelated enough that no real canonical duplicate emerges.", + "canonical_pr_reason": null, + "best_issue_reason": "#42936 is the strongest standalone issue in the group: it is concrete, actionable, and has the most discussion. It is still not actually duplicate-similar to the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Both mention normalization, but one is a model-design question and the other is a config type-hint issue for a different model; not the same bug." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Different failures: Mask2Former ignore_value usage vs. kwargs not being forwarded to a loss function. Similar loss-related area, but not the same code-path problem." + }, + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2Former loss behavior and DeepSeek-V4 expert gating are unrelated model bugs." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "A model-specific ignore_value bug is unrelated to tokenizer regex warning behavior." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "This is a typo in a quantization attribute name, not the same as a Mask2Former loss/ignore-value bug." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Different models and different defect types: unused ignore_value vs incorrect type hint." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Loss-kwargs forwarding bug and ModernBERT config typing issue are unrelated." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "A type-hint bug in one config class is not the same as tokenizer regex warnings across models." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "Incorrect type hint and a typo in a quantization attribute are different, unrelated issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 7, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41084-2", + "nodes": [ + "issue:41084", + "issue:43037" + ], + "soft_pairs": [ + "issue:41084|issue:43037" + ], + "prepared_review_unit_hash": "423c4c103819ef728e836ffa491b43c9b78e413f2629bc38af759475125a5e38", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues do not describe the same problem: one is a vague request about block decoding, while the other is a model-specific support request for DeepSeek v3.2.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #43037 is the better representative issue because it names a concrete target (DeepSeek v3.2) and likely maps to a specific feature request, whereas #41084 is too broad and ambiguous.", + "canonical_pr_reason": null, + "best_issue_reason": "#43037 is the strongest issue in this pair for triage and tracking because it is specific and actionable; #41084 lacks enough detail to serve as a clear canonical reference.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "The titles and apparent intents are unrelated: block decoding vs. DeepSeek v3.2 support. No evidence they refer to the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 8, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "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": "5863a8ab9e6ba87b4e827df3d13434c3fea301d347499c6eb8f98ea360c2780c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related to custom trainers but describe different problems: one is about `remove_unused_column` behavior and trainer customizability, the other about FLOPS estimation for custom trainers. They should not be merged as duplicates.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 43598 is the broader and more representative Trainer customizability request; it is more central to the cluster theme than the FLOPS-estimation-specific issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43598 is the best issue to represent the cluster because it targets a concrete Trainer customization behavior rather than a narrower accounting concern.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "They touch the same general Trainer customization area, but the underlying problems differ: `remove_unused_column` customizability versus FLOPS estimation for custom trainers. Different code paths and fixes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 9, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43684-2", + "nodes": [ + "issue:43684", + "issue:44405" + ], + "soft_pairs": [ + "issue:43684|issue:44405" + ], + "prepared_review_unit_hash": "a78cc3ce4e6872c6377de3b8aac116859c953bc03797487d8d1865c9276e90ee", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related only at a very broad level (both ask for Hugging Face auto-class support for Qwen variants), but they target different model families and different auto tasks, so they are not duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "Issue 43684 is the broader auto-model registration request and is the better anchor for this cluster, but it should not be treated as the same underlying problem as 44405.", + "canonical_pr_reason": null, + "best_issue_reason": "43684 is the best representative issue because it concerns a more general registration gap across AutoModel/AutoModelForConditionalGeneration, while 44405 is a separate, task-specific sequence-classification request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different underlying changes: Qwen3-Omni auto-model/conditional-generation registration vs Qwen3.5 AutoModelForSequenceClassification support. They touch different model families and tasks, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 10, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-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": "03262450dccb701f605020c20d4f532d09e28658fe5c94a1ec130bfdcf631d2e", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Three issues share the exact title 'huggingface' and were opened close together, making them likely duplicates; 'Fett' looks unrelated and should stay separate.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 43845 is the earliest of the repeated 'huggingface' reports and best represents that duplicate set; 43835 has a different title and appears unrelated.", + "canonical_pr_reason": null, + "best_issue_reason": "43845 is the cleanest cluster representative because it is one of the repeated identical-title issues and is the earliest among them; 43859 has a bit more discussion but less body content.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Different titles ('Fett' vs 'huggingface') and no shared target/body evidence suggest unrelated issues." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Different titles and no concrete evidence of the same underlying problem." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "'Fett' does not match the repeated 'huggingface' reports, so this looks like a separate issue." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": true, + "reason": "Exact same title, same topic, and nearly simultaneous creation strongly indicate a duplicate report." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": true, + "reason": "Exact same title and close timestamps make these very likely duplicates of the same issue." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": true, + "reason": "Exact same title and close creation window point to the same underlying report." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded in the packet: 43845/43846/43859 share the same title and are temporally close enough to plausibly form a duplicate set, while 43835 has a different title and no other evidence tying it in. The chosen canonical representative (43845) is the earliest of the repeated 'huggingface' issues." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "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": "5156294e794132c89f28b07323568f6d2c3658806c6d2532db2dc1482e84fb24", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5788, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5788, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All candidates look like unrelated low-signal spam/off-topic issues rather than duplicates of one concrete bug or request. No pair should be merged in the cluster.", + "confidence": 0.98, + "canonical_issue_reason": "No issue stands out as a real canonical report; the set is dominated by spam/off-topic placeholders with no actionable description.", + "canonical_pr_reason": null, + "best_issue_reason": "None of these issues provides a meaningful problem statement or durable target for deduplication, so there is no good global best issue.", + "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 request; same superficial label only." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Both are spam placeholders; no evidence they describe the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam and the other is a generic documentation suggestion; not the same actionable problem." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder vs. off-topic company mention; no shared bug/change target." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "Generic documentation enhancement does not match a spam placeholder as the same issue." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Generic documentation enhancement and spam are unrelated." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Both are spam placeholders, but there is no concrete bug or request to confirm they are the same underlying report." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder vs. unrelated off-topic title; not a duplicate." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder vs. unrelated off-topic title; no same underlying issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45145-2", + "nodes": [ + "issue:45145", + "issue:45151" + ], + "soft_pairs": [ + "issue:45145|issue:45151" + ], + "prepared_review_unit_hash": "237877c796572e50ec158931a7c68160c01dbcd50f206710541d5a3bc8aedee1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two closed issues share the same title and very similar body text, so they are likely duplicate reports of the same energy-optimization proposal.", + "confidence": 0.93, + "canonical_issue_reason": "Issue #45145 is the better canonical issue because it was created earlier and already has an inbound reference, while #45151 adds no distinguishing information.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the strongest representative for the issue cluster due to earliest creation time and slightly more linkage activity.