diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index 9a393a528d35f4982a4310a4ed9d6682dd847287..d941dc8e550ec86dbfef7c3b8532fa16d019e344 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": "20260611T180022Z", - "generated_at": "2026-06-11T18:10:53Z", + "snapshot_id": "20260612T000041Z", + "generated_at": "2026-06-12T00:09:10Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1621,18 +1621,18 @@ ] }, { - "cluster_id": "cluster-45081-3", - "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "cluster_id": "cluster-46305-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46305.", "status": "open", "confidence": 0.8, - "canonical_issue_number": 45081, - "canonical_pr_number": 45317, + "canonical_issue_number": 46305, + "canonical_pr_number": 46546, "issue_numbers": [ - 45081 + 46305 ], "pr_numbers": [ - 45086, - 45317 + 46310, + 46546 ], "evidence_types": [ "closing_reference", @@ -1641,37 +1641,45 @@ ], "pr_comparisons": [ { - "left_pr_number": 45086, - "right_pr_number": 45317, - "code_similarity": 0.589, - "size_similarity": 0.34, + "left_pr_number": 46310, + "right_pr_number": 46546, + "code_similarity": 0.566, + "size_similarity": 0.444, "file_overlap": 0.5, - "area_overlap": 1.0, - "patch_similarity": 0.136, + "area_overlap": 0.867, + "patch_similarity": 0.16, "shared_filenames": [ - "src/transformers/tokenization_utils_tokenizers.py" + "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py" ], "shared_file_areas": [ { - "filename": "src/transformers/tokenization_utils_tokenizers.py", + "filename": "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", "left_ranges": [ [ - 1360, - 1370 - ], - [ - 1374, - 1380 + 356, + 362 ] ], "right_ranges": [ [ - 1360, - 1370 - ], + 355, + 361 + ] + ] + }, + { + "filename": "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py", + "left_ranges": [ [ - 1374, - 1380 + 269, + 275 + ] + ], + "right_ranges": [ + [ + 269, + 275 ] ] } @@ -1680,18 +1688,18 @@ ] }, { - "cluster_id": "cluster-46305-3", - "summary": "Cluster of 1 issues and 2 PRs centered on issue #46305.", + "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": 46305, - "canonical_pr_number": 46310, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, "issue_numbers": [ - 46305 + 45081 ], "pr_numbers": [ - 46310, - 46546 + 45086, + 45317 ], "evidence_types": [ "closing_reference", @@ -1700,45 +1708,37 @@ ], "pr_comparisons": [ { - "left_pr_number": 46310, - "right_pr_number": 46546, - "code_similarity": 0.642, - "size_similarity": 0.571, - "file_overlap": 0.667, - "area_overlap": 0.867, - "patch_similarity": 0.16, + "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/models/moonshine_streaming/modeling_moonshine_streaming.py", - "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py" + "src/transformers/tokenization_utils_tokenizers.py" ], "shared_file_areas": [ { - "filename": "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "filename": "src/transformers/tokenization_utils_tokenizers.py", "left_ranges": [ [ - 356, - 362 - ] - ], - "right_ranges": [ - [ - 355, - 361 - ] - ] - }, - { - "filename": "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py", - "left_ranges": [ + 1360, + 1370 + ], [ - 269, - 275 + 1374, + 1380 ] ], "right_ranges": [ [ - 269, - 275 + 1360, + 1370 + ], + [ + 1374, + 1380 ] ] } @@ -1804,6 +1804,35 @@ } ] }, + { + "cluster_id": "cluster-46504-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 46504, + "issue_numbers": [], + "pr_numbers": [ + 46504, + 46563 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46504, + "right_pr_number": 46563, + "code_similarity": 0.185, + "size_similarity": 0.923, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.0, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, { "cluster_id": "cluster-44018-2", "summary": "Cluster of 2 related pull requests linked by soft_similarity.", @@ -2191,12 +2220,21 @@ }, { "cluster_id": "cluster-46305-3", - "canonical_pr_number": 46310, + "canonical_pr_number": 46546, "duplicate_pr_numbers": [ - 46546 + 46310 ], "target_issue_number": 46305, "reason": "PRs in cluster-46305-3 are treated as duplicates because they converge on issue #46305 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-46504-2", + "canonical_pr_number": 46504, + "duplicate_pr_numbers": [ + 46563 + ], + "target_issue_number": null, + "reason": "PRs in cluster-46504-2 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 3a89e625e4f42736b6727058e51125e147f3fe7a..2af3de24bef931a29d20d7e87fde8b750a54b4fb 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": "20260611T180022Z", - "generated_at": "2026-06-11T18:10:53Z", + "snapshot_id": "20260612T000041Z", + "generated_at": "2026-06-12T00:09:10Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -111,8 +111,8 @@ "issue:43425|issue:44292", "issue:34928|issue:41720", "issue:43828|issue:45161", - "issue:43541|issue:43716", "issue:43828|issue:44292", + "issue:43541|issue:43716", "issue:43716|issue:46468", "issue:43716|issue:45685", "issue:44918|issue:45161", @@ -132,254 +132,419 @@ "issue:43866|issue:46402", "issue:45800|issue:46304" ], - "prepared_review_unit_hash": "c44e57305c40bbea5f7ccd35f6d1ad82fb23a1caf63e3fcc95213155b68185bf", + "prepared_review_unit_hash": "d369b264f3aba78e0b3b00db699ce6cbe44f074605200fc8d0404ea8ea9e3471", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 63, - "serialized_chars": 46526, - "estimated_input_tokens": 11632, - "estimated_eval_tokens": 23520 + "serialized_chars": 46534, + "estimated_input_tokens": 11634, + "estimated_eval_tokens": 23524 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of unrelated issue reports. A few soft pairs do look like true duplicates, especially the DeepSpeed ZeRO-3/BertModel index-out-of-bounds reports; two other pairs appear to share the same low-level compatibility bug, but most pairs are just broad semantic similarity and should stay separate.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43596 is the clearest canonical duplicate target: it states the concrete DeepSpeed ZeRO-3 + BertModel index-out-of-bounds failure succinctly and broadly enough to subsume the near-identical non-pretrained BertModel variant.", + "summary": "This cluster is mostly a similarity-based grab bag of unrelated runtime, model-loading, backend, and docs issues. The soft pairs share broad keywords like DeepSpeed, dtype/device mismatch, or version incompatibility, but they do not look like the same underlying bug or change, so all soft edges are rejected.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45237 is the most central hub in this set of mostly runtime/hardware failures, so it is the best anchor for the cluster, even though it does not truly duplicate the others.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43596 is the best issue representative because it describes a specific, reproducible code-path failure with a clear error signature and strongest duplicate evidence among the set.", + "best_issue_reason": "issue:45237 is the broadest representative of the cluster\u2019s hardware/backend failure reports, making it the best single issue to keep as a reference point.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Different models and failures: missing config field vs streaming chat/completions inference error." + "reason": "Different failures: a config attribute error in Qwen3.5 versus a streaming chat/completions API crash." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Different problems: GPT-OSS on AMD GPUs vs an MPS histogram kernel type error." + "reason": "Different backends and models: AMD GPU failure in GPT-OSS versus an MPS histogram op limitation in MoE." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Both involve model runtime issues, but the model, device, and failure mode are different." + "reason": "Different model paths and root causes: Mistral-3 image dtype mismatch versus GPT-OSS AMD GPU failure." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Different model-specific dtype/autocast bug vs AMD GPU compatibility issue." + "reason": "Different bugs: Phi MoE autocast dtype mismatch versus GPT-OSS AMD GPU incompatibility." }, { "left": "issue:45237", "right": "issue:46468", "accept": false, - "reason": "GPT-OSS AMD GPU failure and Nemotron example breakage are unrelated." + "reason": "One is a GPT-OSS GPU/runtime bug; the other is a broken Nemotron example, so they are not the same issue." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Different underlying bugs: GPU compatibility vs invalid method call in GraniteMoEHybrid." + "reason": "Different model architecture bug: AMD GPU failure versus GraniteMoEHybrid calling an invalid method." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "DeepSpeed collective mismatch and ONNX export failure are not the same code-path." + "reason": "DeepSpeed collective mismatch in Jamba is unrelated to a ModernBERT ONNX export error." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "ModernBERT ONNX export and Qwen3Moe FP8 failure are unrelated." + "reason": "Different models and code paths: ONNX export failure versus Qwen3Moe FP8 failure." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "Different models and failure modes: export-time ONNX issue vs device-map CUDA assert." + "reason": "ONNX export error and Qwen3 auto-device-map cuda assert are unrelated failure modes." }, { "left": "issue:45901", "right": "issue:46015", "accept": false, - "reason": "A task crash and documentation freshness are not the same underlying issue." + "reason": "A table QA crash is not the same as outdated RAG documentation." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are docs regressions, but they reference different removed features and are separate cleanup items." + "reason": "Both are docs cleanup, but they reference different removed topics and are not the same change." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export bug is unrelated to the GenerationMixin import-path breakage." + "reason": "Model export to ONNX is unrelated to the GenerationMixin import error." }, { "left": "issue:44918", "right": "issue:46165", "accept": false, - "reason": "SFT trainer embedding unpacking and resume-training/trust-remote-code problems are different." + "reason": "TRL SFT embedding unpacking is a different training bug from resuming local model training with trust_remote_code." }, { "left": "issue:44805", "right": "issue:45823", "accept": false, - "reason": "Shape mismatch in masked indexing is not the same as Gemma4 PLE device placement mismatch." + "reason": "Mask shape mismatch and Gemma4 device mismatch are different tensor-path failures." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Both mention tensor/index errors, but the loading-time ZeRO-3 Bert bug is different from the mask shape mismatch bug." + "reason": "Both are IndexErrors, but one is ZeRO-3 Bert init and the other is a mask/tensor shape mismatch." }, { "left": "issue:43638", "right": "issue:46165", "accept": false, - "reason": "Different training/loading workflows and different failures." + "reason": "ZeRO-3 Bert initialization failure is unrelated to resume-training with trust_remote_code." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Same broad symptom class, but not the same underlying bug or code path." + "reason": "Different root causes: ZeRO-3 Bert index error versus indexed tensor/mask shape mismatch." }, { "left": "issue:44805", "right": "issue:46294", "accept": false, - "reason": "Different multimodal template/mask issue vs video token mismatch." + "reason": "Mask/tensor shape mismatch is unrelated to Qwen3.5 video token/feature mismatch." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Mask indexing error and GPT-OSS tensor-parallel issue are unrelated." + "reason": "Mask shape mismatch does not match GPT-OSS tensor-parallel failure." }, { "left": "issue:46165", "right": "issue:46468", "accept": false, - "reason": "Training-resume remote-code issue is unrelated to the Nemotron example breakage." + "reason": "Resume-training/trust_remote_code is unrelated to a broken Nemotron example." }, { "left": "issue:44560", "right": "issue:46165", "accept": false, - "reason": "Video StopIteration and resume-training/trust-remote-code are different bugs." + "reason": "Video StopIteration in Qwen3.5 is a different bug from resume-training with local models." }, { "left": "issue:43596", "right": "issue:46165", "accept": false, - "reason": "ZeRO-3 Bert loading bug is unrelated to resume-training behavior." + "reason": "ZeRO-3 Bert index error and resume-training failure are unrelated." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "BarkProcessor voice preset bug is unrelated to ModernBERT ONNX export." + "reason": "BarkProcessor voice_preset and ONNX export are different subsystems and bugs." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Both involve DeepSpeed, but the concrete failures and models differ." + "reason": "Both involve DeepSpeed, but Jamba collective mismatch and Bert ZeRO-3 index error are different failures." }, { "left": "issue:43596", "right": "issue:43638", - "accept": true, - "reason": "Same DeepSpeed ZeRO-3/BertModel index-out-of-bounds bug; one report is just the non-pretrained Bert variant." + "accept": false, + "reason": "Same broad symptom class, but different Bert variants and trigger paths make this not a duplicate." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Liger Kernel cross-entropy invocation and GPT-OSS tensor-parallel failure are unrelated." + "reason": "Liger Kernel cross_entropy usage is unrelated to GPT-OSS tensor-parallel failure." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Tokenizer detokenization regression and LoRA save/reload mismatch are different issues." + "reason": "DeepSeek detokenization and LoRA merge/reload output mismatch are different bugs." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "NVFP4 runtime failure and LoRA merge/save inconsistency are unrelated." + "reason": "Qwen-3-8B NVFP4 runtime failure is unrelated to LoRA save/reload inconsistency." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Different tensor mismatch scenarios in different training setups." + "reason": "Activation-checkpointing/FSDP recomputation mismatch is a different distributed-training bug from a mask shape error." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "Different DeepSpeed errors and different failure points." + "reason": "Both mention DeepSpeed, but a 2-D weight runtime error is not the same as a collective mismatch." }, { "left": "issue:45859", "right": "issue:45902", - "accept": true, - "reason": "Same Qwen3_5MoeTextRotaryEmbedding offload/materialization bug around inv_freq initialization and device movement." + "accept": false, + "reason": "CPU offload incompatibility and uninitialized-memory after to_empty() are different Qwen3_5Moe rotary embedding bugs." }, { - "left": "issue:43425", - "right": "issue:44292", + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "Bert ZeRO-3 init failure is unrelated to GPT-OSS tensor-parallel issues." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "FSDP checkpointing mismatch is unrelated to the CustomTrainer num_items_in_batch API break." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "ZeRO-3 Bert index error and GPT-OSS tensor-parallel failure are different code paths." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Mistral-3 image dtype mismatch and GraniteMoEHybrid invalid method are unrelated model bugs." + }, + { + "left": "issue:45874", + "right": "issue:46294", + "accept": false, + "reason": "Gemma4 reverse-embedding lookup and Qwen3.5 video token mismatch are different multimodal issues." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Qwen-3-8B NVFP4 runtime failure and GPT-OSS AMD GPU failure are different models/backends." + }, + { + "left": "issue:39290", + "right": "issue:45081", "accept": false, - "reason": "Generic Torch 2.10 incompatibility is not the same as Qwen-3-8B-NVFP4 runtime failure." + "reason": "Gemma3 vLLM config regression and Mistral tokenizer regex patch crash are unrelated." }, { "left": "issue:45800", "right": "issue:46515", - "accept": true, - "reason": "Both report the same torch 2.4.1 incompatibility; one just names the transformers version more explicitly." + "accept": false, + "reason": "Both are compatibility complaints, but they involve different version pairs and different breakage surfaces." + }, + { + "left": "issue:45823", + "right": "issue:45874", + "accept": false, + "reason": "Both are Gemma4 issues, but device-map mismatch and expensive reverse embedding lookup are distinct bugs." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "A broad torch-version incompatibility is not the same as a Qwen-3-8B NVFP4 runtime failure." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP activation-checkpointing mismatch is unrelated to Qwen3 auto device mapping cuda assertions." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Phi MoE autocast dtype mismatch and GPT-OSS tensor-parallel failure are different issues." + }, + { + "left": "issue:43828", + "right": "issue:44292", + "accept": false, + "reason": "Different models and bugs: Phi MoE autocast mismatch versus Qwen-3-8B NVFP4 runtime failure." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Torch Dynamo grouped_mm tracing failure is unrelated to Mistral-3 image dtype mismatch." + }, + { + "left": "issue:43716", + "right": "issue:46468", + "accept": false, + "reason": "A Mistral-3 preprocessing dtype bug is not the same as a broken Nemotron example." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Mistral-3 dtype mismatch and an MPS histogram integer limitation are unrelated backend/model bugs." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "TRL SFT embedding unpacking is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:44292", + "right": "issue:44918", + "accept": false, + "reason": "Same broad family only; Qwen-3-8B NVFP4 runtime failure and Qwen3.5 TRL SFT unpacking are distinct bugs." + }, + { + "left": "issue:45507", + "right": "issue:45685", + "accept": false, + "reason": "GraniteMoEHybrid invalid method and MPS histogram not implemented are unrelated failures." + }, + { + "left": "issue:43716", + "right": "issue:44292", + "accept": false, + "reason": "Different models and code paths: Mistral-3 dtype mismatch versus Qwen-3-8B NVFP4 runtime failure." + }, + { + "left": "issue:43039", + "right": "issue:44918", + "accept": false, + "reason": "Liger Kernel cross_entropy behavior is unrelated to TRL SFT embedding unpacking." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": false, + "reason": "Both are ZeRO-3 index errors, but Gemma3 loading and non-pretrained Bert initialization are different triggers." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs issues, but they refer to different removed commands/tasks and should not be merged." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch in Phi MoE is unrelated to an MPS histogram integer limitation." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 SFT embedding unpacking is unrelated to the MPS histogram backend error." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger Kernel cross_entropy and Phi MoE autocast mismatch are different training-stack bugs." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and CustomTrainer kwarg incompatibility are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP activation-checkpointing mismatch is unrelated to the GenerationMixin import failure." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross_entropy and a mask shape mismatch are not the same bug." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "GPU/runtime failure in GPT-OSS is unrelated to outdated RAG documentation." }, { "left": "issue:43866", "right": "issue:46402", "accept": false, - "reason": "Checkpoint corruption and unloaded text encoder weights are different problems." + "reason": "A corrupted Ovis2 checkpoint and missing Sam3Model text encoder weights are different loading failures." }, { "left": "issue:45800", "right": "issue:46304", "accept": false, - "reason": "These are different compatibility regressions with different version combinations and distributed settings." + "reason": "Both are version-compatibility regressions, but they involve different releases and different distributed-config behavior." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative. The accepted duplicate pairs are well supported by title similarity and the pair scores, especially the ZeRO-3/BertModel reports; the rest of the verdicts avoid overstating weak similarity as duplication." - }, - "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 }, @@ -488,244 +653,250 @@ "issue:43541|issue:45161", "issue:43828|issue:45464" ], - "prepared_review_unit_hash": "84de156f6141dec107f7d9ede0bc7f05249ac09da6b5da4cf162f20a366e1c7f", + "prepared_review_unit_hash": "e5e190a0247b02a27048f4e1e7fc8955f5e980b354e048cf91aca413edebcfa0", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 41, - "serialized_chars": 39734, - "estimated_input_tokens": 9934, - "estimated_eval_tokens": 20124 + "serialized_chars": 39729, + "estimated_input_tokens": 9933, + "estimated_eval_tokens": 20122 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous, with one strong duplicate family around the v5 AutoTokenizer token-ID regression and one plausible ZeRO-3 size-0 IndexError pair. No PRs are present.", - "confidence": 0.81, - "canonical_issue_reason": "issue:45920 is the broadest representative of the clear tokenizer-regression family and subsumes the narrower Granite-only report in issue:45812.", + "summary": "Most pairs are only superficially similar (shared error words or broad subsystem names) and do not look like the same bug. The only strong duplicate family here is the AutoTokenizer token-ID regression, where the broader v5 issue subsumes the Granite-specific report.", + "confidence": 0.87, + "canonical_issue_reason": "issue:45920 is the broadest and most central duplicate-family candidate: it describes the same AutoTokenizer v5 token-ID regression across multiple model families, while issue:45812 is a narrower Granite-only manifestation.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 best captures the underlying v5 AutoTokenizer bug across multiple model families, making it the strongest single issue to keep as the umbrella report.", + "best_issue_reason": "issue:45920 is the best global issue representative because it has the widest reported impact, is still open, and cleanly captures the shared tokenizer regression across several models.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43296", "right": "issue:46304", "accept": false, - "reason": "Different failures and subsystems: vLLM model-load error vs distributed-training compatibility break." + "reason": "Different failure domains: vLLM/PaddleOCR-VL load error vs PyTorch distributed build/support regression." }, { "left": "issue:41762", "right": "issue:43596", - "accept": true, - "reason": "Same ZeRO-3 size-0 IndexError during model init/load; different models, but the underlying bug appears to be the same code path." + "accept": false, + "reason": "Both are IndexError-on-ZeRO-3 reports, but the models and triggering paths differ too much to treat as one bug without stronger evidence." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Both mention model/runtime errors, but the concrete failures are unrelated (missing config attribute vs CUDA assert on auto device map)." + "reason": "Gemma3 config attribute missing in vLLM is unrelated to the Qwen3 auto-device-map CUDA assert." }, { "left": "issue:44589", "right": "issue:46304", "accept": false, - "reason": "Unrelated incompatibilities: float8 storage lookup vs USE_DISTRIBUTED=0 support." + "reason": "Float8 storage type lookup failure is unrelated to the USE_DISTRIBUTED=0 support issue." }, { "left": "issue:43901", "right": "issue:44512", "accept": false, - "reason": "Both are docs-related, but they refer to different documentation regressions." + "reason": "Both are docs issues, but they cover different documentation changes and different underlying behavior." }, { "left": "issue:45823", "right": "issue:46294", "accept": false, - "reason": "Different model families and failure modes: device mismatch in Gemma4 PLE vs Qwen3.5 video/chat template mismatch." + "reason": "Gemma4 device-mapping mismatch and Qwen3.5 video chat-template token mismatch are different model-path problems." }, { "left": "issue:45237", "right": "issue:45901", "accept": false, - "reason": "Different tasks and bugs: AMD GPU support for GPT-OSS vs table-question-answering crash." + "reason": "AMD/GPT-OSS backend failure is unrelated to the table-question-answering crash." }, { "left": "issue:44841", "right": "issue:45823", "accept": false, - "reason": "Different model/processor bugs; same broad multimodal area is not enough." + "reason": "Processor failure for Voxtral is a different code path from Gemma4 PLE device mismatch." }, { "left": "issue:43039", "right": "issue:44560", "accept": false, - "reason": "Liger-kernel cross-entropy routing and Qwen3-VL video StopIteration are unrelated." + "reason": "Liger Kernel cross_entropy usage and Qwen3-vl video StopIteration are not the same underlying bug." }, { "left": "issue:43975", "right": "issue:46015", "accept": false, - "reason": "Different topics: detokenization regression vs outdated RAG examples." + "reason": "Tokenizer detokenization regression is unrelated to outdated RAG docs." }, { "left": "issue:44863", "right": "issue:46015", "accept": false, - "reason": "Checkpoint-loading failure vs documentation update; no shared underlying bug." + "reason": "NemotronH checkpoint-loading failure is unrelated to documentation examples." }, { "left": "issue:45020", "right": "issue:46304", "accept": false, - "reason": "Remote_code breakage is not the same as the USE_DISTRIBUTED=0 / PyTorch support regression." + "reason": "Remote_code breakage and USE_DISTRIBUTED=0 support regression are different compatibility problems." }, { "left": "issue:45812", "right": "issue:45920", "accept": true, - "reason": "Same AutoTokenizer token-ID regression in v5; issue:45812 is a narrower Granite-specific instance of the broader multi-model report in issue:45920." + "reason": "Same underlying v5 AutoTokenizer regression: both report wrong token IDs across model families, with 45920 being the broader umbrella." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Corrupted checkpoint vs implementation cannot load checkpoints are different problems." + "reason": "Corrupted Ovis2 checkpoint and NemotronH loading failure are distinct model-specific problems." }, { "left": "issue:43872", "right": "issue:46304", "accept": false, - "reason": "bitsandbytes constructor incompatibility is unrelated to distributed build support." + "reason": "bitsandbytes Int8Params constructor incompatibility is unrelated to the PyTorch distributed toggle regression." }, { "left": "issue:44292", "right": "issue:46015", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime error is unrelated to RAG docs." + "reason": "Qwen-3 NVFP4 runtime failure is unrelated to RAG docs updates." }, { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Both mention sliding-window-related issues, but the concrete errors and model/config paths differ." + "reason": "Both mention sliding_window, but they affect different model families and appear to be different bugs." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Both are IndexErrors, but the actual signatures and contexts differ enough that they do not look like one bug." + "reason": "Mask-shape mismatch during generation and ZeRO-3 BertModel zero-size tensor error are not the same issue." }, { "left": "issue:43792", "right": "issue:45901", "accept": false, - "reason": "Whisper model execution failure and table-question-answering crash are unrelated." + "reason": "Whisper cannot run and table-question-answering crashes are unrelated." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "Unit-test model loading failure vs checkpoint corruption are not the same underlying issue." + "reason": "GLM-4.7-Flash unit-test load failure and Ovis2 checkpoint corruption are different failures." }, { "left": "issue:43425", "right": "issue:45070", "accept": false, - "reason": "Separate version-compatibility problems affecting different layers." + "reason": "Torch version incompatibility and a pydantic PretrainedConfig regression are separate compatibility issues." }, { "left": "issue:43596", "right": "issue:45800", "accept": false, - "reason": "Both involve compatibility/ZeRO-3 style failures, but they are distinct bugs with different triggers and versions." + "reason": "ZeRO-3 BertModel index error and torch 2.4.1/transformers 5.8.0 incompatibility are not the same bug." }, { "left": "issue:43975", "right": "issue:45901", "accept": false, - "reason": "Different feature areas and symptoms; not a shared code-path bug." + "reason": "DeepSeek detokenization regression is unrelated to table-question-answering crashes." }, { "left": "issue:43792", "right": "issue:45070", "accept": false, - "reason": "Whisper loading/runtime issue is unrelated to the pydantic PretrainedConfig regression." + "reason": "Whisper load/run failure and pydantic model regression are unrelated." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Same model family, but different APIs and symptoms; not clearly one concrete bug." + "reason": "Both concern Qwen3.5 chat, but one is a model crash and the other is a streaming chat/completions API failure; not enough to merge." }, { "left": "issue:44292", "right": "issue:45901", "accept": false, - "reason": "Unrelated model runtime error vs table QA crash." + "reason": "Qwen-3 NVFP4 runtime failure is unrelated to table-question-answering." }, { "left": "issue:39290", "right": "issue:45823", "accept": false, - "reason": "Different model/config failures; only a superficial relation to model loading." + "reason": "Gemma3 sliding_window_pattern error and Gemma4 PLE device mismatch are different model-path issues." }, { "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Streaming chat failure and MPS histogram error are unrelated." + "reason": "Qwen3.5 streaming failure is unrelated to the MoE MPS histogram op limitation." }, { "left": "issue:43541", "right": "issue:45685", "accept": false, - "reason": "MoE grouped_mm tracing failure vs MPS int histogram limitation are different bugs." + "reason": "Grouped_mm tracing failure and MPS histogram_mps integer support error are distinct MoE issues." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "MoE tracing/grouped_mm failure and autocast dtype mismatch are different runtime problems." }, { "left": "issue:36010", "right": "issue:39290", "accept": false, - "reason": "ImportError for GenerationMixin and Gemma3/vLLM config regression are unrelated." + "reason": "Importing GenerationMixin and a Gemma3 config attribute error are unrelated." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "Mask-shape IndexError and Qwen3.5 embedding unpacking failure are different bugs." + "reason": "Both are IndexError-ish, but one is a mask/tensor shape mismatch and the other is TRL SFT unpacking failure; different code paths." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Int4 memory/OOM regression is not the same as tied-weights issues on translation models." + "reason": "Int4 OOM regression and tied-weights translation-model regression are different v5 issues." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Grouped_mm tracing failure and GPT-OSS TP issue are different MoE-related bugs, not one duplicate." + "reason": "Grouped_mm tracing error in Mixtral and GPT-OSS tensor-parallel failure are different MoE bugs." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to Qwen3.5 chat streaming failure." + "reason": "Autocast dtype mismatch on Phi-tiny-MoE and Qwen3.5 streaming API failure are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative: it identifies one clear AutoTokenizer v5 regression family, one plausible ZeRO-3 size-0 IndexError pair, and otherwise treats the cluster as heterogeneous. The soft-edge verdicts are generally well supported by the issue titles and do not overstate evidence. No PRs are present, so there is no mergeability concern." + "feedback": "The summary is grounded in the packet, and the only accepted duplicate family\u2014issue:45812 with issue:45920\u2014is well supported by the titles. The remaining soft-edge verdicts stay conservative and do not overclaim equivalence from superficial similarity alone." }, "evaluator_used": true, "retried": false, @@ -828,220 +999,220 @@ "issue:43644|issue:43950", "issue:36331|issue:41762" ], - "prepared_review_unit_hash": "d084c63c466d90166f49a83d31452a7398531c528f6ff9c35db4bf8adb063feb", + "prepared_review_unit_hash": "4a273d5888c6e9d01ad4d9fb4214ea7ad322b355dc824eaacd5a0caa5ac296c0", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36943, - "estimated_input_tokens": 9236, - "estimated_eval_tokens": 18728 + "serialized_chars": 36939, + "estimated_input_tokens": 9235, + "estimated_eval_tokens": 18726 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most candidate links are weak and span different bugs or model families. The only solid duplicate pair is the non-persistent-buffer regression reports (43950/44534).", - "confidence": 0.74, - "canonical_issue_reason": "Issue 43950 is the clearest representative of the only true duplicate group here: it states the non-persistent-buffer corruption regression directly and concretely, with 44534 reading like a later paraphrase of the same bug.", + "summary": "This cluster is mostly a loose grouping of unrelated Transformers bug reports. The only clear duplicate chain is the non-persistent-buffer regression in v5.x, where 43644 and 44534 appear to be the same underlying bug as 43950. Most other pairs are only superficially similar and should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "43950 is the clearest, most specific report of the duplicated v5 non-persistent-buffer corruption bug (`from_pretrained` + `register_buffer(persistent=False)`), and it matches the duplicate reports 43644 and 44534 best.", "canonical_pr_reason": null, - "best_issue_reason": "43950 is the best overall issue because it is the most specific and actionable report of the shared buffer-corruption regression; it captures the root symptom more clearly than the broader follow-up wording in 44534.", + "best_issue_reason": "43950 is the strongest canonical issue candidate because it describes the root regression precisely and has the most actionable reproduction context among the matching reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "Both are v5 loading/config regressions, but one is an unexpected arg TypeError and the other is a PretrainedConfig/pydantic field break; different failure modes." + "reason": "Different bugs: empty-init helper argument handling vs a pydantic `PretrainedConfig` field regression." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "GGUF support for gpt-oss and a Qwen3.5 streaming API failure are unrelated bugs." + "reason": "Different functionality: GGUF architecture support vs a streaming chat/completions failure." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and a config/pydantic regression are not the same code-path problem." + "reason": "Qwen3-MoE sliding-window behavior is unrelated to the pydantic field regression." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "One is remote_code breakage in recent transformers; the other is a torch/transformers version incompatibility." + "reason": "Both are version-compatibility complaints, but they target different failure modes and code paths." }, { "left": "issue:43716", "right": "issue:46294", "accept": false, - "reason": "Image preprocessor dtype mismatch and video feature/token mismatch are different multimodal bugs." + "reason": "Different multimodal bugs: image preprocessor dtype mismatch vs video feature/token mismatch." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Grouped_mm tracing failure and missing Float8 storage type are unrelated runtime errors." + "reason": "Distinct runtime errors in different paths: torch dynamo grouped_mm tracing vs float8 storage lookup." }, { "left": "issue:44479", "right": "issue:46294", "accept": false, - "reason": "Both involve video input regressions in Qwen models, but the titles point to different concrete failures and model scopes." + "reason": "Both mention video input, but one is a broad v5.3 regression and the other is a chat-template mismatch for one model." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Both are index/shape errors, but they occur in different contexts and with different failure mechanics." + "reason": "Both are index errors, but the mask-shape mismatch and ZeRO-3 zero-size tensor issue are different bugs." }, { "left": "issue:43054", "right": "issue:43994", "accept": false, - "reason": "Both mention SigLIP2, but one is about embedding quality and the other about nonsensical AutoModel/pipeline output; not enough to treat as the same bug." + "reason": "Same model family, but one is an embedding-quality complaint and the other is an AutoModel/pipeline correctness issue." }, { "left": "issue:45759", "right": "issue:45902", "accept": false, - "reason": "AutoModel config unwrapping and uninitialized rotary embedding memory are different initialization issues." + "reason": "Different underlying problems: config unwrapping vs uninitialized rotary embedding materialization." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP weight-shape failure and TRL SFT embedding unpacking are distinct training-time bugs." + "reason": "FSDP `'weight' must be 2-D` is unrelated to the TRL SFT embedding-unpacking failure." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Different multimodal/input handling problems with no concrete shared failure path." + "reason": "Different subsystems and symptoms: multimodal preprocessing vs trainer/input unpacking." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Jupyter custom-model initialization and flash-attn fallback gating are unrelated." + "reason": "Notebook model-definition/init regression is unrelated to flash-attn fallback gating." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "High perplexity on gpt-oss and a GLM5 issue are separate model-specific reports." + "reason": "Unrelated reports: perplexity on gpt-oss-20b vs GLM5." }, { "left": "issue:45632", "right": "issue:45698", "accept": false, - "reason": "Both involve custom-code loading, but one is a cache-path collision and the other is loading the wrong module after save_pretrained; too different to merge." + "reason": "Both involve custom code loading, but one is a cache-path collision and the other is loading the wrong saved module." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Different tied-weights problems in different model families and code paths." + "reason": "Tied-weights problems, but on different model families and likely different root causes." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "FineGrainedFP8Config failure and flash-attn fallback checks are unrelated." + "reason": "Fine-grained FP8 model failure is unrelated to flash-attn fallback behavior." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache crash on Qwen3.5 is not the same as translation-model tied-weight issues." + "reason": "PagedAttentionCache linear_attention crash is not the same as the tied-weights translation-model regression." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization/offloading behavior and tied-weight handling are separate concerns." + "reason": "Quantization/offloading behavior is a different bug from tied-weight handling." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "A DSA-indexer discussion and a docs typo are clearly unrelated." + "reason": "A DSA indexer bug and a documentation typo are not duplicates." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading and translation-model tied weights are different bugs." + "reason": "NemotronH checkpoint-loading failure is unrelated to translation tied-weights behavior." }, { "left": "issue:45183", "right": "issue:46463", "accept": false, - "reason": "Both concern input validation messages, but the former is broad XOR validation and the latter is T5 decoder-input validation; not the same concrete bug." + "reason": "Both are input-validation error-message issues, but one is a broad XOR-logic bug and the other is T5-specific; insufficient evidence of the same root cause." }, { "left": "issue:46489", "right": "issue:46491", "accept": false, - "reason": "Different tokenizer regressions affecting different models and symptoms." + "reason": "Different tokenizer regressions affecting different models and different round-trip symptoms." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "TRL input-embedding unpacking and GraniteMoEHybrid invalid method calls are unrelated." + "reason": "Unpacking Qwen3.5 embeddings in TRL is unrelated to GraniteMoEHybrid calling an invalid method." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "A cache-strategy typo and a docs note about a removed command are not duplicates." + "reason": "A cache-strategy typo and a removed-doc-command issue are unrelated documentation bugs." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "DeepSeek-Coder detokenization and Qwen-3 NVFP4 runtime errors are unrelated." + "reason": "Different problems: DeepSeek-Coder detokenization regression vs Qwen NVFP4 runtime failure." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Different failures in different model/runtime paths." + "reason": "Unrelated model/runtime failures." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Both describe the same v5 non-persistent-buffer corruption regression; 44534 is essentially the same bug phrased as a follow-up." + "reason": "Same underlying v5 non-persistent-buffer corruption regression; the titles describe the same bug in different words." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "SigLIP2 docs errors and SigLIP2 embedding quality complaints are related only by model family, not the same defect." + "reason": "Docs example errors and a text-embedding quality regression are not the same issue." }, { - "left": "issue:36331", - "right": "issue:41762", - "accept": false, - "reason": "Unexpected compute_loss keyword argument and ZeRO-3 index-out-of-bounds are unrelated training/loading bugs." + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Both report the same non-persistent-buffer junk/corruption regression in Transformers v5." } ] }, "evaluator_result": { "accept": true, - "feedback": "The result is grounded in the packet and stays conservative overall. The accepted 43950/44534 pair is well supported by the titles, and the summary does not overclaim beyond that. The rejected soft-edge pairs are generally framed cautiously as different failure modes, model families, or code paths." + "feedback": "Grounded overall. The only clearly supported duplicate chain is the v5 non-persistent-buffer corruption regression (43644 / 44534 / 43950), and the remaining soft-edge rejections are conservative based on the titles. Minor caveat: phrases like \"most actionable reproduction context\" are not directly supported by the packet, so keep that wording a bit softer in future summaries." }, "evaluator_used": true, "retried": false, @@ -1133,8 +1304,8 @@ "issue:43643|issue:45020", "issue:41762|issue:42915", "issue:45685|issue:45823", - "issue:43782|issue:44960", "issue:44991|issue:45399", + "issue:43782|issue:44960", "issue:43824|issue:43872", "issue:43881|issue:43883", "issue:45020|issue:45823", @@ -1143,23 +1314,23 @@ "issue:43828|issue:45823", "issue:42491|issue:44368", "issue:45715|issue:46315", - "issue:45020|issue:45356", - "issue:43742|issue:45399" + "issue:43742|issue:45399", + "issue:45020|issue:45356" ], - "prepared_review_unit_hash": "17e1538b6ee53f4ae5379c398bdc6d3a47c487536fbc8f722446b1191416ae98", + "prepared_review_unit_hash": "84d287f55ee3beb1e96fc6ab23d327641f9abaea3682f0b32f9bf76726acac7b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37919, + "serialized_chars": 37920, "estimated_input_tokens": 9480, "estimated_eval_tokens": 19216 }, @@ -1168,220 +1339,217 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The only clearly supported duplicate-like match is the SigLIP2 HF-vs-JAX discrepancy pair; most other soft pairs are false positives. In particular, the Qwen3.5 3D position_ids reports remain separate issues and should not be treated as a strong match.", - "confidence": 0.61, - "canonical_issue_reason": "Issue 45920 is the best umbrella representative: it describes a broad v5 tokenizer regression affecting multiple model families with concrete wrong-ID symptoms.", + "summary": "This cluster is a loose collection of mostly unrelated Transformers bug reports (tokenizers, remote code, MoE/FSDP/precision, model-loading, and performance regressions). The soft pairs are thematically similar but do not look like the same underlying bug or change, so I would not merge any of them as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": "Issue 45020 is the broadest umbrella-style report in the set, centered on recent Transformers version regressions affecting remote-code model loading, which makes it the best single issue to anchor the cluster even though the cluster is not truly one duplicate group.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45920 is the most representative cluster member because it captures the shared cross-model tokenizer regression theme without overfitting to one model-specific report.", + "best_issue_reason": "Issue 45020 is the most representative/high-level issue in this mixed set; it is broader and more central than the narrowly scoped model-specific regressions, so it is the best issue to keep as the cluster anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Different models and failure modes; both are dtype/attention-related but not the same concrete bug." + "reason": "Both mention Qwen3.5/position_ids and dtype/attention failures, but one is a Flash Attention illegal-access crash and the other is an RLHF NaN/SDPA-BF16 regression; different failure modes and code paths." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Both are quantization/storage errors, but the traceback and code paths differ enough to treat as separate bugs." + "reason": "Both are quantization/storage compatibility errors, but they involve different concrete breakages: bitsandbytes Int8Params constructor mismatch vs missing Float8 storage lookup." }, { "left": "issue:45920", "right": "issue:46489", "accept": false, - "reason": "Both concern tokenizer regressions, but this is too broad/model-diverse to be the same concrete underlying issue." + "reason": "Both concern tokenizer regressions, but one is a broad wrong-token-IDs issue across many models while the other is a DeepSeek-Coder v1-specific output bug tied to a particular prior PR gap." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Checkpoint corruption and load-time noise are unrelated symptoms." + "reason": "One is a corrupted checkpoint/model artifact problem; the other is a generic logging/noise issue when loading transformers. They are unrelated." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "One is SigLIP2 embedding quality, the other is a generic tokenization-change report." + "reason": "Both discuss tokenization/model quality changes across versions, but one is a SigLIP2 text-embedding quality complaint and the other is a tokenizer-version behavior change; not the same bug." }, { "left": "issue:43950", "right": "issue:45902", "accept": false, - "reason": "Both involve buffers/materialization, but one is generic from_pretrained corruption and the other is a specific model buffer init bug." + "reason": "One is a from_pretrained buffer persistence corruption regression; the other is uninitialized memory in a specific MoE rotary embedding after meta-to-empty materialization. Different mechanisms." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Same model family, but zero probabilities and worse embeddings are not clearly the same underlying defect." + "reason": "Both are about SigLIP/SigLIP2 outputs, but one reports zero probabilities while the other reports worse text embeddings and a JAX/HF discrepancy; not the same underlying defect." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad tokenizer error messaging vs missing model attribute during load are different issues." + "reason": "Both are loading-time errors, but one is an AutoTokenizer error for Voxtral and the other is a missing attribute on MolmoForCausalLM. Different models and different faults." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Both involve Qwen3.5, but flash-attn packed-sequence crashes and CPU-offload incompatibility are distinct code paths." + "reason": "Both involve Qwen3.5 and rotary/position handling, but one is a Flash Attention crash from 3D position_ids and the other is CPU offload incompatibility in a MoE rotary embedding. Not mergeable as one bug." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Training-memory complaints in different subsystems; no shared concrete bug is evident." + "reason": "Both mention memory/gradient problems with PEFT/LoRA or quantization, but one is gradient checkpointing not taking effect and the other is increased CUDA reserved memory leading to OOM. Different root causes." }, { "left": "issue:45663", "right": "issue:45902", "accept": false, - "reason": "Different models and failure mechanisms; one is FSDP2 shared_kv_states rebuilding, the other is uninitialized rotary buffer materialization." + "reason": "Gemma4 FSDP2 KeyError from per-layer shared_kv_states rebuilding is unrelated to Qwen3_5Moe rotary embedding uninitialized memory." }, { "left": "issue:43054", "right": "issue:43493", - "accept": true, - "reason": "Both point to the same SigLIP2 HF-vs-JAX implementation discrepancy causing degraded text/embedding behavior." + "accept": false, + "reason": "Both are about SigLIP2 discrepancies, but one is a subjective embedding quality regression and the other is a specific HF-vs-JAX implementation mismatch. Related area, not duplicate." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Tokenizer loading UX issue vs model loading failure; not the same bug." + "reason": "Voxtral AutoTokenizer loading and glm-4v-9b loading failures are separate model-specific issues with different error paths." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Generic load noise and a specific model load failure are not duplicates." + "reason": "Generic loading noise versus failure to load a specific model (ScandiBERT). Same broad subsystem, not the same bug." }, { "left": "issue:24643", "right": "issue:28530", "accept": false, - "reason": "Training DeepSpeed weight-shape error and early-stopping metric handling are unrelated." + "reason": "DeepSpeed weight shape error during training is unrelated to the early-stopping metric-for-best-model warning/disablement issue." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Tokenizer detokenization bug and Python 3.13 flash_attn import error are unrelated." + "reason": "One is incorrect detokenization for a specific model; the other is a Python 3.13 import_utils KeyError for flash_attn. Different components." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "A noise/logging complaint is not the same as a concrete model loading failure." + "reason": "Both involve loading/reporting issues, but one is generic load noise and the other is a specific tokenizer/model load failure for cjvt/sleng-bert." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "Both mention memory, but one is checkpointing ineffective and the other is Qwen3 MoE training memory usage." + "reason": "Both are training/memory related, but one is checkpointing not working with LoRA and the other is inefficient memory usage in Qwen3 MoE training. Different fixes." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "Missing fields from trust_remote_code config return is different from broad remote_code loading breakage." + "reason": "Both touch config/loading regressions, but one is missing fields from AutoConfig with trust_remote_code and the other is a broad remote_code compatibility issue. Too broad to call duplicate." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Different models and different DeepSpeed/FP8 failure modes." + "reason": "Gemma3 ZeRO-3 loading IndexError and Qwen3Moe FineGrainedFP8Config failure are distinct model/config-specific problems." }, { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "MPS histogram op failure and Gemma4 device-mismatch are unrelated." + "reason": "MPS histogram_int error in MoE and Gemma4 PLE device mismatch under device_map=auto are unrelated backend/device-placement bugs." }, { - "left": "issue:43782", - "right": "issue:44960", + "left": "issue:44991", + "right": "issue:45399", "accept": false, - "reason": "Qwen3VL weight_only load error and GLM5 are unrelated model-loading problems." + "reason": "Tokenizer loading failure for a specific model and fallback-to-flash-attn2 blocking are different code paths and symptoms." }, { - "left": "issue:44991", - "right": "issue:45399", + "left": "issue:43782", + "right": "issue:44960", "accept": false, - "reason": "Tokenizer load failure and flash-attn fallback gating are unrelated." + "reason": "Qwen3VL weight_only load error and a GLM5 issue are unrelated model-loading failures." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Different model import/load failure vs bitsandbytes parameter-construction incompatibility." + "reason": "Importing Qwen2.5-VL class and bitsandbytes Int8Params constructor mismatch are different failures with no shared fix path." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "Two distinct model-loading failures with different root causes." + "reason": "Both are model-loading errors, but one is glm-4v-9b loading failure and the other is a missing all_tied_weights_keys attribute on MolmoForCausalLM." }, { "left": "issue:45020", "right": "issue:45823", "accept": false, - "reason": "A broad remote_code regression and a Gemma4 device-map bug are not the same concrete issue." + "reason": "Remote-code regressions and Gemma4 PLE device-mismatch are not the same bug; they only share a recent-version symptom surface." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Model-registration edge case and remote_code breakage are different problems." + "reason": "Tokenizer mapping failure when adding a new model-like entry is a narrow registry bug, not the broader remote_code loading regression." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Tokenizer regex crash vs flash-attn fallback gating are unrelated." + "reason": "One is a tokenizer regex patch crash with a backend_tokenizer attribute error; the other is a flash-attn2 fallback gating issue. Different fixes." }, { "left": "issue:43828", "right": "issue:45823", "accept": false, - "reason": "Different model families and failure mechanisms; no shared code-path bug is evident." + "reason": "Both involve model execution on specialized hardware paths, but one is an autocast dtype mismatch in Phi MoE and the other is a Gemma4 device-map placement issue." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "LoRA compatibility across versions and a tied-weights warning are not the same issue." + "reason": "LoRA compatibility across hf4.x/hf5.x for Qwen3 MoE and a warning about tie_word_embeddings in Qwen3.5-27B are not the same underlying bug." }, { "left": "issue:45715", "right": "issue:46315", "accept": false, - "reason": "Both are tokenizer performance bugs, but the specific loops and APIs differ." + "reason": "Both are tokenizer-performance regressions, but one is convert_ids_to_tokens rebuilding all_special_ids per iteration and the other is slow convert_tokens_to_ids for slow tokenizers with many added tokens." }, { - "left": "issue:45020", - "right": "issue:45356", + "left": "issue:43742", + "right": "issue:45399", "accept": false, - "reason": "Generic remote_code breakage and Kimi-K2.5 tokenizer regression are not clearly the same bug." + "reason": "Key error loading a specific model and flash-attn2 fallback blocked by other checks are unrelated." }, { - "left": "issue:43742", - "right": "issue:45399", + "left": "issue:45020", + "right": "issue:45356", "accept": false, - "reason": "Model load key error and flash-attn fallback checks are unrelated." + "reason": "Both are version regressions around loading/tokenization, but one is a broad remote_code breakage and the other is a Kimi-K2.5 tokenizer codec/fix_mistral_regex regression." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The cluster-level summary overreaches a bit: it highlights Qwen3.5 3D position_ids regressions as one of the \u2018strongest matches,\u2019 but that pair was explicitly judged non-duplicate in the soft-edge verdicts. The accepted SigLIP2 pair is grounded, and most false verdicts are conservative, but the summary should stay closer to the actual accepted matches and avoid implying stronger evidence for the rejected Qwen3.5 pair." - }, - "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -1428,6 +1596,7 @@ "issue:44530", "issue:44560", "issue:44589", + "issue:44683", "issue:44805", "issue:44863", "issue:44928", @@ -1440,7 +1609,6 @@ "issue:45588", "issue:45663", "issue:45823", - "issue:45834", "issue:45902", "issue:45923", "issue:46402", @@ -1476,180 +1644,216 @@ "issue:43716|issue:44560", "issue:44155|issue:45584", "issue:43366|issue:44589", - "issue:43950|issue:45834" + "issue:44683|issue:46515" ], - "prepared_review_unit_hash": "ddb13ca4a3f8c912216767ef2dfd89ffa23864850ca8a09c44e468381ddda612", + "prepared_review_unit_hash": "dd81da81ec84b85512966db22ccd9f8e38b566307ccfa1caa8eae0730922f49b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36748, - "estimated_input_tokens": 9187, - "estimated_eval_tokens": 18630 + "serialized_chars": 36752, + "estimated_input_tokens": 9188, + "estimated_eval_tokens": 18632 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly conservative cluster. I\u2019d accept only two duplicate pairs: the v5 docs/pipeline-removal reports and the `_is_hf_initialized` loading regression reports. The Whisper / `_torch_extract_fbank_features()` pair should be rejected because the packet does not show the same concrete bug.", - "confidence": 0.92, - "canonical_issue_reason": "44291 is the clearest canonical issue for the `_is_hf_initialized` loading regression: it states the shared TypeError directly and is broader than the bitsandbytes-specific symptom in 43872.", + "summary": "Mostly heterogeneous issue set. The only strongly supported duplicate pairs are the `_is_hf_initialized` / bitsandbytes loading regression and the repeated mask-shape IndexError. The other same-theme pairs look related at the title level but the packet does not establish a shared underlying defect strongly enough to merge them.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 44291 is the clearest representative bug report: it names the concrete `init_empty_weights` / `_is_hf_initialized` incompatibility and describes the loading failure directly.", "canonical_pr_reason": null, - "best_issue_reason": "44291 is the best overall issue representative because it captures the root cause and likely subsumes the more specific loading failures around it.", + "best_issue_reason": "Issue 44291 is the most actionable and self-contained issue in the cluster, and it best captures the core loading-regression theme.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45488", "right": "issue:46489", "accept": false, - "reason": "Both are tokenizer-related, but they affect different model families and describe different failures; same area is not enough to duplicate." + "reason": "Both are tokenizer-related, but they target different model families and the packet does not show the same root cause." }, { "left": "issue:43827", "right": "issue:44509", - "accept": true, - "reason": "Both report v5 docs still referencing removed pipeline tasks after the pipeline removals; same documentation cleanup." + "accept": false, + "reason": "This is a plausible docs-theme match, but the packet only supports a broad documentation similarity, not a confirmed duplicate defect." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "One is an lm_head tying bug and the other is a non-persistent buffer regression in from_pretrained; different code paths." + "reason": "Different bugs: untied `lm_head` in Mistral3 versus a `from_pretrained()` non-persistent buffer regression." }, { "left": "issue:43872", "right": "issue:44291", "accept": true, - "reason": "Both describe the same `_is_hf_initialized` TypeError during model loading/init_empty_weights; 44291 is the more general formulation." + "reason": "Same `_is_hf_initialized` / `init_empty_weights` loading incompatibility, with matching failure mode and context." }, { "left": "issue:42915", "right": "issue:43872", "accept": false, - "reason": "FineGrainedFP8Config failure vs `_is_hf_initialized`/bitsandbytes loading error; no shared underlying bug." + "reason": "Both involve loading behavior, but the Qwen3MoE FP8 case is a different code path from the bitsandbytes `_is_hf_initialized` bug." }, { "left": "issue:43792", "right": "issue:44220", "accept": false, - "reason": "The packet does not ground them in the same bug: 43792 is Whisper-large-v2 failing to run, while 44220 is about `_torch_extract_fbank_features()`; shared audio context is too broad." + "reason": "Whisper audio issues are related by theme, but the packet does not clearly establish the same underlying feature-extraction defect." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Unexpected `compute_loss` kwarg vs mask/tensor shape mismatch are different failures." + "reason": "A trainer-signature error and a mask/tensor shape mismatch are distinct failures." }, { "left": "issue:42947", "right": "issue:45923", "accept": false, - "reason": "Both involve gradient checkpointing, but one is PEFT LoRA ineffectiveness and the other is a missing capability flag on a trust_remote_code model." + "reason": "Both mention gradient checkpointing, but one is a LoRA behavior issue and the other is a missing flag on a trust_remote_code model." }, { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "Both involve DeepSpeed/Qwen, but one is missing weight conversion and the other is gradient averaging with GAS; not the same bug." + "reason": "Same DeepSpeed/Qwen ecosystem, but the defects differ: weight conversion/loading versus gradient averaging with GAS." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading problems, but for different models and likely different causes." + "reason": "Both are model/tokenizer loading problems, but the packet does not show the same underlying cause." }, { "left": "issue:43940", "right": "issue:45313", "accept": false, - "reason": "Both are ZeRO-3 weight-loading failures on Qwen variants, but the reports point to different models and no shared code-path evidence." + "reason": "Both are ZeRO-3 loading failures, but they affect different Qwen variants and likely different missing-module paths." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "Different models and different loading/test failures; no clear duplicate signal." + "reason": "Different architectures and different load failures; no strong evidence of one bug." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "Gradient explosion under RLHF/position_ids vs flash_attention crashing on `s_aux=None` are unrelated bugs." + "reason": "The Qwen3.5 RLHF/position-id issue and the flash-attention `s_aux=None` crash are unrelated code-path problems." }, { "left": "issue:41093", "right": "issue:44805", - "accept": false, - "reason": "Similar IndexError wording alone is not enough; the reports lack shared model/code-path evidence." + "accept": true, + "reason": "Same mask-vs-tensor length mismatch IndexError, just reported with different concrete sizes." }, { "left": "issue:44479", "right": "issue:44560", "accept": false, - "reason": "Video-input regression across Qwen model families vs Qwen3-vl-embedding StopIteration are related only at a very broad subsystem level." + "reason": "The titles suggest a common Qwen video regression, but the packet does not clearly prove the same defect." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Both are loading issues, but the titles indicate different models and different failure modes." + "reason": "Both concern loading behavior, but one is about a Voxtral error message and the other about a specific model load failure." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "DeepSpeed training weight-shape error vs image processor void segmentation map failure are unrelated." + "reason": "DeepSpeed training shape error and void-segmentation-map processing are unrelated." }, { "left": "issue:44863", "right": "issue:46402", "accept": false, - "reason": "Checkpoint-loading failure vs text encoder weights not loaded are different enough to avoid merging." + "reason": "Different model-loading failures on unrelated architectures." }, { "left": "issue:43596", "right": "issue:46515", "accept": false, - "reason": "Zero3/BERT indexing error and torch 2.4.1 incompatibility are different problems." + "reason": "A BERT ZeRO-3 out-of-bounds error is not the same as a torch 2.4.1 incompatibility." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "PagedAttentionCache linear_attention crash vs NemotronH checkpoint loading failure are unrelated." + "reason": "PagedAttention `linear_attention` crash and NemotronH checkpoint loading are different defects." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "GGUF architecture support request vs autocast dtype mismatch are unrelated." + "reason": "GGUF gpt-oss support and Phi-tiny-MoE autocast dtype mismatch are unrelated." }, { "left": "issue:43638", "right": "issue:46515", "accept": false, - "reason": "Zero3/BERT indexing error and torch 2.4.1 incompatibility are different problems." + "reason": "Non-pretrained BERT ZeRO-3 out-of-bounds and torch-version incompatibility are separate regressions." }, { "left": "issue:45325", "right": "issue:45902", "accept": false, - "reason": "Both involve Qwen/VL rotary-related behavior, but one is temporal position_id scaling and the other is uninitialized memory after `to_empty`." + "reason": "Both involve Qwen rotary/video-ish internals, but the packet does not establish the same failure mechanism." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "GGUF support for gpt-oss is unrelated to the Qwen3-MoE `sliding_window` issue." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Gemma4 FSDP2 shared-state rebuild and PLE device-map mismatch are different bugs." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Mistral-3 image-preprocessor dtype mismatch is not the same as the Qwen3-VL video `StopIteration` regression." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "AudioFlamingo cross-track leakage and Whisper empty-transcription generation are unrelated." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "GGUF gpt-oss support and Float8 storage lookup failure are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:46515", + "accept": false, + "reason": "Compiled flex_attention on torch >=2.9 is a separate torch-compatibility issue." } ] }, "evaluator_result": { "accept": false, - "feedback": "The cluster summary is mostly conservative, but one soft-edge rationale is not grounded in the packet: for issue:43792 vs issue:44220 it says \u201cBoth mention Whisper,\u201d yet issue:44220\u2019s title is only about `_torch_extract_fbank_features()` and does not mention Whisper. That makes the output imprecise enough to reject, even though the two accepted duplicate pairs look plausible." + "feedback": "The summary is mostly grounded, but a few soft-edge accepts are stronger than the packet supports. In particular, the Whisper pair and the Qwen video pair look related by title, but the packet does not clearly establish the same underlying defect, so calling them duplicates is a bit too assertive. The `_is_hf_initialized` and mask-shape pairs are well supported; the rest should be kept more tentative." }, "evaluator_used": true, "retried": true, @@ -1688,6 +1892,7 @@ "issue:43646", "issue:43720", "issue:43824", + "issue:43950", "issue:43957", "issue:43994", "issue:44451", @@ -1695,7 +1900,6 @@ "issue:44534", "issue:44560", "issue:44661", - "issue:44683", "issue:44756", "issue:44805", "issue:44877", @@ -1720,13 +1924,13 @@ "issue:46515" ], "soft_pairs": [ - "issue:44683|issue:46515", + "issue:43950|issue:45834", "issue:44560|issue:45685", "issue:43824|issue:43957", "issue:42915|issue:43957", "issue:43645|issue:43646", - "issue:43550|issue:44910", "issue:44560|issue:44918", + "issue:43550|issue:44910", "issue:42617|issue:43366", "issue:45823|issue:46515", "issue:36010|issue:42915", @@ -1752,247 +1956,240 @@ "issue:45313|issue:45357", "issue:45507|issue:46468", "issue:43644|issue:43645", - "issue:44488|issue:44991", - "issue:39692|issue:43994" + "issue:44488|issue:44991" ], - "prepared_review_unit_hash": "df7a0242f7e6518575fdc4f6a265e5ab5234801bbb54d54567dbebfb63d9807e", + "prepared_review_unit_hash": "172317dfc0020cba52e95b8a415baeaac44d8b2aadb6a3cfaccdc9b2afdaabee", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 37646, - "estimated_input_tokens": 9412, - "estimated_eval_tokens": 19080 + "soft_pair_count": 33, + "serialized_chars": 37432, + "estimated_input_tokens": 9358, + "estimated_eval_tokens": 18972 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of unrelated Transformers regressions, with a few clear duplicate sub-pairs (notably two identical non-persistent buffer bugs, two custom-model-init bugs, and two SigLIP2 output-corruption reports).", - "confidence": 0.76, - "canonical_issue_reason": "issue:43644 is a clear, broad Transformers v5 regression with an exact duplicate (issue:44534) and a precise symptom that is easy to recognize.", + "summary": "Most items are distinct regressions, with only a few clear duplicate pairs around custom model initialization, SigLIP2 output behavior, and non-persistent buffer handling.", + "confidence": 0.79, + "canonical_issue_reason": "issue:43644 is the strongest canonical anchor: it has a precise regression title and a confirmed near-exact duplicate (issue:44534) describing the same non-persistent buffer corruption bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43644 is the cleanest representative issue in the set: it describes a concrete, cross-cutting regression rather than a model-specific or environment-specific symptom.", + "best_issue_reason": "issue:43644 is the best overall issue candidate because it is narrowly scoped, clearly described, and directly represents a duplicated transformers v5 regression.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44683", - "right": "issue:46515", + "left": "issue:43950", + "right": "issue:45834", "accept": false, - "reason": "Different torch-version compatibility problems in different areas; no shared bug or code path." + "reason": "Different failures: buffer corruption during loading vs. Kosmos2.5 OCR index error." }, { "left": "issue:44560", "right": "issue:45685", "accept": false, - "reason": "Unrelated models and failures (Qwen3-VL video error vs MPS Int histogram issue)." + "reason": "Unrelated models and code paths: Qwen3-VL video embedding error vs. MoE MPS histogram issue." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Different failures: missing export for a specific class vs meta-device loading regression." + "reason": "Import/export failure for a missing class is not the same bug as meta-device loading breakage." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "Qwen3-MoE FP8 failure and meta-device loading are distinct code paths/symptoms." + "reason": "FineGrainedFP8Config failure and meta-device loading failure are different root problems." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "Same underlying regression: Transformers v5 breaks custom model initialization; notebook context is just a narrower manifestation." + "reason": "Both report Transformers 5.0 breaking custom model initialization; 43645 is the notebook-specific manifestation of the broader regression." }, { - "left": "issue:43550", - "right": "issue:44910", + "left": "issue:44560", + "right": "issue:44918", "accept": false, - "reason": "Different models and different attention/position-id failures; not the same bug." + "reason": "Different Qwen variants and different failure modes; no concrete shared bug is evident." }, { - "left": "issue:44560", - "right": "issue:44918", + "left": "issue:43550", + "right": "issue:44910", "accept": false, - "reason": "Both mention Qwen3.5-related issues but the actual failures are unrelated (video StopIteration vs SFT embedding unpacking)." + "reason": "Different models and distinct attention/compile code paths; not the same concrete bug." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "Completely different topics (3d_parallel.py runtime vs GGUF gpt-oss support)." + "reason": "3d_parallel.py execution failure is unrelated to GGUF gpt-oss architecture support." }, { "left": "issue:45823", "right": "issue:46515", "accept": false, - "reason": "Gemma4 device-map forwarding mismatch is unrelated to a generic torch incompatibility report." + "reason": "Device-mismatch forward bug is unrelated to a torch version incompatibility report." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "Import path breakage for GenerationMixin is unrelated to Qwen3-MoE FP8 loading." + "reason": "ImportError for GenerationMixin is unrelated to Qwen3Moe FP8 loading." }, { "left": "issue:45812", "right": "issue:46489", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different fixes." + "reason": "Both are tokenizer issues, but they affect different models and likely different loader paths." }, { "left": "issue:44661", "right": "issue:45823", "accept": false, - "reason": "Tokenizer-mapping failure and Gemma4 device placement are unrelated." + "reason": "Model-list generation bug and Gemma4 device mismatch are unrelated." }, { "left": "issue:44560", "right": "issue:44805", "accept": false, - "reason": "Different bugs: video StopIteration vs mask/tensor shape mismatch." + "reason": "Video embedding StopIteration and mask-shape IndexError are unrelated failures." }, { "left": "issue:38175", "right": "issue:43994", "accept": true, - "reason": "Same model family and same symptom class: SigLIP2 inference produces bad/degenerate outputs." + "reason": "Both describe SigLIP2 producing bad outputs; 43994 is the broader AutoModel/pipeline symptom of the same regression." }, { "left": "issue:43632", "right": "issue:45834", "accept": false, - "reason": "_is_hf_initialized flag regression and Kosmos2.5 long-OCR index error are unrelated." + "reason": "_is_hf_initialized flag breakage is unrelated to Kosmos2.5 long-input indexing." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Both involve config validation, but they concern different models and likely different validation rules." + "reason": "Both mention stricter config validation, but the model/config failures are not concrete duplicates." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Zero3 Bert initialization failure is unrelated to tokenizer-mapping behavior." + "reason": "Zero3 BertModel indexing failure is unrelated to tokenizer-mapping model addition logic." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "Different bugs: DeepSpeed zero3 empty-shard index error vs tokenizer-mapping failure." + "reason": "Zero3 init BertModel issue is unrelated to tokenizer mapping failures." }, { "left": "issue:44661", "right": "issue:44805", "accept": false, - "reason": "No shared underlying code path; tokenizer mapping vs tensor shape mismatch." + "reason": "Tokenizer-mapping bug and tensor mask shape mismatch are different bugs." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "Device-map offload loading failure and Strix Halo mmap OOM are different problems." + "reason": "Dequantized model offload failure and mmap OOM on Strix Halo are different loading problems." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "Both mention SigLIP2, but one is inference output corruption and the other is documentation/example issues with quantization/model-processor mismatch." + "reason": "Same model family, but the docs/example issues are not the same concrete runtime bug." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Identical underlying bug: Transformers v5 fills non-persistent buffers with junk." + "reason": "These are the same non-persistent buffer regression, described almost verbatim." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Qwen3-MoE LoRA compatibility and BitNet packed-weight loading are unrelated." + "reason": "Qwen3 MoE LoRA compatibility and BitNet packed-weight loading are unrelated." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Same broad ecosystem but different concrete failures and code paths." + "reason": "Auto device mapping cuda assert and FineGrainedFP8Config failure are different bugs." }, { "left": "issue:42915", "right": "issue:44756", "accept": false, - "reason": "Qwen3-MoE FP8 failure is unrelated to mmap/OOM behavior." + "reason": "Qwen3Moe FP8 failure is unrelated to mmap/OOM behavior." }, { "left": "issue:46459", "right": "issue:46512", "accept": false, - "reason": "Both are dtype-related, but one is loading composite checkpoints and the other is stale config dtype after from_config; different bugs." + "reason": "Both involve dtype handling, but one is checkpoint loading and the other is from_config/config state; not the same bug." }, { "left": "issue:44451", "right": "issue:44488", "accept": false, - "reason": "Similar symptoms around loading specific models, but not enough evidence they share the same root cause." + "reason": "Both fail to load different models, but there is no clear evidence of a shared concrete root cause." }, { "left": "issue:42491", "right": "issue:45276", "accept": false, - "reason": "Different model families and failures (Qwen3-MoE LoRA vs Gemma4 resize_token_embeddings behavior)." + "reason": "LoRA compatibility on qwen3_moe is unrelated to gemma4 token embedding resizing." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "GLM5 and Qwen3.5 flash-attention failures are distinct bugs." + "reason": "GLM5 and Qwen3.5 flash-attention failures are distinct model-specific issues." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "Different Qwen3.5 problems: ZeRO-3 weight loading vs incorrect save_pretrained key serialization." + "reason": "ZeRO-3 weight loading and save_pretrained visual-key regression are different paths and bugs." }, { "left": "issue:45507", "right": "issue:46468", "accept": false, - "reason": "Unrelated model-specific issues (GraniteMoEHybrid invalid method call vs Nemotron example breakage)." + "reason": "GraniteMoEHybrid invalid method call and Nemotron example breakage are unrelated." }, { "left": "issue:43644", "right": "issue:43645", "accept": false, - "reason": "Related v5 regression theme, but one is buffer corruption and the other is custom model initialization in notebooks; not the same bug." + "reason": "They are related to Transformers 5 custom initialization, but 43645 is a notebook-specific symptom rather than the same exact bug." }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "Both are model-loading regressions, but they involve different models and different tokenizer/load failures." - }, - { - "left": "issue:39692", - "right": "issue:43994", - "accept": false, - "reason": "Same model family, but different issue scope: example errors vs actual inference corruption." + "reason": "Both are tokenizer-load failures, but on different models with no clear shared code-path evidence." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays mostly conservative. The canonical choice of issue:43644 is supported by the identical duplicate issue:44534, and the accepted soft pairs are reasonably justified by title-level evidence. I don\u2019t see any overstatement that would require rejection." + "feedback": "Overall grounded and conservative. The selected anchor issue and the accepted duplicate pairs are supported by the titles and similarity in the packet. Minor caution: phrases like \"confirmed near-exact duplicate\" are a bit stronger than the packet explicitly proves, but the inference is still reasonable from the provided evidence." }, "evaluator_used": true, "retried": false, @@ -2031,6 +2228,7 @@ "issue:43873", "issue:43931", "issue:43957", + "issue:43994", "issue:44075", "issue:44155", "issue:44164", @@ -2062,10 +2260,11 @@ "issue:46468" ], "soft_pairs": [ - "issue:43856|issue:43873", "issue:43708|issue:44075", - "issue:43856|issue:44387", + "issue:43856|issue:43873", + "issue:39692|issue:43994", "issue:43854|issue:43931", + "issue:43856|issue:44387", "issue:43257|issue:43854", "issue:42491|issue:45292", "issue:42491|issue:43931", @@ -2079,8 +2278,8 @@ "issue:43404|issue:43454", "issue:44402|issue:44488", "issue:41720|issue:44155", - "issue:42947|issue:45704", "issue:39692|issue:44960", + "issue:42947|issue:45704", "issue:43873|issue:44387", "issue:45663|issue:45759", "issue:44291|issue:44756", @@ -2093,248 +2292,254 @@ "issue:43335|issue:45663", "issue:42947|issue:43381", "issue:44863|issue:46468", - "issue:45276|issue:45292", "issue:45685|issue:46468", + "issue:45276|issue:45292", "issue:44164|issue:44756" ], - "prepared_review_unit_hash": "b8febc961db6abd27fd481f3cd9d7c490f452e286ff3918aafff359f0a2b1caa", + "prepared_review_unit_hash": "e3f4480402b4098cfbb1ec97bfa3645b6e55eb299ce39449f38d4ff3e0c4fc05", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 34, - "serialized_chars": 37365, - "estimated_input_tokens": 9342, - "estimated_eval_tokens": 18940 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38254, + "estimated_input_tokens": 9564, + "estimated_eval_tokens": 19384 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is noisy and mostly not a true duplicate set. The only tight duplicate pocket is around `resize_token_embeddings` failing to propagate to nested embeddings/output heads in Gemma4/T5Gemma variants; most other issues are unrelated model-loading, quantization, checkpointing, tokenizer, or device-mapping bugs.", - "confidence": 0.57, - "canonical_issue_reason": "Issue 45276 is the clearest anchor for the one real duplicate family here: `resize_token_embeddings` not updating nested embedding/output modules. It is more complete than the generic follow-up reports and best represents that specific regression.", + "summary": "Mostly unrelated model- or subsystem-specific bug reports. The only plausible duplicate pair is the Gemma4/general `resize_token_embeddings` regression cluster; the rest should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 45292 is the broadest representative of the resize-token-embeddings regression: it states the core failure on `output_embeddings`, while the Gemma4 report is a narrower variant affecting additional tied embedding modules.", "canonical_pr_reason": null, - "best_issue_reason": "45276 is the best overall issue in this cluster because it has the broadest concrete reproduction of the resize-token-embedding propagation bug and sits at the center of the only accept-worthy duplicate pocket.", + "best_issue_reason": "Issue 45292 is the best global issue candidate because it captures the underlying shared symptom in the most general form and is more likely to aggregate related reports than the model-specific variant.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43856", - "right": "issue:43873", - "accept": false, - "reason": "Both mention memory/offload/quantization pain, but they describe different failure modes and code paths; not the same underlying bug." - }, { "left": "issue:43708", "right": "issue:44075", "accept": false, - "reason": "Trainer checkpoint-resume step calculation is unrelated to optimizer SGD args being ignored." + "reason": "Different training-time bugs: checkpoint resumption/max-steps calculation vs SGD optimizer args not being applied." }, { "left": "issue:43856", - "right": "issue:44387", + "right": "issue:43873", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training memory usage and the other is quantization/offloading behavior; not the same code-path bug." + }, + { + "left": "issue:39692", + "right": "issue:43994", "accept": false, - "reason": "Different issues: one is Qwen3 MoE training memory usage, the other is int4 quantization CUDA reserved-memory/OOM behavior." + "reason": "Same model family, but one is a broken doc example and the other is wrong inference results from AutoModel/pipeline; different underlying failures." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Both are model-loading problems, but they involve different models and distinct load-time failures." + "reason": "Different models and different load failures; no evidence they share the same root cause." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both are memory-related, but one is Qwen3 MoE training and the other is int4 quantization OOM/regression; too broad to merge." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is not the same bug as GLM-4.7-Flash loading in tests." + "reason": "Qwen3 MoE accelerate/deepspeed conversion problem vs GLM-4.7-Flash load issue; unrelated." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "LoRA hf4.x vs hf5.x incompatibility is unrelated to resize_token_embeddings not updating output embeddings." + "reason": "LoRA qwen3_moe hf4.x vs hf5.x compatibility issue is not the same as resize_token_embeddings failing to update embeddings." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "Different concrete bugs: LoRA checkpoint compatibility vs Qwen3-VL shape-mismatch on load." + "reason": "Different model/version-loading problems with no shared concrete code path." }, { "left": "issue:44960", "right": "issue:46050", "accept": false, - "reason": "Different model and different lifecycle bug: GLM5 loading vs sinq save/load corruption." + "reason": "Both concern model/quantization behavior, but one is a GLM5 issue and the other is save/load of sinq-quantized models; distinct bugs." }, { "left": "issue:45488", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and different tokenizer behaviors/code paths." + "reason": "Both are tokenizer regressions, but one is a pre-tokenizer override issue for DeepSeek and the other is wrong token IDs for Granite; different failure modes." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Qwen3.5 chat crash and Gemma4Processor missing `_tokenizer` are unrelated runtime failures." + "reason": "Qwen3.5 chat crash and Gemma4Processor `_tokenizer` attribute error are unrelated." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "Saving/loading extra_state is a different problem from init_empty_weights passing `_is_hf_initialized`." + "reason": "Both involve model loading/saving internals, but extra_state handling and `_is_hf_initialized` arg errors are separate regressions." }, { "left": "issue:45276", "right": "issue:45335", - "accept": true, - "reason": "Same underlying regression: `resize_token_embeddings` fails to update nested embeddings/output heads across composite model variants (Gemma4/T5Gemma)." + "accept": false, + "reason": "Very similar symptom, but they affect different model implementations (`gemma4` vs `t5gemma`) and do not clearly share one concrete fix." }, { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Both touch attention-related code, but one is a Qwen3.5 packed-sequence misinterpretation and the other is an `s_aux=None` crash on sink-less models." + "reason": "Both touch flash attention, but one is packed-sequence/position_ids misinterpretation and the other is `s_aux=None` attribute handling; different bugs." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Torch 2.10 compatibility and meta-device loading failures are separate issues." + "reason": "Torch version incompatibility and meta-device model loading failure are separate issues." }, { "left": "issue:43404", "right": "issue:43454", "accept": false, - "reason": "Both involve weight tying, but they are different models and different manifestations; not a single duplicate bug." + "reason": "Both are untied `lm_head` weights, but in different model classes/code paths; not enough to merge as one bug." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "The reports both concern token/vocab loading, but one is about out-of-vocab special tokens and the other is a specific model failing to load." + "reason": "Both relate to tokenizer/vocab inconsistencies, but the affected models and reported failures differ." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Qwen3 auto device-map CUDA assert and AudioFlamingo3 batched inference token leak are unrelated." + "reason": "Qwen3 device-map CUDA assert vs AudioFlamingo3 track leakage are unrelated." }, { - "left": "issue:42947", - "right": "issue:45704", + "left": "issue:39692", + "right": "issue:44960", "accept": false, - "reason": "Gradient checkpointing/LoRA behavior is unrelated to the T5 apex RMSNorm memory leak." + "reason": "Completely different models and failure types: SigLIP2 docs/load issues vs GLM5/quantization." }, { - "left": "issue:39692", - "right": "issue:44960", + "left": "issue:42947", + "right": "issue:45704", "accept": false, - "reason": "SigLIP2 doc-example errors and GLM5 model-loading issues are unrelated." + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the T5 apex RMSNorm memory leak." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Both are quantization/memory-related, but one is offloading not working and the other is int4 CUDA reserved-memory growth." + "reason": "Both mention quantization/memory, but one is offloading not working and the other is reserved-memory growth leading to OOM; too generic to treat as duplicates." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Gemma-4 FSDP2 KeyError and Qwen multimodal text_config unwrapping are different code paths." + "reason": "Gemma4 FSDP2 attention state bug vs Qwen 3.5/3.6 composite config unwrapping bug are different code paths." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "Unexpected `_is_hf_initialized` during init_empty_weights is unrelated to mmap-induced OOM on Strix Halo." + "reason": "`init_empty_weights` loader arg handling and Strix Halo mmap OOM are unrelated." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "Importing GenerationMixin and Whisper pipeline return_language regressions are separate issues." + "reason": "ImportError for `GenerationMixin` vs Whisper pipeline `return_language` regression are separate issues." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Gemma-4 FSDP2 shared_kv_states bug and Qwen3_5MoeTextRotaryEmbedding CPU-offload incompatibility are different failures." + "reason": "Gemma4 FSDP2 attention-state bug vs Qwen3_5Moe rotary embedding CPU-offload incompatibility are unrelated." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Both mention memory pressure, but the root causes are different and not mergeable as one fix." + "reason": "Different memory/OOM reports with no shared concrete mechanism." }, { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Different quantization formats and symptoms: NVFP4 load failure vs MXFP4 fallback to bf16." + "reason": "Different quantization backends and models; not the same underlying bug." }, { "left": "issue:44155", "right": "issue:45823", "accept": false, - "reason": "AudioFlamingo3 inference leakage and Gemma4 PLE device mismatch are unrelated." + "reason": "Audio token leakage between tracks and Gemma4 device-map mismatch are unrelated." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading breakage and mmap OOM avoidance are different loading mechanisms with different fixes." + "reason": "Meta-device loading regression is not the same as mmap causing OOM on Strix Halo." }, { "left": "issue:43335", "right": "issue:45663", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and Gemma-4 FSDP2 shared_kv_states are unrelated." + "reason": "Sparse-layer creation bug in SwitchTransformers and Gemma4 FSDP2 attention bug are unrelated." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Both reference gradient checkpointing, but one is ineffective with PEFT LoRA and the other forbids eval-mode use; not the same bug." + "reason": "Gradient checkpointing in PEFT LoRA vs eval-mode checkpointing restriction are different issues." }, { "left": "issue:44863", "right": "issue:46468", "accept": false, - "reason": "NemotronH checkpoint loading and the broken official Nemotron example are related only at a high level, not the same concrete defect." - }, - { - "left": "issue:45276", - "right": "issue:45292", - "accept": true, - "reason": "Same `resize_token_embeddings` regression family; the generic output_embeddings report is a narrower instance of the Gemma4 failure." + "reason": "Both are Nemotron-related, but one is checkpoint loading and the other is an example-script breakage; too little evidence of a shared root cause." }, { "left": "issue:45685", "right": "issue:46468", "accept": false, - "reason": "MPS histogram error in MoE is unrelated to the Nemotron example being broken." + "reason": "MOE MPS histogram dtype error and the Nemotron example issue are unrelated." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same underlying resize-token-embeddings regression: the Gemma4 report is a narrower instance of the broader `output_embeddings` failure." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "Extra_state save/load handling and mmap OOM avoidance are distinct bugs." + "reason": "Extra_state save/load bug and Strix Halo mmap OOM are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded and conservative overall. The only accepted duplicate pocket is the resize_token_embeddings propagation regression family centered on issue:45276, and the rejects mostly avoid over-merging unrelated load/quantization/checkpointing/tokenizer reports. The summary stays within the packet evidence, and the accepted pairs look mergeable into one fix." + "feedback": "The clustering is well-grounded in the packet. The 45276/45292 pair is the only clearly mergeable duplicate, and the rest of the soft-edge rejections stay conservative given the titles/body summaries. No overstatement of evidence stands out." }, "evaluator_used": true, "retried": false, @@ -2412,8 +2617,8 @@ "issue:45588|issue:45663", "issue:42796|issue:46326", "issue:43646|issue:43708", - "issue:45292|issue:45335", "issue:43645|issue:44164", + "issue:45292|issue:45335", "issue:43828|issue:43957", "issue:43425|issue:43828", "issue:44402|issue:44451", @@ -2428,228 +2633,228 @@ "issue:44530|issue:45084", "issue:43906|issue:44530", "issue:44479|issue:44805", - "issue:44265|issue:44898", "issue:44589|issue:45084", + "issue:44265|issue:44898", "issue:43901|issue:44509", "issue:43425|issue:44918", "issue:39404|issue:44912", "issue:43994|issue:44079", "issue:36032|issue:43452" ], - "prepared_review_unit_hash": "7e8e2cf01443126303b009e78e163f05d4729b6a1d6ce7aedc0041cfa9b5cb26", + "prepared_review_unit_hash": "236d5d056b1c4b94170a89cc7e2733c2880728c12afa66db12cc461b0290163e", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36865, - "estimated_input_tokens": 9217, - "estimated_eval_tokens": 18690 + "serialized_chars": 36864, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous v5-regression issues across tokenization, model loading, training, and docs; only one soft duplicate pair looks convincing.", - "confidence": 0.83, - "canonical_issue_reason": "issue:45812 is the strongest representative of the recurring v4\u2192v5 tokenizer regression theme and is concrete enough to anchor triage.", + "summary": "Mostly a set of distinct Transformers v5 regressions with broad tokenizer/model-loading overlap, but only one clear duplicate pair emerged: the resize_token_embeddings embedding-update bug. The rest share subsystems or release context, not the same underlying defect.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45292 is the broader, model-agnostic report of the resize_token_embeddings regression; issue:45335 is just the t5gemma-specific manifestation of the same bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the most actionable and broadly representative issue in the set.", + "best_issue_reason": "issue:45292 best represents the duplicate cluster because it states the underlying embedding-resize failure in the most general form and would attract the widest set of affected reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44479", - "right": "issue:44877", + "right": "issue:44805", "accept": false, - "reason": "Different failures: video-input regression for Qwen VL models vs strict-config loading for granite_speech." + "reason": "Both involve model input behavior, but one is a video-regression report and the other is a generic mask-shape index error; different bugs." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "Both are loading regressions, but they concern different models and different code paths (unit-test load failure vs save_pretrained checkpoint correctness)." + "reason": "Unrelated failures: GLM-4.7 unit-test loading issue versus a Qwen3.5 save_pretrained checkpoint regression." }, { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Both touch save/load of custom models, but one is about read-only permission propagation and the other about loading the wrong custom module; not the same bug." + "reason": "Both touch save/from_pretrained for custom models, but one is about read-only source-file permissions and the other about loading the wrong custom module; not the same defect." }, { "left": "issue:43381", "right": "issue:44928", "accept": false, - "reason": "Unrelated: gradient checkpointing in eval mode vs Qwen3.5 RLHF NaN/SDPA fallback behavior." + "reason": "Eval-mode gradient checkpointing incompatibility is unrelated to the Qwen3.5 RLHF NaN caused by 3D position_ids/SDPA fallback." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Different code paths and symptoms: flash_attention s_aux=None crash vs Gemma4 FSDP2 shared_kv_states KeyError." + "reason": "Different code paths: flash_attention sink handling versus Gemma4 FSDP2 shared_kv_states rebuilding." }, { "left": "issue:42796", "right": "issue:46326", "accept": false, - "reason": "Tokenizer-family similarity only; one is BPE marker handling, the other is continue_final_message behavior on mistral_common tokenizers." + "reason": "Both involve tokenizer behavior, but one is BPE marker handling and the other is continue_final_message logic for mistral_common; not the same bug." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Different problems: custom model initialization in notebooks vs resume_from_checkpoint max_steps calculation." - }, - { - "left": "issue:45292", - "right": "issue:45335", - "accept": true, - "reason": "Same underlying resize_token_embeddings bug: output-side embeddings are not updated, with t5gemma showing the same regression in decoder.embed_tokens." + "reason": "Custom model initialization in notebooks is unrelated to incorrect max_steps calculation when resuming Trainer with changed batch size." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Both involve model save/load, but one is notebook custom-model initialization and the other is extra_state handling." + "reason": "Notebook custom-model initialization and extra_state save/load failures are separate issues." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same underlying bug: resize_token_embeddings fails to propagate updated embeddings; 45335 is the t5gemma-specific instance." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "Different dtype/device regressions: autocast mismatch for Phi-tiny-MoE vs meta-device loading breakage for some models." + "reason": "Both are dtype/loading-related, but one is an autocast mismatch in Phi-tiny-MoE and the other is a torch.device('meta') loading regression affecting different models." }, { "left": "issue:43425", "right": "issue:43828", "accept": false, - "reason": "Unrelated issues: Torch 2.10 compatibility vs a model-specific autocast dtype mismatch." + "reason": "Torch version incompatibility is not the same as the Phi-tiny-MoE autocast dtype mismatch." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Both are tokenizer/model-loading complaints, but the concrete bugs differ and no shared failing code path is evident." + "reason": "Tokenizer vocab-size inconsistency and inability to load ScandiBERT may be related in theme, but the reports do not point to the same concrete failure." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Float8 storage lookup error and Strix Halo mmap OOM are unrelated." + "reason": "Float8 storage lookup failure is unrelated to the Strix Halo mmap/OOM workaround." }, { "left": "issue:45812", "right": "issue:46315", "accept": false, - "reason": "Same broad tokenizer area, but one is wrong token IDs for Granite models and the other is a performance regression in convert_tokens_to_ids." + "reason": "Wrong token IDs for Granite models is a correctness regression, while convert_tokens_to_ids slowdown is a performance regression." }, { "left": "issue:46315", "right": "issue:46491", "accept": false, - "reason": "Performance regression vs encode\u2192decode whitespace regression; different bugs." + "reason": "Performance regression in convert_tokens_to_ids and CodeLlama leading-space round-trip bug are different tokenizer issues." }, { "left": "issue:44683", "right": "issue:45823", "accept": false, - "reason": "Compiled flex_attention failure on torch>=2.9 is unrelated to Gemma4 PLE device mismatch." + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device_map auto mismatch are separate execution-path problems." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper pipeline return_language regression and tokenization-version-change report are not the same bug." + "reason": "Whisper pipeline return_language regression is unrelated to a general tokenization change report." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "base_model_prefix loading regression vs custom-model initialization in notebooks; different failure modes." + "reason": "Base_model_prefix loading breakage and notebook custom-model initialization are both v5 custom-model concerns, but they are not the same failure mode." }, { "left": "issue:43425", "right": "issue:45823", "accept": false, - "reason": "Torch version incompatibility is unrelated to Gemma4 device placement during forward." + "reason": "Torch 2.10 incompatibility is broader and distinct from the Gemma4 device mismatch during forward." }, { "left": "issue:43425", "right": "issue:43716", "accept": false, - "reason": "Torch compatibility issue vs Mistral-3 image-preprocessor/model dtype mismatch." + "reason": "Global Torch compatibility issue does not match the Mistral-3 image preprocessor dtype mismatch." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "PagedAttentionCache group-type crash is unrelated to compile-non-template-nodes." + "reason": "PagedAttentionCache linear_attention crash and non-template-node compile error are unrelated compiler/cache failures." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "An isolated reproduction issue is not the same bug as a PagedAttentionCache crash." + "reason": "An isolated reproduction of an earlier issue is not the same as the Qwen3.5 PagedAttentionCache crash." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "Video-input regression and mask shape mismatch are different problems." + "reason": "Video input regression and mask-shape mismatch are different symptoms with different triggers." }, { - "left": "issue:44265", - "right": "issue:44898", + "left": "issue:44589", + "right": "issue:45084", "accept": false, - "reason": "torch.export export failure with torch_compilable_check is unrelated to Perceiver interpolation failure." + "reason": "Float8 storage lookup failure is unrelated to compilation of non-template nodes." }, { - "left": "issue:44589", - "right": "issue:45084", + "left": "issue:44265", + "right": "issue:44898", "accept": false, - "reason": "Different failures and code paths; no concrete duplication." + "reason": "torch.export failure with torch_compilable_check is a different export-path bug than Perceiver non-default-resolution handling." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are docs-related, but they refer to different removed/deprecated pipeline topics." + "reason": "One is a docs mismatch for return_all_scores; the other is broader docs still listing removed tasks." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch compatibility vs Qwen3.5 embedding unpacking with trl SFT trainer." + "reason": "Torch compatibility is unrelated to the Qwen3.5 embedding unpacking problem in trl SFTTrainer." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline language return issue and MXFP4 quantization fallback are unrelated." + "reason": "Whisper return_language pipeline breakage and MXFP4 quantization fallback are unrelated." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "Siglip2 inference quality issue and ModelOutput key assignment bug are different underlying problems." + "reason": "SigLIP2 nonsensical outputs and ModelOutput key assignment are distinct bugs." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "T5Tokenizer add_special_tokens conflict and gguf_file loading breakage are unrelated loading errors." + "reason": "T5Tokenizer add_special_tokens conflict is unrelated to gguf_file breaking AutoTokenizer/AutoModel loading." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall: the summary stays conservative, and the only accepted soft pair (issue:45292 / issue:45335) is well supported by the shared resize_token_embeddings failure. The rejections are generally cautious and tied to distinct model/code paths. No clear overstatement from the packet." + "feedback": "Grounded in the packet. The one accepted duplicate pair, issue:45292 and issue:45335, is supported by the titles as the same resize_token_embeddings bug in a general vs t5gemma-specific form. The remaining soft-edge rejections are conservative and do not overclaim shared causality." }, "evaluator_used": true, "retried": false, @@ -2722,14 +2927,14 @@ "issue:43299|issue:43931", "issue:44671|issue:44991", "issue:43596|issue:44683", - "issue:46050|issue:46326", "issue:44164|issue:45759", + "issue:46050|issue:46326", "issue:43906|issue:46326", "issue:44991|issue:45081", "issue:43856|issue:44756", "issue:44164|issue:44479", - "issue:42796|issue:43404", "issue:44488|issue:45081", + "issue:42796|issue:43404", "issue:44661|issue:44683", "issue:39404|issue:44671", "issue:44488|issue:44987", @@ -2745,219 +2950,219 @@ "issue:44451|issue:44987", "issue:40990|issue:43421", "issue:43541|issue:45084", - "issue:42175|issue:46178", "issue:34567|issue:40444", + "issue:42175|issue:46178", "issue:42915|issue:44530", "issue:43644|issue:43646" ], - "prepared_review_unit_hash": "d2c5a3ff17513afd11976642084f7cc85060b646853ce844b5f3e346a9fb2b75", + "prepared_review_unit_hash": "8665c2fd074987e43281b044a1877835c3167cc7949a50567fb1290bda50dbd7", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36152, - "estimated_input_tokens": 9038, - "estimated_eval_tokens": 18332 + "serialized_chars": 36154, + "estimated_input_tokens": 9039, + "estimated_eval_tokens": 18334 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is dominated by distinct Transformers v5 regression reports, mostly around model/tokenizer loading or initialization. None of the soft pairs look like true duplicates of the same underlying bug/change, so all soft edges are rejected.", - "confidence": 0.74, - "canonical_issue_reason": "issue:43646 is the broadest representative of the cluster: a general Transformers v5 custom-model-initialization regression. It sits near several other loading/initialization breakage reports, but is still the most umbrella-like issue here.", + "summary": "This cluster is mostly a grab-bag of Transformers v5 regression reports, with many model/tokenizer loading failures but no clear duplicate pair strong enough to merge. I found no PRs. The best representative issue is the broad model-loading regression around `base_model_prefix`.", + "confidence": 0.63, + "canonical_issue_reason": "Issue 43611 is the most representative broad v5 regression report here: it targets a core model-loading path and matches the cluster\u2019s general theme better than the more model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the best single issue to represent the cluster because it is generic, version-wide, and closest to the shared theme of v5 regressions affecting model construction/loading.", + "best_issue_reason": "Issue 43611 is the strongest single issue to anchor the cluster because it is generic, clearly regression-focused, and covers a central load-path failure.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43299", "right": "issue:43931", "accept": false, - "reason": "Both are Qwen3/VL loading failures, but one is a MoE version-loading regression and the other is a weight-shape mismatch for a different model; not the same bug." + "reason": "Both are Qwen3-VL loading problems, but one is a generic MoE loading regression and the other is a specific weight-shape mismatch; not enough evidence they share the same concrete bug." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different failure modes (masked LM predictions vs loading a specific tokenizer)." + "reason": "Both are tokenizer/model regressions, but they affect different models and symptoms (CamemBERT predictions vs est-roberta tokenizer loading)." }, { "left": "issue:43596", "right": "issue:44683", "accept": false, - "reason": "These are unrelated: DeepSpeed ZeRO3/BertModel indexing versus torch>=2.9 compiled flex_attention failure." + "reason": "Completely different failures: DeepSpeed/BertModel zero-sized index error vs compiled flex_attention on torch>=2.9." }, { - "left": "issue:46050", - "right": "issue:46326", + "left": "issue:44164", + "right": "issue:45759", "accept": false, - "reason": "One is about save/load breakage for quantized models, the other is a tokenizer behavior bug with mistral_common; different code paths." + "reason": "Different code paths: save/from_pretrained extra_state handling vs AutoModelForCausalLM.from_config text_config unwrapping." }, { - "left": "issue:44164", - "right": "issue:45759", + "left": "issue:46050", + "right": "issue:46326", "accept": false, - "reason": "Both touch save/load or config handling, but one is extra_state serialization and the other is AutoModelForCausalLM config unwrapping for composite Qwen multimodal configs." + "reason": "Unrelated issues: quantized save/load breakage vs mistral_common continue_final_message behavior." }, { "left": "issue:43906", "right": "issue:46326", "accept": false, - "reason": "An isolated reproduction of an earlier issue is not the same as the mistral_common tokenizer bug; different scope and symptoms." + "reason": "The first is a reproduction of an older issue; the second is a tokenizer prompt-format bug. No shared concrete bug." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Tokenizer loading for different models and different failure causes; one ignores tokenizer.json, the other crashes in a Mistral regex patch." + "reason": "Both involve tokenizer loading, but the failures are different models and different code paths (repo tokenizer load vs Mistral regex patch crash)." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "Both are memory-related, but one is Qwen3 MoE training efficiency and the other is mmap/OOM on Strix Halo; not the same bug." + "reason": "Different memory issues in unrelated contexts: Qwen3 MoE training vs Strix Halo mmap/OOM." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "Save/load extra_state handling and a video input regression are unrelated issues." + "reason": "Different regressions: serialization extra_state vs video-input handling for Qwen VL models." }, { - "left": "issue:42796", - "right": "issue:43404", + "left": "issue:44488", + "right": "issue:45081", "accept": false, - "reason": "Tokenizer BPE-marker handling and a tied-lm_head weight bug in Mistral3 are separate failures." + "reason": "Different tokenizer/model loading bugs; one is loading a specific BERT repo, the other is a Mistral regex patch crash." }, { - "left": "issue:44488", - "right": "issue:45081", + "left": "issue:42796", + "right": "issue:43404", "accept": false, - "reason": "Different tokenizer loading problems for different models; the sleng-bert case is not the same as the Mistral regex crash." + "reason": "Tokenizer BPE-marker handling and Mistral3 lm_head tying are unrelated bugs." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "Tokenizer mapping failures and flex_attention compilation failures are unrelated subsystems." + "reason": "Unrelated areas: tokenizer mapping support vs flex_attention compilation." }, { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper return_language behavior and CamemBERT masked LM predictions are unrelated model-specific bugs." + "reason": "Whisper pipeline language return bug vs CamemBERT masked-LM regression; no shared code path." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Both involve loading failures for specific models, but the affected models and root causes differ." + "reason": "Both are model-loading complaints, but the affected models and underlying causes differ." }, { "left": "issue:44462", "right": "issue:46489", "accept": false, - "reason": "Both concern tokenizer loading/behavior, but the first is AutoTokenizer ignoring tokenizer.json while the second is a DeepSeek-Coder regression tied to a prior fix gap." + "reason": "Both mention tokenizer issues, but one is repo tokenizer.json selection and the other is DeepSeek-Coder output correctness; not the same bug." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer initialization; same theme, different bug." + "reason": "Related v5 regression theme, but one is base_model_prefix loading and the other is non-persistent buffer initialization; too different to merge." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer GPU broadcast TypeError and Whisper pipeline return_language are unrelated." + "reason": "Trainer GPU broadcast TypeError and Whisper pipeline return_language regression are unrelated." }, { "left": "issue:44737", "right": "issue:45859", "accept": false, - "reason": "Both mention rotary/CPU movement, but one is XLNet positional encoding on CPU every forward pass and the other is Qwen3.5 MoE rotary embedding offload compatibility." + "reason": "Both are device/offload-related, but they hit different models and different mechanisms (XLNet CPU arange vs Qwen3.5 MoE rotary embedding offload)." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Image preprocessor dtype mismatch and a template-compilation TypeError are unrelated." + "reason": "Different problem classes: image preprocessor dtype mismatch vs Jinja/template compilation error." }, { "left": "issue:46285", "right": "issue:46402", "accept": false, - "reason": "CLIP structure inconsistency and Sam3 text encoder weights not loading are different model-loading regressions." + "reason": "Both are model-loading structure issues, but one is CLIP class inconsistency and the other is Sam3 text encoder weight loading." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and _is_hf_initialized flag breakage are unrelated v5 issues." + "reason": "Unrelated bugs: Apple Silicon allocator warmup TypeError vs `_is_hf_initialized` flag regression." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "Tokenizer method-name collision in T5 and Qwen3 MoE FP8Config failure are different model/tooling bugs." + "reason": "Tokenizer method-name conflict and Qwen3Moe FP8 failure are unrelated." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "AutoConfig trust_remote_code missing fields and PagedAttentionCache linear_attention crash are unrelated." + "reason": "Different bugs: remote-code config field loss vs Qwen3.5 cache/group-type crash." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Two separate model-loading failures for different repositories; not the same underlying bug." + "reason": "Both are load failures, but the concrete failures and model families differ." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "Perplexity quality regression and tokenizer post-processor update behavior are unrelated." + "reason": "Perplexity regression on gpt-oss is unrelated to runtime special-token post-processor updates." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Grouped_mm tracing failure and template-node compilation failure are different runtime/compiler issues." + "reason": "Torch Dynamo grouped_mm tracing error and template-node compilation error are unrelated." }, { - "left": "issue:42175", - "right": "issue:46178", + "left": "issue:34567", + "right": "issue:40444", "accept": false, - "reason": "Missing tensorflow in an extra install and an unknown tensorflow_text backend mapping entry are related only at a high level, not the same bug." + "reason": "TrainerState token counting and multimodal IterableDataset finetuning failures are unrelated." }, { - "left": "issue:34567", - "right": "issue:40444", + "left": "issue:42175", + "right": "issue:46178", "accept": false, - "reason": "TrainerState token counting and Qwen2.5-VL iterable dataset multi-image finetuning are unrelated." + "reason": "Both touch backend packaging, but one is missing tensorflow in a torch extra and the other is an unsupported tensorflow_text backend mapping entry." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "Qwen3 MoE FP8 config failure and PagedAttentionCache linear_attention crash are different code paths." + "reason": "Both involve Qwen models, but one is FP8 configuration and the other is PagedAttentionCache group-type handling." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Both are broad v5 regressions, but one is buffer initialization corruption and the other is custom model initialization; not the same concrete defect." + "reason": "These are both v5 initialization regressions, but one is about non-persistent buffers and the other is a broader custom model-init breakage; not clearly the same bug." } ] }, @@ -3043,21 +3248,21 @@ "issue:44464|issue:45987", "issue:43262|issue:43377", "issue:43646|issue:44534", - "issue:43746|issue:46326", "issue:43645|issue:44534", + "issue:43746|issue:46326", "issue:43883|issue:44991", "issue:44403|issue:45081", "issue:42915|issue:45084", "issue:43404|issue:44466", "issue:44849|issue:45325", "issue:43994|issue:44610", - "issue:43377|issue:43819", "issue:28530|issue:34567", + "issue:43377|issue:43819", "issue:43322|issue:43540", "issue:44451|issue:45081", "issue:44987|issue:45081", - "issue:43089|issue:45987", "issue:42915|issue:44291", + "issue:43089|issue:45987", "issue:44265|issue:44610", "issue:44220|issue:44683", "issue:36032|issue:43742", @@ -3065,232 +3270,238 @@ "issue:42915|issue:44589", "issue:45216|issue:45357" ], - "prepared_review_unit_hash": "7632957dd0d368e41d7e993b6f3e9c2b4310cbeda00bfbd2e1f80368a54890de", + "prepared_review_unit_hash": "3446e2a41de625b731af3cdf0fed298e7f419e58530940d0e1d37a69611aeef6", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37686, - "estimated_input_tokens": 9422, - "estimated_eval_tokens": 19100 + "serialized_chars": 37690, + "estimated_input_tokens": 9423, + "estimated_eval_tokens": 19102 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous and the soft edges are largely false positives. The only clear duplicate-like pairs are the GGUF loading regression reports and the Qwen3.5 save_pretrained serialization reports.", - "confidence": 0.79, - "canonical_issue_reason": "43452 is the broadest, most general regression here: it captures the generic GGUF loader breakage that also explains the more specific Qwen2.5-GGUF failure in 43482.", + "summary": "This set is mostly heterogeneous issue reports, not a tight duplicate cluster. The only strong duplicate-like pair is the Qwen3.5 save_pretrained regression (45216/45357); the rest share broad themes but look like different bugs or different code paths.", + "confidence": 0.41, + "canonical_issue_reason": "Issue 45216 is the broader, earlier report of the Qwen3.5 save_pretrained regression, and 45357 appears to be a later follow-up that the same visual-encoder-key problem still exists in 5.5.3.", "canonical_pr_reason": null, - "best_issue_reason": "43452 is the best representative issue because it is model-agnostic and describes the core from_pretrained/gguf_file regression, making it a good anchor for related reports.", + "best_issue_reason": "Issue 45216 is the best canonical issue among the listed items because it most clearly captures the underlying Qwen3.5 save_pretrained regression and can absorb the later duplicate-like follow-up.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both are Whisper-related, but one is about skip_special_tokens in batch_decode and the other about empty-transcription handling after align_special_tokens; different symptoms and likely different fixes." + "reason": "Both involve Whisper/token handling, but one is about batch_decode skip_special_tokens and the other is an empty-transcription failure after align_special_tokens; different bug surface." }, { "left": "issue:44206", "right": "issue:44220", "accept": false, - "reason": "Both touch audio feature extraction, but 44206 is specifically an unsupported center arg regression while 44220 is too generic to confirm the same underlying bug." + "reason": "Both are audio feature-extraction regressions, but they hit different functions and failure modes (_center arg vs _torch_extract_fbank_features)." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "One is a Molmo tied-weights attribute error; the other is a tokenizer loading failure for a specific model. No shared code-path bug is evident." + "reason": "Both are loading failures, but they concern different models and different missing-key/loader behavior." }, { "left": "issue:40444", "right": "issue:46326", "accept": false, - "reason": "These concern different multimodal/tokenizer paths: multi-image iterable fine-tuning versus continue_final_message behavior for mistral_common tokenizers." + "reason": "Both touch multimodal/chat templating, but one is about multi-image IterableDataset finetuning and the other is about continue_final_message with mistral_common tokenizers." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "A T5Tokenizer special-method conflict and a Jupyter/custom-model initialization regression are distinct issues." + "reason": "One is a T5 tokenizer name conflict; the other is a Transformers 5 custom-model/Jupyter initialization break. Different issues." }, { "left": "issue:43452", "right": "issue:43482", - "accept": true, - "reason": "Both are GGUF loading regressions in v5; 43452 describes the generic loader breakage and 43482 is a concrete Qwen2.5-GGUF manifestation." + "accept": false, + "reason": "Both mention GGUF/loading, but one is a generic gguf_file from_pretrained regression and the other is a specific Qwen2.5-GGUF failure in v5." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "These involve different failure modes: gpu+cpu offload with dequantized models versus GGUF loading for Qwen2.5." + "reason": "Different paths: device_map=auto dequantized model offload versus GGUF loading for Qwen2.5." }, { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Different problems: template-node compilation error versus GraniteMoEHybrid calling an invalid method." + "reason": "These are unrelated model/template bugs; they do not look like the same underlying code-path problem." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Both are model-loading regressions on v5, but they affect different artifacts and likely different loader paths." + "reason": "Both are version-5 tokenizer/model-loading regressions, but for different models and likely different causes." }, { "left": "issue:44464", "right": "issue:45987", "accept": false, - "reason": "Chunked generation under compiled forward is unrelated to StaticCache.get_seq_length returning a tensor instead of an int." + "reason": "Chunked generation/compiled forward inconsistency is a different bug from StaticCache.get_seq_length returning the wrong type." }, { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Audio chat-template sampling-rate defaults and MIMI batched-vs-single padding-mask divergence are unrelated bugs." + "reason": "Audio-related, but one is a default sampling-rate mismatch and the other is missing padding-mask support in MIMI encoder batching." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Custom model initialization breakage and junk-filled non-persistent buffers are separate regressions." + "reason": "Custom model initialization break versus non-persistent buffers being filled with junk; different failures and code paths." }, { - "left": "issue:43746", - "right": "issue:46326", + "left": "issue:43645", + "right": "issue:44534", "accept": false, - "reason": "PEFT adapter loading for GraniteSpeech is unrelated to continue_final_message tokenizer behavior." + "reason": "Jupyter/custom model initialization is unrelated to non-persistent buffer serialization behavior." }, { - "left": "issue:43645", - "right": "issue:44534", + "left": "issue:43746", + "right": "issue:46326", "accept": false, - "reason": "These are different v5 regressions: notebook/custom-model initialization versus buffer serialization corruption." + "reason": "Local checkpoint loading with PEFT adapters is unrelated to continue_final_message behavior for mistral_common tokenizers." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "A missing all_tied_weights_keys attribute in Molmo is not the same bug as loading a specific tokenizer/model package failing on v5." + "reason": "Missing all_tied_weights_keys on Molmo is unrelated to tokenizer loading for EMBEDDIA/est-roberta." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "Loading noise and a Mistral regex patch crash are different issues with different failure points." + "reason": "Noise when loading a transformer is a different issue from a Mistral tokenizer regex patch crash." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to the template compilation error in 45084." + "reason": "Qwen3Moe FP8 config failure and template compilation failure are unrelated." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Both mention lm_head/tied weights, but one is a model-specific tying bug and the other is a device-dependent serialization inconsistency; not the same concrete defect." + "reason": "Both mention lm_head/tied weights, but one is about a model defining untied weights and the other about device-dependent serialization inconsistency." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Different model families and different bugs: output_hidden_states handling versus rope index scaling for still images." + "reason": "Different Qwen model families and different bugs: hidden-states output versus rope-index temporal scaling." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Siglip2 nonsensical outputs and OmDet-Turbo image-size mismatch are separate vision pipeline issues." + "reason": "Siglip2 bad outputs and OmDet-Turbo processor size mismatch are distinct vision bugs." }, { - "left": "issue:43377", - "right": "issue:43819", + "left": "issue:28530", + "right": "issue:34567", "accept": false, - "reason": "MIMI padding-mask inconsistency and DAC from_latents vs forward STE mismatch are different codec/model correctness bugs." + "reason": "Early stopping/eval metric handling is unrelated to TrainerState.num_input_tokens_seen not updating." }, { - "left": "issue:28530", - "right": "issue:34567", + "left": "issue:43377", + "right": "issue:43819", "accept": false, - "reason": "Early stopping metric lookup and num_input_tokens_seen not updating are unrelated TrainerState/Trainer behaviors." + "reason": "MIMI batching/padding-mask support is a different bug from DAC.from_latents not matching forward pass without STE." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Llava Next loading segfault and Qwen3OmniMoe video-input ValueError are different multimodal failures." + "reason": "Both are multimodal loading/processing failures, but they affect different models and different error conditions." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "A specific model-loading failure for sleng-bert does not match a Mistral tokenizer regex-patching crash." + "reason": "A specific model load failure is not the same as a tokenizer regex patch AttributeError." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "These are different loading regressions affecting different code paths and model families." + "reason": "Different models and different failure modes; no shared concrete code-path." }, { - "left": "issue:43089", - "right": "issue:45987", + "left": "issue:42915", + "right": "issue:44291", "accept": false, - "reason": "Generation performance overhead and StaticCache sequence-length type mismatch are not the same bug." + "reason": "FP8 model failure and init_empty_weights unexpected argument are separate loader/initialization bugs." }, { - "left": "issue:42915", - "right": "issue:44291", + "left": "issue:43089", + "right": "issue:45987", "accept": false, - "reason": "FP8 configuration failure for Qwen3Moe and init_empty_weights rejecting _is_hf_initialized are distinct initialization errors." + "reason": "Generation performance overhead is not the same as the StaticCache return-type contract bug." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export exportability with torch_compilable_check is unrelated to OmDet-Turbo's processor image-size mismatch." + "reason": "torch.export exportability issue versus processor image-size mismatch are unrelated." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "An fbank feature-extraction issue and compiled flex_attention failing on torch>=2.9 are different subsystems." + "reason": "Audio fbank feature extraction and compiled flex_attention on torch>=2.9 are separate issues." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "A tokenizer method-name conflict and a KeyError when loading MobileLLM-125M are not the same underlying problem." + "reason": "Tokenizer class-method conflict and MobileLLM key error are different loading regressions." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "AddedToken special-argument duplication is unrelated to the physical-intelligence/fast loading failure." + "reason": "AddedToken special-argument duplication is a tokenizer construction bug, not the same as the model-loading regression in 44987." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Both involve float8-adjacent failures, but one is Qwen3Moe FP8 config and the other is a missing Float8 storage type; not the same bug." }, { "left": "issue:45216", "right": "issue:45357", "accept": true, - "reason": "Both report the same Qwen3.5 save_pretrained serialization bug: incorrect checkpoint contents/visual encoder keys persisting across versions." + "reason": "Both report the same Qwen3.5 save_pretrained regression, with 45357 explicitly indicating the incorrect visual-encoder-key issue persists in a later release." } ] }, "evaluator_result": { "accept": true, - "feedback": "The evaluation is grounded in the packet and the soft-edge calls stay conservative. The GGUF pair (43452/43482) and the Qwen3.5 save_pretrained pair (45216/45357) look like plausible same-bug clusters from the titles alone, and the other pairings are reasonably rejected without overclaiming shared root cause. The summary is cautious and consistent with the evidence provided." + "feedback": "The summary is grounded in the packet and stays appropriately cautious about the mostly heterogeneous set. The only accepted soft pair, 45216/45357, is well-supported by the titles as the same Qwen3.5 save_pretrained regression with a later follow-up report. The other soft-edge rejections are conservative and do not overclaim duplication." }, "evaluator_used": true, "retried": false, @@ -3362,8 +3573,8 @@ "soft_pairs": [ "issue:44479|issue:45325", "issue:44991|issue:46315", - "issue:36032|issue:41720", "issue:43531|issue:44877", + "issue:36032|issue:41720", "issue:43425|issue:44403", "issue:43611|issue:44534", "issue:43296|issue:43531", @@ -3386,235 +3597,235 @@ "issue:43317|issue:43856", "issue:42175|issue:45542", "issue:41553|issue:43582", - "issue:43818|issue:46402", "issue:43531|issue:44291", + "issue:43818|issue:46402", "issue:44079|issue:44534", "issue:44829|issue:45479", "issue:44462|issue:45812" ], - "prepared_review_unit_hash": "b348e060afeee97eb722a51564c65ff0ba523c1481feee05b7a0d33a5fd8ce5c", + "prepared_review_unit_hash": "f273a113f29cc95e06805f483cec77ed920f0a4e830b4f6bd755b7d8533c40f8", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36492, - "estimated_input_tokens": 9123, - "estimated_eval_tokens": 18502 + "serialized_chars": 36496, + "estimated_input_tokens": 9124, + "estimated_eval_tokens": 18504 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issues were grouped by soft similarity. The only convincing duplicate pair is the tokenizer.json / wrong-token-IDs regression affecting Granite models; everything else looks like distinct bugs or regressions in different code paths.", - "confidence": 0.79, - "canonical_issue_reason": "Issue 45812 is the strongest canonical issue: it is the broadest tokenizer regression in the set, has inbound references, and matches the underlying tokenizer.json / wrong-token-ID symptom more directly than the narrower repo-loading report.", + "summary": "The cluster is a mix of several Transformer regression families, with a few plausible near-duplicates but not enough evidence to collapse it into one broad bucket. The clearest canonical issue is the Granite tokenizer regression; the TensorFlow packaging pair looks related enough to keep as a soft duplicate, while the Qwen2.5-VL video/rope-index pair is too weakly supported to merge.", + "confidence": 0.73, + "canonical_issue_reason": "Issue 45812 is the strongest representative: it is a concrete, model-wide Granite tokenizer regression, has downstream references, and best captures the cluster\u2019s most clearly supported duplicate-worthy bug.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45812 is the best representative for the cluster because it is broad, user-visible, and likely to subsume the related tokenizer-loading symptom in 44462.", + "best_issue_reason": "Issue 45812 has the clearest user impact and the most specific evidence of a real regression affecting multiple Granite models, making it the best overall issue in the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44479", "right": "issue:45325", "accept": false, - "reason": "Both involve Qwen VL video/position logic, but one is a broad video-input regression and the other is a specific still-image rope-index scaling bug; not the same concrete defect." + "reason": "Same model family and similar video-positioning terminology, but the packet only shows topical overlap; the underlying bug and fix scope are not clearly the same." }, { "left": "issue:44991", "right": "issue:46315", "accept": false, - "reason": "Tokenizer loading failure vs tokenizer performance regression; different symptoms and likely different fixes." + "reason": "Both concern tokenizers, but one is a load failure and the other is a performance regression in conversion; different defects." }, { - "left": "issue:36032", - "right": "issue:41720", + "left": "issue:43531", + "right": "issue:44877", "accept": false, - "reason": "T5 tokenizer load error and Qwen3 device-map CUDA assert are unrelated bugs." + "reason": "Different model/configuration failures with only broad overlap around loading and config handling." }, { - "left": "issue:43531", - "right": "issue:44877", + "left": "issue:36032", + "right": "issue:41720", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and granite_speech strict-config loading are unrelated." + "reason": "Tokenizer attribute conflict vs auto device-mapping CUDA assert are unrelated code paths." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch 2.10 compatibility and noisy transformer loading messages are different issues." + "reason": "Torch version incompatibility and noisy loading behavior are not the same underlying bug." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "Base-model-prefix loading breakage and non-persistent buffer junk initialization are distinct regressions." + "reason": "Both are v5 regressions, but they affect different internals and would not be fixed by one focused change." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL/vLLM load failure and Qwen3-MoE sliding_window are unrelated." + "reason": "Different models and failure modes; too little evidence of a shared root cause." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "_is_hf_initialized regression and Strix Halo mmap OOM are unrelated." + "reason": "_is_hf_initialized breakage and mmap/OOM on Strix Halo are distinct issues." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Multi-image IterableDataset finetuning failure is not the same bug as the Qwen2.5-VL still-image rope-index scaling issue." + "reason": "Both are Qwen VL-related, but one is multi-image finetuning and the other is rope-index temporal scaling; not the same bug." }, { "left": "issue:45812", "right": "issue:46491", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different behaviors; not clearly the same code-path bug." + "reason": "Different tokenizer regressions on different model families and symptoms; too broad to merge." }, { "left": "issue:42796", "right": "issue:43746", "accept": false, - "reason": "Ministral tokenizer BPE-marker handling and GraniteSpeech PEFT checkpoint loading are unrelated." + "reason": "Tokenizer marker handling and GraniteSpeech PEFT checkpoint loading are unrelated." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "Whisper runtime failure and Float8 storage lookup error are unrelated." + "reason": "Whisper runtime failure and Float8 storage lookup are unrelated defects." }, { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "A logging/noise complaint is not the same as a tokenizer-loading failure." + "reason": "Generic loading noise is not the same as failing to load tokenizer files from the repository." }, { "left": "issue:33666", "right": "issue:40444", "accept": false, - "reason": "Both mention Qwen VL training, but one is multi-GPU training and the other is multi-image prompt finetuning; too broad to treat as duplicates." + "reason": "Both mention Qwen VL, but they describe different training failures and likely different fixes." }, { "left": "issue:43632", "right": "issue:44534", "accept": false, - "reason": "Different v5 regressions with different symptoms and fixes." + "reason": "Separate v5 regressions with different failure mechanisms." }, { "left": "issue:43381", "right": "issue:45923", "accept": false, - "reason": "Eval-mode gradient-checkpointing restriction and missing supports_gradient_checkpointing flag are related to the area but not the same bug." + "reason": "Gradient-checkpointing eval-mode limitation vs missing capability flag on a trust_remote_code model are different bugs." }, { "left": "issue:44079", "right": "issue:45663", "accept": false, - "reason": "ModelOutput key assignment and Gemma4 FSDP2 shared_kv_states KeyError are unrelated." + "reason": "ModelOutput key bookkeeping and Gemma4 FSDP2 shared_kv_states rebuilding are unrelated." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "Unexpected _is_hf_initialized argument and missing Float8 storage object are different TypeErrors from different code paths." + "reason": "init_empty_weights/_is_hf_initialized and Float8 storage lookup do not share the same root cause." }, { "left": "issue:40990", "right": "issue:44811", "accept": false, - "reason": "High perplexity on GPT-OSS and Whisper batch_decode skip_special_tokens behavior are unrelated." + "reason": "GPT-OSS perplexity and Whisper batch_decode special-token handling are unrelated." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "Qwen2.5-GGUF loading under v5 and CLIPTokenizer model_max_length are unrelated." + "reason": "Different model-loading/tokenizer bugs with different symptoms and likely different fixes." }, { "left": "issue:43482", "right": "issue:43525", "accept": false, - "reason": "Qwen2.5-GGUF loading and Llama4Config missing pad_token_id are unrelated." + "reason": "Distinct model-loading failures; no clear shared code-path problem." }, { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "Runtime special-token post-processor sync and mdeberta add_special_tokens BOS/EOS behavior are related only broadly, not clearly the same concrete bug." + "reason": "Runtime post-processor updates vs missing BOS/EOS insertion are different tokenizer behaviors." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values and multi-image finetuning are different features and failures." + "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL multi-image finetuning." }, { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "device_map/offload loading and Qwen3 MoE training memory usage are unrelated." + "reason": "Device-map offload failure and Qwen3 MoE memory inefficiency are different issues." }, { "left": "issue:42175", "right": "issue:45542", - "accept": false, - "reason": "Both concern TensorFlow backend packaging, but one is the torch extra missing TF support while the other is a tensorboard-without-TF runtime backend error; not clearly the same bug." + "accept": true, + "reason": "Both point to the same packaging/backend dependency problem: a torch-only install leaves TensorFlow backend support in a broken state." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Bad AutoTokenizer error message for Voxtral and Apple Silicon caching_allocator_warmup TypeError are unrelated." + "reason": "Bad tokenizer error messaging and an Apple Silicon allocator TypeError are unrelated." }, { - "left": "issue:43818", - "right": "issue:46402", + "left": "issue:43531", + "right": "issue:44291", "accept": false, - "reason": "Video-LLaVA temporal-attention/weight-sharing issue and Sam3Model text-encoder loading issue are unrelated." + "reason": "Sliding_window/Qwen3-MoE and init_empty_weights argument breakage are unrelated." }, { - "left": "issue:43531", - "right": "issue:44291", + "left": "issue:43818", + "right": "issue:46402", "accept": false, - "reason": "Qwen3-MoE sliding_window and init_empty_weights/_is_hf_initialized are unrelated." + "reason": "Video-LLaVA tower initialization and Sam3Model text encoder loading are different model-specific bugs." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "ModelOutput key assignment and non-persistent buffer initialization are different bugs." + "reason": "ModelOutput key assignment and non-persistent buffer filling are unrelated." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "Both touch degenerate training/loss, but one is tied to flash_attention_3 and the other to problem_type=num_labels=1; different root causes." + "reason": "Both involve degenerate classification behavior, but one is flash_attention_3-specific and the other is a num_labels/problem_type edge case." }, { "left": "issue:44462", "right": "issue:45812", - "accept": true, - "reason": "These look like the same tokenizer regression: ignoring tokenizer.json would produce the wrong token IDs reported for Granite models." + "accept": false, + "reason": "Both are tokenizer regressions, but one is about ignoring tokenizer.json while the other is Granite-specific wrong token IDs; too broad to merge confidently." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The cluster summary is grounded in the packet and stays conservative: most pairwise rejections are well-justified as different models/code paths, and the only accepted pair (44462/45812) is a plausible shared tokenizer regression rather than an overclaim. The choice of 45812 as canonical is also supported by its broader Granite-wide impact and inbound references." + "accept": false, + "feedback": "The canonical Granite issue is well supported by the packet, and the TensorFlow packaging pair looks plausibly related. However, the accepted Qwen2.5-VL pair (issue:44479 vs issue:45325) is not conservative enough: the packet only shows topical overlap in video/rope-index handling, not clear evidence they are the same underlying bug or mergeable into one fix. The summary also leans a bit too far by calling the cluster mostly distinct with only a couple of true duplicates, since that judgment isn\u2019t strongly evidenced from titles alone." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -3656,12 +3867,12 @@ "issue:43749", "issue:43792", "issue:43818", + "issue:43819", "issue:43873", "issue:43881", "issue:43957", "issue:43994", "issue:44079", - "issue:44186", "issue:44265", "issue:44291", "issue:44466", @@ -3690,9 +3901,9 @@ "issue:44265|issue:45072", "issue:43749|issue:44079", "issue:42175|issue:43296", - "issue:45325|issue:45381", "issue:40444|issue:45381", "issue:43792|issue:43957", + "issue:45325|issue:45381", "issue:43994|issue:44898", "issue:43454|issue:44466", "issue:45325|issue:45834", @@ -3701,8 +3912,8 @@ "issue:43632|issue:45072", "issue:43701|issue:45325", "issue:44291|issue:44877", - "issue:44479|issue:45823", "issue:43296|issue:45800", + "issue:44479|issue:45823", "issue:43873|issue:45701", "issue:43606|issue:44265", "issue:43262|issue:43746", @@ -3711,223 +3922,226 @@ "issue:43881|issue:44488", "issue:35707|issue:41720", "issue:43454|issue:43550", - "issue:44186|issue:44265" + "issue:43606|issue:43819" ], - "prepared_review_unit_hash": "b42717f1cd88b2472982abb090430f390d780b7b303ad6d7c6a64f0466171f13", + "prepared_review_unit_hash": "23844196764736e8cff8391f6cb4637309825bf720a07c81fd99dd8a2b9e7378", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36758, - "estimated_input_tokens": 9190, - "estimated_eval_tokens": 18636 + "serialized_chars": 36745, + "estimated_input_tokens": 9187, + "estimated_eval_tokens": 18630 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: unrelated model-loading, tokenizer, video, audio, device-mapping, and export failures. I don\u2019t see any soft pair that safely merges to the same concrete bug/change, so all soft edges are rejected. The broad open tokenization issue is the best overall representative of the cluster.", - "confidence": 0.95, - "canonical_issue_reason": "issue:45701 is the broadest, still-open item in the set and has the most cross-reference activity, so it works best as the representative issue for this heterogeneous cluster.", + "summary": "This cluster is mostly heterogeneous and spans many unrelated regressions. The only strong duplicate-like pair is the Qwen2.5-VL visual position-id regression; the rest look like separate bugs in different models/subsystems.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45325 is the most concrete root-cause report in the Qwen2.5-VL position-id regression area (`get_rope_index` / temporal scaling), so it\u2019s the best anchor for that narrow duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45701 is the most general and central issue here (version-dependent tokenization changes), making it the best issue-level anchor despite the cluster not being a true duplicate set.", + "best_issue_reason": "issue:45325 is the most actionable issue in the cluster\u2019s one clear overlap area; the remaining issues are too diverse to make a single global canonical beyond that sub-bucket.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "Different subsystems and failures: MIMI encoder padding-mask batching vs Gemma3 NaN embeddings from sliding-window attention." + "reason": "Both involve batching/masking behavior, but they are different models and different failure modes (MIMI padding mask vs Gemma3 sliding-window NaNs)." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Both are multimodal, but one is about Video-LLaVA tower weights/temporal attention and the other about Qwen3.5 save_pretrained visual encoder keys." + "reason": "Unrelated problems: Video-LLaVA tower architecture/weights vs Qwen3.5 save_pretrained key mapping." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "Unrelated problems: Qwen3 auto device mapping CUDA assert vs Mistral tokenizer regex patch crash." + "reason": "Different subsystems entirely: device-map CUDA assert vs tokenizer regex patch crash." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Different code paths: non-persistent buffer serialization vs ModelOutput key assignment semantics." + "reason": "One is buffer initialization/serialization junk; the other is ModelOutput key assignment logic." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both involve v5 model loading, but the concrete bugs differ: base_model_prefix loading vs custom model initialization." + "reason": "Both are v5 loading regressions, but the concrete bugs differ (base_model_prefix loading vs custom model initialization)." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "Export failures with torch_compilable_check are not the same bug as bf16 dtype mismatches in specific models." + "reason": "Different code paths: torch.export with torch_compilable_check vs bf16 dtype mismatches in inference." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM efficient loading is unrelated to ModelOutput key assignment." + "reason": "FSDP CPU RAM-efficient loading is unrelated to ModelOutput key handling." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "Package/backend dependency issue versus a model-load failure in PaddleOCR-VL; no shared underlying bug." - }, - { - "left": "issue:45325", - "right": "issue:45381", - "accept": false, - "reason": "Both are Qwen2.5-VL position-id regressions, but one is still-image temporal scaling and the other is video vision_position_ids; too different to merge safely." + "reason": "Packaging/backend dependency issue vs a model-loading failure in vLLM." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "IterableDataset with multiple images per prompt is a different failure mode from video vision_position_ids." + "reason": "Both mention multimodal Qwen models, but one is training with multiple images per prompt and the other is a video-position-id regression." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Whisper inference failure and meta-device model loading breakage are unrelated." + "reason": "Whisper runtime failure and meta-device loading breakage are different bugs." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Both describe the same Qwen2.5-VL visual position-id / rope-index regression in transformers 5.3.0; one is the still-image path and the other the video-input path." }, { "left": "issue:43994", "right": "issue:44898", "accept": false, - "reason": "Different models and symptoms: SigLIP2 nonsensical outputs vs Perceiver non-default-resolution failure." + "reason": "Different vision-model failures: SigLIP2 nonsensical outputs vs Perceiver non-default-resolution interpolation failure." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "Both mention tied lm_head weights, but one is a missing tie causing bad generation and the other is device-dependent serialization inconsistency." + "reason": "Related to lm_head/tied weights, but one is a failed tie causing garbage generation while the other is inconsistent serialization depending on device." }, { "left": "issue:45325", "right": "issue:45834", "accept": false, - "reason": "Qwen2.5-VL temporal position-id scaling is unrelated to Kosmos2.5 OCR index errors." + "reason": "Both are vision/multimodal issues, but Qwen2.5-VL position IDs and Kosmos2.5 OCR long-input indexing are unrelated." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Different issues: FP8 config failure for Qwen3Moe vs GGUF loading regression for Qwen2.5." + "reason": "Different regressions: FP8 config failure vs GGUF loading failure." }, { "left": "issue:42796", "right": "issue:45381", "accept": false, - "reason": "Tokenizer BPE marker handling is unrelated to Qwen2.5-VL video position-id bugs." + "reason": "Tokenizer BPE marker handling is unrelated to Qwen video position ids." }, { "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "_is_hf_initialized flag regression is not the same as dtype mismatches during bf16 inference." + "reason": "Different regressions: `_is_hf_initialized` flag handling vs dtype mismatches in inference." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "Checkpoint resume key mismatch is unrelated to Qwen2.5-VL position-id computation." + "reason": "Checkpoint resume key mismatch is not the same as Qwen2.5-VL rope-index/position-id scaling." }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "init_empty_weights argument mismatch and strict config loading are different failures." + "reason": "Both are loading/configuration issues, but one is an unexpected init argument and the other is strict config rejection." }, { - "left": "issue:44479", - "right": "issue:45823", + "left": "issue:43296", + "right": "issue:45800", "accept": false, - "reason": "Both are model regressions, but one is Qwen video-input behavior and the other is Gemma4 device-mapping mismatch." + "reason": "vLLM model-load failure and torch/transformers version incompatibility are separate problems." }, { - "left": "issue:43296", - "right": "issue:45800", + "left": "issue:44479", + "right": "issue:45823", "accept": false, - "reason": "PaddleOCR-VL loading failure is unrelated to torch/transformers version incompatibility." + "reason": "Video regression in Qwen multimodal models vs Gemma4 PLE device mismatch; not the same code path." }, { "left": "issue:43873", "right": "issue:45701", "accept": false, - "reason": "Quantization offloading behavior is unrelated to tokenization changes across versions." + "reason": "Quantized offloading behavior and tokenizer-version changes are unrelated." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "CPU offload device mismatch and torch.export failure with torch_compilable_check are different code-path bugs." + "reason": "CPU offload device mismatch for Bark is unrelated to torch.export / compilable-check failures." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio chat-template sample-rate defaulting is unrelated to GraniteSpeech PEFT checkpoint loading." + "reason": "Audio processor sample-rate defaults and GraniteSpeech checkpoint loading are different bugs." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "High perplexity on a causal LM and Whisper empty-transcription generation failure are unrelated." + "reason": "High perplexity on GPT-OSS is unrelated to Whisper empty-transcription generation failure." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "Bark voice_preset and progressive generation with inputs_embeds/past_key_values are unrelated." + "reason": "Bark voice_preset behavior and progressive generation with inputs_embeds/past_key_values are unrelated." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "Two separate model-loading failures for different models with no shared concrete bug." + "reason": "Different model-loading failures for different models; no shared concrete bug." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "Progressive generation with embeds/past KV is unrelated to a CUDA assert in Qwen3 device mapping." + "reason": "Progressive generation with embeddings/past KV is unrelated to a CUDA assert on auto device mapping." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "Tied-weight loading/serialization issue is unrelated to Bamba torch.compile + SDPA failure." + "reason": "lm_head tying/serialization and Bamba torch.compile SDPA failures are different issues." }, { - "left": "issue:44186", - "right": "issue:44265", + "left": "issue:43606", + "right": "issue:43819", "accept": false, - "reason": "Tokenizer crash on NER padding/truncation is unrelated to torch.export failures." + "reason": "CPU offload device mismatch for Bark is unrelated to DAC.from_latents / missing STE mismatch." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: most items are clearly heterogeneous, and the Qwen2.5-VL position-id/rope-index cluster is the only plausible duplicate sub-bucket. The soft-edge rejections are suitably conservative, and the one accepted pair is reasonably supported by the titles and shared model/version context." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3963,13 +4177,13 @@ "issue:43454", "issue:43526", "issue:43572", - "issue:43606", "issue:43720", "issue:43749", - "issue:43819", "issue:43873", "issue:43881", "issue:43883", + "issue:44186", + "issue:44265", "issue:44387", "issue:44466", "issue:44534", @@ -3993,7 +4207,7 @@ "issue:46402" ], "soft_pairs": [ - "issue:43606|issue:43819", + "issue:44186|issue:44265", "issue:39692|issue:40444", "issue:41720|issue:45823", "issue:42796|issue:46209", @@ -4012,8 +4226,8 @@ "issue:42617|issue:42915", "issue:43749|issue:44534", "issue:41720|issue:43322", - "issue:44466|issue:45357", "issue:43526|issue:43883", + "issue:44466|issue:45357", "issue:26413|issue:28530", "issue:43881|issue:44991", "issue:41720|issue:44977", @@ -4023,208 +4237,208 @@ "issue:44877|issue:44991", "issue:43065|issue:46402" ], - "prepared_review_unit_hash": "0b52291867d079c7814b25807e7a4c96744057d9c31ff48ec59ef97cd8433169", + "prepared_review_unit_hash": "e8833962c9c9cbd4e29c208844e24d3d521f4ac3fd25c8c0ab3759efc0ef32ce", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35940, - "estimated_input_tokens": 8985, - "estimated_eval_tokens": 18226 + "serialized_chars": 35971, + "estimated_input_tokens": 8993, + "estimated_eval_tokens": 18242 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a noisy mix of unrelated Transformers bug reports, with a loose concentration around loading/offloading/quantization regressions; none of the soft pairs are close enough to merge as duplicates.", - "confidence": 0.63, - "canonical_issue_reason": "Issue #43873 is the broadest and most central representative of the cluster\u2019s dominant theme (offloading/quantization regressions) and best fits the shared failure mode seen in several neighboring reports.", + "summary": "The cluster is heterogeneous: the soft pairs share broad themes (tokenizers, loading/offloading, vision models, quantization), but each pair points to a different concrete failure mode, so I reject all duplicate edges. There are no PRs in the cluster; if forced to anchor an issue, 43873 is the broadest umbrella-style report.", + "confidence": 0.65, + "canonical_issue_reason": "Issue 43873 is the best available anchor because it is broad, active, and closest to an umbrella loading/offloading/quantization complaint among the set.", "canonical_pr_reason": null, - "best_issue_reason": "#43873 is a strong cluster anchor because it targets the same high-level code path as several nearby issues and is broad enough to represent the recurring offload/quantization problem space.", + "best_issue_reason": "43873 is the most suitable global issue representative since it has the widest problem framing and enough discussion/inbound refs to serve as a cluster anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43606", - "right": "issue:43819", + "left": "issue:44186", + "right": "issue:44265", "accept": false, - "reason": "Different bugs in different code paths: CPU-offload device mismatch vs DAC latent/forward mismatch." + "reason": "Different bugs: LayoutLMv2 tokenizer crash vs torch.export export compatibility." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "Both are multimodal issues, but one is a SigLIP2 docs example problem and the other is a Qwen2.5-VL iterable dataset failure." + "reason": "Different artifacts and failures: SigLIP2 doc example errors vs Qwen2.5-VL iterable-dataset multi-image finetuning." }, { "left": "issue:41720", "right": "issue:45823", "accept": false, - "reason": "Both involve device_map=auto, but they affect different models with different failure modes." + "reason": "Both involve device mapping, but the symptoms and models differ: cudaErrorAssert on Qwen3 vs forward device mismatch on Gemma4." }, { "left": "issue:42796", "right": "issue:46209", "accept": false, - "reason": "Tokenizer BPE-marker handling and FP8 inference on B200 are unrelated bugs." + "reason": "Tokenizer BPE-marker handling and FP8 inference on B200 are unrelated failures." }, { "left": "issue:46305", "right": "issue:46315", "accept": false, - "reason": "A sliding-window mask off-by-one is unrelated to tokenizer conversion performance." + "reason": "Different code paths: sliding-window mask off-by-one vs slow convert_tokens_to_ids." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "Different model/processor issues: padding-mask batching bug vs image size mismatch." + "reason": "Both are input-shape/processing issues, but one is MIMI padding-mask consistency and the other is OmDet-Turbo resolution mismatch." }, { "left": "issue:44568", "right": "issue:45920", "accept": false, - "reason": "Both are tokenizer regressions, but one is a specific BOS/EOS special-tokens bug and the other is a broader wrong-token-id regression." + "reason": "Tokenizer regressions, but not the same bug: missing BOS/EOS insertion vs wrong token IDs across multiple models." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "Weight-tying corruption and bf16 dtype mismatch are separate problems." + "reason": "Different concrete problems: tied lm_head weights vs bfloat16 dtype mismatches." }, { "left": "issue:43257", "right": "issue:45923", "accept": false, - "reason": "MOE weight conversion during accelerate/deepspeed loading is not the same as a missing gradient-checkpointing flag." + "reason": "Different model-loading concerns: Qwen3 MoE weight conversion vs missing supports_gradient_checkpointing flag." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Related theme, but not the same concrete bug: dequantized model offload failure vs broader offloading-with-quantization report." + "reason": "Related offloading/quantization theme, but different failures and scopes; not the same underlying bug." }, { "left": "issue:42796", "right": "issue:44977", "accept": false, - "reason": "Tokenizer marker handling is unrelated to Qwen3.5 flash-attention generation failure." + "reason": "Tokenizer marker handling vs flash-attention generation failure are unrelated." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "Both concern input handling, but they are different models and different failure mechanisms." + "reason": "Different bug classes: batched-vs-single consistency vs non-default-resolution image classification failure." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "Perplexity/benchmark behavior is unrelated to LoRA gradient-checkpointing ineffectiveness." + "reason": "Perplexity on a specific model/task vs gradient-checkpointing ineffectiveness are unrelated." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "Different loading/serialization bugs: tied weights vs packed-weights unpacking." + "reason": "Different subsystems: weight tying in AyaVision vs packed-weight unpacking during accelerate loading." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "A missing pad_token_idx config field is not the same as failing to unwrap text_config in composite multimodal configs." + "reason": "Different config problems: missing pad_token_idx vs from_config not unwrapping text_config." }, { "left": "issue:43322", "right": "issue:46050", "accept": false, - "reason": "A loading segfault and a save/load quantization regression are distinct issues." + "reason": "Llava Next segmentation fault and sinq save/load breakage are different loading bugs." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "3d_parallel.py runtime failure and FineGrainedFP8Config failure are unrelated." + "reason": "Different failures: 3d_parallel.py runtime issue vs FineGrainedFP8Config failure." }, { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "FSDP CPU RAM loading and non-persistent buffer initialization are different regressions." + "reason": "Both regressions touch loading, but one is FSDP CPU RAM efficient loading and the other is non-persistent buffer initialization." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Different models and different crash signatures; not the same underlying bug." + "reason": "Different models and symptoms: Qwen3 auto device mapping assert vs Llava Next loading segmentation fault." }, { - "left": "issue:44466", - "right": "issue:45357", + "left": "issue:43526", + "right": "issue:43883", "accept": false, - "reason": "Both are serialization-related, but one concerns lm_head weight consistency and the other visual encoder keys." + "reason": "Unrelated bugs: BeitImageProcessorFast label reduction vs Molmo missing all_tied_weights_keys." }, { - "left": "issue:43526", - "right": "issue:43883", + "left": "issue:44466", + "right": "issue:45357", "accept": false, - "reason": "Image label reduction and missing all_tied_weights_keys are unrelated." + "reason": "Both involve serialization/loading of weights, but the concrete bugs differ: lm_head serialization vs visual encoder keys." }, { "left": "issue:26413", "right": "issue:28530", "accept": false, - "reason": "Checkpoint resumption and early-stopping metric resolution are separate training-loop issues." + "reason": "Resume-from-checkpoint epoch iterator failure and early-stopping metric lookup are different trainer issues." }, { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "Model loading failure and tokenizer loading regression are different problem areas." + "reason": "glm-4v-9b loading failure and tokenizer loading regression for EMBEDDIA/est-roberta are unrelated." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Both are Qwen-family reports, but one is device-map crash and the other is flash-attention generation behavior." + "reason": "Same model family name only; one is device-map assert, the other is flash-attention generation bug." }, { "left": "issue:44387", "right": "issue:45704", "accept": false, - "reason": "Int4 quantization OOM and an apex RMSNorm memory leak are different memory issues." + "reason": "Different memory issues: int4 reserved-memory/OOM regression vs apex FusedRMSNorm leak." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "Trainer API keyword mismatch is unrelated to 3d_parallel.py execution failure." + "reason": "CustomTrainer compute_loss signature change and 3d_parallel.py failure are unrelated." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "CustomTrainer loss signature breakage is unrelated to Qwen3Moe FP8 loading." + "reason": "Different runtime errors: compute_loss kwarg mismatch vs Qwen3Moe FineGrainedFP8Config failure." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config loading and tokenizer loading are separate regressions." + "reason": "Strict config blocking granite_speech and tokenizer loading regression for est-roberta are different issues." }, { "left": "issue:43065", "right": "issue:46402", "accept": false, - "reason": "SAM3 pixel-decoder placeholder convolution and missing text-encoder weights are different SAM3 issues." + "reason": "Dummy Conv2d in Sam3PixelDecoder and Sam3Model text-encoder weights not loaded are not the same bug." } ] }, @@ -4311,8 +4525,8 @@ "issue:44186|issue:44610", "issue:44610|issue:44898", "issue:43295|issue:45701", - "issue:44423|issue:45406", "issue:42915|issue:43317", + "issue:44423|issue:45406", "issue:45632|issue:45684", "issue:36010|issue:42617", "issue:41720|issue:45161", @@ -4320,223 +4534,223 @@ "issue:42915|issue:46291", "issue:28530|issue:33666", "issue:44671|issue:44743", - "issue:44186|issue:44361", "issue:36331|issue:41720", + "issue:44186|issue:44361", "issue:43526|issue:43881", "issue:42915|issue:43322", - "issue:43316|issue:44913", "issue:43540|issue:43782", + "issue:43316|issue:44913", "issue:28530|issue:30064", "issue:33357|issue:35707" ], - "prepared_review_unit_hash": "d0e242213171519f687df31118c5bf0941c07e9df27c9a283f92602f60c0229d", + "prepared_review_unit_hash": "edae83e62f798135001a7e9ce26db20657462977266f13712edeb95607af5bed", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36673, - "estimated_input_tokens": 9169, - "estimated_eval_tokens": 18594 + "serialized_chars": 36677, + "estimated_input_tokens": 9170, + "estimated_eval_tokens": 18596 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft edge looks like a true duplicate; the pairings are mostly broad subsystem similarities, not the same concrete bug. If one issue must serve as the cluster representative, issue:45701 is the broadest and most referenced open report, but it is not a duplicate of the others.", - "confidence": 0.76, - "canonical_issue_reason": "issue:45701 is the most representative hub by breadth and inbound references, so it is the best single issue to anchor the cluster despite not matching the others as a true duplicate.", + "summary": "This cluster is highly heterogeneous: the suggested pairs span unrelated bugs in tokenization, multimodal processors, quantization/offload, configs, training, and serve/runtime paths. I don\u2019t see any true duplicates among the soft edges.", + "confidence": 0.96, + "canonical_issue_reason": "No strong canonical issue exists here because the items are not a coherent duplicate set; if one must be picked, issue 45701 is the broadest and most active, but it is still not representative of the rest of the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Open, broadly framed, and one of the most referenced reports in the set; it is the most plausible triage anchor, even though the surrounding issues are not duplicates.", + "best_issue_reason": "Issue 45701 is the most general and externally active item in the list, but only as a weak anchor; the surrounding issues are too unrelated for it to be a meaningful canonical duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Different bugs: video processor batch-shape output vs a tokenization/version-change report." + "reason": "Video processor batch-shape bug vs tokenization/version behavior; different component and failure mode." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "Different code paths: LayoutLMv2 tokenizer crash vs dtype mismatch in bfloat16 inference." + "reason": "LayoutLMv2 tokenizer crash vs dtype mismatch in inference; unrelated subsystems." }, { "left": "issue:40990", "right": "issue:45701", "accept": false, - "reason": "Perplexity regression on a model/eval setup is not the same as a tokenization-change report." + "reason": "Perplexity regression on a specific model is not the same as a generic tokenization-version complaint." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Both involve loading/offload behavior, but one is quantization offloading and the other is mmap/OOM on Strix Halo." + "reason": "Quantization/offload behavior and mmap OOM are different root causes and code paths." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "Both are model loading/runtime failures, but one is SDPA+torch.compile and the other is accelerate packed-weight unpacking." + "reason": "torch.compile/SDPA bug for Bamba vs packed-weight loading bug for BitNet; not the same fix." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Separate failures: Qwen3 auto device-map cuda assert vs a 3d_parallel.py execution problem." + "reason": "Qwen3 device-map crash vs failure to run a 3d_parallel script; no shared underlying bug." }, { "left": "issue:43577", "right": "issue:43708", "accept": false, - "reason": "Unrelated: BLIP2 dtype propagation vs resume_from_checkpoint/max_steps calculation." + "reason": "BLIP2 dtype propagation and trainer checkpoint step math are unrelated." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "Different issues in the same family: untied lm_head weights vs incorrect visual encoder keys on save_pretrained." + "reason": "Weight tying bug in Mistral3 vs incorrect save_pretrained keys in Qwen3.5; different models and issues." }, { "left": "issue:44008", "right": "issue:45823", "accept": false, - "reason": "Different Gemma bugs: variable name collision in forward vs device_map='auto' PLE device mismatch." + "reason": "Gemma3n forward variable collision vs Gemma4 device-map mismatch; unrelated code paths." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Both are device/offload related, but one is torch.compile+SDPA and the other is CPU offload device mismatch." + "reason": "SDPA/compile failure and CPU offload device mismatch are distinct problems." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "Different processors and symptoms: tokenizer crash on NER/padding vs OmDet input-size mismatch." + "reason": "Tokenizer padding/truncation crash vs processor output size mismatch; different components." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Both are vision input-shape problems, but one is processor output size and the other is classification with interpolate_pos_encoding." + "reason": "Both concern image resolution, but they are model-specific processor/mode mismatches with different fixes." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "One is a regression in processor/tokenizer API access; the other is a tokenization-change report without the same failure mode." + "reason": "processor.tokenizer regression and tokenization-version changes are not the same concrete bug." }, { - "left": "issue:44423", - "right": "issue:45406", + "left": "issue:42915", + "right": "issue:43317", "accept": false, - "reason": "Serve-time multimodal crash vs Gemma4Processor missing _tokenizer; different attribute failures." + "reason": "FP8 config failure vs dequantized-model offload loading failure; different quantization paths." }, { - "left": "issue:42915", - "right": "issue:43317", + "left": "issue:44423", + "right": "issue:45406", "accept": false, - "reason": "Both mention model loading, but one is FineGrainedFP8Config and the other is dequantized model offload to gpu+cpu." + "reason": "Continuous-batching multimodal crash and missing _tokenizer in Gemma4Processor are separate serve issues." }, { "left": "issue:45632", "right": "issue:45684", "accept": false, - "reason": "Different filesystem concerns: cache-path collision vs permission propagation from source files." + "reason": "Cache-path collision and read-only permission propagation are unrelated save/load filesystem issues." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to a failing script run." + "reason": "ImportError for GenerationMixin is unrelated to a 3d_parallel runtime failure." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Both touch parallelism, but one is device-map cuda assert and the other is tensor-parallel-only support for GPT-OSS MoE." + "reason": "Qwen3 device-map assert and GPT-OSS MoE tensor-parallel support are different distributed inference bugs." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to StableLmConfig missing pad_token_idx." + "reason": "Qwen3-MoE sliding-window behavior and StableLm pad_token_idx config are unrelated." }, { "left": "issue:42915", "right": "issue:46291", "accept": false, - "reason": "FineGrainedFP8 model failure is unrelated to kernels dependency version/revision import errors." + "reason": "FP8 model loading and kernels package version/revision import failure are separate issues." }, { "left": "issue:28530", "right": "issue:33666", "accept": false, - "reason": "Early stopping metric selection is unrelated to Qwen2-VL multi-GPU training." + "reason": "Trainer early-stopping metric selection vs Qwen2-VL multi-GPU training are unrelated." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "Different model behaviors: masked LM prediction regression vs recurrent-state reset with cache." + "reason": "CamemBERT masked-LM regression and Qwen3.5 recurrent-state reset are different model bugs." }, { - "left": "issue:44186", - "right": "issue:44361", + "left": "issue:36331", + "right": "issue:41720", "accept": false, - "reason": "Both are tokenizer/task issues, but they involve different tokenizer classes and different failure mechanisms." + "reason": "CustomTrainer compute_loss signature mismatch vs Qwen3 device-map crash; no common code path." }, { - "left": "issue:36331", - "right": "issue:41720", + "left": "issue:44186", + "right": "issue:44361", "accept": false, - "reason": "Trainer compute_loss signature mismatch is unrelated to a Qwen3 auto device-map cuda assert." + "reason": "Both are tokenizer-related, but they affect different tokenizers and different failures." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug is unrelated to glm-4v-9b loading failure." + "reason": "BeitImageProcessorFast label reduction bug and glm-4v loading failure are unrelated." }, { "left": "issue:42915", "right": "issue:43322", "accept": false, - "reason": "Different concrete failures: Qwen3Moe FP8 config vs Llava Next loading segfault." + "reason": "FP8 config issue vs Llava Next segmentation fault; different models and root causes." }, { - "left": "issue:43316", - "right": "issue:44913", + "left": "issue:43540", + "right": "issue:43782", "accept": false, - "reason": "Both are config-related, but one is an API discrepancy and the other is rotary_pct not persisting on reload." + "reason": "Qwen3OmniMoe video processing error and Qwen3VL weight_only loading error are not the same bug." }, { - "left": "issue:43540", - "right": "issue:43782", + "left": "issue:43316", + "right": "issue:44913", "accept": false, - "reason": "Different Qwen3Omni/Qwen3VL video vs weight_only loading problems; not the same bug." + "reason": "Gemma3TextConfig API mismatch and GPTNeoX rotary_pct reload issue are different config regressions." }, { "left": "issue:28530", "right": "issue:30064", "accept": false, - "reason": "Training early-stopping logic is unrelated to image processor void segmentation-map handling." + "reason": "Early stopping metric lookup and void segmentation-map processing are unrelated." }, { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "Different runtime failures: Mac bus error with a CLIP model vs progressive generation with inputs_embeds/past_key_values." + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are different failure modes." } ] }, @@ -4614,8 +4828,8 @@ "issue:30064|issue:34634", "issue:43550|issue:43653", "issue:43606|issue:43720", - "issue:43388|issue:45059", "issue:43873|issue:45704", + "issue:43388|issue:45059", "issue:44387|issue:45701", "issue:44492|issue:44509", "issue:41720|issue:43746", @@ -4626,231 +4840,231 @@ "issue:42490|issue:43329", "issue:34634|issue:41720", "issue:34567|issue:35707", - "issue:44451|issue:44991", "issue:43531|issue:44589", + "issue:44451|issue:44991", "issue:45759|issue:45859", "issue:44077|issue:44625", "issue:44625|issue:45479", "issue:44898|issue:45072", - "issue:44062|issue:44451", "issue:45588|issue:45759", + "issue:44062|issue:44451", "issue:44987|issue:45042", "issue:44514|issue:44545", "issue:44368|issue:45357", "issue:44918|issue:44936", "issue:39692|issue:44610" ], - "prepared_review_unit_hash": "a0a37ea827f2b0101a88965a669baaf257eaf1461ee32942d18057b5f268b2a1", + "prepared_review_unit_hash": "1da5fc515a78481ca246e56d1cd5c81ee3d152879d4d90753bff80e241f3c603", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36673, - "estimated_input_tokens": 9169, - "estimated_eval_tokens": 18594 + "serialized_chars": 36679, + "estimated_input_tokens": 9170, + "estimated_eval_tokens": 18596 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous rather than one duplicate set. The only accepted soft edge is the near-duplicate Qwen2_5_VLProcessor crash pair (44514/44545); the rest are distinct bugs, models, or code paths.", - "confidence": 0.84, - "canonical_issue_reason": "issue:44514 is the best canonical issue because it has a near-duplicate in 44545 and the report is specific and well-scoped.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers issues that only look superficially similar by subsystem. The only clear duplicate is the Qwen2_5_VLProcessor batched apply_chat_template crash; the rest should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "issue:44514 is the earlier, more complete report of the same Qwen2_5_VLProcessor.apply_chat_template crash; issue:44545 is effectively the same bug with nearly identical title and should fold into it.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44514 is the strongest representative issue in the set due to the near-duplicate report in 44545; otherwise the cluster is too diverse to have a broader canonical bug.", + "best_issue_reason": "issue:44514 is the best canonical issue because it is the earlier, direct statement of the bug and matches the duplicate report exactly.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:29159", "right": "issue:36032", "accept": false, - "reason": "Both are tokenizer bugs, but one is slow/fast behavior inconsistency and the other is a T5 tokenizer load conflict; different failure modes." + "reason": "Both are tokenizer-related, but one is slow-vs-fast inconsistency and the other is a T5 tokenizer class/method naming conflict; different failure modes and fixes." }, { "left": "issue:30064", "right": "issue:34634", "accept": false, - "reason": "Image processor void segmentation maps and BarkProcessor voice_preset are unrelated processor bugs." + "reason": "Both are processor bugs, but one is about void segmentation maps in image processing and the other is Bark voice_preset handling; unrelated code paths." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "Different subsystems: Bamba torch.compile/SDPA vs BigBirdTokenizer special-token registration." + "reason": "Different models and different bugs: torch.compile/SDPA failure versus BigBirdTokenizer special-token decoding." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "Both involve loading/offload, but one is Bark CPU offload device mismatch and the other is BitNet packed-weight unpacking during accelerate loading." + "reason": "Both involve loading/offload, but one is a device-mismatch CPU offload bug and the other is unpacking BitNet packed weights during accelerate loading; not the same change." }, { - "left": "issue:43388", - "right": "issue:45059", + "left": "issue:43873", + "right": "issue:45704", "accept": false, - "reason": "Metric gathering label truncation is unrelated to SAM3 text/box behavior." + "reason": "Quantization offloading versus an apex FusedRMSNorm memory leak in T5 are distinct problems." }, { - "left": "issue:43873", - "right": "issue:45704", + "left": "issue:43388", + "right": "issue:45059", "accept": false, - "reason": "Quantization offloading and a T5 RMSNorm memory leak are different code paths and symptoms." + "reason": "One is gather_for_metrics truncating tuple labels; the other is SAM3 PCS behavior with text and boxes. No shared code path." }, { "left": "issue:44387", "right": "issue:45701", "accept": false, - "reason": "Int4 quantization OOM is unrelated to tokenization changes." + "reason": "Both mention memory/tokenization regressions, but one is int4 CUDA reserved memory/OOM and the other is tokenization changing across versions; too broad and not the same bug." }, { "left": "issue:44492", "right": "issue:44509", "accept": false, - "reason": "A docs typo in cache strategies is not the same issue as stale pipeline task docs." + "reason": "A typo in cache strategies is not the same as docs still referencing removed pipeline tasks." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "Qwen3 auto device mapping CUDA assert and GraniteSpeech PEFT checkpoint loading are unrelated." + "reason": "Both touch model loading, but one is Qwen3 auto device mapping cudaErrorAssert and the other is GraniteSpeech local checkpoint/PEFT adapter loading." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Both are Qwen-related, but one is sliding_window MoE behavior and the other is composite config unwrapping in from_config." + "reason": "sliding_window on Qwen3-MoE is a generation/runtime issue, while from_config not unwrapping text_config is a config-construction issue." }, { "left": "issue:45406", "right": "issue:45464", "accept": false, - "reason": "Different models and failures: Gemma4Processor missing _tokenizer vs Qwen3.5 streaming chat/completions failure." + "reason": "Gemma4Processor missing _tokenizer is a processor initialization bug; chat/completions streaming on Qwen3.5-0.8B is an API/streaming failure." }, { "left": "issue:45357", "right": "issue:46402", "accept": false, - "reason": "Both touch multimodal save/load, but visual encoder key saving and text encoder weights loading are different bugs." + "reason": "Both are Qwen3.5 multimodal loading-related, but one is save_pretrained key serialization and the other is Sam3Model text encoder weights loading; different artifacts and fixes." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "Bark CPU offload device mismatch is unrelated to BigBirdTokenizer mask token registration." + "reason": "CPU offload device mismatch is unrelated to BigBirdTokenizer mask token registration/empty decode output." }, { "left": "issue:42490", "right": "issue:43329", "accept": false, - "reason": "Trailing-slash import warning and multimodal token counting video-branch NameErrors are unrelated." + "reason": "Trailing-slash local path import warning is a path parsing issue; undefined video-patch helpers in multimodal token counting is a separate processor bug." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping are different models and code paths." + "reason": "BarkProcessor voice_preset failure and Qwen3 auto device-mapping cudaErrorAssert are different models and different runtime paths." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "TrainerState token counting and progressive generation with inputs_embeds/past_key_values are unrelated trainer behaviors." + "reason": "TrainerState.num_input_tokens_seen not updating is unrelated to progressive generation with inputs_embeds and past_key_values." }, { - "left": "issue:44451", - "right": "issue:44991", + "left": "issue:43531", + "right": "issue:44589", "accept": false, - "reason": "Both are load failures, but different models and different tokenizer/model loading issues." + "reason": "Qwen3-MoE sliding_window and Float8 storage lookup are unrelated inference/runtime failures." }, { - "left": "issue:43531", - "right": "issue:44589", + "left": "issue:44451", + "right": "issue:44991", "accept": false, - "reason": "Sliding-window MoE behavior is unrelated to the Float8 storage TypeError." + "reason": "Loading one specific model failing and tokenizer loading for another model are both loading symptoms, but there is no evidence they share the same underlying defect." }, { "left": "issue:45759", "right": "issue:45859", "accept": false, - "reason": "Composite config unwrapping and CPU-offload rotary embedding compatibility are distinct Qwen3.5 issues." + "reason": "One is from_config/text_config handling; the other is rotary embedding not compatible with CPU offload. Different layers, different fix." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "Patchtsmixer post_init validation and Qwen3.5 num_labels propagation are unrelated." + "reason": "patchtsmixer optional post_init validation and Qwen3.5 num_labels propagation are unrelated config issues." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Both involve num_labels/classification, but one is config propagation and the other is degenerate loss behavior across models." + "reason": "num_labels propagation in Qwen3.5 is different from the zero-loss bug when problem_type=single_label_classification and num_labels=1." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Perceiver resolution handling and bfloat16 dtype mismatches are unrelated model bugs." + "reason": "Both are dtype/resolution mismatches, but one is Perceiver interpolation at non-default resolution and the other is bfloat16 inference dtype mismatches for SwitchTransformers/TimmWrapperModel." }, { - "left": "issue:44062", - "right": "issue:44451", + "left": "issue:45588", + "right": "issue:45759", "accept": false, - "reason": "AddedToken special-argument conflict is a tokenizer constructor bug, not the same as ScandiBERT loading failure." + "reason": "Flash attention s_aux=None crash in sink-less models is unrelated to Qwen multimodal config unwrapping." }, { - "left": "issue:45588", - "right": "issue:45759", + "left": "issue:44062", + "right": "issue:44451", "accept": false, - "reason": "Flash-attention s_aux=None crash and Qwen3.5 config unwrapping are different issues." + "reason": "AddedToken special-argument conflict and a specific model loading failure are not the same underlying tokenizer bug from the available evidence." }, { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Physical-intelligence/fast loading failure is unrelated to PIL backend requiring torchvision." + "reason": "A model loading failure for physical-intelligence/fast and PIL backend processors unnecessarily requiring torchvision are different problems." }, { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Near-duplicate reports of Qwen2_5_VLProcessor.apply_chat_template crashing on batched input with padding=False." + "reason": "Same exact Qwen2_5_VLProcessor.apply_chat_template crash on batched input with padding=False; nearly identical titles and likely duplicate reports." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "A warning about tie_word_embeddings during LoRA fine-tuning is unrelated to incorrect visual encoder keys in save_pretrained." + "reason": "tie_word_embeddings warning during LoRA fine-tuning is a config warning; save_pretrained saving wrong visual encoder keys is a serialization bug." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Both mention trainer flows, but one is Qwen3.5 input-embedding unpacking with TRL SFT and the other is evaluate-after-train behavior." + "reason": "Unpacking Qwen3.5 input embeddings with TRL SFT trainer and trainer.evaluate() failing after trainer.train() are different trainer lifecycle bugs." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "SigLIP2 docs errors and OmDet-Turbo processor/model size mismatch are different multimodal issues." + "reason": "SigLIP2 docs example errors and OmDet-Turbo processor/model input-size mismatch are different issues, even though both involve multimodal processors." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The cluster-level conclusion is mostly grounded, but the canonical-issue rationale is slightly overstated: 44514 and 44545 are near-identical reports, not an exact title match. Please soften that wording to \"near-duplicate\" or \"near-exact duplicate.\" The soft-edge verdicts themselves are conservative, and the single accepted pair (44514/44545) looks justified." + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative. The one accepted duplicate pair (issue:44514 / issue:44545) is supported by near-identical titles and timing, and the rest of the soft-edge rejections are appropriately cautious rather than overstated. I did not see an unsupported merge or an overclaim that would warrant rejection." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -4925,8 +5139,8 @@ "issue:43531|issue:43645", "issue:41720|issue:45704", "issue:45704|issue:45859", - "issue:42886|issue:44843", "issue:43782|issue:45381", + "issue:42886|issue:44843", "issue:46032|issue:46463", "issue:45663|issue:45923", "issue:43421|issue:43653", @@ -4952,241 +5166,238 @@ "issue:43531|issue:43643", "issue:43653|issue:43819" ], - "prepared_review_unit_hash": "9f1b0bb53c9523fb20c1d62d54edb221b93ce968deca303c7c5ca50c875bc16a", + "prepared_review_unit_hash": "afb4e08a5dc54c129a989ed7cd89b7f552ffccccaee638e66ce93f240434bf88", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37993, - "estimated_input_tokens": 9499, - "estimated_eval_tokens": 19254 + "serialized_chars": 37990, + "estimated_input_tokens": 9498, + "estimated_eval_tokens": 19252 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly false positives across unrelated bugs. I kept issue:45081 as the strongest anchor for the tokenizer-regex cluster, but I do not merge it with issue:45356: the overlap is too broad and the evidence only supports a related regression family, not the same underlying bug.", - "confidence": 0.71, - "canonical_issue_reason": "issue:45081 is the most concrete and actionable report in the set, tied to a specific `_patch_mistral_regex` crash path with a clear repro.", + "summary": "These issues are broadly heterogeneous: a few tokenizer/offline-loading reports are semantically adjacent, but most pairs are clearly different bugs across models, modalities, and distributed-training paths. I did not find any pair that is safe to merge as the same concrete fix.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43421 is the best representative of the tokenizer-special-token subtheme: it describes a concrete backend/post-processor update bug and is the most central issue among the tokenizer-related reports in this set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45081 is the best issue candidate because it names the exact failing helper and a specific tokenizer-loading crash, making it the clearest anchor for triage.", + "best_issue_reason": "issue:43421 is the strongest standalone issue in this group because it has a clear repro, a specific code-path target, and the broadest overlap with the few tokenizer-adjacent candidates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Different bugs: eval-mode gradient-checkpointing restrictions vs. abnormal gradient growth during ZeRO-3 training." + "reason": "Gradient-checkpointing eval-mode behavior and ZeRO-3 gradient growth are different code paths and different failure modes." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Both mention token IDs/special tokens, but one is runtime post-processor syncing and the other is save/load config loss." + "reason": "Both involve token handling, but one is a runtime post-processor update bug and the other is save/load corruption of custom token IDs." }, { "left": "issue:45704", "right": "issue:45923", "accept": false, - "reason": "Unrelated: T5 RMSNorm memory leak vs. missing gradient-checkpointing metadata on a trust_remote_code model." + "reason": "Apex RMSNorm memory leak and missing supports_gradient_checkpointing are unrelated model/runtime issues." }, { "left": "issue:43746", "right": "issue:46050", "accept": false, - "reason": "Both are load/save problems, but the concrete failures differ: PEFT local checkpoint loading vs. quantized sinq persistence." + "reason": "Both are loading/saving problems, but PEFT local-checkpoint loading and sinq quantized save/load breakage are distinct mechanisms." }, { "left": "issue:43531", "right": "issue:43645", "accept": false, - "reason": "Different scopes and code paths: Qwen3-MoE sliding_window behavior vs. Jupyter custom-model initialization regression." + "reason": "Qwen3-MoE sliding_window and Jupyter custom-model initialization are unrelated regressions." }, { "left": "issue:41720", "right": "issue:45704", "accept": false, - "reason": "Different failures in unrelated areas: Qwen3 device-mapping crash vs. T5 norm leak." + "reason": "Auto device-mapping cuda assert and T5 apex RMSNorm leak are unrelated." }, { "left": "issue:45704", "right": "issue:45859", "accept": false, - "reason": "Different model/path: T5 RMSNorm leak vs. Qwen3 MoE rotary-embedding CPU-offload incompatibility." + "reason": "Different problems: Apex RMSNorm leak versus CPU-offload incompatibility in Qwen3_5MoeTextRotaryEmbedding." }, { - "left": "issue:42886", - "right": "issue:44843", + "left": "issue:43782", + "right": "issue:45381", "accept": false, - "reason": "Both hit offline/cache-related behavior, but one is cache loading and the other is a Mistral-regex patch calling `model_info()`." + "reason": "Different Qwen VL failures: weight_only loading error versus incorrect video vision_position_ids." }, { - "left": "issue:43782", - "right": "issue:45381", + "left": "issue:42886", + "right": "issue:44843", "accept": false, - "reason": "Different model families and symptoms: Qwen3VL weight-only load error vs. qwen2.5-vl video position-id mismatch." + "reason": "Both are offline/tokenizer-loading themed, but 44843 is a specific Mistral-regex model_info bug while 42886 is a broader cache/offline failure; not enough to treat as the same fix." }, { "left": "issue:46032", "right": "issue:46463", "accept": false, - "reason": "Different bug classes: incorrect cache behavior in Mamba2Mixer vs. an error-message validation issue in T5." + "reason": "Mamba2 use_cache correctness and T5 decoder-input error messaging are unrelated." }, { "left": "issue:45663", "right": "issue:45923", "accept": false, - "reason": "Unrelated: Gemma-4 FSDP2 shared-state rebuild bug vs. missing gradient-checkpointing flag." + "reason": "Gemma4 FSDP2 shared_kv_states and missing gradient-checkpointing support are different issues." }, { "left": "issue:43421", "right": "issue:43653", "accept": false, - "reason": "Both are tokenizer-special-token issues, but the concrete failures are different: runtime post-processor updates vs. BigBird mask-token registration." + "reason": "Both touch tokenizer special-token handling, but one is runtime post-processor refresh and the other is BigBird mask-token registration; different code paths." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "Different models and failure modes: SigLIP2 pipeline output issue vs. CamemBERT masked-LM regression." + "reason": "SigLIP2 model/processor mismatch and add-new-model-like mapping failure are unrelated." }, { "left": "issue:39692", "right": "issue:43873", "accept": false, - "reason": "One is a documentation/example bug plus quantization failure; the other is a broader quantization/offloading behavior report." + "reason": "SigLIP2 documentation/model mismatch and quantization-offloading behavior are not the same bug." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Different loading scenarios: PEFT checkpoint loading vs. quantization/offloading behavior." + "reason": "Local checkpoint loading with PEFT and quantization offloading are different persistence/offload issues." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Both touch tokenizer handling, but the concrete failures are different: post-processor updates vs. LayoutLMv2 NER/padding crashes." + "reason": "Both involve tokenizers, but the reported crashes are different: post-processor updates versus NER/padding/truncation failures." }, { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "Unrelated index/initialization bugs in different components." + "reason": "ZeRO-3/BertModel indexing and fbank feature extraction are unrelated." }, { "left": "issue:45081", "right": "issue:45356", "accept": false, - "reason": "Related Mistral/Kimi tokenizer-regression family, but the reports describe different concrete failures and the evidence is not strong enough to call them the same duplicate." + "reason": "Both are tokenizer-regex related, but one is an AttributeError in fix_mistral_regex while the other is a Kimi codec regression with a warning issue; distinct fixes." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Completely different domains: image segmentation processing vs. a transformers import regression." + "reason": "ImageProcessor void segmentation maps and GenerationMixin import failure are unrelated." }, { "left": "issue:45356", "right": "issue:45907", "accept": false, - "reason": "Different code paths: Mistral/Kimi tokenizer regex handling vs. hub revision forwarding in `list_repo_files()`." + "reason": "Kimi tokenizer regression and list_repo_files revision forwarding are unrelated." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "Different BitNet accelerate-loading issue vs. DAC latent reconstruction mismatch." + "reason": "BitNet packed-weight loading and DAC from_latents/STE mismatch are different model bugs." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "Tokenizer special-token registration vs. BitNet packed-weight unloading; no shared underlying bug." + "reason": "BigBird special-token registration and BitNet accelerate-loading unpacking are unrelated." }, { "left": "issue:43540", "right": "issue:46294", "accept": false, - "reason": "Both involve multimodal/video models, but the concrete issues are different: Qwen3OmniMoe video processing vs. Qwen3.5 video chat-template token/feature mismatch." + "reason": "Both are video-input issues, but they hit different stages: Qwen3OmniMoe video processing versus apply_chat_template token/feature mismatch in Qwen3.5-9B." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Different problems in unrelated code paths: fbank feature extraction vs. add-new-model-like mapping failure." + "reason": "fbank feature extraction and TOKENIZER_MAPPING_NAMES add-new-model-like behavior are unrelated." }, { "left": "issue:43295", "right": "issue:45834", "accept": false, - "reason": "Both are model-loading/input issues, but one is a processor.tokenizer regression and the other is a long-OCR index error in Kosmos2.5." + "reason": "Custom model processor/tokenizer regression and Kosmos2.5 long-OCR index error are different failures." }, { "left": "issue:43618", "right": "issue:43697", "accept": false, - "reason": "Different regressions: missing CLIP attentions assignment vs. changed RTDetrV2 outputs." + "reason": "CLIPOutput attentions regression and RTDetrV2 output drift are different model behavior changes." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "Different inference issues: Bamba torch.compile/SDPA bug vs. dtype mismatches in other models." + "reason": "Torch.compile/SDPA failure in Bamba and dtype mismatch in SwitchTransformers/TimmWrapperModel are distinct inference bugs." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "Different tokenizer bugs: AddedToken special-argument conflict vs. model-loading failure for cjvt/sleng-bert." + "reason": "AddedToken special-argument conflict and sleng-bert loading failure are unrelated." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Unrelated model/runtime bugs with different failure modes and code paths." + "reason": "Different model-path problems: Bamba compile/SDPA versus DAC latent reconstruction mismatch." }, { "left": "issue:43540", "right": "issue:44008", "accept": false, - "reason": "Different multimodal models and failures: video-input processing vs. Gemma3n variable-name collision." + "reason": "Qwen3OmniMoe video processing and Gemma 3n audio_mel_mask variable collision are unrelated multimodal bugs." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "Different model/input issues: LayoutLMv2 tokenizer crashes vs. Perceiver non-default-resolution image classification failure." + "reason": "LayoutLMv2 tokenizer padding/NER crash and Perceiver non-default-resolution failure are unrelated." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Different topics: Qwen3-MoE sliding-window behavior vs. trust_remote_code config field loss." + "reason": "Qwen3-MoE sliding_window and trust_remote_code AutoConfig field loss are unrelated." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "Different bugs in different components: BigBird tokenizer special-token handling vs. DAC latent conversion mismatch." + "reason": "BigBird mask-token registration and DAC STE mismatch are unrelated issues." } ] }, - "evaluator_result": { - "accept": false, - "feedback": "The summary is mostly grounded, but it is not conservative enough on the Mistral tokenizer pair. The packet marks issue:45081 vs. issue:45356 as deterministic_accept=false, yet the analyst set that soft-edge verdict to accept=true and described them as the same underlying regression family. That overstates the evidence. The rest of the pairwise rejections look cautious, but this one mismatch is enough to reject." - }, - "evaluator_used": true, - "retried": true, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -5274,190 +5485,190 @@ "issue:44485|issue:44509", "issue:43606|issue:45072", "issue:34634|issue:36010", - "issue:44479|issue:45834", - "issue:43540|issue:45381", "issue:44805|issue:44936", + "issue:44479|issue:45834", "issue:43577|issue:43818", + "issue:43540|issue:45381", "issue:43782|issue:44977" ], - "prepared_review_unit_hash": "39b649ad31cd1a2f0b652ec4eaef13cfe14e86dfaf0256c0f9130a9fba8c23d6", + "prepared_review_unit_hash": "d1d30f8c6d01bd57472f16670edf8da2d88d9459dcfe9ccfb876d76ddbf3eb6e", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 25, - "serialized_chars": 35554, - "estimated_input_tokens": 8889, - "estimated_eval_tokens": 18034 + "serialized_chars": 35561, + "estimated_input_tokens": 8891, + "estimated_eval_tokens": 18038 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "All soft edges look like superficially related but distinct bugs; none appear to be the same underlying issue or mergeable into one fix. No PRs are present.", + "summary": "The cluster is not a true duplicate set: the soft pairs are mostly superficial topical matches across unrelated model, tokenizer, dtype, config, and training bugs. None of the candidate pairs look mergeable into a single fix.", "confidence": 0.96, - "canonical_issue_reason": "No single canonical issue is warranted here: the items span unrelated subsystems and the soft links do not indicate true duplicates.", + "canonical_issue_reason": "No canonical issue: the items span many distinct regressions and the pairwise similarities are only broad topic overlap, not the same underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "No clear global best issue for deduplication; issue 44843 is the most concrete standalone report, but it only covers its own Mistral/offline regression.", + "best_issue_reason": "No single issue cleanly represents the set; the cluster is too heterogeneous to justify a representative duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "Different models and failure modes: sparse-layer config bug vs RoPE layer-drop complaint." + "reason": "Different bugs: sparse-layer creation in SwitchTransformers config vs Smollm3 RoPE-layer count mismatch." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "Both involve config loading, but one is missing fields with trust_remote_code and the other is strict-config rejection for granite_speech." + "reason": "Both involve config loading, but one is missing fields from trust_remote_code and the other is strict config rejection for granite_speech." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Completely unrelated subsystems: image processor segmentation maps vs trainer compute_loss signature mismatch." + "reason": "Unrelated subsystems: image processor segmentation-map handling vs CustomTrainer compute_loss signature change." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Same Mistral regex patch area, but different concrete bugs: offline model_info call vs backend_tokenizer AttributeError." + "reason": "Both touch Mistral tokenizer patching, but one is an offline model_info call and the other is a backend_tokenizer AttributeError." }, { "left": "issue:41720", "right": "issue:44991", "accept": false, - "reason": "Different components and symptoms: Qwen3 device-mapping CUDA assert vs tokenizer loading failure for est-roberta." + "reason": "Different areas entirely: Qwen3 device-mapping CUDA assert vs tokenizer loading for EMBEDDIA/est-roberta." }, { "left": "issue:43322", "right": "issue:45685", "accept": false, - "reason": "Segfault in Llava loading is unrelated to MPS histogram/int errors in moe code." + "reason": "Segfault when loading Llava Next is unrelated to the MPS histogram/int error in moe." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "AutoImageProcessor URL loading issue is unrelated to the template compilation TypeError." + "reason": "AutoImageProcessor URL loading and template-node compilation are separate failures with no shared code path." }, { "left": "issue:45183", "right": "issue:45479", "accept": false, - "reason": "Validation error-message bug vs degenerate zero-loss classification bug; different code paths and fixes." + "reason": "XOR validation for input_ids/inputs_embeds is unrelated to the single-label-classification zero-loss bug." }, { "left": "issue:43388", "right": "issue:46032", "accept": false, - "reason": "Metric gathering on tuple labels is unrelated to Mamba2 use_cache corruption." + "reason": "Different code paths: gather_for_metrics last-batch label trimming vs Mamba2Mixer use_cache incorrect results." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Dtype mismatch between preprocessor/model is unrelated to Gemma3n's variable-name collision crash." + "reason": "Both are model-specific regressions, but dtype mismatch in Mistral-3 is unrelated to Gemma3n's variable-name collision." }, { "left": "issue:39692", "right": "issue:45834", "accept": false, - "reason": "Docs/model mismatch and quantization example errors are unrelated to Kosmos2.5 long-OCR index errors." + "reason": "Docs/example errors for SigLIP2 do not match Kosmos2.5 long-input OCR index failures." }, { "left": "issue:43950", "right": "issue:45987", "accept": false, - "reason": "Both are dtype/contract-adjacent, but one corrupts non-persistent buffers and the other returns the wrong shape from get_seq_length()." + "reason": "Both involve state/buffer semantics, but one is a from_pretrained non-persistent-buffer regression and the other is a StaticCache return-type contract bug." }, { "left": "issue:44186", "right": "issue:44442", "accept": false, - "reason": "Tokenizer crash on NER/padding is unrelated to AutoTokenizer failing for a different tokenizer class." + "reason": "Tokenizer crashes in LayoutLMv2 and AutoTokenizer loading FastSpeech2ConformerTokenizer are distinct bugs." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Both are output regressions in v5, but they affect different models with no shared code-path evidence." + "reason": "Both are output regressions in v5, but Pegasus output drift and CamemBERT masked-LM prediction errors are different model-specific issues." }, { "left": "issue:45584", "right": "issue:45834", "accept": false, - "reason": "Whisper empty-transcription failure is unrelated to Kosmos2.5 OCR long-input indexing." + "reason": "Whisper empty-transcription generation failure is unrelated to Kosmos2.5 long OCR input indexing." }, { "left": "issue:43644", "right": "issue:43749", "accept": false, - "reason": "Non-persistent buffer corruption is a different regression from FSDP CPU RAM efficient loading." + "reason": "Non-persistent-buffer corruption in from_pretrained is unrelated to FSDP_CPU_RAM_EFFICIENT_LOADING being broken." }, { "left": "issue:43582", "right": "issue:46512", "accept": false, - "reason": "Apple Silicon warmup TypeError is unrelated to stale config.dtype after from_config(dtype=...)