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Likely duplicate / probably the same issue: the titles match exactly and the bodies are highly similar, but the evidence is only soft_similarity, so this is a cautious accept rather than a strong one." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The packet supports a likely duplicate relationship, but the soft-edge verdict is a bit too strong for the evidence provided. Here the only pair evidence is `soft_similarity` with no deterministic accept, so the verdict should stay conservative (e.g. \"likely duplicate\" or \"probably the same issue\") rather than stating it as a strong indication. The canonical issue choice and earlier-creation rationale are grounded." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45295-2", + "nodes": [ + "issue:45295", + "issue:45373" + ], + "soft_pairs": [], + "prepared_review_unit_hash": "f5fd95fada2f530ab0999e343c17c16266cde0c7182bb22509972ca44827cd49", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues describe the same missing Gemma 4 sequence-classification support; the later, more specific open issue is the better cluster representative.", + "confidence": 0.94, + "canonical_issue_reason": "issue #45373 is the clearest canonical tracker: it is open, explicitly names the missing class `Gemma4ForSequenceClassification`, and states the module-level gap relative to Gemma 2/3.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #45373 best captures the concrete request and likely remaining work, while #45295 is a shorter earlier duplicate with less detail and is already closed.", + "best_pr_reason": null, + "soft_edge_verdicts": [] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45672-2", + "nodes": [ + "issue:45672", + "issue:45674" + ], + "soft_pairs": [ + "issue:45672|issue:45674" + ], + "prepared_review_unit_hash": "2ba7602c785333c840314e61a346d9a26ff5cc50d3f7b6dc56df93d5f1188c1a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate 4-bit inference bugs, not duplicates: one is a Gemma4 audio-module uint8/torch.finfo TypeError, the other is a BitsAndBytesConfig skip_modules regression that removes the default lm_head exclusion and triggers an AssertionError.", + "confidence": 0.97, + "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 failure modes and code paths: Gemma4 audio modules hit a uint8/torch.finfo TypeError, while BitsAndBytesConfig llm_int8_skip_modules changes lm_head exclusion handling and causes an AssertionError. Shared 4-bit context is too broad 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": 15, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45762-2", + "nodes": [ + "issue:45762", + "issue:45763" + ], + "soft_pairs": [ + "issue:45762|issue:45763" + ], + "prepared_review_unit_hash": "a8394b428735c9e783329c63a2139565f11b957698cd20f3680fd1721cfd8be5", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1476, + "estimated_input_tokens": 369, + "estimated_eval_tokens": 994 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues look unrelated and appear to be standalone name/date entries rather than duplicate bug reports. There is no clear canonical duplicate pair here.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a strong canonical candidate; both are sparse, unrelated, and not obviously part of the same underlying report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "The titles reference different people and different dates, with no shared issue content or concrete underlying bug/change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 16, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45764-2", + "nodes": [ + "issue:45764", + "issue:45765" + ], + "soft_pairs": [ + "issue:45764|issue:45765" + ], + "prepared_review_unit_hash": "f1fe72dfe55770f793b35d1e21d0613313ff5255af477cf17d119f1bfaab41be", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two closed issues appear unrelated: both are empty, title-only items with different personal-name/date strings and no shared evidence of a duplicate relationship.", + "confidence": 0.98, + "canonical_issue_reason": "No clear canonical issue exists here because the items do not look like duplicate reports of the same problem; they have different titles and no body or linkage evidence.", + "canonical_pr_reason": null, + "best_issue_reason": "If forced to pick a representative issue, 45764 is only marginally cleaner, but neither issue is a good cluster representative because the pair does not appear to describe the same underlying bug or request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "Different title content with no body, filenames, targets, or references tying them to the same underlying issue; this looks unrelated rather than duplicate." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 17, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "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": "0b838f156c852154e84638da6665d39f35460a44b0ea8a5f1b34588a5accc3b3", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5729, + "estimated_input_tokens": 1433, + "estimated_eval_tokens": 3122 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5729, + "estimated_input_tokens": 1433, + "estimated_eval_tokens": 3122 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster contains two clear duplicate pairs: 45870/45871 and 45872/45873. Issue 45869 is only loosely related by the financial-data theme and should not be merged with the others.", + "confidence": 0.92, + "canonical_issue_reason": "45872 is the best canonical representative: it is the earlier of the exact duplicate pair 45872/45873 and has the most descriptive, project-specific title in the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "45872 is the strongest standalone issue to keep because it clearly states the DataForge financial time-series + scraping-service use case and is duplicated verbatim by 45873.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Related topic only; one is a dataset/sample listing and the other is an example request. They do not look like the same underlying issue." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Related topic only; the dataset bundle request is not the same concrete issue as the DataForge example request." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Exact duplicate titles and same example request; these clearly refer to the same issue." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Same general financial/DataForge theme, but different concrete requests: an example vs a scraping-service integration." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Same general financial/DataForge theme, but not the same underlying request as the scraping-service issue." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "The example request and the scraping-service request are different enough that they should not be merged as duplicates." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Different concrete requests despite overlapping domain keywords; not the same underlying issue." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Exact duplicate titles and same DataForge financial time-series + scraping-service request; clearly the same issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the only exact title duplicates are 45870/45871 and 45872/45873, and the conservative non-duplicate calls for 45869 and the cross-theme pairs are consistent with the available evidence. No soft-edge verdict overstates the evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-39895-14", + "nodes": [ + "pull_request:39895", + "pull_request:41855", + "pull_request:42668", + "pull_request:43251", + "pull_request:43254", + "pull_request:43424", + "pull_request:43823", + "pull_request:43995", + "pull_request:44827", + "pull_request:45170", + "pull_request:45213", + "pull_request:45655", + "pull_request:46048", + "pull_request:46067" + ], + "soft_pairs": [ + "pull_request:39895|pull_request:43424", + "pull_request:41855|pull_request:43823", + "pull_request:41855|pull_request:43424", + "pull_request:42668|pull_request:43823", + "pull_request:43251|pull_request:46067", + "pull_request:43251|pull_request:43823", + "pull_request:43251|pull_request:44827", + "pull_request:43424|pull_request:46067", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:43995", + "pull_request:43823|pull_request:46067", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:46067", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:46048", + "pull_request:44827|pull_request:45655", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "8d8b9254ca5de51faa645168084c5b0725f57e437c9adb98026097a7e9448f2c", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 22, + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 22, + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a mix of unrelated PRs; the only true duplicate-style overlap is the pair of fixed_cross_entropy kwargs fixes (43251 and 43254), which target the same loss-function change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43251 is the best canonical PR because it is the clearer, more explicit version of the shared cross-entropy fix: it targets issue #43240 and makes the concrete signature/body change in src/transformers/loss/loss_utils.py.", + "best_issue_reason": null, + "best_pr_reason": "PR 43251 is the strongest representative of the actual duplicate pair: it directly addresses the same cross_entropy kwargs support bug, has the explicit issue link, and is narrowly scoped.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "VideoPrism model addition vs executorch dynamic-shape export test; unrelated changes." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Mistral tokenizer API addition vs new MobileLLM model implementation; different code paths." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer method addition vs executorch export test; no shared underlying bug or change." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor robustness work across multimodal autos vs adding MobileLLM; not the same fix." + }, + { + "left": "pull_request:43251", + "right": "pull_request:46067", + "accept": false, + "reason": "Cross-entropy kwargs support vs base_model_prefix conversion loading fixes; unrelated subsystems." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Loss-function kwargs fix vs adding MobileLLM; completely different artifacts." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Cross-entropy API fix vs Mistral4 test/model changes; not the same underlying issue." + }, + { + "left": "pull_request:43424", + "right": "pull_request:46067", + "accept": false, + "reason": "Executorch dynamic-shape test vs conversion loading prefix handling; unrelated." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Executorch export test vs Mistral4 fixes; different model/code-path concerns." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Executorch export test vs MobileLLM addition; no duplicate relationship." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "New MobileLLM model vs config-resolution ordering fix; unrelated changes." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "MobileLLM model addition vs Falcon output-interface refactor; different model work." + }, + { + "left": "pull_request:43823", + "right": "pull_request:46067", + "accept": false, + "reason": "MobileLLM addition vs conversion-loading prefix fix; not the same bug/change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon output-collection refactor vs model-creation skill/docs; only loosely connected by tooling, not same change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon refactor vs Mistral4 tests; unrelated model-specific work." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46067", + "accept": false, + "reason": "Mistral4 test fixes vs conversion-loading prefix handling; different issues." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 tests vs layrnorm->layernorm rename/conversion mapping; not the same concrete bug." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 test fixes vs TDT loss kernel integration; unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 tests vs config-resolution ordering fix; different code paths." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 tests vs model-creation skill; shared tooling references are too broad to be a duplicate." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "MobileLLM model addition vs model-creation skill; one is a model implementation, the other is a meta workflow/doc PR." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs fix the same fixed_cross_entropy bug by adding supported kwargs to nn.functional.cross_entropy in src/transformers/loss/loss_utils.py; they could plausibly be merged into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted pair, 43251/43254, is well-supported by shared issue target 43240 and the same loss-utils change, and the other soft-edge rejections do not overstate similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-40546-2", + "nodes": [ + "pull_request:40546", + "pull_request:42781" + ], + "soft_pairs": [ + "pull_request:40546|pull_request:42781" + ], + "prepared_review_unit_hash": "c9b1a8a5b0418d288265bb5cd5adae09dd6e9139727e32d849b582d64af5f92e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7225, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7225, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related VibeVoice work, but they are not the same change: PR #40546 adds the core VibeVoice model and docs, while PR #42781 is a draft focused on a realtime variant plus acoustic tokenizer support. They overlap in shared package plumbing, but the underlying changes are different enough that they should not be merged as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #40546 is the better canonical VibeVoice PR because it is the non-draft, broader base implementation of VibeVoice with docs and auto-class wiring, and it has substantial review activity.", + "best_issue_reason": null, + "best_pr_reason": "PR #40546 is the best representative PR for this cluster since it covers the primary VibeVoice model integration and appears more mature than the draft realtime follow-up.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Shared VibeVoice filenames mostly reflect common framework plumbing, but the scope differs: #40546 implements the base VibeVoice model, while #42781 adds a realtime-specific variant and acoustic tokenizer. They are not the same concrete change and would not plausibly be merged into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 3, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-40755-2", + "nodes": [ + "pull_request:40755", + "pull_request:41224" + ], + "soft_pairs": [ + "pull_request:40755|pull_request:41224" + ], + "prepared_review_unit_hash": "0f4db8cdf10389cd74f76d700276d27fca334e305ff9cc1584fe9562c0b5f90a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two pull requests are unrelated: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. They should not be deduplicated or clustered as the same change.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different models, different code paths, and different features: TimesFM forecasting with covariates vs 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": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41356-2", + "nodes": [ + "pull_request:41356", + "pull_request:44339" + ], + "soft_pairs": [ + "pull_request:41356|pull_request:44339" + ], + "prepared_review_unit_hash": "aead1308c167b0bae6ff11bf9358340baa5e76f816618383418b505a10a9cd01", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two DEIMv2 pull requests target the same model family, but they are not the same change: PR 41356 is an early/basic model + image processor add, while PR 44339 is a much broader follow-up that adds loss code, auto-mappings, conversion tooling, modular files, and more integration work.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44339 is the more complete and representative DEIMv2 implementation; PR 41356 looks like an earlier partial version of the same model bring-up.", + "best_issue_reason": null, + "best_pr_reason": "PR 44339 is the strongest single artifact for the cluster because it subsumes the basic model work and includes the broader integration needed for a full Transformers addition.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41356", + "right": "pull_request:44339", + "accept": false, + "reason": "Same model area, but not the same concrete change: 44339 expands the implementation substantially and adds new code paths beyond the basic PR 41356. They look like related iterations, not mergeable duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 5, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41797-3", + "nodes": [ + "pull_request:41797", + "pull_request:45186", + "pull_request:45630" + ], + "soft_pairs": [ + "pull_request:41797|pull_request:45186", + "pull_request:45186|pull_request:45630" + ], + "prepared_review_unit_hash": "8ad1d4ed5ef5efce201970aeeef7666c7bd3394e7a9f26b8f6ef3ece691c42b2", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10435, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10435, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three distinct model-addition PRs that share repository boilerplate but target different models and codepaths, so they should not be clustered as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "No true canonical PR: DeepSeek OCR, Isaac, and Kimi2-6 are separate model additions with different implementation files, tests, and model-specific docs; the overlap is mostly shared registration/scaffolding.", + "best_issue_reason": null, + "best_pr_reason": "If one representative had to be picked, PR 45630 is the broadest and most actively reviewed of the three, but it is still not a duplicate of the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different model additions (DeepSeek OCR vs Isaac). Shared auto/docs files are generic registration boilerplate, not the same underlying change or code-path fix." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Different model additions (Isaac vs Kimi2-6) with distinct model/config/processing implementations. Similar framework plumbing is insufficient to treat them 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": 6, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42975-2", + "nodes": [ + "pull_request:42975", + "pull_request:42976" + ], + "soft_pairs": [ + "pull_request:42975|pull_request:42976" + ], + "prepared_review_unit_hash": "8b47e684eee483916e5331f58abc6b83fb36a67bac12ebd2c5a6caa694645f2c", + "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 open PRs both modernize GitHub Actions workflows, but they target different actions and different compatibility upgrades, so they should not be treated as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 42975 is the clearer representative because it has a specific, well-scoped goal: updating core workflow actions for Node 24 compatibility across multiple workflows.", + "best_issue_reason": null, + "best_pr_reason": "PR 42975 is the better canonical PR for this cluster since its title and diff describe a concrete compatibility-driven workflow update, while 42976 is a separate action-version bump focused on Docker build actions.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "Reject: both are GitHub Actions maintenance PRs, but they change different action families and do not appear to fix the same concrete problem. 42975 is about Node 24 compatibility for checkout/cache/upload-artifact; 42976 upgrades docker/build-push-action and related workflow versions. They could be separate PRs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42978-4", + "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": "f9f4c75b2f258575158e263e761d69c2f64710f80973951f16c632d07ae30a37", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14204, + "estimated_input_tokens": 3551, + "estimated_eval_tokens": 7358 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14204, + "estimated_input_tokens": 3551, + "estimated_eval_tokens": 7358 + }, + "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, CTSM). They share the usual Transformers scaffolding files, but they target different models, different configs/processors, and different implementation paths, so they are not duplicates of one another.