." + "reason": "Apple Silicon caching_allocator_warmup TypeError and stale model.config.dtype after from_config are different bugs." }, { "left": "issue:44485", "right": "issue:44509", "accept": false, - "reason": "RoPE implementation discussion is unrelated to docs still mentioning removed pipeline tasks." + "reason": "RoPE implementation discussion is not the same as docs mentioning removed pipeline tasks." }, { "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "CPU offload device mismatch in Bark is unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "CPU-offload device mismatch for bark-small is unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." }, { "left": "issue:34634", "right": "issue:36010", "accept": false, - "reason": "BarkProcessor voice_preset bug and GenerationMixin import error are unrelated." + "reason": "BarkProcessor voice_preset and GenerationMixin import failure are unrelated." }, { - "left": "issue:44479", - "right": "issue:45834", + "left": "issue:44805", + "right": "issue:44936", "accept": false, - "reason": "Different models and failures: Qwen video-input regression vs Kosmos2.5 long OCR indexing." + "reason": "Mask/shape index errors are not the same as trainer.evaluate() failing after train()." }, { - "left": "issue:43540", - "right": "issue:45381", + "left": "issue:44479", + "right": "issue:45834", "accept": false, - "reason": "Both are video-input issues, but they hit different models and different bugs (processing ValueError vs wrong vision_position_ids)." + "reason": "Qwen video-input regression and Kosmos2.5 OCR long-input index error are different model paths." }, { - "left": "issue:44805", - "right": "issue:44936", + "left": "issue:43577", + "right": "issue:43818", "accept": false, - "reason": "Mask-shape IndexError is unrelated to trainer.evaluate failing after train()." + "reason": "BLIP2 dtype staying float32 is unrelated to Video-LLaVA's missing temporal attention/shared weights." }, { - "left": "issue:43577", - "right": "issue:43818", + "left": "issue:43540", + "right": "issue:45381", "accept": false, - "reason": "BLIP2 dtype propagation bug and Video-LLaVA temporal-attention/weight-sharing issue are different problems." + "reason": "Both mention vision/video inputs, but Qwen3OmniMoe video ValueError and qwen2.5-vl vision_position_ids mismatch are not the same bug." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Qwen3VL weight_only loading error is unrelated to Qwen3.5 flash-attention generation failure." + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation failure are separate load-time and runtime bugs." } ] }, @@ -5505,6 +5716,7 @@ "issue:44291", "issue:44368", "issue:44561", + "issue:44589", "issue:44610", "issue:44671", "issue:44811", @@ -5513,11 +5725,11 @@ "issue:44991", "issue:45003", "issue:45020", + "issue:45042", "issue:45072", "issue:45081", "issue:45084", "issue:45325", - "issue:45372", "issue:45406", "issue:45538", "issue:45701", @@ -5526,7 +5738,6 @@ "issue:46285", "issue:46291", "issue:46294", - "issue:46326", "issue:46463" ], "soft_pairs": [ @@ -5537,9 +5748,9 @@ "issue:44561|issue:45020", "issue:44610|issue:45072", "issue:43720|issue:45072", - "issue:45987|issue:46192", "issue:43577|issue:43582", "issue:28218|issue:29159", + "issue:45987|issue:46192", "issue:43825|issue:43827", "issue:43749|issue:44291", "issue:44671|issue:45701", @@ -5555,150 +5766,150 @@ "issue:43742|issue:45020", "issue:44060|issue:44368", "issue:45084|issue:45538", - "issue:45372|issue:46326" + "issue:44589|issue:45042" ], - "prepared_review_unit_hash": "fbf557b422cdc89dc47c7018ef52a32b0f3f7c0eec37a58ab20342fc145f91de", + "prepared_review_unit_hash": "d0747a64de03b9ddf0042a7f65d164ac778df1301b16c357947be2f31e744796", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 26, - "serialized_chars": 35325, - "estimated_input_tokens": 8832, - "estimated_eval_tokens": 17920 + "serialized_chars": 35256, + "estimated_input_tokens": 8814, + "estimated_eval_tokens": 17884 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose mix of unrelated regression reports across tokenization, processors, loading, dtype, and environment-specific failures. I did not find any soft pair that looks like the same underlying bug or change, so I rejected all duplicate candidates.", - "confidence": 0.93, - "canonical_issue_reason": "No single issue cleanly represents this cluster: the items span many different code paths and failure modes rather than one duplicated bug.", + "summary": "Mostly a heterogeneous bucket of unrelated issue reports (tokenization, model loading, processor behavior, dtype, CI/backends). None of the soft pairs look like true duplicates or mergeable same-bug reports, so this is not a strong duplicate cluster.", + "confidence": 0.41, + "canonical_issue_reason": "issue:45701 is the broadest umbrella-style report in the set and the closest thing to a cluster anchor, but it still does not actually duplicate the other issues.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45701 is the broadest tokenizer-regression report in the set, but it is still too generic to be a strong canonical duplicate target for this mixed cluster.", + "best_issue_reason": "issue:45701 is the best representative issue because it is the most generic version-regression/tokenization report and has the broadest potential overlap, though the cluster is too mixed for a true canonical match.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "Both mention pipeline-related behavior, but one is an error-message/feature-support complaint and the other is a model output quality issue; different bugs." + "reason": "One is about a misleading pipeline error message for translation support; the other is about a specific model producing bad outputs with AutoModel/pipeline. Different bugs and code paths." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Runtime special-token post-processing updates are not the same as Whisper batch_decode ignoring skip_special_tokens; different tokenizer code paths." + "reason": "Both involve tokenization-related behavior, but one is about runtime special-token/post-processor updates and the other about Whisper batch_decode skipping special tokens. Not the same underlying bug." }, { "left": "issue:44991", "right": "issue:46291", "accept": false, - "reason": "Tokenizer loading failure for a specific model and kernel import/version failure are unrelated." + "reason": "Tokenizer loading failure for one model versus a kernels dependency/version requirement error on importing pipelines. Unrelated failures." }, { "left": "issue:46294", "right": "issue:46463", "accept": false, - "reason": "Video chat-template mismatch and T5 decoder-input validation error are different models and different failure modes." + "reason": "Video chat-template mismatch versus T5 decoder-input validation messaging. Different models and failure modes." }, { "left": "issue:44561", "right": "issue:45020", "accept": false, - "reason": "Both involve remote_code/trust_remote_code breakage, but the concrete causes differ (removed helper vs broader compatibility breakage)." + "reason": "Both mention remote_code breakage, but one is specifically about removing is_torch_fx_available and the other is a broader recent-version regression report. Not concrete enough to be one PR." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "Processor input-size mismatch and dtype mismatch in inference are unrelated." + "reason": "Processor input-size mismatch versus bf16 dtype mismatch. Different subsystems and fixes." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "Packed-weight loading failure in BitNet and bfloat16 dtype mismatch are separate problems." - }, - { - "left": "issue:45987", - "right": "issue:46192", - "accept": false, - "reason": "Cache sequence-length return type bug and text_kwargs/return_tensors handling bug are different API issues." + "reason": "Packed-weight unpacking during loading is unrelated to bfloat16 dtype mismatch in inference." }, { "left": "issue:43577", "right": "issue:43582", "accept": false, - "reason": "BLIP2 dtype propagation and Apple Silicon allocator warmup TypeError are unrelated." + "reason": "Blip2 dtype propagation bug and Apple Silicon allocator TypeError are separate issues." }, { "left": "issue:28218", "right": "issue:29159", "accept": false, - "reason": "Extra space after added token is a specific tokenizer regression; inconsistent slow/fast tokenizer behavior is broader and not the same bug." + "reason": "Tokenization whitespace after added tokens may be one symptom, but the second report is a broader slow/fast tokenizer inconsistency. Not clearly the same defect." + }, + { + "left": "issue:45987", + "right": "issue:46192", + "accept": false, + "reason": "Cache sequence-length return-type bug and return_tensors being ignored are unrelated API issues." }, { "left": "issue:43825", "right": "issue:43827", "accept": false, - "reason": "Error-message wording and documentation references around pipeline removal are related contextually but not the same underlying issue." + "reason": "One is an incorrect runtime error message; the other is stale docs referencing removed pipeline behavior. Same theme, but different artifacts and fixes." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "FSDP CPU RAM efficient loading breakage and init_empty_weights unexpected-argument error are different loading paths." + "reason": "Both are model-loading related, but one is FSDP CPU RAM efficient loading and the other is an unexpected _is_hf_initialized argument with init_empty_weights. Different code paths." }, { "left": "issue:44671", "right": "issue:45701", "accept": false, - "reason": "CamemBERT masked-LM regression and generic tokenization changes are not the same concrete defect." + "reason": "CamemBERT masked-LM regression and a generic tokenization-change report are not the same underlying bug." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "Torch backend install packaging and Mistral regex patch crash are unrelated." + "reason": "Backend installation packaging problem versus a Mistral tokenizer regex crash. No shared code-path." }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "Qwen3VL weight_only loading error and Qwen2.5-VL rope-index scaling bug affect different models and code paths." + "reason": "Different Qwen model families and different bugs: weight_only loading error versus rope-index scaling." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "Apple Silicon allocator TypeError and CUDA AMP crash in LwDetrImageLoss are unrelated." + "reason": "AppleSilicon TypeError in allocator warmup versus CUDA float16 AMP crash. Separate platform-specific problems." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "BarkProcessor voice_preset bug and CustomTrainer.compute_loss signature break are unrelated." + "reason": "BarkProcessor voice_preset failure and CustomTrainer.compute_loss signature incompatibility are unrelated." }, { "left": "issue:45003", "right": "issue:46285", "accept": false, - "reason": "Unsafe sys.modules access and CLIPTextModel structure inconsistency are different issues in different subsystems." + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to CLIP text model structure inconsistency." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sampling-rate default bug and DAC.from_latents forward-pass mismatch are unrelated." + "reason": "Audio processor chat-template sampling-rate default versus DAC.from_latents/STE mismatch. Different components and fixes." }, { "left": "issue:43653", @@ -5710,37 +5921,37 @@ "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "BeitImageProcessorFast reduce_labels bug and RTDetrV2 output mismatch are different model/processor regressions." + "reason": "Label-reduction bug in an image processor and RTDetrV2 output differences are different regression classes." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Tokenizer regex patch crash and Gemma4Processor missing _tokenizer crash are different attribute/path failures." + "reason": "Both involve processor/tokenizer internals, but they are model-specific attribute errors in different classes. Not mergeable as one fix." }, { "left": "issue:43742", "right": "issue:45020", "accept": false, - "reason": "MobileLLM loading KeyError and generic remote_code breakage are not the same concrete bug." + "reason": "Specific MobileLLM loading KeyError versus generic remote_code regressions. Too broad to be the same bug." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "Incorrect tied-weights warning across layers and repeated tie_word_embeddings warning during LoRA fine-tuning are not the same defect." + "reason": "Tied-weights warning in Qwen3-Next and LoRA tie_word_embeddings warning in Qwen3.5-27B are distinct configuration issues." }, { "left": "issue:45084", "right": "issue:45538", "accept": false, - "reason": "Template-node compilation TypeError and CLIPTokenizer max-length constant issue are unrelated." + "reason": "Template compilation TypeError and CLIPTokenizer model_max_length default are unrelated." }, { - "left": "issue:45372", - "right": "issue:46326", + "left": "issue:44589", + "right": "issue:45042", "accept": false, - "reason": "Gemma 4 processor import failure from mistral_common and continue_final_message behavior for mistral_common tokenizers are different problems." + "reason": "Float8 storage lookup failure versus torchvision dependency requirement in PIL image processors. Separate issues." } ] }, @@ -5785,7 +5996,6 @@ "issue:43696", "issue:43720", "issue:43746", - "issue:43749", "issue:43819", "issue:43854", "issue:43906", @@ -5794,265 +6004,258 @@ "issue:44387", "issue:44479", "issue:44554", - "issue:44589", "issue:44610", "issue:44617", - "issue:44756", "issue:44811", "issue:44912", "issue:44936", "issue:44977", "issue:44991", "issue:45042", + "issue:45372", "issue:45459", "issue:45600", "issue:45759", "issue:45803", "issue:46209", "issue:46291", + "issue:46326", "issue:46421", "issue:46424" ], "soft_pairs": [ "issue:43906|issue:46209", - "issue:44589|issue:45042", + "issue:45372|issue:46326", "issue:43526|issue:44554", "issue:43329|issue:45600", "issue:43638|issue:44936", "issue:43452|issue:44991", "issue:43550|issue:44912", - "issue:36010|issue:43645", "issue:43819|issue:43927", + "issue:36010|issue:43645", "issue:43257|issue:43329", "issue:42175|issue:46291", "issue:36010|issue:36331", - "issue:43720|issue:44610", - "issue:43746|issue:44811", "issue:43531|issue:45042", + "issue:43746|issue:44811", + "issue:43720|issue:44610", "issue:38175|issue:44610", "issue:44387|issue:44617", "issue:39692|issue:44912", "issue:43854|issue:44977", - "issue:45459|issue:45803", "issue:42490|issue:43257", "issue:46421|issue:46424", + "issue:45459|issue:45803", "issue:43650|issue:43906", "issue:43643|issue:45759", "issue:43262|issue:44811", "issue:43257|issue:43696", "issue:43329|issue:44479", "issue:43329|issue:43450", - "issue:43257|issue:43931", - "issue:43749|issue:44756" + "issue:43257|issue:43931" ], - "prepared_review_unit_hash": "49c9de98f0c9e4800276ba9578bd0b2ee3d49ab603dd5c06c8f4ddfa7d3e4f0d", + "prepared_review_unit_hash": "1c7a3662f4e9f36dab0c51b84573aca9e4ff6d506120c2cae4ef44f3ff258bfa", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36789, - "estimated_input_tokens": 9198, - "estimated_eval_tokens": 18652 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36053, + "estimated_input_tokens": 9014, + "estimated_eval_tokens": 18284 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The soft pairs are heterogeneous and do not describe the same underlying bug/change. They span different models, APIs, and failure modes, so all candidate duplicate links should be rejected.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "This is not a duplicate cluster: the items span unrelated bugs across imports, training, quantization, cache behavior, multimodal/video processing, and packaging. None of the soft-edge pairs look like the same underlying change or defect, so all should be rejected.", + "confidence": 0.74, + "canonical_issue_reason": "No strong canonical issue exists because the set is heterogeneous. If an anchor is required, issue 43329 is the broadest concrete code-path bug here (_get_num_multimodal_tokens video branch with undefined names) and is the most actionable representative.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43749 is the strongest standalone report here: it has the highest discussion/inbound-reference activity and a clearly described, high-impact loading regression.", + "best_issue_reason": "Issue 43329 is the best single issue to anchor on because it describes a specific failing branch with a clear fix surface and broader relevance across multiple VLM processors.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43906", "right": "issue:46209", "accept": false, - "reason": "One is an isolated reproduction of issue 38071; the other is an FP8 inference failure on B200/SM100. Different bugs and code paths." + "reason": "A reproduction of a prior issue is not the same bug as an FP8 inference OOM on B200; different symptoms and code paths." }, { - "left": "issue:44589", - "right": "issue:45042", + "left": "issue:45372", + "right": "issue:46326", "accept": false, - "reason": "Float8 storage lookup error versus PIL backend image processors incorrectly requiring torchvision. Unrelated failures." + "reason": "ImportError in Gemma 4 processor loading from mistral_common is unrelated to continue_final_message behavior in tokenizer chat handling." }, { "left": "issue:43526", "right": "issue:44554", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug versus an MPS attention correctness issue. Different subsystems." + "reason": "BEIT label reduction bug and MPS attention correctness issue are unrelated subsystems and failure modes." }, { "left": "issue:43329", "right": "issue:45600", "accept": false, - "reason": "Undefined variables in multimodal video token counting versus stale auto_mappings references to removed Sam3LiteText configs. Not the same bug." + "reason": "Multimodal video token counting bug vs stale auto_mappings config references; different files and defects." }, { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "DeepSpeed ZeRO3 index error in non-pretrained BERT training versus trainer.evaluate() failing after train(). Different regressions." + "reason": "Zero3/non-pretrained Bert index error is a training initialization bug, not the same as evaluate() failing after train()." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "gguf_file handling breaks from_pretrained loaders, while the tokenizer issue is specific to EMBEDDIA/est-roberta. Different load paths and causes." + "reason": "gguf_file loading breakage and tokenizer load failure for EMBEDDIA/est-roberta are separate model-loading problems." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "Bamba torch.compile/SDPA failure versus MXFP4 quantization fallback on git-oss-20b. Different models and failures." + "reason": "torch.compile+SDPA failure in Bamba and MXFP4 quantization fallback in git-oss are different model/backend bugs." }, { - "left": "issue:36010", - "right": "issue:43645", + "left": "issue:43819", + "right": "issue:43927", "accept": false, - "reason": "GenerationMixin import error versus Jupyter notebook custom-model initialization breakage. Different API breaks." + "reason": "DAC forward/from_latents mismatch and DiaConfig save/load token ID loss are different model families and bug types." }, { - "left": "issue:43819", - "right": "issue:43927", + "left": "issue:36010", + "right": "issue:43645", "accept": false, - "reason": "DAC.from_latents mismatch due to missing STE versus DiaConfig losing custom token IDs on save/load. Different model behaviors." + "reason": "GenerationMixin import error and Jupyter custom-model initialization breakage are both import-adjacent but not the same defect." }, { "left": "issue:43257", "right": "issue:43329", "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed versus undefined helper variables in multimodal video token counting. Not the same issue." + "reason": "Qwen3 MoE weight conversion/loading with accelerate+deepspeed is unrelated to video token counting undefined names." }, { "left": "issue:42175", "right": "issue:46291", "accept": false, - "reason": "Missing tensorflow in the torch extra versus pipelines import failing on kernels v0.15 version/revision requirements. Unrelated dependency issues." + "reason": "Missing tensorflow in the torch extra is packaging-related; kernels v0.15 import failure is a separate dependency/API issue." }, { "left": "issue:36010", "right": "issue:36331", "accept": false, - "reason": "Importing GenerationMixin versus CustomTrainer.compute_loss() receiving an unexpected kwarg. Different API surface and symptom." + "reason": "GenerationMixin import failure and unexpected num_items_in_batch in CustomTrainer.compute_loss are unrelated bugs." }, { - "left": "issue:43720", - "right": "issue:44610", + "left": "issue:43531", + "right": "issue:45042", "accept": false, - "reason": "Packed BitNet weights not unpacked during accelerate loading versus OmDet-Turbo processor/model input-size mismatch. Different bugs." + "reason": "Qwen3-MoE sliding_window bug and PIL backend torchvision dependency regression are different subsystems." }, { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "GraniteSpeech local PEFT checkpoint loading versus Whisper batch_decode skip_special_tokens being ignored. Different processors and code paths." + "reason": "GraniteSpeech PEFT local checkpoint loading and Whisper batch_decode skip_special_tokens handling do not share a concrete code path." }, { - "left": "issue:43531", - "right": "issue:45042", + "left": "issue:43720", + "right": "issue:44610", "accept": false, - "reason": "Qwen3-MoE sliding_window bug versus PIL backend requiring torchvision. No meaningful overlap." + "reason": "BitNet packed-weight unpacking during accelerate loading and OmDet input-size mismatch are unrelated issues." }, { "left": "issue:38175", "right": "issue:44610", "accept": false, - "reason": "Unexpected zero probabilities in SigLIP2 versus OmDet-Turbo input resolution mismatch. Different vision issues." + "reason": "Unexpected zero probabilities in SigLIP2 is not the same bug as OmDet processor producing 640px inputs." }, { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "Int4 quantization reserved-memory/OOM regression versus Sam3Video OOM. Both memory-related, but not the same underlying cause." + "reason": "Both are OOM-related, but one is an int4 reserved-memory regression and the other is a Sam3Video memory exhaustion report." }, { "left": "issue:39692", "right": "issue:44912", "accept": false, - "reason": "SigLIP2 documentation example errors versus git-oss-20b MXFP4 loading fallback. Docs/example bugs are unrelated to quantization loading." + "reason": "Documentation/example errors for SigLIP2 are not the same as MXFP4 quantization loading fallback for git-oss." }, { "left": "issue:43854", "right": "issue:44977", "accept": false, - "reason": "GLM-4.7-Flash unit-test loading failure versus Qwen3.5 flash-attention generation bug. Different models and failure modes." - }, - { - "left": "issue:45459", - "right": "issue:45803", - "accept": false, - "reason": "Exception handling around protobuf decode versus a bare except in FuyuBatchFeature.convert_to_tensors(). Similar theme, different code and symptom." + "reason": "GLM-4.7 Flash load failure in tests and Qwen3.5 flash-attention generation bug are different models and code paths." }, { "left": "issue:42490", "right": "issue:43257", "accept": false, - "reason": "Local path trailing-slash import warning versus Qwen3 MoE accelerate/deepspeed conversion problem. Different loading issues." + "reason": "Trailing-slash local-model import warning and Qwen3 MoE conversion/loading failure are distinct loading edge cases." }, { "left": "issue:46421", "right": "issue:46424", "accept": false, - "reason": "Chunked-prefill recompilation from cache lazy init versus max_cache_length only honored for Quantized Cache. Related area, but not the same bug." + "reason": "Both concern cache behavior, but lazy-init recompilation with chunked prefill is not the same as max_cache_length being ignored except for Quantized Cache." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Hidden tokenizer decode errors when protobuf is absent and bare-except swallowing interrupts/errors are different exception-handling bugs." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "A nonspecific request ('ADD THE DATA') versus an isolated reproduction of issue 38071. No duplicate relationship." + "reason": "A nonspecific 'ADD THE DATA' issue is unrelated to the isolated reproduction of issue 38071." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "AutoConfig.from_pretrained with trust_remote_code dropping fields versus AutoModelForCausalLM.from_config not unwrapping text_config. Similar domain, different API behavior." + "reason": "Missing fields from trust_remote_code AutoConfig and unwrapped text_config in composite multimodal configs are separate config-loading bugs." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Audio apply_chat_template defaulting to 16kHz versus Whisper batch_decode skip_special_tokens handling. Different processor methods." + "reason": "Audio apply_chat_template sample-rate defaulting and Whisper batch_decode skip_special_tokens are unrelated processor behaviors." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Qwen3 MoE weight conversion failure versus GPT-oss-20b CUDA OOM. Completely different issues." + "reason": "Qwen3 MoE loading/conversion bug and GPT-oss-20b CUDA OOM are unrelated model/runtime failures." }, { "left": "issue:43329", "right": "issue:44479", "accept": false, - "reason": "Undefined video-patch helper variables in token counting versus a broader video-input regression affecting multiple Qwen VL/MoE models. Different concrete bugs." + "reason": "Both mention video, but one is undefined helpers in token counting and the other is a broader v5.3.0 video input regression; not the same concrete bug." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Undefined helpers in multimodal token counting versus batched video processor shape errors. Related to video, but not the same failure." + "reason": "Undefined video token-count helpers and batched video processor shape errors are different failures in the video stack." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Qwen3 MoE loading/conversion issue versus Qwen3-VL-30B-A3B weight-shape mismatch. Same family name, different model and bug." - }, - { - "left": "issue:43749", - "right": "issue:44756", - "accept": false, - "reason": "FSDP CPU RAM efficient loading regression versus mmap disablement to avoid OOM on Strix Halo. Different loading/memory problems." + "reason": "Qwen3 MoE accelerate/deepspeed conversion issues and Qwen3-VL weight-shape mismatches are separate model-loading problems." } ] }, @@ -6106,6 +6309,7 @@ "issue:44479", "issue:44560", "issue:44568", + "issue:44756", "issue:44821", "issue:44849", "issue:44912", @@ -6121,22 +6325,22 @@ "issue:45538", "issue:45701", "issue:45859", - "issue:45907", "issue:46291", "issue:46512" ], "soft_pairs": [ + "issue:43749|issue:44756", "issue:43454|issue:43577", "issue:42915|issue:43742", - "issue:43606|issue:45859", "issue:44291|issue:45042", + "issue:43606|issue:45859", "issue:34567|issue:40990", "issue:42915|issue:44821", "issue:43525|issue:46291", "issue:29159|issue:34567", "issue:44568|issue:45356", - "issue:43404|issue:45127", "issue:44987|issue:45538", + "issue:43404|issue:45127", "issue:43927|issue:44912", "issue:43697|issue:45357", "issue:42915|issue:45538", @@ -6144,181 +6348,186 @@ "issue:44849|issue:45701", "issue:42175|issue:43824", "issue:43577|issue:46512", - "issue:43317|issue:45538", "issue:28282|issue:34689", - "issue:43452|issue:43582", + "issue:43317|issue:45538", "issue:45137|issue:45405", + "issue:43452|issue:43582", "issue:33357|issue:34567", "issue:44062|issue:44991", "issue:43742|issue:44291", "issue:41720|issue:43540", + "issue:28282|issue:29127", "issue:43540|issue:44560", "issue:44479|issue:45381", - "issue:28282|issue:29127", "issue:43540|issue:44479", - "issue:42915|issue:43749", - "issue:44987|issue:45907" + "issue:42915|issue:43749" ], - "prepared_review_unit_hash": "30bdfee10f742ad8233e0959a7aa83b4f25d5c035e335b55bce84077c8f2fa2f", + "prepared_review_unit_hash": "e36f6e7da34d1029457985b4619dffc878a5daaa402ebdb98b5c3e9fb6a418a7", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37358, - "estimated_input_tokens": 9340, - "estimated_eval_tokens": 18936 + "serialized_chars": 37312, + "estimated_input_tokens": 9328, + "estimated_eval_tokens": 18912 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a heterogeneous grab-bag of unrelated tokenizer, model-loading, dtype, video, and distributed-training issues. No pair is a clear duplicate; the only plausible umbrella is a broad tokenizer/version-regression report, but even that is a weak fit.", - "confidence": 0.36, - "canonical_issue_reason": "issue:45701 is the broadest, most umbrella-like report here (version-dependent tokenization changes), so it is the least-bad canonical issue if one must be chosen.", + "summary": "This set is highly heterogeneous: most items are unrelated model-loading, tokenizer, dtype, offload, or backend-specific bugs. None of the soft pairs look like the same underlying change or defect, so all soft edges are rejected. No PRs are present.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 45701 is the broadest, most central-sounding item in the set (generic tokenization regression with multiple inbound refs), but the overall cluster is too mixed for a truly strong canonical issue.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45701 is the most general and still-open issue, making it the best representative anchor for this otherwise fragmented set.", + "best_issue_reason": "Issue 45701 is the best available representative by breadth and linkage activity, though it does not actually unify the heterogeneous bugs in this cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Both relate to memory/offload behavior, 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:43454", "right": "issue:43577", "accept": false, - "reason": "Different bugs: lm_head weight tying in Mistral3 vs dtype propagation in Blip2/Qformer." + "reason": "One is lm_head weight tying in Mistral3, the other is dtype propagation in BLIP2; unrelated bugs." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Different failures in different models: FineGrainedFP8Config vs a key error loading MobileLLM-125M." + "reason": "Different models and symptoms: FineGrainedFP8Config failure vs a key error loading MobileLLM-125M." }, { - "left": "issue:43606", - "right": "issue:45859", + "left": "issue:44291", + "right": "issue:45042", "accept": false, - "reason": "Both mention CPU offload, but the concrete code paths and models differ (bark-small device mismatch vs Qwen3_5Moe rotary embedding)." + "reason": "Init-empty-weights argument handling is unrelated to PIL backend torchvision requirements." }, { - "left": "issue:44291", - "right": "issue:45042", + "left": "issue:43606", + "right": "issue:45859", "accept": false, - "reason": "Unrelated issues: init_empty_weights/_is_hf_initialized handling vs PIL backend image processors requiring torchvision." + "reason": "Both mention CPU offload, but they involve different models and distinct runtime errors." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "TrainerState token counting is unrelated to the GPT-OSS perplexity regression." + "reason": "TrainerState token counting is unrelated to perplexity on gpt-oss-20b." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "Different subsystems: Qwen3Moe FP8 loading vs AutoImageProcessor URL loading." + "reason": "FineGrainedFP8Config loading failure is unrelated to AutoImageProcessor URL loading." }, { "left": "issue:43525", "right": "issue:46291", "accept": false, - "reason": "Unrelated: missing pad_token_id on Llama4Config vs kernels v0.15 revision/version requirement." + "reason": "Missing pad_token_id in Llama4Config is unrelated to kernels version/revision import failure." }, { "left": "issue:29159", "right": "issue:34567", "accept": false, - "reason": "Tokenizer slow/fast inconsistency is unrelated to TrainerState num_input_tokens_seen updates." + "reason": "Tokenizer consistency vs TrainerState token accounting are different subsystems and bugs." }, { "left": "issue:44568", "right": "issue:45356", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different root causes." + "reason": "Both are tokenizer regressions, but they concern different tokenizers and different concrete failures." }, { - "left": "issue:43404", - "right": "issue:45127", + "left": "issue:44987", + "right": "issue:45538", "accept": false, - "reason": "Different bugs: Mistral3 weight tying vs LoRA merge collapse with extended vocabulary." + "reason": "Model-loading failure for a specific model is unrelated to CLIPTokenizer model_max_length behavior." }, { - "left": "issue:44987", - "right": "issue:45538", + "left": "issue:43404", + "right": "issue:45127", "accept": false, - "reason": "Loading a specific model on >=5.1.0 is unrelated to CLIPTokenizer's model_max_length behavior." + "reason": "Both involve tied embeddings/lm_head, but one is a missing tie in a specific model and the other is LoRA merge collapse after vocab extension; not the same bug." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "Different issues: DiaConfig token IDs lost on save/load vs MXFP4 quantization falling back to bf16." + "reason": "Custom token ID persistence on save/load is unrelated to MXFP4 quantization fallback." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "Different model families and symptoms: RTDetrV2 output change vs Qwen3.5 save_pretrained key mismatch." + "reason": "RTDetrV2 output drift and Qwen3.5 visual encoder key saving are unrelated issues." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Qwen3Moe FP8 loading and CLIPTokenizer max_length are unrelated." + "reason": "FP8 model load failure is unrelated to tokenizer max-length settings." }, { - "left": "issue:43425", + "left": "issue:43452", "right": "issue:46291", "accept": false, - "reason": "Torch 2.10 incompatibility is unrelated to kernels import/revision handling." + "reason": "gguf_file loading breakage is unrelated to kernels version/revision import requirements." }, { "left": "issue:44849", "right": "issue:45701", "accept": false, - "reason": "Qwen3.5 hidden-state output bug is unrelated to the generic tokenization-change report." + "reason": "Qwen3.5 hidden-state output behavior is unrelated to a broad tokenization-change report." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Backend packaging issue for TensorFlow is unrelated to the Qwen2_5_VL import error." + "reason": "TensorFlow backend packaging and missing Qwen2.5-VL class import are unrelated." }, { "left": "issue:43577", "right": "issue:46512", "accept": false, - "reason": "Both are dtype-related, but one is incorrect module weight dtype and the other is stale config dtype after from_config()." - }, - { - "left": "issue:43317", - "right": "issue:45538", - "accept": false, - "reason": "Device-map/offload loading failure is unrelated to CLIPTokenizer's model_max_length." + "reason": "BLIP2 dtype propagation and config.dtype staleness are related only superficially; different affected code paths and models." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "Missing PyTorch import error is unrelated to Llama 3.2 Vision loading breakage." + "reason": "PyTorch-not-found import error is unrelated to Llama 3.2 vision model loading regression." }, { - "left": "issue:43452", - "right": "issue:43582", + "left": "issue:43317", + "right": "issue:45538", "accept": false, - "reason": "GGUF loading failures are unrelated to the Apple Silicon caching_allocator_warmup TypeError." + "reason": "Device-map/offload loading failure is unrelated to tokenizer max-length behavior." }, { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error is unrelated to the PEFT version bump." + "reason": "DeepSpeed ZeRO3 deque error is unrelated to a PEFT version pin issue." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf loading and Apple Silicon allocator warmup TypeError are unrelated bugs." }, { "left": "issue:33357", @@ -6330,55 +6539,49 @@ "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "AddedToken construction error is not the same bug as loading the EMBEDDIA/est-roberta tokenizer." + "reason": "Different tokenizer errors with different root causes; not the same underlying bug." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "Different failures: MobileLLM key error vs init_empty_weights unexpected argument." + "reason": "Key error loading a model and init_empty_weights argument handling are unrelated." }, { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Different Qwen issues: CUDA assert during auto device mapping vs video-input processing ValueError." + "reason": "CUDA assert under auto device mapping and video-input ValueError in Qwen3OmniMoe are different failures." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "Missing PyTorch import error is unrelated to LayoutLMv3 box-information error messaging." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Both are video-input failures, but they are different models and distinct failure modes (ValueError vs StopIteration)." + "reason": "Both involve video inputs, but the reported errors and likely code paths differ." }, { "left": "issue:44479", "right": "issue:45381", "accept": false, - "reason": "Related theme, but one is a broad multi-model video regression and the other is a specific qwen2.5-vl vision_position_ids bug; not enough to treat as one duplicate." - }, - { - "left": "issue:28282", - "right": "issue:29127", - "accept": false, - "reason": "Missing PyTorch import error is unrelated to LayoutLMv3 box-information validation." + "reason": "Related family and version area, but one is a broad video-input regression across multiple models and the other is a specific position-id bug in qwen2.5-vl." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Both concern video inputs, but they target different models and likely different root causes." + "reason": "Qwen3OmniMoe video-processing failure and the multi-model v5.3.0 video regression are not concrete duplicates." }, { "left": "issue:42915", "right": "issue:43749", "accept": false, - "reason": "Qwen3Moe FP8 loading is unrelated to FSDP CPU RAM efficient loading." - }, - { - "left": "issue:44987", - "right": "issue:45907", - "accept": false, - "reason": "Model loading failure and list_repo_files revision forwarding are unrelated bugs." + "reason": "FineGrainedFP8Config failure is unrelated to FSDP CPU RAM-efficient loading." } ] }, @@ -6432,18 +6635,18 @@ "issue:44821", "issue:44898", "issue:44964", + "issue:44987", "issue:45042", "issue:45072", "issue:45081", "issue:45325", - "issue:45357", "issue:45405", "issue:45479", "issue:45488", "issue:45588", "issue:45600", "issue:45823", - "issue:45920", + "issue:45907", "issue:45987", "issue:46399", "issue:46463", @@ -6451,213 +6654,219 @@ "issue:46491" ], "soft_pairs": [ - "issue:38175|issue:44898", + "issue:44987|issue:45907", "issue:24643|issue:29159", "issue:42175|issue:44488", "issue:44568|issue:45488", + "issue:38175|issue:44898", "issue:44737|issue:46463", "issue:43493|issue:43994", "issue:44291|issue:44821", + "issue:27957|issue:28282", "issue:34567|issue:36331", "issue:43257|issue:43756", - "issue:27957|issue:28282", "issue:43577|issue:45072", "issue:44062|issue:45081", "issue:44792|issue:44964", "issue:43906|issue:46468", "issue:45042|issue:45081", "issue:44568|issue:46491", + "issue:43742|issue:44589", "issue:43526|issue:45479", "issue:44737|issue:45588", - "issue:43742|issue:44589", "issue:43756|issue:46399", - "issue:43404|issue:45987", "issue:43756|issue:45325", + "issue:43404|issue:45987", "issue:45823|issue:46463", "issue:43531|issue:44821", "issue:43452|issue:44488", "issue:45405|issue:45600", - "issue:42796|issue:43906", - "issue:45357|issue:45920" + "issue:42796|issue:43906" ], - "prepared_review_unit_hash": "b8e2a75d58be69766dc31a54d7045a827fb3f76e950a5efd618311c58aa67ce7", + "prepared_review_unit_hash": "d72472684c4a4aaa5fe92e36aea7be14faf117a4066693658298310734650d9b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35922, - "estimated_input_tokens": 8981, - "estimated_eval_tokens": 18218 + "serialized_chars": 35854, + "estimated_input_tokens": 8964, + "estimated_eval_tokens": 18184 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous; only the two SigLIP2 reports look like the same underlying bug. The other soft pairs describe different models, different code paths, or different failure modes and should stay separate.", - "confidence": 0.8, - "canonical_issue_reason": "Issue #43994 is the clearest canonical for the one true duplicate pair (#43493/#43994): both describe the same SigLIP2 regression, and #43994 is the broader user-facing report.", + "summary": "This cluster is mostly heterogeneous: tokenizer regressions, model-loading failures, dtype/device issues, and trainer bugs across unrelated code paths. I only see one plausible duplicate pair in the SigLIP2 reports; the rest should stay separate.", + "confidence": 0.76, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue #45920 is the most generally representative issue in the cluster overall: it is open, recent, and describes a broad tokenizer regression affecting multiple models.", + "best_issue_reason": "issue:43994 is the best single issue to anchor on because it is a concrete, user-facing SigLIP2 regression with a clear repro, and it is the closest match to the only likely duplicate pair in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:38175", - "right": "issue:44898", + "left": "issue:44987", + "right": "issue:45907", "accept": false, - "reason": "Both mention SigLIP-related failures, but one is a zero-probabilities inference bug and the other is a Perceiver image-classification resize/interpolation issue; different code paths." + "reason": "Different problems: model loading failure for physical-intelligence/fast vs list_repo_files not forwarding revision in CI." }, { "left": "issue:24643", "right": "issue:29159", "accept": false, - "reason": "DeepSpeed training shape error vs tokenizer fast/slow inconsistency; unrelated subsystems." + "reason": "Unrelated subsystems: DeepSpeed training weight shape error vs slow/fast tokenizer inconsistency." }, { "left": "issue:42175", "right": "issue:44488", "accept": false, - "reason": "Package/backend dependency complaint vs a model-loading failure for a specific checkpoint; not the same bug." + "reason": "Different bugs: packaging/backend dependency issue vs model load failure for cjvt/sleng-bert." }, { "left": "issue:44568", "right": "issue:45488", "accept": false, - "reason": "Both are tokenizer regressions, but one is missing BOS/EOS insertion for mdeberta-v3-base and the other is a LlamaTokenizer pre-tokenizer override affecting DeepSeek; different failures." + "reason": "Both are tokenizer regressions, but on different models and code paths: mdeberta special-token handling vs DeepSeek/Llama tokenizer pre-tokenizer override." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model failures: SigLIP2 zero probabilities vs Perceiver non-default resolution interpolation failure." }, { "left": "issue:44737", "right": "issue:46463", "accept": false, - "reason": "XLNet CPU allocation bug vs T5 decoder input validation error; no shared code-path." + "reason": "Distinct model/code-path issues: XLNet CPU arange bug vs T5 decoder input validation message." }, { "left": "issue:43493", "right": "issue:43994", "accept": true, - "reason": "Both describe the same SigLIP2 implementation regression producing incorrect outputs; #43493 is the HF-vs-JAX diagnostic and #43994 is the broader user-facing symptom." + "reason": "Same model family and same underlying SigLIP2 wrong-output regression; the JAX-vs-HF discrepancy and nonsensical AutoModel/pipeline outputs plausibly stem from the same implementation bug." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "Unexpected `_is_hf_initialized` argument during empty-weight loading vs AutoImageProcessor URL loading; unrelated." + "reason": "Different loading failures: unexpected _is_hf_initialized argument vs AutoImageProcessor-from-URL behavior." + }, + { + "left": "issue:27957", + "right": "issue:28282", + "accept": false, + "reason": "Different issues: Flash Attention 2 with XLM-Roberta vs missing PyTorch import error." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "TrainerState counter not updating vs CustomTrainer signature break; different trainer internals." + "reason": "Trainer metrics bug vs custom compute_loss signature change; related area but not the same defect." }, { "left": "issue:43257", "right": "issue:43756", "accept": false, - "reason": "Qwen3 MoE weight conversion with accelerate+deepspeed vs Smollm3 RoPE-layer count regression; unrelated model-specific bugs." - }, - { - "left": "issue:27957", - "right": "issue:28282", - "accept": false, - "reason": "Flash Attention 2 support for XLMRoberta vs missing-PyTorch ImportError; not the same issue." + "reason": "Both mention model internals, but one is MoE weight conversion under accelerate/deepspeed and the other is RoPE layer dropping in SmolLM3." }, { "left": "issue:43577", "right": "issue:45072", "accept": false, - "reason": "Both are dtype-related, but they affect different model families and different forward/loading paths." + "reason": "Separate dtype issues in different models and contexts: BLIP2 loading vs SwitchTransformers/TimmWrapperModel bfloat16 inference." }, { "left": "issue:44062", "right": "issue:45081", "accept": false, - "reason": "AddedToken keyword duplication vs `_patch_mistral_regex` crashing on `backend_tokenizer`; distinct tokenizer bugs." + "reason": "Tokenizer bugs, but different failures: AddedToken special-arg duplication vs Mistral regex patch crash." }, { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "Janus image-generation test failure vs Phi-4 multimodal load failure; different checkpoints and failures." + "reason": "Different multimodal failures: Janus test failure vs Phi-4 multimodal load failure." }, { "left": "issue:43906", "right": "issue:46468", "accept": false, - "reason": "A reproduction of an older issue vs a broken Nemotron example; not enough evidence of the same underlying bug." + "reason": "No clear overlap; one is a reproduction of a prior issue, the other is a Nemotron example breakage." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "PIL backend now requiring torchvision vs Mistral regex patch crash; unrelated." + "reason": "Different subsystems: PIL image processor torchvision dependency vs Mistral tokenizer regex patching." }, { "left": "issue:44568", "right": "issue:46491", "accept": false, - "reason": "mdeberta special-token insertion bug vs CodeLlama leading-space round-trip regression; different tokenizer semantics." + "reason": "Both tokenizer regressions, but different models and symptoms: missing BOS/EOS vs CodeLlama round-trip leading-space loss." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "Different load-time failures: MobileLLM key error vs missing Float8 storage object." }, { "left": "issue:43526", "right": "issue:45479", "accept": false, - "reason": "BEiT image-processor label reduction bug vs sequence-classification zero-loss bug; different components." + "reason": "BeitImageProcessorFast label reduction bug is unrelated to sequence-classification zero-loss behavior across multiple models." }, { "left": "issue:44737", "right": "issue:45588", "accept": false, - "reason": "XLNet CPU tensor creation issue vs flash-attention sinkless-model AttributeError; unrelated." + "reason": "XLNet CPU tensor allocation bug vs flash_attention crash on s_aux=None are unrelated." }, { - "left": "issue:43742", - "right": "issue:44589", + "left": "issue:43756", + "right": "issue:46399", "accept": false, - "reason": "MobileLLM key error vs missing Float8 storage object; unrelated load/runtime failures." + "reason": "Different bugs: SmolLM3 RoPE-layer dropping vs Qwen3.5 kernel registration crash." }, { "left": "issue:43756", - "right": "issue:46399", + "right": "issue:45325", "accept": false, - "reason": "Smollm3 RoPE-layer regression vs Qwen3.5 kernel registration error; different code paths." + "reason": "Both touch RoPE-related code, but the model families and defect types differ (layer dropping vs still-image temporal scaling)." }, { "left": "issue:43404", "right": "issue:45987", "accept": false, - "reason": "Mistral3 lm_head tying bug vs StaticCache return-type bug; unrelated." + "reason": "Mistral3 lm_head tying issue is unrelated to StaticCache.get_seq_length returning a tensor." }, { - "left": "issue:45405", - "right": "issue:45600", + "left": "issue:45823", + "right": "issue:46463", "accept": false, - "reason": "Version-pin complaint for peft vs stale auto_mappings references; different maintenance issues." + "reason": "Gemma4 device-map mismatch and T5 decoder input validation are different problems." }, { "left": "issue:42796", "right": "issue:43906", "accept": false, - "reason": "Tokenizer BPE-marker handling bug vs a reproduction of another issue; no shared underlying defect." - }, - { - "left": "issue:45357", - "right": "issue:45920", - "accept": false, - "reason": "Qwen3.5 save_pretrained key mismatch vs broad AutoTokenizer token-id regression across many models; different failure modes." + "reason": "Ministral tokenizer BPE-marker handling is unrelated to the isolated reproduction of a separate prior issue." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet. The representative issue choice (issue:45920) is defensible because it is open, recent, and broad. The canonical issue choice for the only clearly duplicate pair (issue:43493 / issue:43994) is also supported. The soft-edge verdicts are appropriately conservative overall, and I don\u2019t see any overstated evidence or mergeability problems." + "feedback": "The summary is grounded in the packet: it stays conservative about the cluster\u2019s heterogeneity, and the single accepted SigLIP2 pair is a reasonable soft match rather than an overclaim. I don\u2019t see any soft-edge verdicts that cross into unsupported certainty." }, "evaluator_used": true, "retried": false, @@ -6713,12 +6922,13 @@ "issue:45005", "issue:45200", "issue:45216", - "issue:45356", + "issue:45357", "issue:45372", "issue:45375", "issue:45381", "issue:45542", "issue:45823", + "issue:45920", "issue:46050", "issue:46121", "issue:46170", @@ -6727,6 +6937,7 @@ "issue:46491" ], "soft_pairs": [ + "issue:45357|issue:45920", "issue:44913|issue:46121", "issue:44625|issue:44877", "issue:43720|issue:44898", @@ -6745,216 +6956,208 @@ "issue:44936|issue:45823", "issue:44361|issue:44442", "issue:38175|issue:43295", - "issue:45372|issue:46170", "issue:43653|issue:44361", "issue:45542|issue:46178", + "issue:45372|issue:46170", "issue:42898|issue:46491", "issue:44484|issue:44704", "issue:44514|issue:45381", - "issue:43452|issue:43577", - "issue:43295|issue:45356", - "issue:43577|issue:43697" + "issue:43452|issue:43577" ], - "prepared_review_unit_hash": "26575c6a6cb09647d8647fa03a4842ec07c0fa9eb6c3f2371b62291e169cf6c9", + "prepared_review_unit_hash": "e1e2351b45af73adc07994a6ce9c4da5f591d572843b530604752ccd536d2c8d", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 27, - "serialized_chars": 35645, - "estimated_input_tokens": 8912, - "estimated_eval_tokens": 18080 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35930, + "estimated_input_tokens": 8983, + "estimated_eval_tokens": 18222 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous set of unrelated issue reports across tokenization, config, loading, and trainer paths. The closest pair is 44466/45005, but the evidence is only title-level and should be treated as a likely related tied-weights/save_pretrained regression, not a confirmed duplicate.", - "confidence": 0.73, - "canonical_issue_reason": "If this cluster is anchored anywhere, issue 45005 is the best anchor because it is the broadest tied-weights/save_pretrained report; 44466 reads like a narrower device-dependent symptom, but the linkage is only suggestive.", + "summary": "Most items are unrelated single-bug reports across different model families and code paths. The only plausible duplicate pair is the tied-weights/save_pretrained regression in v5 (issues 44466 and 45005), with 45005 as the broader umbrella.", + "confidence": 0.63, + "canonical_issue_reason": "Issue 45005 is the best umbrella representative for the tied-weights regression cluster: it is broader, explicitly about tied weights in v5, and can subsume the narrower device-dependent serialization symptom reported in 44466.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45005 is the strongest overall issue candidate because it describes the broader v5 tied-weights regression and can plausibly subsume 44466 if the pair is treated as related.", + "best_issue_reason": "45005 is the strongest issue candidate because it captures the shared underlying save_pretrained/tied-weights problem at a higher level than the narrower 44466 report.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Different regressions: Qwen3.5 visual-encoder key serialization vs AutoTokenizer returning wrong token IDs across multiple models." + }, { "left": "issue:44913", "right": "issue:46121", "accept": false, - "reason": "Both are config-related, but the bugs differ: GPTNeoX rotary_pct reload vs RoPE validation of a JSON-loaded list." + "reason": "Both touch config/rope-related code, but one is a silent reload defaulting bug and the other is a TypeError on JSON-loaded lists; different root cause." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "Different bugs and code paths: Qwen3.5 num_labels propagation vs strict config blocking granite_speech loading." + "reason": "Both are config-loading issues, but they affect different models and code paths: num_labels propagation vs strict config rejection." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "Unrelated models and failure modes: BitNet packed-weight loading vs Perceiver interpolation/image-classification failure." + "reason": "Unrelated model bugs: BitNet packed-weight loading vs Perceiver interpolation handling." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "Different subsystems and symptoms: RTDetrV2 output divergence vs Qwen3.5 save_pretrained regression." + "reason": "Different failure modes and models: RTDetrV2 output regression vs Qwen3.5 save_pretrained checkpoint corruption." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both involve config fields, but the underlying problems differ: trust_remote_code field loss vs a specific missing field in Qwen3_5MoeVisionConfig." + "reason": "Both involve config fields, but one is a generic AutoConfig trust_remote_code issue and the other is a model-specific strict-field omission." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed training weight-dimension runtime error and a LayoutLMv3 error-message clarity issue are not the same bug." + "reason": "Completely different bugs: DeepSpeed weight shape error vs LayoutLMv3 box-information error message." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "Gemma3 DeepSpeed ZeRO-3 loading failure and trainer.evaluate() after train are unrelated code paths." + "reason": "Different code paths: Gemma3 loading with ZeRO-3 vs trainer.evaluate failing after train." }, { "left": "issue:45200", "right": "issue:46463", "accept": false, - "reason": "Different models and defects: Gemma 4 mm_token_type_ids handling vs T5 decoder-input validation messaging." + "reason": "One is a missing-default input requirement for Gemma 4 fine-tuning; the other is a T5 decoder-input validation message issue." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "AutoModel missing PyTorch import failure and Trainer _gpu_broadcast_one NoneType error are unrelated." + "reason": "Different errors and subsystems: missing PyTorch import vs Trainer NoneType broadcast failure." }, { "left": "issue:44493", "right": "issue:46050", "accept": false, - "reason": "Unexpected position-id keys during loading and sinq quantized save/load breakage are different issues." + "reason": "Different regressions: unexpected position-id keys vs quantized model save/load breakage." }, { "left": "issue:44466", "right": "issue:45005", - "accept": false, - "reason": "They are likely related tied-weights/save_pretrained reports, but the evidence is only title-level and deterministic_accept is false, so this is not strong enough to call a duplicate." + "accept": true, + "reason": "Both point to the same v5 tied-weights/save_pretrained serialization problem; 45005 is the broader umbrella and 44466 looks like a device-dependent manifestation." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "Missing PyTorch import and MacOS bus error on a CLIP model are unrelated failures." + "reason": "Different environment failures: PyTorch import missing vs macOS bus error with a CLIP model." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Both involve loading, but one is gguf/from_pretrained breakage and the other is offline model_info() behavior in _patch_mistral_regex." + "reason": "Both are loading-related, but one is gguf file handling and the other is an offline-mode mistral regex patch bug." }, { "left": "issue:43688", "right": "issue:43749", "accept": false, - "reason": "Auxiliary-loss normalization and FSDP CPU RAM efficient loading are different training/load problems." + "reason": "Different problems: auxiliary-loss normalization vs broken FSDP CPU RAM-efficient loading." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC latent decoding mismatch and Whisper batch_decode skip_special_tokens behavior are unrelated." + "reason": "Unrelated: DAC latent decoding mismatch vs Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:44936", "right": "issue:45823", "accept": false, - "reason": "trainer.evaluate() failing after train is not the same as Gemma4 device_map='auto' forward device mismatch." + "reason": "Different runtime failures: trainer lifecycle issue vs Gemma4 device-map mismatch." }, { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Both are tokenizer loading failures, but for different tokenizer classes and different failure modes." + "reason": "Both are tokenizer-related, but they involve different tokenizers and different failure modes." }, { "left": "issue:38175", "right": "issue:43295", "accept": false, - "reason": "Siglip2 zero probabilities and processor.tokenizer access regression are distinct issues." - }, - { - "left": "issue:45372", - "right": "issue:46170", - "accept": false, - "reason": "Gemma 4 processor import failure from mistral_common and Windows ROCm torch.distributed dependency errors are unrelated." + "reason": "Different regressions: SIGLIP2 zero probabilities vs processor.tokenizer/custom model regression." }, { "left": "issue:43653", "right": "issue:44361", "accept": false, - "reason": "BigBird special-token registration and MLukeTokenizer AttributeError are different tokenizer bugs." + "reason": "Different tokenizer bugs with different models and symptoms." }, { "left": "issue:45542", "right": "issue:46178", "accept": false, - "reason": "Both involve backend handling, but they affect different optional backends and different failure paths." + "reason": "Both concern backend availability, but they are different backend-mapping failures (tensorflow vs tensorflow_text) with different causes." + }, + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Different dependency/import failures: mistral_common ReasoningEffort vs Windows ROCm torch.distributed dependency." }, { "left": "issue:42898", "right": "issue:46491", "accept": false, - "reason": "Both are tokenizer behavior regressions, but one is a v5 cleanup-space policy change and the other is a CodeLlama round-trip leading-space loss." + "reason": "Both involve tokenization behavior, but one is a v5 API-default change and the other is a CodeLlama whitespace round-trip regression." }, { "left": "issue:44484", "right": "issue:44704", "accept": false, - "reason": "max_shard_size documentation/question and AutoProcessor cached_file kwargs forwarding are unrelated." + "reason": "Different save/load API issues: shard-size question vs kwargs not passed to cached_file." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Different Qwen2.5-VL bugs: batched chat template crash vs incorrect vision_position_ids for video input." + "reason": "Related model family, but distinct bugs: batched chat-template crash vs wrong vision_position_ids for video input." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "gguf/from_pretrained breakage and Blip2 dtype propagation are unrelated." - }, - { - "left": "issue:43295", - "right": "issue:45356", - "accept": false, - "reason": "Custom processor/tokenizer regression and Kimi-K2.5 codec handling regression are different bugs." - }, - { - "left": "issue:43577", - "right": "issue:43697", - "accept": false, - "reason": "Blip2 dtype loading and RTDetrV2 output differences are unrelated." + "reason": "Different loading regressions: gguf handling vs dtype not propagating to Blip2 submodules." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The overall grouping is mostly grounded, but the one accepted soft pair is stated too strongly. In the packet, 44466\u219445005 is only a title-level soft-similarity pair and its deterministic_accept is false; calling it a \"true duplicate\" and the \"same underlying tied-weights/save_pretrained regression\" overstates what the evidence shows. It would be safer to describe this as a possible/likely related pair rather than a confirmed duplicate. The other negative verdicts are generally conservative and well-supported by the titles." + "accept": true, + "feedback": "The cluster call is grounded in the packet. The only accepted soft pair, issue:44466 \u2194 issue:45005, is a conservative and plausible duplicate/umbrella match around v5 tied-weights save_pretrained serialization. The rest of the pair judgments also stay appropriately cautious and do not overclaim shared root causes. The summary is consistent with the evidence and does not overstate similarity." }, "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": false, + "retried": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -6976,9 +7179,7 @@ "issue:33357", "issue:33666", "issue:36010", - "issue:36331", "issue:41720", - "issue:43262", "issue:43295", "issue:43504", "issue:43531", @@ -6989,6 +7190,7 @@ "issue:43646", "issue:43653", "issue:43688", + "issue:43697", "issue:43720", "issue:43742", "issue:43749", @@ -7010,6 +7212,7 @@ "issue:44977", "issue:44991", "issue:45290", + "issue:45356", "issue:45440", "issue:45491", "issue:45701", @@ -7021,266 +7224,266 @@ "issue:46402" ], "soft_pairs": [ + "issue:43295|issue:45356", "issue:43531|issue:43742", + "issue:43577|issue:43697", "issue:43653|issue:44568", - "issue:43646|issue:43749", "issue:44464|issue:44743", + "issue:43646|issue:43749", "issue:44843|issue:44991", "issue:43688|issue:45704", "issue:29159|issue:33357", "issue:43756|issue:46402", - "issue:43688|issue:45440", "issue:43653|issue:43927", + "issue:43688|issue:45440", "issue:43550|issue:43927", "issue:43688|issue:45701", "issue:44479|issue:44625", - "issue:45491|issue:46002", "issue:43550|issue:44361", + "issue:45491|issue:46002", "issue:44464|issue:44977", "issue:33666|issue:36010", "issue:43742|issue:44821", "issue:32090|issue:33666", "issue:41720|issue:44464", "issue:43643|issue:44561", - "issue:43504|issue:43720", + "issue:44466|issue:46050", "issue:43504|issue:43550", + "issue:43504|issue:43720", "issue:44514|issue:45290", - "issue:44466|issue:46050", "issue:43577|issue:44554", "issue:43504|issue:44610", "issue:43720|issue:43927", - "issue:44554|issue:46365", "issue:43606|issue:43927", - "issue:44977|issue:46209", - "issue:33666|issue:36331", - "issue:43262|issue:43295" + "issue:44554|issue:46365", + "issue:44977|issue:46209" ], - "prepared_review_unit_hash": "d54f49cb3a7c85ead9102bb7660e8c00991ffb6773c7d561fe922a7965f8f813", + "prepared_review_unit_hash": "6094adf0ec07a5ea6bf793b656004f269fb7494d5128a61f38a418b448f19154", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37678, - "estimated_input_tokens": 9420, - "estimated_eval_tokens": 19096 + "serialized_chars": 37676, + "estimated_input_tokens": 9419, + "estimated_eval_tokens": 19094 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the items span tokenizer regressions, model-loading failures, training/runtime errors, and architecture-specific bugs. None of the soft-edge pairs look like the same underlying bug/change, so the safe triage outcome is to keep them separate.", - "confidence": 0.34, - "canonical_issue_reason": "issue:45701 is the broadest umbrella in the set; it is open, has multiple inbound references, and describes a version-dependent tokenization regression, which makes it the most reasonable anchor despite the cluster\u2019s overall heterogeneity.", + "summary": "This cluster is broadly about Transformers regressions and compatibility failures, with many tokenizer- and model-loading-related reports, but the soft pairs mostly describe different concrete bugs, not duplicates.", + "confidence": 0.42, + "canonical_issue_reason": "issue:45701 is the broadest umbrella report in the set (tokenization changes across Transformers versions) and best fits the shared regression theme; the other issues are narrower model- or code-path-specific failures.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45701 is the best representative issue for this cluster because it is active/open and broad enough to serve as a practical triage anchor for the tokenizer-regression side of the set.", + "best_issue_reason": "issue:45701 is the best global representative because it captures the general tokenizer/version-regression pattern seen across the cluster, even though the cluster is heterogeneous and not a tight duplicate group.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but one is about processor.tokenizer/images in custom code and the other is Kimi-K2.5 codec handling; different bug paths." + }, { "left": "issue:43531", "right": "issue:43742", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior vs MobileLLM-125M loading key error; different models and different failure modes." + "reason": "Qwen3-MoE sliding_window behavior and a MobileLLM-125M load key error are unrelated failures." }, { - "left": "issue:43653", - "right": "issue:44568", + "left": "issue:43577", + "right": "issue:43697", "accept": false, - "reason": "Both concern tokenizer special-token behavior, but the bugs are different: BigBird mask token registration vs mdeberta BOS/EOS omission." + "reason": "Blip2 dtype propagation and RTDetrV2 output mismatch are different model-specific issues." }, { - "left": "issue:43646", - "right": "issue:43749", + "left": "issue:43653", + "right": "issue:44568", "accept": false, - "reason": "Custom model initialization breakage is unrelated to FSDP CPU RAM efficient loading." + "reason": "Both involve special tokens, but BigBird mask-token registration and mdeberta BOS/EOS omission are distinct tokenizer bugs." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Both involve generation/cache/compiled paths, but one is chunked generation inconsistency and the other is recurrent-state reset in modular_qwen3_5." + "reason": "Chunked generation with compiled forward and Qwen3.5 recurrent-state reset are different generation-path bugs." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Custom model initialization breakage and FSDP CPU RAM efficient loading are separate startup/loading regressions." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading regressions, but one is HF_HUB_OFFLINE/mistral regex patching and the other is a specific tokenizer load failure for EMBEDDIA/est-roberta." + "reason": "Offline model_info patching and est-roberta tokenizer loading failures are both tokenizer-related but not the same underlying issue." }, { "left": "issue:43688", "right": "issue:45704", "accept": false, - "reason": "Aux-loss normalization in MoE models is unrelated to T5 using apex.FusedRMSNorm and leaking memory." + "reason": "Aux-loss normalization and T5 apex RMSNorm memory leak are unrelated model implementation bugs." }, { "left": "issue:29159", "right": "issue:33357", "accept": false, - "reason": "Slow/fast tokenizer inconsistency is unrelated to a MacOS bus error loading a CLIP model." + "reason": "Tokenizer slow/fast inconsistency and a MacOS CLIP bus error are not the same failure mode." }, { "left": "issue:43756", "right": "issue:46402", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch and Sam3Model text-encoder weights not loading are different model-specific issues." + "reason": "Smollm3 RoPE-layer mismatch and Sam3 text-encoder weights not loading are different model config/loading problems." }, { - "left": "issue:43688", - "right": "issue:45440", + "left": "issue:43653", + "right": "issue:43927", "accept": false, - "reason": "Both reference model discrepancies, but the underlying bugs are different: auxiliary-loss normalization vs DeepSeekV3 implementation divergence." + "reason": "Missing BigBird special-token registration and DiaConfig losing custom token IDs after save/load are distinct tokenizer/state bugs." }, { - "left": "issue:43653", - "right": "issue:43927", + "left": "issue:43688", + "right": "issue:45440", "accept": false, - "reason": "BigBird tokenizer special-token registration is not the same as DiaConfig losing custom token IDs on save/load." + "reason": "MoE auxiliary-loss normalization and DeepseekV3MoE divergence are separate algorithm/implementation issues." }, { "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "Bamba torch.compile/SDPA failure is unrelated to DiaConfig save/load token ID loss." + "reason": "torch.compile with SDPA in Bamba and DiaConfig token-ID loss are unrelated." }, { "left": "issue:43688", "right": "issue:45701", "accept": false, - "reason": "MoE auxiliary-loss normalization is not the same as a broad tokenization-regression report." + "reason": "One is a MoE loss-scaling bug; the other is a broad tokenization regression report." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Video-input regression in Qwen VL models is unrelated to num_labels propagation in Qwen3.5 config." + "reason": "Video-input regression and num_labels propagation are different Qwen3.5 problems." }, { - "left": "issue:45491", - "right": "issue:46002", + "left": "issue:43550", + "right": "issue:44361", "accept": false, - "reason": "Gemma3 NaN embeddings from sliding-window attention are different from NaNs in classification heads after checkout/init." + "reason": "Bamba SDPA/compile failure and MLukeTokenizer AttributeError have no shared code path." }, { - "left": "issue:43550", - "right": "issue:44361", + "left": "issue:45491", + "right": "issue:46002", "accept": false, - "reason": "Bamba torch.compile SDPA failure is unrelated to MLukeTokenizer AttributeError." + "reason": "Gemma3 sliding-window NaNs and classification-head NaNs on checkout/init are separate numerical issues." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Compiled chunked-generation inconsistency and Qwen3.5 flash-attention generation failure are different code paths and symptoms." + "reason": "Compiled-forward chunked generation and flash-attention generation failure are different generation bugs." }, { "left": "issue:33666", "right": "issue:36010", "accept": false, - "reason": "Multi-GPU training problems are unrelated to the GenerationMixin import error." + "reason": "Multi-GPU Qwen2-VL training and GenerationMixin import error are unrelated." }, { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "MobileLLM loading key error and AutoImageProcessor URL loading failure are different loader bugs." + "reason": "MobileLLM key error and AutoImageProcessor URL loading are different loader failures." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Trainer _gpu_broadcast_one TypeError is unrelated to Qwen2-VL multi-GPU training." + "reason": "NoneType gpu broadcast in Trainer and Qwen2-VL multi-GPU training are not the same issue." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "Qwen3 auto device-mapping cudaErrorAssert is not the same as chunked generation inconsistency with compiled forward." + "reason": "Qwen3 auto device mapping cudaErrorAssert and chunked generation with compiled forward are different runtime paths." }, { "left": "issue:43643", "right": "issue:44561", "accept": false, - "reason": "Both touch trust_remote_code, but one is missing fields in AutoConfig and the other is is_torch_fx_available removal breaking remote-code models." + "reason": "Missing config fields with trust_remote_code and removal of is_torch_fx_available are separate trust_remote_code regressions." }, { - "left": "issue:43504", - "right": "issue:43720", + "left": "issue:44466", + "right": "issue:46050", "accept": false, - "reason": "BEiT legacy-field load failure and BitNet packed-weight unpacking failure are unrelated loading bugs." + "reason": "lm_head.weight serialization inconsistency and sinq quantization save/load breakage are different persistence bugs." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "BEiT pretrained loading and Bamba torch.compile/SDPA are different subsystems and failure modes." + "reason": "Beit legacy-field loading and Bamba SDPA compile failure are unrelated." }, { - "left": "issue:44514", - "right": "issue:45290", + "left": "issue:43504", + "right": "issue:43720", "accept": false, - "reason": "Both are apply_chat_template crashes, but they target different inputs and bugs: batched padding=False vs tool-call assistant messages with no content." + "reason": "Legacy-field model loading and packed-weight unpacking during accelerate loading are different model-loading bugs." }, { - "left": "issue:44466", - "right": "issue:46050", + "left": "issue:44514", + "right": "issue:45290", "accept": false, - "reason": "lm_head.weight serialization depends on model device, while sinq quantized save/load is a different persistence bug." + "reason": "Both concern apply_chat_template, but one is batched padding=False on Qwen2.5-VL and the other is tool-call assistant messages with no content." }, { "left": "issue:43577", "right": "issue:44554", "accept": false, - "reason": "Blip2 dtype propagation is unrelated to the MPS attention correctness issue." + "reason": "Blip2 dtype propagation and an MPS attention correctness issue are unrelated." }, { "left": "issue:43504", "right": "issue:44610", "accept": false, - "reason": "BEiT legacy-field loading and OmDet-Turbo processor image-size mismatch are unrelated." + "reason": "Beit legacy-field loading and OmDet-Turbo processor image-size mismatch are different processor/model configuration issues." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "BitNet accelerate-loading unpacking and DiaConfig token-ID persistence are different save/load problems." - }, - { - "left": "issue:44554", - "right": "issue:46365", - "accept": false, - "reason": "MPS attention correctness is unrelated to the Trainer import breakage caused by datasets." + "reason": "BitNet packed-weight loading and DiaConfig custom token-ID loss are distinct." }, { "left": "issue:43606", "right": "issue:43927", "accept": false, - "reason": "CPU offload device mismatch for bark-small is unrelated to DiaConfig save/load token-ID loss." - }, - { - "left": "issue:44977", - "right": "issue:46209", - "accept": false, - "reason": "Qwen3.5 flash-attention generation failure is unrelated to FP8 inference failure on B200." + "reason": "CPU offload device mismatch for bark-small and DiaConfig token-ID loss are unrelated." }, { - "left": "issue:33666", - "right": "issue:36331", + "left": "issue:44554", + "right": "issue:46365", "accept": false, - "reason": "Qwen2-VL multi-GPU training and CustomTrainer.compute_loss signature mismatch are different training issues." + "reason": "Attention correctness on MPS and Trainer import affected by datasets are entirely different problems." }, { - "left": "issue:43262", - "right": "issue:43295", + "left": "issue:44977", + "right": "issue:46209", "accept": false, - "reason": "Audio apply_chat_template sampling-rate defaulting and processor.tokenizer/image-handling regression are separate processor API bugs." + "reason": "Qwen3.5 flash-attention generation failure and FP8 inference on B200 are different hardware/backend issues." } ] }, @@ -7307,21 +7510,22 @@ "issue:28218", "issue:33357", "issue:33666", + "issue:36331", "issue:39692", "issue:43257", + "issue:43262", "issue:43278", "issue:43295", "issue:43299", "issue:43334", "issue:43452", "issue:43493", - "issue:43504", "issue:43525", "issue:43540", "issue:43572", - "issue:43606", "issue:43646", "issue:43688", + "issue:43701", "issue:43720", "issue:43824", "issue:43825", @@ -7330,7 +7534,6 @@ "issue:43994", "issue:44361", "issue:44451", - "issue:44464", "issue:44466", "issue:44488", "issue:44568", @@ -7350,10 +7553,11 @@ "issue:46050", "issue:46192", "issue:46209", - "issue:46424", "issue:46489" ], "soft_pairs": [ + "issue:43262|issue:43295", + "issue:33666|issue:36331", "issue:43525|issue:43572", "issue:43452|issue:44451", "issue:43278|issue:43688", @@ -7361,8 +7565,8 @@ "issue:45823|issue:46050", "issue:43994|issue:46192", "issue:43493|issue:43825", - "issue:39692|issue:43493", "issue:44964|issue:45405", + "issue:39692|issue:43493", "issue:44849|issue:46209", "issue:44625|issue:45584", "issue:28218|issue:33357", @@ -7371,228 +7575,233 @@ "issue:43278|issue:43844", "issue:43493|issue:43844", "issue:44488|issue:44843", - "issue:44568|issue:46489", "issue:43646|issue:44466", - "issue:43540|issue:44977", + "issue:44568|issue:46489", "issue:33357|issue:33666", + "issue:43540|issue:44977", "issue:44589|issue:45520", "issue:43257|issue:43299", "issue:44568|issue:45812", "issue:43334|issue:43572", - "issue:43824|issue:44589", "issue:43927|issue:44361", - "issue:44464|issue:46424", - "issue:43334|issue:45759", - "issue:43504|issue:43606" + "issue:43824|issue:44589", + "issue:43278|issue:43701", + "issue:43334|issue:45759" ], - "prepared_review_unit_hash": "dbc29d608f01a551e2fdf0830b9899900c22416e0cecd2eb77a1ef7bd45a4710", + "prepared_review_unit_hash": "9b812e7c7abcd1b6dde8cd674ac938e5f0d40354f8531f4c001de45172cb8327", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36787, - "estimated_input_tokens": 9197, - "estimated_eval_tokens": 18650 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36515, + "estimated_input_tokens": 9129, + "estimated_eval_tokens": 18514 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: it mixes tokenizer regressions, config/attribute breaks, model-loading failures, and training/runtime bugs. None of the soft edges look like true duplicates or mergeable PR-equivalent changes.", - "confidence": 0.93, - "canonical_issue_reason": "No single canonical issue stands out because the items span many unrelated bugs; if a representative is needed, issue 45812 is the clearest broad tokenizer-regression report, but it does not cover the rest of the cluster.", + "summary": "The cluster is mostly heterogeneous: tokenizer regressions, model-loading failures, config attribute issues, and unrelated training/runtime bugs. None of the soft pairs look like true duplicates under a strict triage standard. The clearest representative issue is the Granite tokenizer regression (#45812).", + "confidence": 0.82, + "canonical_issue_reason": "#45812 is the most concrete, high-signal, and broadly scoped regression in the set: it describes a silent tokenizer bug affecting all Granite models with a clear v4\u2192v5 change. It is a better canonical representative than the narrower or more model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45812 is the strongest representative issue: it describes a broad, user-visible tokenizer regression affecting all Granite models, with a clear symptom and likely shared root cause.", + "best_issue_reason": "#45812 is the best global representative issue because it is specific enough to be actionable, yet broad enough to represent the main tokenizer-regression theme present in the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Different subsystems and failure modes: audio chat-template sampling-rate defaulting vs processor/tokenizer access and image-tokenizer regression." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Qwen2-VL multi-GPU training is unrelated to a CustomTrainer compute_loss signature mismatch." + }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Both mention missing pad token fields after v5, but they affect different configs/models and different attribute names; not the same concrete bug." + "reason": "Both mention missing pad token fields, but they affect different models and different config attributes; not the same underlying bug." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "One is a gguf_file regression in loading APIs, the other is a specific model load failure for ScandiBERT; no shared code-path evidence." + "reason": "Both are loading failures, but one is a gguf/auto-loader regression and the other is a specific model load issue for ScandiBERT." }, { "left": "issue:43278", "right": "issue:43688", "accept": false, - "reason": "BF16/FP32 dtype drift during eval is unrelated to auxiliary-loss normalization in MoE training." + "reason": "Training/eval dtype mismatch is unrelated to auxiliary-loss normalization." }, { "left": "issue:43720", "right": "issue:44361", "accept": false, - "reason": "Packed-weight loading in BitNet and a tokenizer AttributeError in MLuke are unrelated failures." + "reason": "BitNet packed-weight loading and MLukeTokenizer AttributeError are unrelated code paths." }, { "left": "issue:45823", "right": "issue:46050", "accept": false, - "reason": "Different models and different failure modes: device_map forward mismatch versus save/load breakage after quantization." + "reason": "Device-map forward mismatch is not the same as sinq quantized model save/load corruption." }, { "left": "issue:43994", "right": "issue:46192", "accept": false, - "reason": "SigLIP2 inference quality problems are unrelated to return_tensors being ignored when text_kwargs is passed." + "reason": "SigLIP2 inference quality issues are unrelated to return_tensors being ignored when text_kwargs is passed." }, { "left": "issue:43493", "right": "issue:43825", "accept": false, - "reason": "A SigLIP2 implementation discrepancy is not the same bug as an incorrect pipeline error message about translation tasks." + "reason": "A SigLIP2 implementation discrepancy is unrelated to a pipeline error message bug." }, { - "left": "issue:39692", - "right": "issue:43493", + "left": "issue:44964", + "right": "issue:45405", "accept": false, - "reason": "Both involve SigLIP2, but one is documentation/example breakage and the other is an implementation mismatch; not the same issue." + "reason": "Model-loading failure and a dependency-version bump are not the same bug." }, { - "left": "issue:44964", - "right": "issue:45405", + "left": "issue:39692", + "right": "issue:43493", "accept": false, - "reason": "Model-loading failure for Phi-4 multimodal is unrelated to an unsafely bumped PEFT version floor." + "reason": "Docs example errors are not the same underlying problem as the SigLIP2 implementation discrepancy." }, { "left": "issue:44849", "right": "issue:46209", "accept": false, - "reason": "Qwen3.5 hidden-states behavior and FP8/B200 inference are different runtime paths with no clear overlap." + "reason": "Qwen3.5 hidden-state behavior and FP8 inference on B200 are unrelated issues." }, { "left": "issue:44625", "right": "issue:45584", "accept": false, - "reason": "A missing num_labels propagation bug is unrelated to Whisper empty-transcription generation after token alignment." + "reason": "Qwen3.5 num_labels propagation and Whisper empty-transcription generation are different model/component bugs." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer spacing after added tokens and a MacOS bus error in CLIP loading are unrelated." + "reason": "Tokenizer spacing after added tokens is unrelated to a MacOS bus error in CLIP loading." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Both are Qwen3.5, but hidden_states behavior and incorrect visual encoder key serialization are different concrete bugs." + "reason": "Both involve Qwen3.5, but one is hidden-states output and the other is save_pretrained visual-encoder key serialization." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Documentation example issues do not match the processor.tokenizer / image-tokenization regression in custom model code." + "reason": "A SigLIP2 docs example bug is unrelated to the processor.tokenizer/image-tokenizer regression." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Eval dtype changes and ZeRO-3 gradient growth are separate training-path problems." + "reason": "Embedding dtype changes and gradient explosion with ZeRO-3 are different training pathologies." }, { "left": "issue:43493", "right": "issue:43844", "accept": false, - "reason": "SigLIP2 implementation mismatch is unrelated to ZeRO-3 gradient instability." + "reason": "SigLIP2 implementation mismatch is unrelated to ZeRO-3 gradient growth." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "A specific model load failure and an offline-mode Mistral-regex regression are different loading issues." - }, - { - "left": "issue:44568", - "right": "issue:46489", - "accept": false, - "reason": "Both are tokenizer regressions, but they affect different tokenizers and symptoms; the code paths are not shown to be the same." + "reason": "A specific model load failure is not the same as AutoTokenizer offline-mode breakage in _patch_mistral_regex." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Custom model initialization breakage and lm_head serialization differences are unrelated." + "reason": "Custom model initialization breaks broadly, but inconsistent lm_head serialization is a different save/load bug." }, { - "left": "issue:43540", - "right": "issue:44977", + "left": "issue:44568", + "right": "issue:46489", "accept": false, - "reason": "Video-input processing in Qwen3OmniMoe is not the same as flash-attention generation behavior in Qwen3.5." + "reason": "Both are tokenizer regressions, but they concern different models and different failure mechanisms; too broad to merge." }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "MacOS CLIP bus error and Qwen2-VL multi-GPU training are unrelated subsystems." + "reason": "MacOS bus error in CLIP loading and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Video-input processing in Qwen3OmniMoe is unrelated to Qwen3.5 flash-attention generation behavior." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage lookup failure and flash_attn import mapping on Python 3.13 are separate compatibility issues." + "reason": "Float8 storage lookup failure is unrelated to a flash_attn import_utils KeyError on Python 3.13." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Both involve Qwen3 MoE loading, but one is accelerate/deepspeed weight conversion and the other is composite Qwen3VL model loading; not enough to treat as one bug." + "reason": "Both are Qwen3-MoE-related, but one is weight conversion with accelerate/deepspeed and the other is multimodal loading in dev; not the same bug." }, { "left": "issue:44568", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they concern different models and different incorrect-token behaviors." + "reason": "Both are tokenizer regressions, but one is mDeBERTa special-token handling while the other is Granite token IDs; different models and fixes." }, { "left": "issue:43334", "right": "issue:43572", "accept": false, - "reason": "Missing pad_token_id in Qwen3-VL text config and missing pad_token_idx in StableLmConfig are similar in theme but not the same concrete defect." - }, - { - "left": "issue:43824", - "right": "issue:44589", - "accept": false, - "reason": "An import symbol missing for Qwen2.5-VL is unrelated to a Float8 storage-object error." + "reason": "Missing pad_token_id in Qwen3-VL text config is not the same as missing pad_token_idx in StableLmConfig." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "DiaConfig save/load token-ID loss and MLukeTokenizer task AttributeError are unrelated." + "reason": "DiaConfig custom token IDs lost on save/load is unrelated to MLukeTokenizer task AttributeError." }, { - "left": "issue:44464", - "right": "issue:46424", + "left": "issue:43824", + "right": "issue:44589", "accept": false, - "reason": "Chunked-generation inconsistency with compiled forward is not the same as max_cache_length being ignored outside quantized cache." + "reason": "An import error for Qwen2.5-VL is unrelated to a Float8 storage type error." }, { - "left": "issue:43334", - "right": "issue:45759", + "left": "issue:43278", + "right": "issue:43701", "accept": false, - "reason": "Both are multimodal Qwen config issues, but one is a missing pad_token_id attribute and the other is improper unwrapping of text_config; different failures." + "reason": "Embedding dtype drift in evaluation is unrelated to a resume_from_checkpoint key mismatch." }, { - "left": "issue:43504", - "right": "issue:43606", + "left": "issue:43334", + "right": "issue:45759", "accept": false, - "reason": "Beit legacy-field loading and suno/bark CPU-offload device mismatch are unrelated bugs." + "reason": "pad_token_id missing in Qwen3-VL config is unrelated to AutoModelForCausalLM.from_config not unwrapping text_config." } ] }, @@ -7626,7 +7835,6 @@ "issue:38175", "issue:42490", "issue:43065", - "issue:43278", "issue:43493", "issue:43504", "issue:43540", @@ -7663,10 +7871,12 @@ "issue:46050", "issue:46121", "issue:46285", - "issue:46369" + "issue:46369", + "issue:46424" ], "soft_pairs": [ - "issue:43278|issue:43701", + "issue:43504|issue:43606", + "issue:44464|issue:46424", "issue:43575|issue:45405", "issue:43572|issue:44291", "issue:44792|issue:45405", @@ -7682,208 +7892,214 @@ "issue:43844|issue:44964", "issue:45588|issue:46050", "issue:43065|issue:46369", - "issue:29159|issue:32090", "issue:42490|issue:43696", + "issue:29159|issue:32090", "issue:44423|issue:44734", "issue:45084|issue:45520", "issue:28218|issue:34567", "issue:44554|issue:45446", - "issue:45406|issue:46121", "issue:15354|issue:38175", - "issue:43540|issue:45200", + "issue:45406|issue:46121", "issue:43493|issue:43701", + "issue:43540|issue:45200", "issue:45072|issue:45198" ], - "prepared_review_unit_hash": "3e71404ae4f1b713febe2f8df5581c3473d5278513ed8eaae11e9c8e49b952f7", + "prepared_review_unit_hash": "2fe8a7f6565e003e322f74ae69b6bf73cbea4a6693acfdf74c7f0a46dd7b1d28", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 35826, - "estimated_input_tokens": 8957, - "estimated_eval_tokens": 18170 + "soft_pair_count": 28, + "serialized_chars": 36113, + "estimated_input_tokens": 9029, + "estimated_eval_tokens": 18314 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair appears to be the same underlying bug/change; the candidates span unrelated tokenizer, serving, quantization, model-loading, and training issues. If a representative issue is needed, issue 43873 is the broadest and most actively discussed, but it is not a confirmed duplicate of the others.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 43873 is the broadest, most active open issue in the set and reads like a central user-facing regression report, so it is the best representative issue here.", + "summary": "This cluster is over-broad: the items span unrelated tokenizer, model-loading, serving, quantization, training, and CI failures. I rejected all soft duplicate edges because they do not describe the same concrete bug/change.", + "confidence": 0.96, + "canonical_issue_reason": "No single issue is a reliable duplicate hub; the set is too heterogeneous to justify one canonical issue.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43873 has the strongest mix of scope and discussion activity, making it the best standalone issue candidate in this cluster.", + "best_issue_reason": "Issue 43873 is the closest representative of the few related quantization/offloading reports, but it still is not a true duplicate target for the rest.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43278", - "right": "issue:43701", + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy field loading failure vs CPU offload device mismatch; different code paths and symptoms." + }, + { + "left": "issue:44464", + "right": "issue:46424", "accept": false, - "reason": "Different failures: dtype drift in eval vs checkpoint key mismatch during resume." + "reason": "Compiled-forward generation inconsistency vs cache-length config behavior; related area, but not the same bug." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "OOM during tensor-parallel model load is unrelated to a PEFT version bump/release issue." + "reason": "Tensor-parallel OOM vs PEFT version pin; unrelated issues." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "One is a missing config field; the other is a constructor/init_empty_weights API mismatch." + "reason": "Missing config field vs init_empty_weights TypeError; different loading failures." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "Janus image-generation test failure has no relation to the PEFT version problem." + "reason": "Janus image-generation test failure vs PEFT version bump; no overlap in underlying problem." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Text-only fine-tuning defaults vs multi-GPU inference OOM are different code paths and bugs." + "reason": "Missing multimodal token IDs for fine-tuning vs multi-GPU inference OOM; distinct bugs." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "Both mention devices, but they are different model/test failures with different root causes." + "reason": "Both mention device behavior, but one is a model offload mismatch and the other is a stale CI test." }, { "left": "issue:44493", "right": "issue:46285", "accept": false, - "reason": "Unexpected position-id keys and CLIP structural inconsistency are distinct API regressions." + "reason": "Unexpected position-id key handling vs CLIP text-model structure change; not the same issue." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "Legacy-field loading for BEiT and BigBird tokenizer special-token registration are unrelated." + "reason": "Pretrained loading field issue vs tokenizer special-token registration; unrelated." }, { "left": "issue:45229", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma4 inference issues, but OOM and device mismatch are different underlying bugs." + "reason": "OOM during multi-GPU inference vs PLE device mismatch in Gemma4 forward; different failure modes." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Compiled-forward chunked generation mismatch is unrelated to Whisper empty-transcription handling." + "reason": "Compiled forward generation inconsistency vs Whisper empty-transcription failure after token alignment; different models and code paths." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 bad outputs and position-id key warnings are different symptoms and code paths." + "reason": "SigLIP2 bad outputs vs position-id key warnings; not the same underlying bug." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "Flash Attention 2 test failures are unrelated to MLflow reporting after training failure." + "reason": "Flash Attention CI tests vs MLflow failure-reporting bug; unrelated." }, { "left": "issue:43873", "right": "issue:46050", "accept": false, - "reason": "Quantization/offloading behavior and sinq save/load breakage are not the same defect." + "reason": "Both involve quantization/offloading, but one is offload behavior and the other is save/load after sinq quantization; not mergeable as one fix." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "ZeRO-3 gradient growth is unrelated to Phi-4 multimodal model loading." + "reason": "ZeRO-3 gradient anomaly vs Phi-4 multimodal load failure; different problems." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "flash_attention.py sink-less crash and sinq persistence issues involve different components." + "reason": "Flash-attention sink-less AttributeError vs quantized save/load breakage; unrelated code paths." }, { "left": "issue:43065", "right": "issue:46369", "accept": false, - "reason": "A dummy Conv2d in Sam3PixelDecoder is not the same as memory overflow in sam3-litetext." + "reason": "Dummy Conv2d implementation vs memory overflow in sam3-litetext; one is structural, the other is a runtime OOM." }, { - "left": "issue:29159", - "right": "issue:32090", + "left": "issue:42490", + "right": "issue:43696", "accept": false, - "reason": "Tokenizer slow/fast inconsistency and NoneType GPU broadcast errors are unrelated." + "reason": "Import warning for local paths vs GPT-oss OOM; no relation." }, { - "left": "issue:42490", - "right": "issue:43696", + "left": "issue:29159", + "right": "issue:32090", "accept": false, - "reason": "Trailing-slash import warning and GPT-oss OOM are completely different issues." + "reason": "Tokenizer fast/slow inconsistency vs Trainer broadcast NoneType error; unrelated subsystems." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Both are serve/runtime crashes, but one is multimodal .to() handling and the other is KV-cache indexing." + "reason": "Serve crashes in different multimodal/KV-cache paths; same surface area, but not the same concrete bug." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Template-node compilation error and Python 3.13 flash_attn import error are unrelated." + "reason": "Template-node compilation error vs flash_attn import KeyError on Python 3.13; unrelated." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Added-token spacing and token counter updates are different tokenizer/trainer behaviors." + "reason": "Added-token spacing bug vs TrainerState token counter not updating; different functionality." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "MPS attention correctness differs from a PyTorch version guard bug in flex_attention." - }, - { - "left": "issue:45406", - "right": "issue:46121", - "accept": false, - "reason": "Gemma4Processor missing _tokenizer and rope validation type handling are unrelated." + "reason": "MPS attention correctness issue vs PyTorch version check for AuxRequest import; unrelated." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero-probability outputs are different problems." + "reason": "TorchScript GeneratorExp export failure vs SigLIP2 zero-probability outputs; different models and failures." }, { - "left": "issue:43540", - "right": "issue:45200", + "left": "issue:45406", + "right": "issue:46121", "accept": false, - "reason": "Video-input processing in Qwen3OmniMoe is unrelated to mm_token_type_ids defaults." + "reason": "Gemma4Processor missing _tokenizer vs rope-parameter JSON list TypeError; unrelated." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 HF-vs-JAX discrepancy and checkpoint key mismatch are not the same bug." + "reason": "SigLIP2 implementation discrepancy vs resume_from_checkpoint key mismatch; different bug classes." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video-input processing error vs Gemma4 text-only token defaults; not the same issue." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "bfloat16 dtype mismatches and Wav2Vec2 save/tokenization failure are unrelated." + "reason": "Bfloat16 dtype mismatch in inference vs Wav2Vec2 save/tokenization failure; unrelated." } ] }, @@ -7960,8 +8176,8 @@ "issue:45200|issue:45823", "issue:43295|issue:43450", "issue:44186|issue:46192", - "issue:45823|issue:46192", "issue:43232|issue:44849", + "issue:45823|issue:46192", "issue:43295|issue:44610", "issue:43295|issue:43873", "issue:44521|issue:45290", @@ -7971,29 +8187,29 @@ "issue:39692|issue:46192", "issue:44554|issue:45356", "issue:43572|issue:43957", - "issue:45381|issue:46209", "issue:37428|issue:43824", - "issue:44466|issue:45127", "issue:43450|issue:43575", - "issue:44898|issue:45412", + "issue:45381|issue:46209", + "issue:44466|issue:45127", "issue:43950|issue:46285", "issue:42490|issue:43493", - "issue:44792|issue:45563", + "issue:44898|issue:45412", "issue:43334|issue:43525", + "issue:44792|issue:45563", "issue:42915|issue:43824", "issue:38175|issue:43493", "issue:43653|issue:44112", "issue:43452|issue:46512", "issue:28218|issue:35141" ], - "prepared_review_unit_hash": "de8e0149abfd136a1a95efa4d5d2227cfb0145549bb32ed9326ad2837cdc9b2b", + "prepared_review_unit_hash": "37621dac511fdb9db72157b42ba84ee71158b9080f7bd952b2e990a5a4f1f6c8", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, @@ -8008,186 +8224,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "None of the proposed soft pairs look like true duplicates; they span unrelated regressions, model families, and code paths. No PRs were provided.", - "confidence": 0.34, - "canonical_issue_reason": "Issue 44521 is the clearest open, user-facing bug with solid activity and a well-scoped reproduction around multimodal chat masking.", + "summary": "This set is heterogeneous: the soft-similarity pairs mostly share broad subsystem words (loading, tokenizer, multimodal, config) but not the same concrete bug. I reject all proposed duplicate links; the strongest standalone issue in the set is a core `from_pretrained()` regression.", + "confidence": 0.88, + "canonical_issue_reason": "issue:43950 is the strongest representative report here: it describes a concrete, high-impact `from_pretrained()` regression with a clear failure mode and broad loading impact, unlike the more model-specific or UI/doc-related reports.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 is the strongest standalone representative: active, clearly described, and directly impacts a common public API path.", + "best_issue_reason": "issue:43950 is the best issue to anchor the cluster because it is the most general and actionable library regression in the set, with the broadest downstream effect.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Different failures: flash-attention import symbol missing vs. meta-device model loading breakage. Not the same underlying bug." + "reason": "Different root causes: missing flash-attn import symbol vs. meta-device loading breakage." }, { "left": "issue:45200", "right": "issue:45823", "accept": false, - "reason": "Different symptoms and paths: mm_token_type_ids defaulting for text-only fine-tuning vs. device mismatch under device_map=auto." + "reason": "Gemma4 fine-tuning input defaulting vs. PLE device mismatch during forward; different code paths." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Different components: processor/tokenizer regression vs. batched video processor shape bug." + "reason": "Processor/tokenizer image-handling regression vs. batched video-shape bug; not the same failure." }, { "left": "issue:44186", "right": "issue:46192", "accept": false, - "reason": "Tokenizer crash on NER/batched padding is unrelated to return_tensors being ignored when text_kwargs is passed." + "reason": "NER padding/truncation crash vs. `return_tensors` being ignored when `text_kwargs` is passed." }, { - "left": "issue:45823", - "right": "issue:46192", + "left": "issue:43232", + "right": "issue:44849", "accept": false, - "reason": "One is a device placement bug, the other is an argument-handling bug in processor inputs." + "reason": "Generation kwargs after `sync_gpus` vs. Qwen3.5 hidden-state output bug; unrelated." }, { - "left": "issue:43232", - "right": "issue:44849", + "left": "issue:45823", + "right": "issue:46192", "accept": false, - "reason": "Generation-state update bug after sync_gpus is unrelated to Qwen3.5 output_hidden_states behavior." + "reason": "Device-map/PLE mismatch is unrelated to `return_tensors` handling." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "Custom processor/tokenizer regression is unrelated to OmDet-Turbo\u2019s image size mismatch." + "reason": "Different multimodal processor bugs: tokenizer/image path vs. OmDet-Turbo size mismatch." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Processor API regression and quantization/offloading behavior are different code paths." + "reason": "Processor/tokenizer regression is not the same as quantization offloading behavior." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both involve apply_chat_template, but one is zero assistant masks for multimodal inputs while the other crashes on tool-call messages. Not the same bug." + "reason": "Both hit chat templates, but one is all-zero assistant masks and the other is a tool-call/no-content crash." }, { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "Missing pad_token_idx in one config class is unrelated to loading physical-intelligence/fast on newer transformers." + "reason": "Missing `pad_token_idx` in StableLmConfig vs. model-loading failure for `physical-intelligence/fast`; different bugs." }, { "left": "issue:45520", "right": "issue:46178", "accept": false, - "reason": "Different backend import errors: flash_attn on Python 3.13 vs tensorflow_text missing from BACKENDS_MAPPING." + "reason": "Import registry failure for `flash_attn` vs. unsupported `tensorflow_text` backend mapping entry." }, { "left": "issue:43550", "right": "issue:43844", "accept": false, - "reason": "torch.compile + SDPA failure in Bamba is unrelated to ZeRO-3 gradient growth in DeepSpeed." + "reason": "`torch.compile`+SDPA bug in Bamba vs. DeepSpeed ZeRO-3 gradient inflation; different runtime/training issues." }, { "left": "issue:39692", "right": "issue:46192", "accept": false, - "reason": "SigLIP2 documentation mistakes are unrelated to the return_tensors/text_kwargs API regression." + "reason": "Doc example/model-processor mismatch and quantization failure vs. tokenizer argument handling; unrelated." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "MPS attention correctness issue is unrelated to Kimi-K2.5 tokenizer regression." + "reason": "MPS attention correctness issue vs. Kimi-K2.5 tokenizer codec regression; different layers." }, { "left": "issue:43572", "right": "issue:43957", "accept": false, - "reason": "Missing pad_token_idx and meta-device loading failures are separate regressions." - }, - { - "left": "issue:45381", - "right": "issue:46209", - "accept": false, - "reason": "Qwen2.5-VL video position IDs bug is unrelated to FP8 inference failure on B200." + "reason": "Config field regression vs. meta-device model loading failure; different code paths." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "Different missing imports for different symbols and model families." + "reason": "Both are import errors, but for different missing symbols/modules and different models." }, { - "left": "issue:44466", - "right": "issue:45127", + "left": "issue:43450", + "right": "issue:43575", "accept": false, - "reason": "Serialization inconsistency of tied weights is not the same as LoRA merge collapse with extended vocabulary." + "reason": "Video processor batch-shape bug vs. tensor-parallel OOM while loading Qwen2; unrelated." }, { - "left": "issue:43450", - "right": "issue:43575", + "left": "issue:45381", + "right": "issue:46209", "accept": false, - "reason": "Batched video processor shape bug and tensor-parallel OOM are unrelated." + "reason": "Qwen2.5-VL video position IDs bug vs. FP8 inference failure on B200; no shared concrete fix." }, { - "left": "issue:44898", - "right": "issue:45412", + "left": "issue:44466", + "right": "issue:45127", "accept": false, - "reason": "Perceiver input-size failure and RT-DETR memory-release issue affect different model behavior." + "reason": "Serialization inconsistency of tied weights vs. LoRA merge collapse after vocab extension; related domain, different bug." }, { "left": "issue:43950", "right": "issue:46285", "accept": false, - "reason": "Non-persistent buffer corruption on from_pretrained is unrelated to CLIP text model structural inconsistency." + "reason": "Non-persistent buffer corruption in `from_pretrained()` vs. CLIP class structure inconsistency; different symptoms and code paths." }, { "left": "issue:42490", "right": "issue:43493", "accept": false, - "reason": "Local-path trailing-slash warning is unrelated to SigLIP2 implementation discrepancy." + "reason": "Trailing-slash import warning vs. SigLIP2 HF-vs-JAX discrepancy; unrelated." }, { - "left": "issue:44792", - "right": "issue:45563", + "left": "issue:44898", + "right": "issue:45412", "accept": false, - "reason": "Janus image-generation test failure is unrelated to stale num_return_sequences warning in paged generate()." + "reason": "Perceiver interpolation failure vs. RT-DETR memory not releasing; different subsystems." }, { "left": "issue:43334", "right": "issue:43525", "accept": false, - "reason": "Both mention missing pad_token_id, but they are different model configs and different load paths; not enough evidence of one shared bug." + "reason": "Same symptom class (`pad_token_id` missing), but different model configs and likely separate model-specific fixes." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure vs. stale paged-generate warning; not the same bug." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to Qwen2.5-VL import failure." + "reason": "Qwen3Moe FP8 failure vs. Qwen2_5_VL import error; different model/runtime problems." }, { "left": "issue:38175", "right": "issue:43493", "accept": false, - "reason": "Both are SigLIP2-related, but one reports zero probabilities while the other reports a JAX/HF discrepancy; insufficient evidence they are the same bug." + "reason": "Both concern SigLIP2, but one reports zero probabilities and the other a broader HF-vs-JAX discrepancy; not enough to treat as one bug." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "BigBirdTokenizer special-token registration and GraniteSpeech CI device-override failure are unrelated." + "reason": "BigBird tokenizer special-token registration bug vs. GraniteSpeech CI stale override test; unrelated." }, { "left": "issue:43452", "right": "issue:46512", "accept": false, - "reason": "GGUF loading failures are unrelated to stale config.dtype after from_config(dtype=...)." + "reason": "GGUF `from_pretrained()` breakage vs. `from_config(dtype=...)` leaving stale config dtype; different loading paths." }, { "left": "issue:28218", "right": "issue:35141", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization after resize with untied weights." + "reason": "Tokenizer added-space behavior vs. resized-embedding reinit bug; different components entirely." } ] }, @@ -8263,222 +8479,222 @@ "soft_pairs": [ "issue:43278|issue:43856", "issue:43824|issue:43975", - "issue:45200|issue:45325", "issue:43388|issue:43526", "issue:43526|issue:46192", + "issue:45200|issue:45325", "issue:44442|issue:45198", "issue:43262|issue:43450", - "issue:45520|issue:46170", - "issue:38175|issue:45412", "issue:43819|issue:44112", "issue:41720|issue:46209", + "issue:38175|issue:45412", "issue:44561|issue:46234", - "issue:45759|issue:46285", "issue:43827|issue:43901", + "issue:45759|issue:46285", "issue:44610|issue:44857", "issue:44206|issue:45584", "issue:43819|issue:45083", - "issue:44857|issue:45593", "issue:43335|issue:43550", - "issue:43611|issue:46402", + "issue:44857|issue:45593", + "issue:45520|issue:46170", "issue:43606|issue:45083", "issue:37428|issue:43425", + "issue:43611|issue:46402", "issue:45759|issue:46512", "issue:44514|issue:44521", "issue:29942|issue:32090", "issue:43335|issue:43653", "issue:46431|issue:46566" ], - "prepared_review_unit_hash": "31af2fbc1c47eb546cf8ce69efb9dadd830a776177b51d0b22c4cc9a60321aec", + "prepared_review_unit_hash": "3fe922a61254b95f4251432ec28d3efa2cb7413eed97730afdf0c9b2c384779b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 36009, - "estimated_input_tokens": 9003, - "estimated_eval_tokens": 18262 + "serialized_chars": 36006, + "estimated_input_tokens": 9002, + "estimated_eval_tokens": 18260 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is not a true duplicate cluster: it mixes unrelated model-loading, processor, training, and docs bugs across many subsystems. None of the soft edges look mergeable into a single issue, so all candidate links should be rejected.", - "confidence": 0.95, - "canonical_issue_reason": "No single canonical issue: the items are heterogeneous and do not share one underlying bug. The closest thematic subgroup is around chat_template/processor behavior, but it is still not a duplicate set.", + "summary": "The pairs are only superficially similar by keywords; they span unrelated bugs across audio, multimodal chat templates, import/version breaks, training, and device/runtime issues. I would not merge any of the soft edges.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 44521 is the most active and well-scoped item in the set, with multiple inbound references and discussion, so it is the best representative issue even though it is not a duplicate of the others.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 is the broadest, most central current processor/chat_template bug among the listed items, but it still does not represent the full set of unrelated reports.", + "best_issue_reason": "Issue 44521 is the strongest standalone issue to keep as the cluster representative because it is concrete, reproducible, and has the most external references/activity in this group.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43278", "right": "issue:43856", "accept": false, - "reason": "BF16/eval dtype mismatch vs Qwen3 MoE memory usage; different bugs and code paths." + "reason": "Different bugs: dtype mismatch during evaluation vs MoE training memory usage." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "Qwen2.5-VL import error vs DeepSeek detokenization regression; unrelated failures." - }, - { - "left": "issue:45200", - "right": "issue:45325", - "accept": false, - "reason": "Both involve multimodal models, but one is mm_token_type_ids defaults and the other is RoPE position scaling; not the same bug." + "reason": "Unrelated failures: missing Qwen2.5-VL import vs incorrect detokenization in DeepSeek Coder." }, { "left": "issue:43388", "right": "issue:43526", "accept": false, - "reason": "Both touch labels, but one is gather_for_metrics batching and the other is BeitImageProcessorFast reduce_labels." + "reason": "Both concern labels, but one drops tuple labels in metrics and the other truncates BEiT image labels." }, { "left": "issue:43526", "right": "issue:46192", "accept": false, - "reason": "Image processor label reduction bug vs return_tensors being ignored when text_kwargs is passed; unrelated." + "reason": "Different processor issues: label reduction bug vs silently ignored return_tensors with text_kwargs." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Different multimodal Qwen bugs: missing token-type defaults vs rope index scaling error." }, { "left": "issue:44442", "right": "issue:45198", "accept": false, - "reason": "Different tokenizer classes and different failure modes: AutoTokenizer loading vs save_pretrained/tokenization." + "reason": "Both are tokenizer/speech related, but they affect different models and different failure modes." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Chat template sampling-rate default vs video processor batched shape; different processor pipelines." - }, - { - "left": "issue:45520", - "right": "issue:46170", - "accept": false, - "reason": "Python 3.13 flash_attn import key error vs Windows ROCm torch.distributed dependency issue; unrelated environment breakages." - }, - { - "left": "issue:38175", - "right": "issue:45412", - "accept": false, - "reason": "Unexpected zero probabilities vs RT-DETR memory leak; different models and symptoms." + "reason": "Different modalities and code paths: audio chat-template sampling rate vs video batch output shape." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "DAC forward/from_latents mismatch vs GraniteSpeech CI stale device override test; not the same code-path problem." + "reason": "Distinct problems: DAC latent reconstruction vs a GraniteSpeech CI device-override test." }, { "left": "issue:41720", "right": "issue:46209", "accept": false, - "reason": "Different hardware-specific failures: cudaErrorAssert on A800 vs FP8 inference on B200." + "reason": "Both are hardware/runtime failures, but they involve different models, precisions, and root causes." }, { - "left": "issue:44561", - "right": "issue:46234", + "left": "issue:38175", + "right": "issue:45412", "accept": false, - "reason": "Both are import_utils symbol removals, but they involve different missing helpers and different breakages." + "reason": "Completely different issues: zero probabilities in SigLIP2 vs RT-DETR memory reclamation." }, { - "left": "issue:45759", - "right": "issue:46285", + "left": "issue:44561", + "right": "issue:46234", "accept": false, - "reason": "Both are config/structure issues, but Qwen multimodal text_config unwrapping is unrelated to CLIP .text_model inconsistency." + "reason": "Both are import-utils breakages, but they are for different missing helpers and different downstream failures." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are docs regressions, but they refer to different APIs and different deprecated options." + "reason": "Both are docs cleanup around v5 pipelines, but they target different docs and different deprecated APIs." + }, + { + "left": "issue:45759", + "right": "issue:46285", + "accept": false, + "reason": "Different model-configuration bugs: unwrapping text_config vs inconsistent CLIP text model structure." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "OmDet-Turbo processor size mismatch vs LwDetrImageLoss AMP crash; unrelated detection issues." + "reason": "Different detection-related bugs: wrong processor image size vs AMP/CUDA crash in a loss implementation." }, { "left": "issue:44206", "right": "issue:45584", "accept": false, - "reason": "Unsupported center arg in a feature extractor vs Whisper empty-transcription generation bug; different audio paths." + "reason": "Different audio bugs: unsupported feature-extractor arg vs Whisper empty-transcription generation failure." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "DAC latent mismatch vs qwen3_omni_moe feature-length helper behavior; different helpers and models." + "reason": "Both mention feature-extraction helpers, but they are separate code paths and separate model families." }, { - "left": "issue:44857", - "right": "issue:45593", + "left": "issue:43335", + "right": "issue:43550", "accept": false, - "reason": "AMP crash in LwDetrImageLoss vs D-FINE auxiliary-loss logic when denoising is off; different training bugs." + "reason": "Unrelated: SwitchTransformers config layer construction vs Bamba torch.compile/SDPA failure." }, { - "left": "issue:43335", - "right": "issue:43550", + "left": "issue:44857", + "right": "issue:45593", "accept": false, - "reason": "SwitchTransformers sparse-layer creation bug vs Bamba torch.compile SDPA failure; unrelated model-specific issues." + "reason": "Different detection model bugs: mixed-precision crash vs missing auxiliary losses when denoising is off." }, { - "left": "issue:43611", - "right": "issue:46402", + "left": "issue:45520", + "right": "issue:46170", "accept": false, - "reason": "Generic base_model_prefix loading regression vs Sam3Model text encoder weight loading; not the same bug." + "reason": "Different platform/version issues: Python 3.13 flash_attn lookup vs Windows ROCm distributed import failure." }, { "left": "issue:43606", "right": "issue:45083", "accept": false, - "reason": "CPU-offload device mismatch for bark-small vs qwen3_omni_moe output-length helper; unrelated." + "reason": "Different subsystems and symptoms: CPU-offload device mismatch vs Qwen3 Omni feature-length helper behavior." }, { "left": "issue:37428", "right": "issue:43425", "accept": false, - "reason": "Missing flash_supports_window_size import vs Torch 2.10 incompatibility; different compatibility problems." + "reason": "Version compatibility problems, but one is a missing flash-attention symbol and the other is Torch 2.10 incompatibility." + }, + { + "left": "issue:43611", + "right": "issue:46402", + "accept": false, + "reason": "Both are model-loading regressions, but they affect different models and different loading stages." }, { "left": "issue:45759", "right": "issue:46512", "accept": false, - "reason": "Both involve from_config, but one is text_config unwrapping and the other is stale config.dtype after dtype construction." + "reason": "Both involve from_config in multimodal setups, but one is about unwrapping text_config and the other about stale dtype metadata." }, { "left": "issue:44514", "right": "issue:44521", "accept": false, - "reason": "Both are apply_chat_template bugs, but one crashes on batched padding=False input while the other produces all-zero assistant masks." + "reason": "Related to apply_chat_template, but one is batched padding behavior and the other is assistant mask generation." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Flash Attention 2 test failures vs Trainer _gpu_broadcast_one NoneType error; unrelated." + "reason": "Completely different failures: flash-attention tests vs Trainer GPU broadcast TypeError." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "Sparse layer creation in SwitchTransformers vs BigBirdTokenizer special-token decode issue; unrelated." + "reason": "Different tokenizer/config bugs with no shared code path." }, { "left": "issue:46431", "right": "issue:46566", "accept": false, - "reason": "Processor missing chat template vs DiffusionGemma staying on meta device; different failure classes." + "reason": "Unrelated: missing chat template vs weights stuck on meta with auto device mapping." } ] }, @@ -8558,8 +8774,8 @@ "issue:44855|issue:45083", "issue:43335|issue:43504", "issue:37428|issue:44351", - "issue:45127|issue:45245", "issue:45200|issue:45588", + "issue:45127|issue:45245", "issue:43278|issue:43825", "issue:29127|issue:32090", "issue:43335|issue:45083", @@ -8567,243 +8783,243 @@ "issue:45588|issue:45859", "issue:39692|issue:43526", "issue:43819|issue:44493", - "issue:26413|issue:31501", "issue:44206|issue:45823", "issue:43335|issue:44112", + "issue:26413|issue:31501", "issue:43335|issue:45593", "issue:45310|issue:45478", "issue:45459|issue:46234", "issue:34689|issue:35141", - "issue:39692|issue:44493", "issue:42490|issue:43854", + "issue:39692|issue:44493", "issue:43606|issue:44857", "issue:44448|issue:44779", "issue:44360|issue:44561", - "issue:44857|issue:45072", "issue:28218|issue:32090", + "issue:44857|issue:45072", "issue:45290|issue:45381", "issue:38175|issue:44493" ], - "prepared_review_unit_hash": "82540a8467b592611b21e643cce13d0e3e5632c1f6530af0bd15230904fb4d1e", + "prepared_review_unit_hash": "aa848f6ee382b2f9c94a5717810ed6e0ba3d62600f59760022a3ffe14f302b62", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37237, - "estimated_input_tokens": 9310, - "estimated_eval_tokens": 18876 + "serialized_chars": 37244, + "estimated_input_tokens": 9311, + "estimated_eval_tokens": 18878 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most candidates are clearly unrelated despite some broad subsystem overlap. The only strong duplicate pair is the two reports of the same Qwen3.5 MoE `from_pretrained` regression; everything else differs by model, code path, or symptom.", - "confidence": 0.91, - "canonical_issue_reason": "issue:45310 is the more detailed of the duplicate Qwen3.5 MoE regression reports and is the better anchor for the cluster.", + "summary": "This cluster is mostly heterogeneous and does not form one duplicate set. The only clear duplicate pair is the Qwen3.5 MoE from_pretrained regression: issue 45310 and 45478 describe the same failure.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 45310 is the best canonical issue for the only true duplicate pair here: it matches 45478 on title and reported failure, and it is the earlier/more detailed report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the clearest standalone representative: it has the fuller description and matches the duplicate report in issue:45478.", + "best_issue_reason": "Issue 45310 is the strongest representative artifact in this cluster because it is the clearest exact duplicate target among the items and has the most complete description for that specific bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45290", "right": "issue:46326", "accept": false, - "reason": "Both concern tokenizer/chat-template behavior, but one is tool-call handling and the other is `continue_final_message` for mistral_common; not the same bug." + "reason": "Both are tokenizer/chat-template bugs, but they concern different APIs and failure modes; not the same underlying issue." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "One is a Bamba torch.compile/SDPA failure, the other is a GraniteSpeech CI stale-device test; different models and failure modes." + "reason": "Both mention CI/device-related failures, but they affect different models and code paths with different root causes." }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Pegasus v4/v5 output drift is unrelated to Qwen3.5 recurrent-state reset in cached seq_len>1 paths." + "reason": "Different regressions in different model implementations; no shared concrete bug path." }, { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "Python 3.13 IndentationError in DebertaV2 import is unrelated to a qwen3_omni_moe feature-length helper bug." + "reason": "One is a Python 3.13 import/parsing failure, the other is a helper output-length behavior issue; unrelated." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "SwitchTransformers config layer creation bug and BEiT legacy-field loading bug affect different code paths." + "reason": "SwitchTransformers sparse-layer config and BEiT pretrained loading are unrelated bugs." }, { "left": "issue:37428", "right": "issue:44351", "accept": false, - "reason": "A missing flash-attention import helper is unrelated to the missing `HybridCache` symbol." + "reason": "Both are import errors, but they involve different missing symbols/modules and different causes." }, { - "left": "issue:45127", - "right": "issue:45245", + "left": "issue:45200", + "right": "issue:45588", "accept": false, - "reason": "LoRA/extended-vocab collapse is unrelated to the categorical-cardinality runtime error." + "reason": "Different models and different failure surfaces; not the same underlying bug." }, { - "left": "issue:45200", - "right": "issue:45588", + "left": "issue:45127", + "right": "issue:45245", "accept": false, - "reason": "Gemma4 `mm_token_type_ids` defaulting and flash-attention `s_aux=None` crashes are different model-specific defects." + "reason": "LoRA merge collapse and category-cardinality limits are unrelated problems." }, { "left": "issue:43278", "right": "issue:43825", "accept": false, - "reason": "BF16-to-FP32 dtype drift in evaluation is unrelated to an error-message wording issue for unsupported translation tasks." + "reason": "Training/eval dtype drift is unrelated to the pipeline translation-task error message." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "LayoutLMv3 box-validation messaging and `_gpu_broadcast_one` rejecting None are unrelated failures." + "reason": "Error-message clarity in LayoutLMv3 is unrelated to the Trainer NoneType broadcast failure." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "Sparse-layer creation in SwitchTransformers is not the same bug as qwen3_omni_moe feature-length calculation." + "reason": "Different subsystems and different bug classes; no duplicate evidence." }, { "left": "issue:43720", "right": "issue:44112", "accept": false, - "reason": "BitNet packed-weight loading and GraniteSpeech stale-device CI are different issues with no shared code path." + "reason": "Both are CI failures, but one is packed-weight loading and the other is a stale device override test; not the same bug." }, { "left": "issue:45588", "right": "issue:45859", "accept": false, - "reason": "Both involve Qwen3.5/GPU offload topics, but one is flash-attention `s_aux` handling while the other is rotary embedding CPU-offload compatibility; not mergeable as one fix." + "reason": "Both involve flash/offload-related code, but the concrete failure paths and affected components differ." }, { "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "SigLIP2 docs/example problems are unrelated to BeitImageProcessorFast's `reduce_labels` return-value bug." + "reason": "Documentation/example failures and a BeitImageProcessorFast label bug are unrelated." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "DAC latent decoding mismatch is unrelated to position-id key warnings across models." - }, - { - "left": "issue:26413", - "right": "issue:31501", - "accept": false, - "reason": "Resume-from-checkpoint epoch_iterator failure and FineWeb training not starting are different training pipeline problems." + "reason": "DAC latent/forward mismatch and unexpected position-id keys are different issues." }, { "left": "issue:44206", "right": "issue:45823", "accept": false, - "reason": "Unsupported `center` arg in LasrFeatureExtractor is unrelated to Gemma4 device-map forward mismatch." + "reason": "Unsupported center argument in a feature extractor is unrelated to the Gemma4 PLE device mismatch." }, { "left": "issue:43335", "right": "issue:44112", "accept": false, - "reason": "SwitchTransformers sparse-layer config and GraniteSpeech device-override CI are unrelated." + "reason": "A configuration bug in SwitchTransformers is unrelated to a CI test failing due to stale device overrides." + }, + { + "left": "issue:26413", + "right": "issue:31501", + "accept": false, + "reason": "Both affect training startup, but the checkpoint resume iterator error and the FineWeb SLM startup failure are different root causes." }, { "left": "issue:43335", "right": "issue:45593", "accept": false, - "reason": "SwitchTransformers layer-creation bug and D-FINE auxiliary-loss behavior are unrelated." + "reason": "Sparse-layer configuration and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:45310", "right": "issue:45478", "accept": true, - "reason": "Same underlying regression: Qwen3.5 MoE `from_pretrained` fails in transformers>=5.4.0; the titles and context match." + "reason": "Identical reported bug: both describe the same transformers>=5.4.0 Qwen3.5 MoE from_pretrained error, with matching title and failure context." }, { "left": "issue:45459", "right": "issue:46234", "accept": false, - "reason": "Tokenizer protobuf error masking and missing `is_falcon_mamba_ssm_available` utility are unrelated import/tokenizer issues." + "reason": "Tokenizer protobuf-error masking and a missing import_utils availability helper are unrelated." }, { "left": "issue:34689", "right": "issue:35141", "accept": false, - "reason": "Llama 3.2 vision model loading regression and embedding reinitialization after resize are different code paths." + "reason": "Model-loading breakage for Llama 3.2 Vision and embedding reinitialization after resize are different code-path problems." }, { - "left": "issue:39692", - "right": "issue:44493", + "left": "issue:42490", + "right": "issue:43854", "accept": false, - "reason": "SigLIP2 docs/quantization issues do not match the broad position-id warning regression." + "reason": "Local-path slash import warning and GLM-4.7 Flash test-loading failure are different issues." }, { - "left": "issue:42490", - "right": "issue:43854", + "left": "issue:39692", + "right": "issue:44493", "accept": false, - "reason": "Trailing-slash local-path import warning and GLM-4.7-Flash unit-test load failure are different loading problems." + "reason": "SigLIP2 example errors and unexpected position-id keys are unrelated." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "CPU offload device mismatch in Bark is unrelated to float16 AMP crash in LwDETR loss." + "reason": "CPU offload device mismatch and float16 AMP CUDA crash are different failure modes." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "Pegasus v4/v5 output mismatch and DeepSeek tokenizer v5 regression are different model/tokenizer behaviors." + "reason": "Pegasus v4/v5 output regression and Deepseek tokenizer regression are unrelated." }, { "left": "issue:44360", "right": "issue:44561", "accept": false, - "reason": "DSA indexer activation issue and removal of `is_torch_fx_available` breaking trust_remote_code are unrelated." + "reason": "DSA indexer ReLU discussion and removal of is_torch_fx_available breaking trust_remote_code are unrelated." }, { - "left": "issue:44857", - "right": "issue:45072", + "left": "issue:28218", + "right": "issue:32090", "accept": false, - "reason": "Different dtype-related failures in different models/code paths; not the same underlying bug." + "reason": "Tokenizer added-space behavior and Trainer broadcast NoneType error are unrelated." }, { - "left": "issue:28218", - "right": "issue:32090", + "left": "issue:44857", + "right": "issue:45072", "accept": false, - "reason": "Tokenizer added-space behavior and Trainer GPU broadcast NoneType error are unrelated." + "reason": "Different dtype-related crashes in different models/components; not the same underlying bug." }, { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Chat-template tool-call handling and Qwen2.5-VL video `vision_position_ids` are different bugs." + "reason": "Tool-call chat-template crash and qwen2.5-vl video position-id bug are different issues." }, { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "Zero probabilities in SigLIP2 is unrelated to position-id warnings across models." + "reason": "Zero probabilities in SigLIP2 and unexpected position-id keys are not the same bug." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet: issue:45310 and issue:45478 are an exact title match and clearly the strongest duplicate pair. The soft-edge verdicts are conservative and mostly separate by model, code path, or symptom, with no obvious overstatement of evidence." + "feedback": "The summary is well grounded in the packet: 45310 and 45478 have identical titles and very similar failure descriptions, while the other soft pairs are conservatively rejected with issue-specific reasons. The canonical choice is also reasonable and not overstated." }, "evaluator_used": true, "retried": false, @@ -8879,16 +9095,16 @@ "issue:32090|issue:33357", "issue:29942|issue:35141", "issue:43550|issue:45083", - "issue:44855|issue:46399", "issue:42490|issue:43575", "issue:44448|issue:44977", + "issue:44855|issue:46399", "issue:44977|issue:45229", "issue:43927|issue:45698", - "issue:10000|issue:15354", "issue:27957|issue:34689", + "issue:10000|issue:15354", "issue:43746|issue:46209", - "issue:43278|issue:43329", "issue:37428|issue:44987", + "issue:43278|issue:43329", "issue:44857|issue:45588", "issue:44561|issue:44779", "issue:43278|issue:43381", @@ -8896,215 +9112,215 @@ "issue:44977|issue:45563", "issue:41720|issue:45229", "issue:43976|issue:45440", - "issue:44568|issue:45491", "issue:25251|issue:32090", "issue:32090|issue:33290", + "issue:44568|issue:45491", "issue:45588|issue:45823", "issue:44448|issue:44466", "issue:43412|issue:45588" ], - "prepared_review_unit_hash": "166c663382ef54558121912ba6e8f564c2ff2ebff91e7bc536d98c4000c26734", + "prepared_review_unit_hash": "94eeded5e9b831a9ead200450df1c1a22b988174a5bd547175f3366341106c2b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35962, - "estimated_input_tokens": 8991, - "estimated_eval_tokens": 18238 + "serialized_chars": 35960, + "estimated_input_tokens": 8990, + "estimated_eval_tokens": 18236 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No duplicate pairs should be merged here; the cluster is a heterogeneous mix of tokenizer, model-loading, flash-attention, training, export, and device-map bugs. If forced to pick a representative issue, issue:45920 is the broadest concrete regression, but it does not subsume the rest of the cluster.", - "confidence": 0.93, - "canonical_issue_reason": "issue:45920 is the closest representative because it is a concrete v5 tokenizer regression affecting multiple models, but the cluster is still too mixed to have a true canonical issue.", + "summary": "This cluster does not contain clear duplicates: the soft pairs span unrelated models, features, and failure modes. I would not merge any of the suggested edges; the most representative issue is the broad tokenizer regression report.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45920 is the strongest representative because it is an open, high-discussion regression affecting multiple tokenizers/models, making it broader and more central than the narrowly scoped bug reports elsewhere in the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 is the best single issue to represent the cluster\u2019s strongest recurring theme (tokenizer/output regressions across models), though it is not a real umbrella for the other bugs.", + "best_issue_reason": "issue:45920 is the best cluster anchor for triage: it has the widest blast radius and the strongest sign of an actual platform-wide regression, so it is the most useful issue to keep visible.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "Checkpoint resume key mismatch is unrelated to qwen3_omni_moe feature-length calculation." + "reason": "Checkpoint key mismatch and a Qwen3 Omni feature-length helper bug are unrelated code paths." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "Multi-image VL fine-tuning failure and Beit label reduction bug are different components and failure modes." + "reason": "One is a Qwen2.5-VL training/data-loading failure; the other is a BEiT image processor label-reduction bug." }, { "left": "issue:45920", "right": "issue:46192", "accept": false, - "reason": "Both involve tokenizers, but one is a wrong-token-id regression and the other is return_tensors being ignored when text_kwargs is passed." + "reason": "Both touch tokenization, but one is a broad v5 token-ID regression while the other is a specific return_tensors argument-handling bug." }, { "left": "issue:32090", "right": "issue:33357", "accept": false, - "reason": "Trainer broadcast NoneType error and MacOS bus error in CLIP loading are unrelated." + "reason": "Trainer broadcast NoneType error and a MacOS CLIP bus error are different failures in different execution environments." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Flash Attention 2 test failures do not match the token-embedding reinitialization bug." + "reason": "Flash Attention 2 test failures and token-embedding reinitialization in post_init are unrelated issues." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "Bamba torch.compile/SDPA failure is a different code path from qwen3_omni_moe length computation." - }, - { - "left": "issue:44855", - "right": "issue:46399", - "accept": false, - "reason": "Both mention parsing/decorator issues, but they affect different models and fail in different places." + "reason": "A torch.compile/SDPA failure in Bamba-9B-v2 does not match a Qwen3 Omni feature-length helper bug." }, { "left": "issue:42490", "right": "issue:43575", "accept": false, - "reason": "Trailing-slash import warning is not the same as Qwen2-57B-A14B-Instruct OOM during tensor-parallel load." + "reason": "A trailing-slash import warning and a tensor-parallel OOM on Qwen2-57B are different problems." }, { "left": "issue:44448", "right": "issue:44977", "accept": false, - "reason": "Pegasus v4/v5 output drift and Qwen3.5 flash-attention generation failure are distinct bugs." + "reason": "Pegasus v4/v5 output drift and Qwen3.5 flash-attention generation failure are different model-specific regressions." + }, + { + "left": "issue:44855", + "right": "issue:46399", + "accept": false, + "reason": "Both involve import/decorator parsing, but they affect different models and different Torch/JIT failure modes." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Flash-attention generation bug and Gemma4 multi-GPU CUDA OOM are not the same underlying problem." + "reason": "Flash-attention generation corruption and a multi-GPU OOM are not the same underlying bug." }, { "left": "issue:43927", "right": "issue:45698", "accept": false, - "reason": "Saving custom token IDs and loading the wrong custom module after save_pretrained are both serialization-related but not the same bug." + "reason": "A saved config losing custom token IDs is distinct from loading the wrong custom module after save_pretrained." }, { - "left": "issue:10000", - "right": "issue:15354", + "left": "issue:27957", + "right": "issue:34689", "accept": false, - "reason": "German DistilBertModel issue and ViT torch.jit GeneratorExp export failure are unrelated." + "reason": "XLMRoberta Flash Attention 2 and Llama 3.2 vision model loading failures are unrelated." }, { - "left": "issue:27957", - "right": "issue:34689", + "left": "issue:10000", + "right": "issue:15354", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support and Llama 3.2 Vision model loading regression are different failures." + "reason": "German DistilBertModel and torch.jit.script GeneratorExp export issues are different models and different root causes." }, { "left": "issue:43746", "right": "issue:46209", "accept": false, - "reason": "PEFT local-checkpoint loading and FP8 inference on B200 are unrelated." + "reason": "PEFT local-checkpoint loading and FP8 inference on B200 are unrelated concerns." }, { - "left": "issue:43278", - "right": "issue:43329", + "left": "issue:37428", + "right": "issue:44987", "accept": false, - "reason": "BF16-to-FP32 dtype change in eval is unrelated to the undefined video-branch helpers in multimodal token counting." + "reason": "A missing flash-attention utility import and a model-loading failure are too generic to be the same bug." }, { - "left": "issue:37428", - "right": "issue:44987", + "left": "issue:43278", + "right": "issue:43329", "accept": false, - "reason": "Missing internal flash-attention helper import and loading physical-intelligence/fast are different regression symptoms." + "reason": "Embedding dtype drift in eval and undefined video-token helper names are unrelated." }, { "left": "issue:44857", "right": "issue:45588", "accept": false, - "reason": "AMP loss crash and flash_attention.py crashing on s_aux=None are distinct code paths." + "reason": "float16 AMP loss crashes and a sink-less Flash Attention attribute error are different attention/precision bugs." }, { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "trust_remote_code breakage from removed API is unrelated to Deepseek tokenizer regression." + "reason": "Removal of is_torch_fx_available breaking trust_remote_code is unrelated to DeepSeek tokenizer regression." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Embedding dtype mismatch and gradient-checkpointing-in-eval are different issues." + "reason": "Embedding dtype changes and gradient checkpointing in eval mode are separate issues." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Quantization offloading behavior and multi-GPU CUDA OOM are not the same defect." + "reason": "Quantization offloading behavior and Gemma4 multi-GPU CUDA OOM are not the same failure." }, { "left": "issue:44977", "right": "issue:45563", "accept": false, - "reason": "Generation failure with flash-attention and a stale num_return_sequences warning are different severities and behaviors." + "reason": "Flash-attention generation failure and a stale num_return_sequences warning are unrelated." }, { "left": "issue:41720", "right": "issue:45229", "accept": false, - "reason": "Auto device mapping cudaErrorAssert and Gemma4 multi-GPU OOM do not share the same bug." + "reason": "Qwen3 device-mapping cudaErrorAssert and Gemma4 multi-GPU OOM are distinct distributed-inference failures." }, { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Python version compatibility and DeepseekV3MoE divergence are unrelated." - }, - { - "left": "issue:44568", - "right": "issue:45491", - "accept": false, - "reason": "Tokenizer special-token handling and Gemma3 sliding-window NaNs are separate components." + "reason": "A Python version compatibility issue and a DeepSeekV3MoE implementation divergence are different kinds of problems." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Pipeline top_k nesting change and Trainer _gpu_broadcast_one NoneType error are unrelated." + "reason": "Pipeline top_k nesting behavior and Trainer GPU broadcast errors are unrelated." }, { "left": "issue:32090", "right": "issue:33290", "accept": false, - "reason": "Trainer broadcast failure and Adafactor/DeepSpeed OOM are different training-time problems." + "reason": "Trainer broadcast NoneType and deepspeed Adafactor OOM are different runtime failures." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer special-token behavior and Gemma3 sliding-window NaNs are unrelated bugs." }, { "left": "issue:45588", "right": "issue:45823", "accept": false, - "reason": "flash_attention.py AttributeError on sink-less models is unrelated to Gemma4 PLE device mismatch." + "reason": "A flash-attention attribute crash and a Gemma4 PLE device mismatch are different forward-pass issues." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "v4/v5 Pegasus output drift and lm_head.weight serialization inconsistency are both v5 regressions but not the same bug." + "reason": "Pegasus output drift and lm_head.weight serialization inconsistency are both v5 regressions, but not the same underlying bug." }, { "left": "issue:43412", "right": "issue:45588", "accept": false, - "reason": "Executorch export failure and flash_attention.py s_aux=None crash affect different subsystems and code paths." + "reason": "Executorch export/training-guard failures and flash-attention sink-less crashes are unrelated." } ] }, @@ -9180,49 +9396,49 @@ "issue:45198|issue:45803", "issue:45372|issue:45588", "issue:42490|issue:43873", + "issue:33357|issue:39692", "issue:45245|issue:45563", "issue:29127|issue:34567", - "issue:33357|issue:39692", "issue:45372|issue:45406", "issue:43611|issue:46121", "issue:43653|issue:44743", - "issue:43065|issue:45412", - "issue:44361|issue:45198", "issue:43278|issue:43575", + "issue:44361|issue:45198", "issue:44871|issue:44977", "issue:42915|issue:42994", "issue:44186|issue:45198", "issue:44206|issue:44479", + "issue:43065|issue:45412", "issue:43065|issue:43994", "issue:43381|issue:43701", "issue:45459|issue:45712", "issue:29942|issue:30064", "issue:45412|issue:46369", "issue:43756|issue:45440", - "issue:43994|issue:45412", "issue:44265|issue:45198", "issue:45372|issue:45459", "issue:43653|issue:44442", + "issue:43994|issue:45412", "issue:43066|issue:44462", "issue:27957|issue:29942", "issue:29127|issue:33357", "issue:45440|issue:45593", "issue:43421|issue:45491" ], - "prepared_review_unit_hash": "e852b99cd17dff3f8c5b7f42dc24c618725d504749414a9f79a283a259e9f16f", + "prepared_review_unit_hash": "7562140989c232c3cb497573df58e41edb1be2b16dabd70a38bbea5ba36b9c5a", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36390, + "serialized_chars": 36391, "estimated_input_tokens": 9098, "estimated_eval_tokens": 18452 }, @@ -9231,198 +9447,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is a noisy mix of unrelated issue reports across many models and utilities. None of the soft pairs look like the same underlying bug/change, so I would not merge any of them into duplicates. There is no clear canonical issue or PR for this set.", - "confidence": 0.94, - "canonical_issue_reason": null, + "summary": "The set is highly heterogeneous; most soft pairs only share broad themes like tokenizers, quantization, or model loading, but not the same concrete bug. I would not merge any of the proposed soft edges. The closest thing to a hub is the Gemma 4 processor-loading issue, but even that only loosely relates to two other Gemma 4 reports.", + "confidence": 0.43, + "canonical_issue_reason": "issue:45372 is the most central concrete bug report in this set: it has a clear failure mode, related inbound references, and sits at the center of the only small sub-thread (Gemma 4 processor loading).", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45372 is the best overall representative issue because it is the most actionable and specific bug report, while the rest are mostly isolated reports from unrelated subsystems.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45198", "right": "issue:45803", "accept": false, - "reason": "Different bugs: Wav2Vec2 save/tokenization failure vs Fuyu batch tensor conversion swallowing errors." + "reason": "Both are tokenizer/data-handling bugs, but one is Wav2Vec2 save/tokenization failure and the other is a bare-except bug in Fuyu conversion; different code paths and fixes." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma4 processor import failure vs flash-attention crash on `s_aux=None`; different code paths and symptoms." + "reason": "Both mention Gemma 4-ish loading/runtime issues, but one is a missing ReasoningEffort import during processor loading and the other is a flash-attention crash on sink-less models; not the same bug." }, { "left": "issue:42490", "right": "issue:43873", "accept": false, - "reason": "Local-path trailing-slash import warning is unrelated to quantization offloading behavior." + "reason": "Local-path import warning vs quantization/offloading behavior; same broad loading area but different underlying failures and fixes." }, { - "left": "issue:45245", - "right": "issue:45563", + "left": "issue:33357", + "right": "issue:39692", "accept": false, - "reason": "Category-count runtime error vs stale `num_return_sequences` warning; not the same issue." + "reason": "Community CLIP bus error on macOS and SigLIP2 docs example errors are unrelated failures; one is runtime stability, the other is documentation/example correctness." }, { - "left": "issue:29127", - "right": "issue:34567", + "left": "issue:45245", + "right": "issue:45563", "accept": false, - "reason": "LayoutLMv3 box validation messaging is unrelated to TrainerState token-count tracking." + "reason": "Category-count runtime error and a stale generate warning are unrelated issues with different triggers and code paths." }, { - "left": "issue:33357", - "right": "issue:39692", + "left": "issue:29127", + "right": "issue:34567", "accept": false, - "reason": "MacOS CLIP bus error and SigLIP2 docs/model mismatch are separate problems." + "reason": "LayoutLMv3 error messaging and TrainerState token-count tracking are unrelated subsystems and bugs." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "Both touch Gemma4, but one is an import dependency break and the other is a serve-time missing `_tokenizer` attribute; not clearly the same fix." + "reason": "Both involve Gemma 4 processor loading, but one is an import dependency problem and the other is a missing _tokenizer attribute in serve; distinct fixes." }, { "left": "issue:43611", "right": "issue:46121", "accept": false, - "reason": "Base-model-prefix loading regression is unrelated to ROPE-validation list typing." + "reason": "Base-model-prefix loading regression and rope-validation TypeError are separate configuration/serialization bugs." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "BigBird tokenizer special-token decode bug vs Qwen3.5 recurrent-state reset bug; different components." + "reason": "Tokenizer special-token registration and Qwen3.5 recurrent-state reset are unrelated model/tokenization behaviors." }, { - "left": "issue:43065", - "right": "issue:45412", + "left": "issue:43278", + "right": "issue:43575", "accept": false, - "reason": "Dummy Conv2d placeholder cleanup is unrelated to RT-DETR memory being retained after deletion." + "reason": "Mixed precision dtype drift during eval and tensor-parallel OOM are different problems in different execution phases." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "MLuke tokenizer task AttributeError is not the same as Wav2Vec2 save/tokenization failure." - }, - { - "left": "issue:43278", - "right": "issue:43575", - "accept": false, - "reason": "BF16-to-FP32 eval dtype drift is unrelated to Qwen2-57B-A14B TP OOM." + "reason": "MLukeTokenizer AttributeError and Wav2Vec2 save/tokenization failure are different tokenizer implementations with no shared concrete path." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "Gemma3 EOS config mismatch and Qwen3.5 flash-attention generation failure are different issues." + "reason": "Gemma-3 eos_token_id mismatch and Qwen3.5 flash-attention generation failure are unrelated model-specific bugs." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "FineGrainedFP8Config failure and quantized model saving failure are separate quantization bugs." + "reason": "Both touch quantization, but one is FP8 config loading for Qwen3Moe and the other is quantized model saving; different stages and likely different fixes." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 tokenizer NER/padding crash is unrelated to Wav2Vec2 save/tokenization failure." + "reason": "LayoutLMv2 tokenization crashes and Wav2Vec2 tokenization/save failure are separate tokenizers and failure modes." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Unsupported `center` arg in LasrFeatureExtractor vs video-input regression for Qwen VL models." + "reason": "Feature extractor argument regression and Qwen video-input regression are unrelated code paths." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d placeholder in Sam3PixelDecoder and RT-DETR memory not released are not the same bug." }, { "left": "issue:43065", "right": "issue:43994", "accept": false, - "reason": "Dummy Conv2d placeholder cleanup does not match SigLIP2 nonsense-output behavior." + "reason": "A dummy layer implementation issue and SigLIP2 nonsensical outputs are unrelated." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient-checkpointing eval-mode restriction is unrelated to resume-from-checkpoint key mismatch." + "reason": "Eval-mode gradient checkpointing restriction and resume_from_checkpoint key mismatch are different training-loop bugs." }, { "left": "issue:45459", "right": "issue:45712", "accept": false, - "reason": "Tokenizer error masking without protobuf and leftover dummy classes are unrelated maintenance bugs." + "reason": "Hidden tokenizer errors from missing protobuf and leftover dummy classes are both cleanup/error-handling issues, but not the same underlying problem." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 test failures and void segmentation-map processing are different failures." + "reason": "Flash Attention 2 test failures and void segmentation-map processing are unrelated issues." }, { "left": "issue:45412", "right": "issue:46369", "accept": false, - "reason": "RT-DETR GC/memory-retention issue is not the same as sam3-litetext memory overflow." + "reason": "Both are memory-related reports, but RT-DETR deletion GC behavior and SAM3-litetext overflow are different models and failure sources." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch vs DeepseekV3MoE implementation divergence are unrelated model-specific bugs." - }, - { - "left": "issue:43994", - "right": "issue:45412", - "accept": false, - "reason": "SigLIP2 output corruption is unrelated to RT-DETR object lifetime/memory retention." + "reason": "Smollm3 RoPE-layer mismatch and DeepseekV3MoE divergence are separate architecture/implementation discrepancies." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "torch.export/torch_compilable_check failure is unrelated to Wav2Vec2 tokenization/save behavior." + "reason": "torch.export exportability and Wav2Vec2 tokenization/save failure do not share a concrete code path." }, { "left": "issue:45372", "right": "issue:45459", "accept": false, - "reason": "Gemma4 import error and protobuf-gated tokenizer error masking are different tokenizer-loading problems." + "reason": "Gemma 4 processor loading failure and a protobuf-related tokenizer exception hiding bug are unrelated." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "BigBird special-token decode bug is unrelated to AutoTokenizer failing to load FastSpeech2ConformerTokenizer." + "reason": "BigBirdTokenizer special-token registration and AutoTokenizer loading FastSpeech2ConformerTokenizer are different tokenizer-loading problems." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 inference corruption and RT-DETR memory leak are different model/runtime issues." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Wrong tokenizer decoder type in v5 and ignoring repository tokenizer.json are distinct loading regressions." + "reason": "Wrong tokenizer decoder type in v5 and AutoTokenizer ignoring tokenizer.json are both tokenizer-loading related, but they are distinct bugs and fixes." }, { "left": "issue:27957", "right": "issue:29942", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support and failing FA2 tests are not clearly the same code-path bug." + "reason": "XLMRoberta Flash Attention 2 support and failing Flash Attention 2 tests are related only at a high level; not the same concrete issue." }, { "left": "issue:29127", "right": "issue:33357", "accept": false, - "reason": "LayoutLMv3 error-message clarity and MacOS CLIP bus error are unrelated." + "reason": "LayoutLMv3 error clarity and CLIP bus error on macOS are entirely different bugs." }, { "left": "issue:45440", "right": "issue:45593", "accept": false, - "reason": "DeepseekV3MoE implementation divergence and D-FINE auxiliary-loss behavior are unrelated model bugs." + "reason": "DeepSeekV3MoE implementation divergence and D-FINE auxiliary-loss omission are unrelated model-specific issues." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Runtime special-token post-processor updates and Gemma3 NaN embeddings are different problems." + "reason": "Runtime special-token post-processor updates and Gemma3 NaN embeddings from sliding-window attention are distinct tokenizer/model-runtime bugs." } ] }, @@ -9497,25 +9713,25 @@ ], "soft_pairs": [ "issue:43526|issue:45834", - "issue:43994|issue:46369", "issue:29127|issue:35141", "issue:43575|issue:43701", - "issue:43526|issue:44186", + "issue:43994|issue:46369", "issue:42490|issue:43381", - "issue:43824|issue:44279", + "issue:43526|issue:44186", "issue:44442|issue:45072", "issue:43575|issue:44387", + "issue:43824|issue:44279", "issue:42915|issue:43931", "issue:43526|issue:44857", - "issue:46121|issue:46399", "issue:43334|issue:43531", + "issue:46121|issue:46399", "issue:43761|issue:45859", "issue:27957|issue:29127", "issue:44977|issue:45245", "issue:27957|issue:33290", "issue:43262|issue:44206", - "issue:45588|issue:45600", "issue:33290|issue:34689", + "issue:45588|issue:45600", "issue:44464|issue:45491", "issue:43646|issue:44077", "issue:33290|issue:35141", @@ -9528,220 +9744,214 @@ "issue:44062|issue:45907", "issue:34689|issue:43611" ], - "prepared_review_unit_hash": "dcf9e533eaf46e6746ec5b7ea14eebcb6a104c8de1691f5a16f53898ab93d95e", + "prepared_review_unit_hash": "37983b0f1d58e5f8b212240f74b981bf672b05accdb41001eb474b918c10f4f7", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36857, - "estimated_input_tokens": 9215, - "estimated_eval_tokens": 18686 + "serialized_chars": 36862, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: mostly unrelated transformer/model bugs with no convincing duplicate pairs. I rejected all soft edges; the only plausible commonality is broad subsystem overlap, not the same underlying fix.", - "confidence": 0.9, - "canonical_issue_reason": "No strong canonical duplicate emerges from the set; the issues span unrelated model-loading, tokenizer, vision/audio, and training/runtime bugs.", + "summary": "These items are mostly unrelated, model-specific bug reports with only superficial lexical overlap. None of the soft-similarity pairs look like the same underlying bug, so I would not merge any of them as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": "issue:43873 is the strongest representative issue: it is broad, user-facing, still open, and has inbound references; the rest are mostly narrow, model-specific regressions or errors.