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 43451 is the broadest and most complete addition in this set (modeling, processing, image/video support, auto-mapping, docs, and many tests), 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 models and codepaths: ViT NEPA vs Molmo. Shared files like docs and auto registration are expected in separate model-addition PRs, not evidence of a duplicate fix." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "Different models and codepaths: ViT NEPA vs Molmo2. They add separate architectures and support surfaces, so they cannot be merged as one duplicate PR." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Different models and codepaths: ViT NEPA vs CTSM. The overlap is only in generic Transformers plumbing, not the same underlying change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are related by name but are distinct model families with different files, processors, and tests. This is not the same concrete change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo vs CTSM are unrelated model additions. Shared auto/model-doc boilerplate is too generic to treat them as duplicates." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM implement different models with different support components (including video processing for Molmo2). They do not fix or add the same thing." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 8, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43291-4", + "nodes": [ + "pull_request:43291", + "pull_request:43665", + "pull_request:44815", + "pull_request:46102" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43665|pull_request:46102", + "pull_request:43665|pull_request:44815" + ], + "prepared_review_unit_hash": "e81228c371ff23e30cd8b19a7b9c749bd448d24cca52bfdf6cb6995efebbde13", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 4, + "serialized_chars": 9391, + "estimated_input_tokens": 2348, + "estimated_eval_tokens": 4952 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 4, + "serialized_chars": 9391, + "estimated_input_tokens": 2348, + "estimated_eval_tokens": 4952 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is heterogeneous: a Whisper test/tokenizer fix, a test-only placeholder patch for CLIP/ViT, an FP8 dequantization fix, and an FSDP tied-weights cleanup. They do not look like duplicates of one underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:44815 is the strongest standalone change: it fixes a concrete FP8 dequantization/code-loading path and includes multiple implementation touchpoints plus tests.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:44815 is the most substantive and coherent fix in the set; the others are either test-only or address unrelated model/runtime paths.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Both are test-oriented, but they target different models and different changes; one updates Whisper tokenizer behavior, the other injects placeholder asserts into CLIP/ViT tests. Not the same underlying bug." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper decode/timestamp handling is unrelated to FP8 dequantization and model loading. Different subsystems and concrete fixes." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46102", + "accept": false, + "reason": "A dummy test patch for CLIP/ViT is unrelated to FSDP tied-weights cleanup. They do not fix the same code-path problem." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "The CLIP/ViT test-only assertions are not the same change as the FP8 dequant/loading fix; shared test activity 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": 9, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43488-5", + "nodes": [ + "pull_request:43488", + "pull_request:43532", + "pull_request:45932", + "pull_request:46100", + "pull_request:46114" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:46114", + "pull_request:43488|pull_request:43532", + "pull_request:43532|pull_request:46114", + "pull_request:45932|pull_request:46100", + "pull_request:46100|pull_request:46114" + ], + "prepared_review_unit_hash": "3f540bc0dc3f077f65a36c7178c2a528aa381b020704a20e90971b85d5fecea0", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Most PRs are unrelated test/CI noise in ViT or repo-bot/Grafana experiments. The only true duplicate-looking pair is the two identical ViT test-migration PRs (46100 and 46114).", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46100 is the best canonical representative because it is the earlier, cleaner instance of the exact ViT test-migration diff that is duplicated by 46114.", + "best_issue_reason": null, + "best_pr_reason": "PR 46100 best represents the duplicate cluster: its diff matches 46114 exactly in the same ViT test file, while the other PRs are clearly different changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:46114", + "accept": false, + "reason": "Both are repo-test PRs touching ViT tests, but the actual changes are different: 43488 is a setup/version/dependency bot check, while 46114 is a failing-test migration snippet. Not the same underlying change." + }, + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Shared ViT test file is not enough; 43488 is a packaging/version check, whereas 43532 changes CI workflows and helpers. Different concrete problems." + }, + { + "left": "pull_request:43532", + "right": "pull_request:46114", + "accept": false, + "reason": "43532 is a workflow/CI plumbing change with multiple files, while 46114 is only a ViT test stub change. They do not fix the same bug or implement the same change." + }, + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "45932 is a broad observability/model config PR across many files; 46100 is a small ViT test migration. Same general test context, but not the same underlying change." + }, + { + "left": "pull_request:46100", + "right": "pull_request:46114", + "accept": true, + "reason": "These are effectively the same PR change: identical diff in the same ViT test file, adding the same failing test assertions. They could be merged as one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: 46100 and 46114 have identical ViT test diffs and are a plausible duplicate pair, while the other pairings are reasonably rejected as different changes. The summary stays conservative and does not overstate evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 10, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43612-2", + "nodes": [ + "pull_request:43612", + "pull_request:43613" + ], + "soft_pairs": [ + "pull_request:43612|pull_request:43613" + ], + "prepared_review_unit_hash": "bd01a73da5613cd1d14347800a847c0caa0f2ebd0b4005f9ec1b299eebd2a590", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6081, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6081, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs add separate pipeline features for different promptable segmentation tasks. They share scaffolding but target different APIs, docs, model integrations, and tests, so they are related siblings rather than duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: 43612 and 43613 implement different pipelines (concept vs visual segmentation) with different codepaths and model coverage.", + "best_issue_reason": null, + "best_pr_reason": "43613 is the broader, more integrated PR (more model adapters/tests) and is the better representative of this feature family, but it is still distinct from 43612.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Shared pipeline/docs scaffolding is not enough; they expose different pipeline classes and work on different model stacks, so they are not the same underlying change and should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 11, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43651-2", + "nodes": [ + "pull_request:43651", + "pull_request:43663" + ], + "soft_pairs": [ + "pull_request:43651|pull_request:43663" + ], + "prepared_review_unit_hash": "5ff24214fdb38904c51aa47b76c00348e76928a022bb481f6ad3abeb06fb4288", + "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 unrelated trainer.py pull requests: one changes gradient-accumulation loss scaling behavior, the other extracts signature-column logic for overrideability. They share a file but not the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: 43651 addresses loss scaling in training_step, while 43663 adds a separate override point for signature column filtering; they are distinct changes.", + "best_issue_reason": null, + "best_pr_reason": "If a representative PR is needed, 43651 is the more concrete behavior fix, but it should not be treated as a duplicate of 43663.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Different trainer extension points and different code paths: GA loss scaling vs. signature-column filtering. Sharing trainer.py is too broad to imply duplication." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 12, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43656-2", + "nodes": [ + "pull_request:43656", + "pull_request:43842" + ], + "soft_pairs": [ + "pull_request:43656|pull_request:43842" + ], + "prepared_review_unit_hash": "405f758a3657c7cdafd1a46c2dc701f7d8437adcd3cc47c67460702516581f77", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs touch the same CLI module and both deal with import-time annotation/name-resolution problems, but they fix different concrete points in `serve.py` rather than the same exact bug.", + "confidence": 0.63, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43656 is the broader, earlier fix for the CLI NameError/annotation issue and has more discussion activity, so it is the better cluster representative.", + "best_issue_reason": null, + "best_pr_reason": "PR #43842 is the most explicit and narrowly scoped fix: it directly targets the `TypeAdapter` NameError path and states the pydantic-absent case clearly.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Both modify `src/transformers/cli/serve.py`, but one fixes broader type-annotation/runtime issues with `__future__ annotations` while the other specifically handles `TypeAdapter` import/type-checking when pydantic is absent; they are related follow-ups, not the same concrete change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 13, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43779-2", + "nodes": [ + "pull_request:43779", + "pull_request:43816" + ], + "soft_pairs": [ + "pull_request:43779|pull_request:43816" + ], + "prepared_review_unit_hash": "e7c674adcc031aa8883109a5dfc678b4407fb20b10764302092e384d031ba83a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests appear to implement the same SwanLab integration change: adding run id/resume support in `integration_utils.py` for issue #43698, with one variant also expanding the docs. They are strong duplicate PR candidates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43816 is the cleaner representative: it carries the same SwanLab id/resume support plus the accompanying documentation updates, making it the more complete version of the shared change.", + "best_issue_reason": null, + "best_pr_reason": "PR #43816 is the best global representative because it appears to include the full feature change and docs for SwanLab run id/resume support, while #43779 is the narrower implementation-focused variant.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43779", + "right": "pull_request:43816", + "accept": true, + "reason": "Same underlying SwanLab callback enhancement for id/resume support in the same file and same issue target; the doc differences do not change the core code-path change, so these are duplicate PRs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs target issue #43698, touch the same file, and the previews indicate the same SwanLab id/resume support with #43816 adding docs. The duplicate/mergeable verdict is conservative enough given the shared target and overlapping code change." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44024", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44066|pull_request:44071", + "pull_request:44018|pull_request:44068", + "pull_request:44066|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44019|pull_request:44722", + "pull_request:44086|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44722", + "pull_request:44068|pull_request:44116", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44116", + "pull_request:44018|pull_request:44086", + "pull_request:44018|pull_request:44085", + "pull_request:44068|pull_request:44086", + "pull_request:44071|pull_request:44085", + "pull_request:44059|pull_request:44086", + "pull_request:44085|pull_request:44086", + "pull_request:44098|pull_request:44101", + "pull_request:44059|pull_request:44722", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:43997|pull_request:44073", + "pull_request:44071|pull_request:44086", + "pull_request:44073|pull_request:44101", + "pull_request:44017|pull_request:44071", + "pull_request:44059|pull_request:44071", + "pull_request:44056|pull_request:44101", + "pull_request:44073|pull_request:44161", + "pull_request:44059|pull_request:44068", + "pull_request:44018|pull_request:44059", + "pull_request:44101|pull_request:44161", + "pull_request:43997|pull_request:44017", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44018", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44010|pull_request:44073", + "pull_request:44018|pull_request:44129", + "pull_request:43997|pull_request:44074", + "pull_request:43997|pull_request:44129", + "pull_request:44071|pull_request:44129", + "pull_request:44010|pull_request:44018", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154", + "pull_request:44024|pull_request:44076" + ], + "prepared_review_unit_hash": "ccea786088197cbc5b66c738a90be73b9b9e57d2dbfa1f3be50f612ebf494e19", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 + }, + "budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 72, + "serialized_chars": 76934, + "estimated_input_tokens": 19234, + "estimated_eval_tokens": 38724 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a broad standardized-output-tracing refactor across many model-specific PRs. Most pairs only share the migration pattern and are not the same underlying fix; the only clearly grounded duplicate-style linkage is the GPT-J/CodeGen follow-up around PR 44722.", + "confidence": 0.79, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44722 is the best representative because it is the clearest later GPT-J/CodeGen output-tracing refactor and the strongest grounded follow-up in the cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 44722 is the best single PR to anchor the cluster: it is the most concrete GPT-J/CodeGen standardized-output follow-up and the closest to the mergeable duplicate candidates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Same broad tracing theme, but FNet/CVT and GPT-J are different model paths." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Different model families and code paths; shared refactor style is not enough." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": false, + "reason": "Both are GPT-J-series refactors, but this looks like separate variant work rather than one mergeable bugfix." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "Different vision backbones and files; same migration pattern only." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet and GPT-J/CodeGen are unrelated code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen vs EfficientNet is only thematic overlap." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa v2 are unrelated implementations." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen and MGP-STR are different model code paths." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen and MPT are different architectures." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-Neo pair is only soft similarity; not clearly the same mergeable bugfix." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen vs GPT-Neo are separate model implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are different code paths." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet and GPT-J/CodeGen are unrelated model paths." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen are unrelated implementations." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet and GPT-J are different backbones." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet only share the migration theme." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are unrelated model paths." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are not the same concrete fix." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert/XLM are different models." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are unrelated code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are unrelated implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are different model families." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are separate model families." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44086", + "accept": false, + "reason": "MPT and MGP-STR are different implementations." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-2 and MGP-STR are unrelated models." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J and MGP-STR are unrelated code paths." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert/XLM are different model families." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-2 and GPT-J/CodeGen are not the same fix." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different models." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "Segformer and GPT-J/CodeGen are unrelated code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "Segformer and LongT5 are different model families." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "Segformer and VisualBert are unrelated implementations." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are unrelated code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet and VisualBert are different model families." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT and ImageGPT are unrelated architectures." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44101", + "accept": false, + "reason": "VisualBert and Flaubert/XLM are different implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and Segformer are different vision stacks." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are unrelated model paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are different architectures." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are unrelated implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are different models." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert/XLM are unrelated code paths." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44071", + "accept": false, + "reason": "RegNet and MPT are different architectures." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are unrelated models." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert and VisualBert are different model families." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are unrelated code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are different model families." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44129", + "accept": false, + "reason": "RegNet and SpeechT5 are unrelated implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are unrelated code paths." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBert and GPT-Neo are unrelated implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "FNet/CVT and LongT5 are different model paths." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "FNet/CVT and Segformer are unrelated models." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert and DeBERTa v2 are different model families." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa v2 and TextNet are unrelated models." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "Same GPT-J/CodeGen output-tracing refactor; 44722 looks like the grounded later follow-up on the same code path." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Same GPT-J output-tracing change on the same file path; 44722 plausibly supersedes it." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-J and SpeechT5 are unrelated model implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": false, + "reason": "FNet/CVT and RegNet/ResNet are different model families." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": false, + "reason": "FNet/CVT and SpeechT5 are unrelated code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": false, + "reason": "DeBERTa v2 and SpeechT5 are different models." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44154", + "accept": false, + "reason": "DeBERTa v2 and VITS are unrelated architectures." + }, + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "FocalNet and ImageGPT are unrelated code paths." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster-level summary is mostly grounded, but the soft-edge verdicts are not conservative enough. In the packet, the GPT-Neo pair (44018, 44068) and the GPT-J pair (44066, 44085) are only supported by soft similarity; they are not deterministic accepts and do not clearly look like the same mergeable bugfix. Marking those pairs as true overstates the evidence. The accepted PR pairs that are clearly grounded are the GPT-J/CodeGen follow-ups involving 44722, but the analyst extends the duplicate call beyond what the packet supports." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 27, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "942f15d66d4839bfcc115c8d896aa8d01e93fa1bcbb4a1372bc53a189fed7df4", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 5, + "serialized_chars": 13069, + "estimated_input_tokens": 3268, + "estimated_eval_tokens": 6792 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These PRs are all model-specific migrations to the new standardized output tracing/capture system, not duplicates of one another. They share a refactor theme, but each touches a different code path and model implementation.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #44056 is a clear exemplar of the shared migration pattern: it adds the capture_outputs plumbing and explicit _can_record_outputs mapping in a self-contained model implementation.", + "best_issue_reason": null, + "best_pr_reason": "PR #44056 is the strongest representative of the cluster because it most directly expresses the standardized output-tracing refactor, but it still targets MPNet specifically rather than duplicating the other PRs.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "Both are output-tracing refactors, but they apply to different models (VisualBert vs TextNet) and different code paths. Same migration theme is not enough to treat them as one PR." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "These are separate model migrations (MPNet vs VisualBert). They follow the same refactor pattern, but they are not the same underlying change and would not be merged as one PR." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "Different model implementations with model-specific output-capture wiring. Shared decorator usage does not make them duplicates." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are distinct model code paths. Both migrate toward standardized outputs, but they fix different implementations and are not mergeable as a single concrete change." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert and SpeechT5 are unrelated model internals. Despite similar output-capture refactoring, they are separate changes with different affected layers and behavior." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 15, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43998-14", + "nodes": [ + "pull_request:43998", + "pull_request:43999", + "pull_request:44000", + "pull_request:44001", + "pull_request:44002", + "pull_request:44003", + "pull_request:44004", + "pull_request:44025", + "pull_request:44026", + "pull_request:44027", + "pull_request:44028", + "pull_request:44029", + "pull_request:44030", + "pull_request:46106" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44028", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030", + "pull_request:44027|pull_request:46106", + "pull_request:44026|pull_request:46106", + "pull_request:44001|pull_request:46106", + "pull_request:44002|pull_request:46106", + "pull_request:43998|pull_request:46106" + ], + "prepared_review_unit_hash": "30bb104ebea4a9a53c6f2cd55caa1684c607d23e165798c0bd9092691976fd72", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 45, + "serialized_chars": 43774, + "estimated_input_tokens": 10944, + "estimated_eval_tokens": 22144 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 45, + "serialized_chars": 43774, + "estimated_input_tokens": 10944, + "estimated_eval_tokens": 22144 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are mostly independent model-specific output-tracing refactors plus one unrelated docs toctree fix; they share a theme but not the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: each PR targets a different model/code path, so the cluster is not one duplicate change.", + "best_issue_reason": null, + "best_pr_reason": "No PR is a true cluster representative; if forced, the broadest code change is 44003, but it still isn't the same fix as the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models (`timm_backbone` vs `upernet`); same refactor theme, not the same code path." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Separate model implementations (`timm_backbone` vs `vision_text_dual_encoder`)." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Different model families (`timm_backbone` vs `univnet`); not one merged fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "`timm_backbone` and `mamba`/`falcon_mamba` are unrelated model code paths." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Different architectures (`timm_backbone` vs `rwkv`); only the refactor pattern is shared." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "`mobilenet_v1` and `univnet` are unrelated models." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Separate model-specific changes (`mobilenet_v1` vs `upernet`)." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "`mobilenet_v1` vs `depth_anything` are different code paths and tasks." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Different model wrappers (`mobilenet_v1` vs `vision_encoder_decoder`)." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "Different model wrappers (`mobilenet_v1` vs `speech_encoder_decoder`)." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Different models (`vision_text_dual_encoder` vs `univnet`); not the same bug." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Separate model code paths (`vision_text_dual_encoder` vs `upernet`)." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model families (`vision_text_dual_encoder` vs `mamba`)." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "`vision_text_dual_encoder` and `depth_anything` are distinct models with different fixes." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "`univnet` and `upernet` are unrelated model implementations." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different code paths (`univnet` vs `mamba`/`falcon_mamba`)." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "Separate model-specific refactors (`univnet` vs `depth_anything`)." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families (`upernet` vs `rwkv`)." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "`upernet` and `mamba` are different implementations; only the output-tracing theme matches." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "Different models (`mamba` vs `superpoint`) and different forward logic." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "`mamba` and `codegen` touch unrelated code paths despite both refactoring outputs." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Different models (`timm_backbone` vs `codegen`)." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "Separate model implementations (`univnet` vs `codegen`)." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "`upernet` and `codegen` are unrelated model code paths." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Different models (`vision_text_dual_encoder` vs `codegen`)." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "`depth_anything`/`prompt_depth_anything` vs `vision_encoder_decoder` are different model wrappers." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Different model families (`depth_anything` vs `speech_encoder_decoder`)." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "`depth_anything` and `superpoint` are separate vision models." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families (`depth_anything` vs `rwkv`)." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "`depth_anything` and `dpr` are unrelated model code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "`vision_encoder_decoder` and `rwkv` are unrelated models." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "Different encoder-decoder wrappers (`vision_encoder_decoder` vs `speech_encoder_decoder`)." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "Different models (`vision_encoder_decoder` vs `superpoint`)." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "`vision_encoder_decoder` and `dpr` target different model code." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "Different models (`speech_encoder_decoder` vs `superpoint`)." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families (`speech_encoder_decoder` vs `rwkv`)." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "`speech_encoder_decoder` and `dpr` are separate model implementations." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "Different models (`superpoint` vs `rwkv`)." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "`superpoint` and `dpr` are unrelated code paths." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "Different models (`rwkv` vs `dpr`) and different forward semantics." + }, + { + "left": "pull_request:44027", + "right": "pull_request:46106", + "accept": false, + "reason": "`speech_encoder_decoder` code change vs docs-only toctree edit; not the same fix." + }, + { + "left": "pull_request:44026", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs navigation change is unrelated to `vision_encoder_decoder` output tracing." + }, + { + "left": "pull_request:44001", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs-only toctree fix is unrelated to `univnet`." + }, + { + "left": "pull_request:44002", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs-only toctree fix is unrelated to `upernet`." + }, + { + "left": "pull_request:43998", + "right": "pull_request:46106", + "accept": false, + "reason": "Docs-only toctree fix is unrelated to `timm_backbone`." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 16, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44662-3", + "nodes": [ + "pull_request:44662", + "pull_request:45075", + "pull_request:45401" + ], + "soft_pairs": [ + "pull_request:44662|pull_request:45401", + "pull_request:44662|pull_request:45075", + "pull_request:45075|pull_request:45401" + ], + "prepared_review_unit_hash": "5deb9c40b1e27830f8992722fb8802a53dc777220ecd20a6f8b2818654d720b6", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10912, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10912, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These three pull requests are unrelated model-addition implementations for different architectures (PenguinVL, DeepSeek-OCR-2, and Voxtral TTS). They share only generic Transformers boilerplate areas, so none should be treated as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No duplicate PR exists here; if a representative is needed, #45075 is the most complete/production-ready of the three because it includes model, image processor, converter, docs, and tests for a fully scoped model addition.", + "best_issue_reason": null, + "best_pr_reason": "#45075 is the strongest standalone PR to represent this cluster because it appears more complete than the draft TTS PR and is a full model implementation with conversion and test coverage. The other two PRs target entirely different models and should remain separate.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different model families and tasks: PenguinVL vs Voxtral TTS. Shared auto/docs scaffolding is generic and not evidence of the same bug or change." + }, + { + "left": "pull_request:44662", + "right": "pull_request:45075", + "accept": false, + "reason": "Distinct model additions for different architectures (PenguinVL vs DeepSeek-OCR-2). Overlap is only in standard Transformers registration files and docs, which is too broad to merge as duplicates." + }, + { + "left": "pull_request:45075", + "right": "pull_request:45401", + "accept": false, + "reason": "OCR model implementation vs TTS model implementation are unrelated changes. They touch similar framework entry points, but they do not fix the same concrete code-path problem and should not be merged together." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44979-2", + "nodes": [ + "pull_request:44979", + "pull_request:45363" + ], + "soft_pairs": [ + "pull_request:44979|pull_request:45363" + ], + "prepared_review_unit_hash": "4a8eaf094cbb11deb704cd010c930b3548679762ca8547ec00d4f16feecff58d", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related only at a theme level: one PR adds a generic module-fusion API, while the other adds n-to-1 kernel fusion through KernelConfig and integration hooks. They are not the same underlying change and should not be merged as duplicates.", + "confidence": 0.9, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #44979 is the most canonical anchor for the cluster because it introduces the reusable Module Fusion API itself; the other PR is a separate downstream kernel-fusion application built around KernelConfig.", + "best_issue_reason": null, + "best_pr_reason": "PR #44979 is the best representative PR here because it defines the core fusion abstraction and has focused tests, whereas PR #45363 is a broader, separate feature implementation on top of that idea.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "Different changes: #44979 adds a new module-fusion API layer, while #45363 implements KernelConfig-driven n-to-1 kernel fusion across model loading/integration paths. Shared fusion terminology is not enough to treat them as the same fix/feature." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45086-2", + "nodes": [ + "pull_request:45086", + "pull_request:45317" + ], + "soft_pairs": [ + "pull_request:45086|pull_request:45317" + ], + "prepared_review_unit_hash": "ab6d792972de836fa639a654a699f10af62ec7d582386bcd826a5c059b5e3873", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs target the same Mistral regex patch bug on the same code path in `tokenization_utils_tokenizers.py`; the second PR is a slightly broader follow-up with tests, not a distinct fix.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45317 is the better canonical PR because it addresses the same AttributeError fix as 45086 and adds test coverage, making it the more complete representation of the change.", + "best_issue_reason": null, + "best_pr_reason": "PR 45317 is the best PR overall since it fixes the same concrete bug as 45086 but is more complete and validated by tests.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying bug and same edited code path; 45317 appears to be an expanded version of 45086 with added tests, so they are 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": 19, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45115-2", + "nodes": [ + "pull_request:45115", + "pull_request:45569" + ], + "soft_pairs": [ + "pull_request:45115|pull_request:45569" + ], + "prepared_review_unit_hash": "8a8155fb74b461cb1a89636b667b781c56936eee7c6604de12c9f6df4b02e4f0", + "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": "Two related NemotronH refactor PRs, but they are not duplicates: one is a narrower conversion/inheritance cleanup, the other is a broader model split and documentation/test update. The shared files indicate overlap in the same area, not the same concrete change.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45569 is the broader, more complete NemotronH change: it introduces the dense/sparse model split, updates auto/model registrations, docs, and tests. PR 45115 is a narrower refactor focused on inheritance and conversion mapping, so 45569 is the better canonical representative for the cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR 45569 best represents the main architectural change because it covers the full model-family restructuring and user-facing integration, whereas 45115 is a smaller supporting refactor.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Both touch NemotronH internals, but they are not the same underlying fix. 45115 is a targeted refactor of conversion mapping and class inheritance, while 45569 adds the dense/sparse model variants plus docs, auto-mapping, and tests. They are related but not mergeable as one duplicate PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 20, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45332-2", + "nodes": [ + "pull_request:45332", + "pull_request:45333" + ], + "soft_pairs": [ + "pull_request:45332|pull_request:45333" + ], + "prepared_review_unit_hash": "b21b903d8d03230fc585efbd0d2d34a063c6b3497aa3f65e67aa2fbcbabe7bd1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5398, + "estimated_input_tokens": 1350, + "estimated_eval_tokens": 2956 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5398, + "estimated_input_tokens": 1350, + "estimated_eval_tokens": 2956 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are about heterogeneous layer-wise configuration support, but one focuses on config plumbing while the other adds modeling/cache behavior on top of it. They are related, yet not the same concrete change and would not be safely merged as a single PR.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45332 is the broader umbrella change: it introduces heterogeneous model support across modeling/cache code, while PR #45333 is a narrower config-only foundation that appears to feed into it.", + "best_issue_reason": null, + "best_pr_reason": "PR #45332 is the best representative for the cluster because it spans the larger end-to-end heterogeneous-model feature, not just the configuration layer.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Related feature area, but not the same fix/change: #45333 adds heterogeneous config support, while #45332 adds modeling and cache handling. They share files and concept, yet are separable PRs rather than one mergeable 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": 21, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45415-2", + "nodes": [ + "pull_request:45415", + "pull_request:45425" + ], + "soft_pairs": [ + "pull_request:45415|pull_request:45425" + ], + "prepared_review_unit_hash": "c3b7f9a9cc8591ee478a8a646824841570296a966a261575a47504140b10583b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5513, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5513, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are both typing-related, but they are not the same change: #45415 is a broad repo-wide type-checking sweep, while #45425 is a narrower typing/protocol update centered on modeling_utils/PEFT-related code. They should not be merged as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR for a duplicate pair: #45415 is a broad type-checking sweep across many modules, while #45425 is a focused typing/protocol patch in a different area.", + "best_issue_reason": null, + "best_pr_reason": "#45425 is the better representative PR for this cluster because it is more self-contained and specific, whereas #45415 is a much broader cleanup effort.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Different underlying changes: #45415 is a broad typing sweep across many files, while #45425 adds typing aliases/protocols and related modeling_utils/PEFT updates. Same theme, not the same concrete fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 22, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45487-2", + "nodes": [ + "pull_request:45487", + "pull_request:46101" + ], + "soft_pairs": [ + "pull_request:45487|pull_request:46101" + ], + "prepared_review_unit_hash": "c456e2c7aec2dc7138b1427c0ebb47f9f214603b54c2d4167357185d08428a9a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5131, + "estimated_input_tokens": 1283, + "estimated_eval_tokens": 2822 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5131, + "estimated_input_tokens": 1283, + "estimated_eval_tokens": 2822 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are both device-placement/model-parallel fixes, but they target different model families and different failing code paths, so they are not duplicates. PR 45487 is the broader and more representative fix in this cluster.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45487 is the stronger canonical PR candidate because it addresses a concrete model-parallel device mismatch across multiple affected models/files and includes test coverage; the other PR is a separate Qwen2-Audio fix.", + "best_issue_reason": null, + "best_pr_reason": "PR 45487 is the best representative artifact for the cluster since it is broader, multi-model, and clearly fixes the underlying device mismatch bug with explicit code and test changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45487", + "right": "pull_request:46101", + "accept": false, + "reason": "Both are parallel/device fixes, but 45487 changes token_type_ids handling for AltCLIP/ChineseCLIP/etc. while 46101 fixes audio feature masking in Qwen2-Audio. Different models, different functions, and not plausibly one merged PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 23, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45513-3", + "nodes": [ + "pull_request:45513", + "pull_request:46084", + "pull_request:46098" + ], + "soft_pairs": [ + "pull_request:45513|pull_request:46084", + "pull_request:46084|pull_request:46098" + ], + "prepared_review_unit_hash": "8287a8b5f71d5eb45b1b80536a1c9871ee84b88626b07f9e78a4d29fe6f6f220", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7426, + "estimated_input_tokens": 1857, + "estimated_eval_tokens": 3970 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7426, + "estimated_input_tokens": 1857, + "estimated_eval_tokens": 3970 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster splits into one unrelated PR ([Qwen3.5] GDN linear attention caching) and a Mamba2 cache bug pair. The Mamba2 PRs (46084 and 46098) target the same seq_len>1 use_cache incorrect-results issue; 46098 is the broader, more complete version. ", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46098 is the best representative because it states the full Mamba2 seq_len>1 use_cache bug, explicitly targets the tracking issue, and covers both CPU/GPU paths plus tests/docs.", + "best_issue_reason": null, + "best_pr_reason": "PR 46098 is the strongest candidate to anchor the cluster: it is the most complete fix for the shared Mamba2 bug and is more self-contained than 46084.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45513", + "right": "pull_request:46084", + "accept": false, + "reason": "Different models and code paths: Qwen3.5/OLMO hybrid linear-attention caching versus Mamba2 use_cache handling. No evidence of the same underlying bug." + }, + { + "left": "pull_request:46084", + "right": "pull_request:46098", + "accept": true, + "reason": "Same underlying Mamba2 seq_len>1 cache correctness bug on the same model/file family; 46098 appears to be the broader superset fix, so these are plausibly mergeable into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 45513 is clearly a different model/code path, and 46084/46098 both target the same Mamba2 seq_len>1 use_cache bug with shared issue target/files. The mergeability judgment for 46084 and 46098 is conservative and supported by the evidence. " + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45548-2", + "nodes": [ + "pull_request:45548", + "pull_request:45662" + ], + "soft_pairs": [ + "pull_request:45548|pull_request:45662" + ], + "prepared_review_unit_hash": "56cff4d840ed84fc50028e58a774d045b1da2e477aad75715346b71023bf24f3", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related only at a very high level (both involve expert parallelism and model loading), but they fix different concrete bugs: one is about preserving/loading distributed_config for Accelerate + ZeRO-3, the other is about expert parameters being overwritten under FSDP2/rank-0 broadcast.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR #45662 is the stronger canonical candidate for this cluster because it addresses a specific, well-scoped loading corruption bug across the relevant code path and has the broader set of coordinated changes; #45548 is a narrower serialization/flag-preservation fix for a different failure mode.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Both touch EP/loading, but they target different failure modes and code paths: #45548 preserves distributed_config and adds has_ep for Accelerate/ZeRO-3 loading, while #45662 fixes expert params being overwritten during FSDP2 post-shard handling. They would not plausibly merge 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": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45639-2", + "nodes": [ + "pull_request:45639", + "pull_request:45645" + ], + "soft_pairs": [ + "pull_request:45639|pull_request:45645" + ], + "prepared_review_unit_hash": "1b0ef316b0cca0fce3565772a8ba41f3f9cc00e3f7bec69cc3866c415abc979b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4590, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4590, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same xdist collision problem around captured_info artifacts in Transformers testing/debug logging. PR #45645 is the broader, more complete fix and subsumes the narrower code-only change in #45639.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45645 is the better canonical PR because it includes the core xdist-safe file change plus the CI/log-collection updates needed to preserve and read all captured_info shards.", + "best_issue_reason": null, + "best_pr_reason": "PR #45645 is the strongest overall candidate: it fixes the same underlying captured_info collision bug and covers the operational CI path, while #45639 only changes the library-side helper.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Both PRs target the same underlying xdist-safety bug for captured_info logging and touch the same concrete files; #45645 is an expanded superset of #45639 rather than a different change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 26, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45737-2", + "nodes": [ + "pull_request:45737", + "pull_request:45738" + ], + "soft_pairs": [ + "pull_request:45737|pull_request:45738" + ], + "prepared_review_unit_hash": "cc2ade73a49ec6a8883a7a24c433988645588bcd049e8aad2bce56b1b24daa1c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same MusicgenMelody cache initialization bug in `forward()`: replacing `EncoderDecoderCache(DynamicCache, DynamicCache)` with a single `DynamicCache`. PR #45737 is the more complete duplicate because it bundles the fix with a regression test, while #45738 is the fix-only variant.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45737 is the best canonical PR because it contains the same code fix as #45738 plus a regression test, making it the most complete representation of the bug fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #45737 is the strongest overall candidate: it fixes the concrete cache bug and adds a targeted regression test, whereas #45738 only changes the implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45737", + "right": "pull_request:45738", + "accept": true, + "reason": "Same underlying code-path bug in `musicgen_melody` cache setup; the diffs match on the core fix and the test-only addition in #45737 could plausibly live in the same PR as #45738." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: both PRs make the same `musicgen_melody` cache initialization change, and #45737 only adds a regression test on top of that fix. The soft-edge accept is conservative enough because the diffs are mergeable into one PR for the same bugfix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 27, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46069-2", + "nodes": [ + "pull_request:46069", + "pull_request:46073" + ], + "soft_pairs": [ + "pull_request:46069|pull_request:46073" + ], + "prepared_review_unit_hash": "d069e51ef43de099666188efc9294786b2ba4a6f440bddb1b14fcb9e5385bf71", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4394, + "estimated_input_tokens": 1099, + "estimated_eval_tokens": 2454 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4394, + "estimated_input_tokens": 1099, + "estimated_eval_tokens": 2454 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are both about adding SDPA attention support, but they target different models (ShieldGemma2 vs CTRL) and different code paths, so they are not duplicates or mergeable into one change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "pull_request:46069", + "right": "pull_request:46073", + "accept": false, + "reason": "Different model implementations and test/docs updates; they share a broad attention-dispatch theme but not the same concrete bug or code path, so they should 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 + } + ] +} diff --git a/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/manifest.json b/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..d334686c5de763255a62216a223bfeca321b94a5 --- /dev/null +++ b/snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260520t120132z", + "artifacts": { + "hybrid": "snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-05-20T12:09:58Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260520T120132Z", + "variant": "hybrid" +} diff --git a/snapshots/20260520T120132Z/manifest.json b/snapshots/20260520T120132Z/manifest.json index dda7ab76698e188752b8eb587b8b7a332d0ded4e..863ccdf45a9a23560c4ce93f5367c7b514c99cd1 100644 --- a/snapshots/20260520T120132Z/manifest.json +++ b/snapshots/20260520T120132Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 1095 }, "extracted_at": "2026-05-20T12:01:32Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260520t120132z", + "runs": { + "hybrid-model-20260520t120132z": { + "analysis_id": "hybrid-model-20260520t120132z", + "artifacts": { + "hybrid": "snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260520T120132Z/analysis-runs/hybrid-model-20260520t120132z/manifest.json", + "model": null, + "published_at": "2026-05-20T12:09:58Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260520T120132Z", "watermark": {