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the broadest and most actively discussed report here, with inbound references and a general quantization/offloading memory symptom, so it is the best anchor if one issue must represent the cluster.", + "best_issue_reason": "issue:43873 is the best overall issue candidate because it has the widest apparent scope and the most ongoing signal among this mixed set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43526", "right": "issue:45834", "accept": false, - "reason": "Different bugs in different models: BeitImageProcessorFast label reduction vs Kosmos2.5 OCR long-input indexing." - }, - { - "left": "issue:43994", - "right": "issue:46369", - "accept": false, - "reason": "Different model families and failures: SigLIP2 bad outputs vs SAM3 LiteText memory overflow." + "reason": "Both are vision/OCR-related, but one is a label-reduction bug in BeitImageProcessorFast and the other is an OCR long-input index error in Kosmos2.5; different failure modes and code paths." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "Unrelated problems: LayoutLMv3 error messaging vs token embedding reinit after resize." + "reason": "LayoutLMv3 error-message clarity and token-embedding reinitialization are unrelated bugs in different components." }, { "left": "issue:43575", "right": "issue:43701", "accept": false, - "reason": "Different code paths: tensor-parallel load OOM vs resume_from_checkpoint key mismatch." + "reason": "One is an OOM when loading a large model with tensor parallelism, the other is a checkpoint key mismatch; no shared code-path bug." }, { - "left": "issue:43526", - "right": "issue:44186", + "left": "issue:43994", + "right": "issue:46369", "accept": false, - "reason": "Different tokenizer/processor bugs in different models; no shared concrete fix." + "reason": "SigLIP2 wrong outputs via AutoModel/pipeline and SAM3-LiteText memory overflow are different model-specific issues." }, { "left": "issue:42490", "right": "issue:43381", "accept": false, - "reason": "Local-path import warning is unrelated to gradient checkpointing eval-mode behavior." + "reason": "A local-path trailing-slash import warning is unrelated to gradient checkpointing being disallowed in eval mode." }, { - "left": "issue:43824", - "right": "issue:44279", + "left": "issue:43526", + "right": "issue:44186", "accept": false, - "reason": "Import error for a specific Qwen class vs a generic dependency issue; not the same bug." + "reason": "BeitImageProcessorFast label reduction and LayoutLMv2Tokenizer NER/padding crashes are separate preprocessing bugs." }, { "left": "issue:44442", "right": "issue:45072", "accept": false, - "reason": "AutoTokenizer loading failure vs dtype mismatch in inference; separate problems." + "reason": "FastSpeech2ConformerTokenizer loading and dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated." }, { "left": "issue:43575", "right": "issue:44387", "accept": false, - "reason": "Both mention OOM/quantization, but one is model load with tensor parallelism and the other is int4 reserved-memory regression." + "reason": "Both mention OOM, but one is a specific tensor-parallel load failure and the other is a quantization memory regression in 5.x; different problems." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Missing Qwen2.5-VL import symbol versus a vague dependency issue; not the same concrete bug." }, { "left": "issue:42915", "right": "issue:43931", "accept": false, - "reason": "Different failure modes and model families: FineGrainedFP8Config vs weight-shape mismatch." + "reason": "FineGrainedFP8Config failure in Qwen3-MoE and shape-mismatch loading in Qwen3-VL are different model/config issues." }, { "left": "issue:43526", "right": "issue:44857", "accept": false, - "reason": "Image processor label reduction bug is unrelated to LwDetrImageLoss AMP/CUDA crash." + "reason": "Label reduction in an image processor and AMP/CUDA crash in LwDetrImageLoss are unrelated." }, { - "left": "issue:46121", - "right": "issue:46399", + "left": "issue:43334", + "right": "issue:43531", "accept": false, - "reason": "RoPE validation TypeError and Qwen3.5 kernel registration ValueError are different issues." + "reason": "Qwen3-VL missing pad_token_id and Qwen3-MoE sliding_window behavior are distinct model-family regressions." }, { - "left": "issue:43334", - "right": "issue:43531", + "left": "issue:46121", + "right": "issue:46399", "accept": false, - "reason": "Missing pad_token_id on Qwen3-VL load is unrelated to sliding_window behavior in Qwen3-MoE." + "reason": "Rope-validation TypeError from JSON-loaded lists and a missing-decorator crash in Qwen3.5 kernels are different code paths." }, { "left": "issue:43761", "right": "issue:45859", "accept": false, - "reason": "CLIPVisionModel hidden_states regression vs Qwen3.5 MoE CPU-offload incompatibility; not the same bug." + "reason": "CLIPVisionModel hidden_states regression and Qwen3_5MoeTextRotaryEmbedding CPU-offload incompatibility are unrelated." }, { "left": "issue:27957", "right": "issue:29127", "accept": false, - "reason": "FlashAttention 2 support request for XLM-Roberta is unrelated to LayoutLMv3 error handling." + "reason": "Flash Attention 2 support for XLMRoberta and LayoutLMv3 error-message clarity are unrelated issues." }, { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Flash-attention generation bug and category-cardinality RuntimeError are unrelated." + "reason": "Qwen3.5 flash-attention generation bug and a categorical-limit runtime error are different failures." }, { "left": "issue:27957", "right": "issue:33290", "accept": false, - "reason": "XLM-Roberta Flash Attention 2 request does not match an Adafactor+DeepSpeed OOM." + "reason": "XLMRoberta Flash Attention 2 support and DeepSpeed Adafactor OOM do not share the same underlying bug." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Audio chat-template sampling-rate default is unrelated to LasrFeatureExtractor's unsupported center arg." + "reason": "Sampling-rate defaulting in audio chat templates and an unsupported `center` arg in LasrFeatureExtractor are separate audio bugs." }, { - "left": "issue:45588", - "right": "issue:45600", + "left": "issue:33290", + "right": "issue:34689", "accept": false, - "reason": "Flash-attention sink-less AttributeError and stale auto_mappings config references are different failures." + "reason": "AdaFactor/DeepSpeed OOM and Llama 3.2 Vision model-loading breakage are different failures despite both being resource/loading related." }, { - "left": "issue:33290", - "right": "issue:34689", + "left": "issue:45588", + "right": "issue:45600", "accept": false, - "reason": "Adafactor/DeepSpeed OOM is not the same as Llama 3.2 Vision model-loading regression." + "reason": "A flash-attention runtime AttributeError and stale auto_mappings references are unrelated; one is execution-time, the other is CI/config drift." }, { "left": "issue:44464", "right": "issue:45491", "accept": false, - "reason": "Chunked generation inconsistency under compiled forward is unrelated to Gemma3 sliding-window NaNs." + "reason": "Compiled-forward chunked generation inconsistency and Gemma3 NaNs from sliding-window/all-padding windows are different bugs." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Both involve initialization semantics, but one is a general v5 custom-model break and the other is a specific model's optional post_init policy." + "reason": "Custom model initialization breakage in Transformers 5.0.0 and patchtsmixer's optional post_init policy are not the same code-path problem." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "Training OOM in deepspeed/Adafactor is unrelated to embedding reinitialization after resize." + "reason": "DeepSpeed OOM with Adafactor is unrelated to output-embedding reinitialization after resizing token embeddings." }, { "left": "issue:33290", "right": "issue:34567", "accept": false, - "reason": "OOM during training and num_input_tokens_seen not updating are distinct issues." + "reason": "An optimizer memory issue and TrainerState token-count tracking bug are unrelated." }, { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Eval-mode gradient checkpointing restriction is unrelated to quantization offloading behavior." + "reason": "Gradient checkpointing in eval mode and quantization offloading failures are different training/runtime behaviors." }, { "left": "issue:44743", "right": "issue:44811", "accept": false, - "reason": "Qwen3.5 recurrent-state reset and Whisper batch_decode skip_special_tokens are unrelated." + "reason": "Qwen3.5 recurrent-state reset and Whisper batch_decode skip_special_tokens handling are unrelated model/processor bugs." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Tokenizer loading failure and Perceiver image-classification resolution bug are different subsystems." + "reason": "Tokenizer loading failure and Perceiver interpolation failure are different subsystems and failure modes." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "sys.modules access issue in modeling_utils is unrelated to DeepSeekV3 implementation divergence." - }, - { - "left": "issue:42490", - "right": "issue:44743", - "accept": false, - "reason": "Local-path import warning has no relation to Qwen3.5 recurrent-state cache behavior." + "reason": "Unsafe sys.modules access in modeling_utils and DeepseekV3MoE divergence from remote implementation are unrelated." }, { "left": "issue:44062", "right": "issue:45907", "accept": false, - "reason": "AddedToken constructor conflict and list_repo_files revision forwarding are unrelated." + "reason": "AddedToken keyword conflict and list_repo_files revision forwarding are separate API bugs." }, { "left": "issue:34689", "right": "issue:43611", "accept": false, - "reason": "These are both model-loading regressions, but they affect different releases and code paths; not the same concrete fix." + "reason": "Both are loading regressions, but they concern different release versions and distinct loading mechanisms; not the same bug." } ] }, @@ -9820,9 +10030,9 @@ "issue:43720|issue:44442", "issue:43761|issue:44079", "issue:44077|issue:44479", - "issue:43232|issue:43388", "issue:43761|issue:44188", "issue:43493|issue:43827", + "issue:43232|issue:43388", "issue:44987|issue:46178", "issue:42907|issue:45698", "issue:44442|issue:45803", @@ -9830,8 +10040,8 @@ "issue:43931|issue:45216", "issue:43316|issue:45759", "issue:43749|issue:43931", - "issue:43701|issue:46506", "issue:44855|issue:45137", + "issue:43701|issue:46506", "issue:43643|issue:44913", "issue:44464|issue:46421", "issue:44190|issue:44496", @@ -9842,196 +10052,196 @@ "issue:42617|issue:43575", "issue:43606|issue:43701" ], - "prepared_review_unit_hash": "a2af47fc3e0de21b02a73ef64e88eb6752762da4fd50458a65a2fccbf5161b8c", + "prepared_review_unit_hash": "ce99dc6583b0835a8afa407e4de0eb78c4b2b894342f78a3e4626156cabad617", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 27, - "serialized_chars": 35394, - "estimated_input_tokens": 8849, - "estimated_eval_tokens": 17954 + "serialized_chars": 35399, + "estimated_input_tokens": 8850, + "estimated_eval_tokens": 17956 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: it mixes unrelated config, loading, generation, tokenizer, and backend/dependency issues. None of the soft pairs looks like a true duplicate of the same underlying bug/change, so they should stay separate.", - "confidence": 0.95, - "canonical_issue_reason": "No strong duplicate center exists; if one issue must anchor the cluster, issue 45698 is the broadest save/load serialization report and has the most discussion/inbound references, but it is still not a true duplicate of the others.", + "summary": "This set is highly heterogeneous; the soft links mostly connect superficially similar issues rather than true duplicates, so no PRs are present and no soft edge should be merged.", + "confidence": 0.69, + "canonical_issue_reason": "issue:43701 is the most central node in the soft-link graph and is a concrete, reproducible bug, but it only serves as a weak anchor because the surrounding issues span unrelated failure modes.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45698 is the most broadly framed and actively discussed issue in the set, so it is the least-bad representative despite the cluster being mixed.", + "best_issue_reason": "issue:43701 is the best single representative for triage due to its higher connectivity and clear bug report, though it is not a true duplicate of the other items.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43316", "right": "issue:43525", "accept": false, - "reason": "Both mention config/API mismatches, but one is Gemma3TextConfig field parity and the other is Llama4Config missing pad_token_id; different models and fixes." + "reason": "Both are config/API mismatches, but they concern different models and different missing fields; not the same underlying bug." }, { "left": "issue:42175", "right": "issue:46234", "accept": false, - "reason": "Different backend/import problems: one is torch install packaging not including TensorFlow, the other is a missing import-utils symbol for falcon mamba availability." + "reason": "One is about backend installation extras, the other about a missing import-utils symbol; different code paths and fixes." }, { "left": "issue:43723", "right": "issue:45479", "accept": false, - "reason": "Tokenizer loading in v5 vs degenerate zero-loss sequence-classification behavior; unrelated code paths." + "reason": "Tokenizer loading in v5 and degenerate classification loss are unrelated failures in different layers." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "BitNet accelerate-loading of packed weights is unrelated to AutoTokenizer failing on a FastSpeech2Conformer tokenizer." + "reason": "Accelerate packed-weight loading vs. tokenizer loading; no shared code-path bug." }, { "left": "issue:43761", "right": "issue:44079", "accept": false, - "reason": "One is CLIPVisionModel not returning hidden_states, the other is generic ModelOutput key assignment after None; related theme but not the same concrete bug." + "reason": "A model forward missing hidden states is distinct from ModelOutput key assignment behavior." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "PatchTSMixer post_init API validation and Qwen video-input regression are different model-specific problems." - }, - { - "left": "issue:43232", - "right": "issue:43388", - "accept": false, - "reason": "Generation kwargs after sync_gpus and metric gathering of tuple labels are separate training/eval issues." + "reason": "Optional post_init enforcement in one model has nothing to do with the Qwen video-input regression." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "Hidden-states forwarding bug vs attention-kernel divergence under torch.compile; different failure modes and fixes." + "reason": "Hidden-states output regression and torch.compile attention-kernel divergence are different problems." }, { "left": "issue:43493", "right": "issue:43827", "accept": false, - "reason": "SigLIP2 implementation parity is unrelated to stale pipeline references in docs after v5 removals." + "reason": "SigLIP2 implementation parity and stale docs referencing pipeline() are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs syncing and gather_for_metrics label dropping affect different subsystems and behaviors." }, { "left": "issue:44987", "right": "issue:46178", "accept": false, - "reason": "A model load regression for physical-intelligence/fast is not the same as a missing tensorflow_text backend mapping." + "reason": "Model loading failure and missing tensorflow_text backend mapping are separate dependency/import issues." }, { "left": "issue:42907", "right": "issue:45698", "accept": false, - "reason": "Both involve save/load, but one is failing to save dequantized models while the other loads the wrong custom module after save_pretrained; distinct bugs." + "reason": "One concerns dequantized model saving, the other custom-module resolution after save/load; similar theme but different bugs." }, { "left": "issue:44442", "right": "issue:45803", "accept": false, - "reason": "Tokenizer class resolution failure is unrelated to Fuyu tensor conversion swallowing KeyboardInterrupt/errors." + "reason": "Tokenizer loading and a broad exception in Fuyu tensor conversion are unrelated." }, { "left": "issue:43873", "right": "issue:44871", "accept": false, - "reason": "Quantized offloading behavior and Gemma-3 eos_token_id config inconsistency do not share the same code path." + "reason": "Quantization/offloading behavior is not the same as eos_token_id configuration mismatch." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Weight-shape mismatch on model load versus incorrect checkpoint saving after save_pretrained; related family but different bug surfaces." + "reason": "Weight-shape mismatch on load differs from incorrect checkpoint serialization; similar model family but distinct failures." }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Gemma3TextConfig API discrepancy is separate from AutoModelForCausalLM.from_config not unwrapping text_config for composite Qwen multimodal configs." + "reason": "Config API discrepancy and multimodal text_config unwrapping are different configuration bugs." }, { "left": "issue:43749", "right": "issue:43931", "accept": false, - "reason": "FSDP CPU RAM efficient loading and Qwen3-VL weight-shape mismatch are both loading issues but not the same underlying defect." + "reason": "FSDP CPU RAM efficient loading and a specific model weight mismatch are not the same defect." }, { - "left": "issue:43701", - "right": "issue:46506", + "left": "issue:44855", + "right": "issue:45137", "accept": false, - "reason": "resume_from_checkpoint key mismatch and PPDocLayoutV3 CPU/CUDA output differences are unrelated." + "reason": "A Python 3.13 TorchScript parsing issue is unrelated to a ZeRO3 deque underflow." }, { - "left": "issue:44855", - "right": "issue:45137", + "left": "issue:43701", + "right": "issue:46506", "accept": false, - "reason": "Python 3.13 import/torch.jit parsing error is unrelated to DeepSpeed ZeRO3 deque underflow." + "reason": "Resume-checkpoint key mismatch and CPU/CUDA result divergence are different bug classes with no clear common fix." }, { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "Missing fields when trust_remote_code=True and rotary_pct resetting on reload are both config persistence problems, but they affect different mechanisms and fixes." + "reason": "Missing fields from AutoConfig with trust_remote_code and GPTNeoX rotary_pct reload behavior are separate config-loading issues." }, { "left": "issue:44464", "right": "issue:46421", "accept": false, - "reason": "Chunked generation inconsistency with compiled forward is distinct from cache lazy initialization causing recompilation." + "reason": "Both mention chunked generation/prefill, but one is inconsistent outputs and the other is spurious recompilation; not the same code-path bug." }, { "left": "issue:44190", "right": "issue:44496", "accept": false, - "reason": "A local dataset loading script bug is unrelated to an unrecognized model_type in config.json." + "reason": "Local dataset loading in a training script and unrecognized model config are unrelated." }, { "left": "issue:46234", "right": "issue:46291", "accept": false, - "reason": "Both touch import utilities, but they concern different missing dependencies/symbols and different package entry points." + "reason": "A missing utility function and an import failure due to kernels version constraints are different import-time problems." }, { "left": "issue:42796", "right": "issue:45290", "accept": false, - "reason": "Tokenizer BPE-marker handling and chat-template tool-call crashing are different tokenizer/chat-template bugs." + "reason": "Tokenizer BPE marker handling and chat-template tool-call crashes are distinct tokenizer behaviors." }, { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "torch.compile + SDPA failure in Bamba is unrelated to resume_from_checkpoint key mismatch." + "reason": "torch.compile SDPA failure in Bamba is unrelated to checkpoint resume key mismatch." }, { "left": "issue:25251", "right": "issue:28282", "accept": false, - "reason": "Pipeline top_k nesting behavior and AutoModel PyTorch ImportError are unrelated." + "reason": "Pipeline top_k nesting behavior and an AutoModel PyTorch import error are completely unrelated." }, { "left": "issue:42617", "right": "issue:43575", "accept": false, - "reason": "3d_parallel.py execution failure and Qwen2 TP out-of-memory are both parallelism-related but not the same bug." + "reason": "3d_parallel.py runtime failure and model-load OOM under tensor parallelism are different issues." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device mismatch and resume_from_checkpoint key mismatch are unrelated." + "reason": "CPU-offload device mismatch and resume_from_checkpoint key mismatch do not share the same underlying bug." } ] }, @@ -10119,10 +10329,10 @@ "issue:44162|issue:44792", "issue:43646|issue:44913", "issue:44871|issue:46050", - "issue:45803|issue:46170", "issue:43650|issue:46202", - "issue:43065|issue:44617", + "issue:45803|issue:46170", "issue:41950|issue:43475", + "issue:43065|issue:44617", "issue:43540|issue:44514", "issue:43452|issue:44496", "issue:43404|issue:43441", @@ -10131,14 +10341,14 @@ "issue:45127|issue:45200", "issue:43572|issue:46178" ], - "prepared_review_unit_hash": "be95edc6e06ed7bac4f61c718c2237504a832aa7526f1f6051b501210fa684c8", + "prepared_review_unit_hash": "485c92a4353b26cb6d4309b1847f07053a8761787873ad251268f031ad779d31", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 47, @@ -10153,168 +10363,168 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are mostly unrelated standalone issues, so there is no strong single duplicate cluster. The closest theme is general v5/v5.3 regression and API compatibility breakage, but the soft pairs do not look like the same underlying bugs.", - "confidence": 0.33, - "canonical_issue_reason": "No clear universal canonical exists across these heterogeneous issues; if one must be chosen, issue 43295 is the broadest representative because it captures a backward-compatibility regression affecting existing custom model/processor code.", + "summary": "This cluster is a set of soft-similarity false positives rather than true duplicates: the paired issues cover unrelated models, pipelines, config regressions, and import/runtime failures. No PRs are present, and no single issue is a clean canonical target.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the soft pairs are mostly unrelated bugs across different models and subsystems, so keeping one representative would be misleading.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43295 is the best umbrella representative for the overall set because it is a concrete, user-facing regression with broad impact, while the rest are narrower or unrelated model-specific failures.", + "best_issue_reason": "No global best issue: there is no single issue that plausibly subsumes the rest of this mixed cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43295", "right": "issue:43825", "accept": false, - "reason": "Both are v5 regressions, but one is a processor/tokenizer API break and the other is an error-message/task-support issue; different bugs and fixes." + "reason": "Different regressions: processor.tokenizer/image handling vs an error-message mismatch about translation tasks." }, { "left": "issue:43575", "right": "issue:43742", "accept": false, - "reason": "OOM while loading a large Qwen model with tensor parallelism is unrelated to a key error loading MobileLLM-125M." + "reason": "Different loading failures for different models; one is TP/OOM, the other is a key error on MobileLLM." }, { "left": "issue:27957", "right": "issue:32090", "accept": false, - "reason": "FlashAttention support for XLMRoberta and a Trainer `_gpu_broadcast_one` NoneType error are separate code paths." + "reason": "Unrelated problems: FlashAttention support for XLMRoberta vs Trainer broadcast failing on NoneType." }, { "left": "issue:16998", "right": "issue:22355", "accept": false, - "reason": "A model_max_length question for DeBERTa-V3 is unrelated to a missing `transformers.onnx` module import error." + "reason": "A model max_length question is unrelated to a missing transformers.onnx module import error." }, { "left": "issue:43316", "right": "issue:43716", "accept": false, - "reason": "Gemma3Config API discrepancy and Mistral-3 image preprocessor dtype mismatch are different model-family bugs." + "reason": "Different code paths and symptoms: Gemma3TextConfig API mismatch vs image preprocessor/model dtype mismatch." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "Processor/tokenizer regression and Gemma eos_token_id inconsistency are distinct configuration/API issues." + "reason": "Both are regressions, but they affect different APIs and model families; not the same underlying bug." }, { "left": "issue:42175", "right": "issue:43316", "accept": false, - "reason": "Missing TensorFlow in a Torch install extra is unrelated to a Gemma3TextConfig API mismatch." + "reason": "Backend packaging issue for torch extras is unrelated to a Gemma config API discrepancy." }, { "left": "issue:44871", "right": "issue:45584", "accept": false, - "reason": "Gemma eos_token_id mismatch and Whisper empty-transcription failure after special-token alignment are different bugs." + "reason": "Gemma eos_token_id inconsistency and Whisper empty-transcription failure are different bugs." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "Optional post_init allowance in patchtsmixer and non-persistent buffer junk in v5 are unrelated regressions." + "reason": "One is about disallowing optional post_init on a specific model; the other is a buffer initialization regression." }, { "left": "issue:45632", "right": "issue:46424", "accept": false, - "reason": "Trust-remote-code cache-path collision and max_cache_length handling for quantized cache are separate cache features." + "reason": "Both mention cache, but one is a local trust_remote_code path collision and the other is max_cache_length semantics." }, { "left": "issue:45803", "right": "issue:46365", "accept": false, - "reason": "A bare except swallowing errors in Fuyu conversion is unrelated to importing Trainer failing with datasets installed." + "reason": "Bare except swallowing errors is unrelated to a Trainer import conflict with datasets." }, { "left": "issue:44610", "right": "issue:46369", "accept": false, - "reason": "OmDet-Turbo input-size mismatch and sam3-litetext memory overflow are not the same issue." + "reason": "Different models and failures: OmDet-Turbo image size mismatch vs Sam3-LiteText memory overflow." }, { "left": "issue:44162", "right": "issue:44792", "accept": false, - "reason": "ESM2 breakage and a janus image-generation test failure affect different models and failure modes." + "reason": "ESM2 breakage is unrelated to a Janus image generation test failure." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Custom model initialization regression and GPTNeoX rotary_pct not persisting on reload are different configuration problems." + "reason": "A broad Transformers 5 initialization regression is not the same as GPTNeoX rotary_pct not persisting on reload." }, { "left": "issue:44871", "right": "issue:46050", "accept": false, - "reason": "Gemma eos_token_id inconsistency and sinq quantization save/load corruption are unrelated." - }, - { - "left": "issue:45803", - "right": "issue:46170", - "accept": false, - "reason": "Fuyu's exception handling bug and a Windows ROCm torch.distributed import/type error do not share a concrete root cause." + "reason": "Gemma eos_token_id config mismatch and sinq save/load quantization breakage are different issues." }, { "left": "issue:43650", "right": "issue:46202", "accept": false, - "reason": "The first is a non-informative issue body; the second is a specific Gemma pipeline roundtrip failure." + "reason": "The placeholder issue title is unrelated to a Gemma pipeline roundtrip failure." }, { - "left": "issue:43065", - "right": "issue:44617", + "left": "issue:45803", + "right": "issue:46170", "accept": false, - "reason": "A dummy Conv2d in Sam3PixelDecoder and Sam3Video CUDA OOM are related only by model family, not by the same bug." + "reason": "Exception handling in Fuyu conversion is unrelated to the Windows/ROCm torch.distributed import problem." }, { "left": "issue:41950", "right": "issue:43475", "accept": false, - "reason": "Video-classification pipeline processor lookup and a missing `fpn_position_embeddings` attribute are different SAM3/video issues." + "reason": "Both are video-adjacent, but one is a pipeline processor lookup bug and the other is a missing output attribute in Sam3Video." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Same family name, but one is a dummy Conv2d implementation detail and the other is a CUDA OOM symptom." }, { "left": "issue:43540", "right": "issue:44514", "accept": false, - "reason": "Both involve Qwen multimodal/video processing, but one is a video-input validation error in Qwen3OmniMoe and the other is a batched chat-template padding crash in Qwen2_5_VLProcessor." + "reason": "Different Qwen video/chat-template failures with different inputs and failure modes." }, { "left": "issue:43452", "right": "issue:44496", "accept": false, - "reason": "Broken `gguf_file` loading and an unrecognized model missing `model_type` are separate model-loading failures." + "reason": "Tokenizer/model loading with gguf_file is unrelated to an unrecognized model_type in config.json." }, { "left": "issue:43404", "right": "issue:43441", "accept": false, - "reason": "LM head weight tying in Mistral-3 and FlashAttention failure in Ministral-3 are different code paths." + "reason": "Both are Mistral-family, but one is tied weights and the other is FlashAttention compatibility; distinct bugs." }, { "left": "issue:43761", "right": "issue:44479", "accept": false, - "reason": "CLIPVisionModel hidden_states regression and Qwen VL video-input regression are unrelated." + "reason": "CLIPVision hidden_states regression is unrelated to the Qwen video-input regression." }, { "left": "issue:43388", "right": "issue:43688", "accept": false, - "reason": "A metrics-gathering bug with tuple labels and auxiliary-loss normalization are different trainer/model-training issues." + "reason": "Metric gathering dropping labels is unrelated to auxiliary-loss normalization." }, { "left": "issue:45127", "right": "issue:45200", "accept": false, - "reason": "LoRA merge collapse with extended vocab and Gemma 4 mm_token_type_ids defaults are not the same defect." + "reason": "LoRA/vocab-merge collapse and Gemma4 mm_token_type_ids defaulting are different model behaviors." }, { "left": "issue:43572", "right": "issue:46178", "accept": false, - "reason": "StableLmConfig missing pad_token_idx and a backend-mapping error for tensorflow_text are unrelated configuration/import issues." + "reason": "StableLm pad_token_idx config and tensorflow_text backend registration are unrelated." } ] }, @@ -10389,8 +10599,8 @@ ], "soft_pairs": [ "issue:43673|issue:44743", - "issue:43504|issue:45698", "issue:45542|issue:46234", + "issue:43504|issue:45698", "issue:43525|issue:43531", "issue:44792|issue:44977", "issue:44734|issue:45406", @@ -10399,237 +10609,234 @@ "issue:44534|issue:44913", "issue:44190|issue:44964", "issue:43673|issue:46421", - "issue:45381|issue:46369", "issue:43931|issue:44060", - "issue:43761|issue:44242", + "issue:45381|issue:46369", "issue:35141|issue:42490", + "issue:43761|issue:44242", "issue:45563|issue:45584", "issue:44855|issue:45588", "issue:43827|issue:43994", "issue:44479|issue:44734", "issue:43065|issue:44610", - "issue:41720|issue:45245", "issue:43644|issue:44913", + "issue:41720|issue:45245", "issue:44933|issue:45362", - "issue:44464|issue:45563", "issue:43976|issue:44855", + "issue:44464|issue:45563", "issue:43550|issue:44190", "issue:44617|issue:46369", - "issue:44077|issue:44877", - "issue:25251|issue:27957" + "issue:25251|issue:27957", + "issue:44077|issue:44877" ], - "prepared_review_unit_hash": "b1078b719101eccb60cd9b4acd33587738fbdfab622f410380ea17b46f85b691", + "prepared_review_unit_hash": "220fc7a238e7af3b81be61df4d16c7f351a728805d6fa1279dd8f30d9bd2c09a", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36211, - "estimated_input_tokens": 9053, - "estimated_eval_tokens": 18362 + "serialized_chars": 36200, + "estimated_input_tokens": 9050, + "estimated_eval_tokens": 18356 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated issue reports that only share broad Transformers/v5 or model-family themes. The one likely duplicate pair is the chunked-prefill cache regression pair; the rest should stay separate.", - "confidence": 0.91, - "canonical_issue_reason": "issue:46421 is the best canonical issue: it describes the underlying chunked-prefill cache initialization problem directly, and it is the most current open report in that line of bugs.", + "summary": "These items are mostly heterogeneous issue reports across different bugs and subsystems. All supplied soft pairs look like false positives; there is no pair here that clearly describes the same underlying bug/change. The only obvious duplicate theme in the item list is the v5 non-persistent-buffer regression, but that duplicate pair was not provided as a soft edge.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 43644 is the best canonical anchor among this set because it describes a concrete, broad regression and is the clearest representative of the only obvious duplicate theme in the list.", "canonical_pr_reason": null, - "best_issue_reason": "issue:46421 is the strongest overall representative because it names the code path and failure mode most concretely (lazy cache init with chunked prefill), making it the best target for deduping follow-up reports.", + "best_issue_reason": "Issue 43644 is the strongest overall representative: specific enough to identify the bug, yet broad enough to stand in for the related v5 buffer-corruption reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43673", "right": "issue:44743", "accept": false, - "reason": "Both mention cache/recurrent-state behavior, but one is a v5.0.0 chunked_prefill cache-missing bug and the other is a Qwen3 modular recurrent-state reset bug; different models and different failure modes." + "reason": "Both mention cache-related generation behavior, but one is a missing GenerationMixin cache in chunked prefill and the other is Qwen3.5 recurrent-state reset; different bugs and code paths." }, { - "left": "issue:43504", - "right": "issue:45698", + "left": "issue:45542", + "right": "issue:46234", "accept": false, - "reason": "Different problems: Beit preset loading breaks on a legacy field, while from_pretrained/save_pretrained loads the wrong custom module." + "reason": "Completely different problems: a TensorBoard/TensorFlow backend import issue versus a missing import_utils helper for falcon-mamba availability." }, { - "left": "issue:45542", - "right": "issue:46234", + "left": "issue:43504", + "right": "issue:45698", "accept": false, - "reason": "Unrelated packaging/import issues: one is a TensorBoard-without-TensorFlow backend error, the other is a missing import utility symbol." + "reason": "Unrelated failures: BEiT semantic segmentation legacy-field loading versus from_pretrained selecting the wrong custom module after save_pretrained." }, { "left": "issue:43525", "right": "issue:43531", "accept": false, - "reason": "Both involve Qwen/Llama config objects, but one is a missing pad_token_id attribute and the other is a sliding_window behavior issue in Qwen3-MoE." + "reason": "Different model bugs: missing pad_token_id on Llama4Config versus a sliding_window issue in Qwen3-MoE." }, { "left": "issue:44792", "right": "issue:44977", "accept": false, - "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation issues are different model paths and likely different bugs." + "reason": "Both are model-generation failures, but on different models and different mechanisms: Janus test failure versus Qwen3.5 flash-attention behavior." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "Both are runtime crashes in serve/chat-like paths, but one is KV-cache continuation tensor indexing and the other is a Gemma4Processor missing _tokenizer attribute." + "reason": "Serve KV-cache tensor indexing crash and Gemma4Processor missing _tokenizer are unrelated code paths and symptoms." }, { "left": "issue:44617", "right": "issue:45584", "accept": false, - "reason": "Sam3Video OOM is a memory issue; Whisper empty transcription after align_special_tokens is a generation edge-case bug." + "reason": "CUDA OOM in Sam3Video versus Whisper empty-transcription failure; different models and failure modes." }, { "left": "issue:44792", "right": "issue:45600", "accept": false, - "reason": "A failing janus image-generation test is not the same as stale auto_mappings references to removed Sam3LiteText configs." + "reason": "A Janus generate-images test failure and a CI break from removed Sam3LiteText configs are not the same bug." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "One is non-persistent buffers being junk-filled in v5; the other is GPTNeoX rotary_pct not persisting on reload." + "reason": "One is the v5 non-persistent-buffer corruption regression; the other is GPTNeoXConfig rotary_pct not persisting on reload." }, { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "A local dataset loading issue in a training script is unrelated to a multimodal model loading error for Phi-4." + "reason": "Local dataset loading in a training script is unrelated to model-loading failure for microsoft/Phi-4-multimodal-instruct." }, { "left": "issue:43673", "right": "issue:46421", - "accept": true, - "reason": "Both describe the same chunked-prefill cache initialization regression in v5: one as a missing GenerationMixin cache and the other as lazy cache init causing spurious recompilation." - }, - { - "left": "issue:45381", - "right": "issue:46369", "accept": false, - "reason": "Both are video/model-specific regressions, but one is Qwen2.5-VL video position IDs and the other is memory overflow in sam3-litetext." + "reason": "Both involve chunked prefill/cache, but one is missing cache in v5.0.0 and the other is lazy cache init causing recompilation; distinct defects." }, { "left": "issue:43931", "right": "issue:44060", "accept": false, - "reason": "Different Qwen3-VL vs Qwen3-Next loading/tied-weights issues; not the same code-path bug." + "reason": "Qwen3-VL weight-shape mismatch and Qwen3-Next tied-weights warning are separate model/configuration issues." }, { - "left": "issue:43761", - "right": "issue:44242", + "left": "issue:45381", + "right": "issue:46369", "accept": false, - "reason": "CLIPVision hidden_states regression and MoE load-balancing loss omission are separate training/inference bugs." + "reason": "Qwen2.5-VL video position IDs regression and Sam3-LiteText memory overflow are different models and problems." }, { "left": "issue:35141", "right": "issue:42490", "accept": false, - "reason": "Resizing token embeddings reinitialization and trailing-slash local model import warnings are unrelated loader behaviors." + "reason": "Embedding resize reinitialization and local-path trailing-slash import warning are unrelated." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "CLIPVision hidden_states regression and missing load-balancing loss when output_router_logits=False affect different components." }, { "left": "issue:45563", "right": "issue:45584", "accept": false, - "reason": "A stale warning for paged generate num_return_sequences is not the same as Whisper failing on empty transcription after token alignment." + "reason": "A stale warning in paged generate versus Whisper failing on empty transcription after align_special_tokens; different bug surfaces." }, { "left": "issue:44855", "right": "issue:45588", "accept": false, - "reason": "Python 3.13 TorchScript parsing/IndentationError is unrelated to flash_attention.py crashing on s_aux=None." + "reason": "Python 3.13 IndentationError in a scripted model file and a flash_attention AttributeError on s_aux=None are unrelated." }, { "left": "issue:43827", "right": "issue:43994", "accept": false, - "reason": "Docs still referencing pipeline() after removals is not the same as a model producing nonsensical outputs with AutoModel/pipeline." + "reason": "Docs still referencing pipeline() and SigLIP2 nonsensical outputs with AutoModel/pipeline are not the same issue." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "A v5.3.0 video-input regression across Qwen VL variants is broader and different from a serve-side KV cache continuation indexing crash." + "reason": "Video-input regression across Qwen VL variants and a serve KV-cache indexing crash are different code paths." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d and OmDet-Turbo processor input-size mismatch are different model/processor configuration problems." + "reason": "Dummy Conv2d in Sam3PixelDecoder versus OmDet-Turbo processor output-size mismatch; not the same bug." }, { - "left": "issue:41720", - "right": "issue:45245", + "left": "issue:43644", + "right": "issue:44913", "accept": false, - "reason": "Qwen3 auto device mapping cudaErrorAssert and the 2^24 categories limit are unrelated runtime errors." + "reason": "The v5 non-persistent-buffer junk issue is unrelated to GPTNeoX rotary_pct not surviving reload." }, { - "left": "issue:43644", - "right": "issue:44913", + "left": "issue:41720", + "right": "issue:45245", "accept": false, - "reason": "Junk-filled non-persistent buffers and GPTNeoX rotary_pct resetting on reload are different persistence bugs." + "reason": "Qwen3 auto device mapping cuda assert and the 2^24 category limit error are unrelated runtime failures." }, { "left": "issue:44933", "right": "issue:45362", "accept": false, - "reason": "Missing import from image_utils and Qwen3.5-35B chat crashes are unrelated." + "reason": "A missing import from image_utils and Qwen3.5-35B crashing in transformers chat are unrelated." }, { - "left": "issue:44464", - "right": "issue:45563", + "left": "issue:43976", + "right": "issue:44855", "accept": false, - "reason": "Chunked generation inconsistencies with compiled forward are not the same as a stale warning about num_return_sequences in paged generate." + "reason": "A Python-version support regression and a Python 3.13 parsing/IndentationError in a scripted function are different problems." }, { - "left": "issue:43976", - "right": "issue:44855", + "left": "issue:44464", + "right": "issue:45563", "accept": false, - "reason": "Python-version compatibility for Transformers 5.1.0 is unrelated to a TorchScript decorator/comment parsing error." + "reason": "Compiled-forward chunked generation inconsistency and a stale num_return_sequences warning do not share the same underlying bug." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "Bamba torch.compile/SDPA failure is unrelated to loading a local dataset in a training script." + "reason": "Bamba-9B-v2 compile/SDPA failure and local dataset loading failure in an example script are unrelated." }, { "left": "issue:44617", "right": "issue:46369", "accept": false, - "reason": "Both are Sam3-related but one is OOM in Sam3Video and the other is memory overflow in sam3-litetext; different models and likely different fixes." + "reason": "Both are memory-related, but one is Sam3Video OOM and the other is Sam3-LiteText memory overflow; different models and causes." }, { - "left": "issue:44077", - "right": "issue:44877", + "left": "issue:25251", + "right": "issue:27957", "accept": false, - "reason": "PatchTSMixer post_init policy and strict config blocking granite_speech are separate configuration-policy issues." + "reason": "Top-k nested-list output in a pipeline and XLMRoberta Flash Attention 2 support are unrelated." }, { - "left": "issue:25251", - "right": "issue:27957", + "left": "issue:44077", + "right": "issue:44877", "accept": false, - "reason": "Pipeline top_k nesting behavior and XLMRoberta Flash Attention 2 support are unrelated issues." + "reason": "An optional post_init policy issue for patchtsmixer and a strict-config loading failure for granite_speech are separate configuration bugs." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: issue:46421 and issue:43673 are clearly the closest duplicate pair by title/content, and the rest of the soft-edge rejections are conservative and supported by the issue titles. I don\u2019t see overstatement or an unsafe merge of unrelated reports." - }, - "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 }, @@ -10649,6 +10856,7 @@ "issue:36246", "issue:38617", "issue:41628", + "issue:41950", "issue:42175", "issue:42796", "issue:43116", @@ -10658,8 +10866,8 @@ "issue:43540", "issue:43643", "issue:43854", + "issue:44016", "issue:44112", - "issue:44186", "issue:44190", "issue:44246", "issue:44279", @@ -10706,220 +10914,227 @@ "issue:44938|issue:45542", "issue:43295|issue:45276", "issue:44462|issue:45907", - "issue:45276|issue:45584", "issue:45632|issue:46234", + "issue:45276|issue:45584", "issue:42796|issue:44521", "issue:36246|issue:41628", - "issue:45245|issue:45584", - "issue:46082|issue:46121", "issue:36246|issue:38617", + "issue:46082|issue:46121", + "issue:45245|issue:45584", "issue:44246|issue:45003", "issue:42175|issue:44279", - "issue:44683|issue:45446", "issue:43643|issue:44496", "issue:44295|issue:46506", + "issue:44683|issue:45446", "issue:44964|issue:45600", - "issue:44186|issue:44734", + "issue:44112|issue:44964", + "issue:43854|issue:44964", "issue:43404|issue:45216", - "issue:45632|issue:46421", - "issue:43854|issue:44964" + "issue:41950|issue:44016", + "issue:45632|issue:46421" ], - "prepared_review_unit_hash": "5404eb41e97a61ef7f90d1577f8121086a8195cdd51dad434d1de21605f0a629", + "prepared_review_unit_hash": "25f4afec95aea52d861ba402e2e70b40079134177296c7aaf4bc58ec5cbd3dfa", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 28, - "serialized_chars": 35782, - "estimated_input_tokens": 8946, - "estimated_eval_tokens": 18148 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36642, + "estimated_input_tokens": 9161, + "estimated_eval_tokens": 18578 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "All soft-pair candidates look like false positives: they touch different code paths, different model families, or different failure modes. This cluster is heterogeneous rather than a true duplicate set.", - "confidence": 0.97, - "canonical_issue_reason": "No single canonical issue stands out because the items span unrelated import, tokenizer, config, CI, model-loading, and runtime bugs.", + "summary": "This cluster is heterogeneous: mostly unrelated Transformers issues spanning model loading, tokenizers, multimodal processing, config validation, CI, and import/runtime bugs. I don\u2019t see any pair that clearly shares the same concrete underlying defect, so all soft edges should be rejected. If a single issue must be picked as the closest hub, the multimodal chat-template bug is the most central, but it is not a true duplicate anchor.", + "confidence": 0.84, + "canonical_issue_reason": "issue:44521 is the closest representative because it is a broad multimodal-input bug with multiple inbound references and sits near several other multimodal/processor-themed reports, but the cluster as a whole is not a single duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 is the most active/open and has the most references, but it still represents only one of many unrelated bugs in this cluster.", + "best_issue_reason": "issue:44521 has the strongest \u201chub\u201d characteristics in this set: open, well-discussed, and centered on multimodal processing behavior that overlaps thematically with several neighboring reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45127", "right": "issue:45357", "accept": false, - "reason": "Different Qwen bugs: LoRA/extended-vocab collapse vs incorrect visual encoder keys in save_pretrained." + "reason": "Both involve Qwen3/Qwen2.5-family model behavior, but one is LoRA merge/tied embeddings collapse and the other is save_pretrained checkpoint key serialization; different code paths." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "Unrelated paths: SwitchTransformers sparse-layer config bug vs image-classification example dataset loading." + "reason": "Sparse-layer config creation and local dataset loading in an example script are unrelated bugs." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "Both multimodal, but one is assistant mask generation and the other is video vision_position_ids; not the same bug." + "reason": "Both are multimodal Qwen issues, but one is chat-template assistant mask generation and the other is video position-id computation; not the same defect." }, { "left": "issue:44112", "right": "issue:45600", "accept": false, - "reason": "GraniteSpeech CI device-override test failure is unrelated to auto_mappings referencing removed Sam3LiteText configs." + "reason": "These are separate CI failures in different areas: GraniteSpeech device-override testing vs auto-mappings referencing removed configs." }, { "left": "issue:43540", "right": "issue:44734", "accept": false, - "reason": "Different failures: Qwen3OmniMoe video input ValueError vs serve KV-cache continuation tensor indexing." + "reason": "Qwen3OmniMoe video input validation and serve KV-cache continuation indexing are different runtime failures." }, { "left": "issue:44779", "right": "issue:46489", "accept": false, - "reason": "Both DeepSeek tokenizer-related, but one is a v5 regression and the other is a specific DeepSeek-Coder v1 gap in a prior fix." + "reason": "Both are DeepSeek tokenizer regressions, but they target different model families/issues and the evidence does not show the same concrete tokenizer bug." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Different example scripts and symptoms: empty multi-label classification results vs failing to load a local dataset." + "reason": "Example-script failures in classification and image-classification dataset loading are unrelated." }, { "left": "issue:44737", "right": "issue:45803", "accept": false, - "reason": "XLNet CPU device placement bug is unrelated to FuyuBatchFeature's bare-except error handling." + "reason": "XLNet CPU allocation behavior and Fuyu tensor conversion error handling are unrelated." }, { "left": "issue:44938", "right": "issue:45542", "accept": false, - "reason": "Different dependency/import problems: Python 3.14 load failure vs tensorboard-only install missing TensorFlow backend." + "reason": "Python 3.14 import failure and missing TensorFlow backend dependency are different environment/setup problems." }, { "left": "issue:43295", "right": "issue:45276", "accept": false, - "reason": "Processor/tokenizer API regression is unrelated to Gemma4 embedding-resize propagation." + "reason": "Processor/tokenizer regression and Gemma4 embedding resize behavior are different model-code issues." }, { "left": "issue:44462", "right": "issue:45907", "accept": false, - "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to list_repo_files not forwarding revision." + "reason": "AutoTokenizer ignoring tokenizer.json and list_repo_files forwarding revision are unrelated repository-access behaviors." }, { - "left": "issue:45276", - "right": "issue:45584", + "left": "issue:45632", + "right": "issue:46234", "accept": false, - "reason": "Different model-specific bugs: embedding resize not propagating vs Whisper empty transcription after special-token alignment." + "reason": "trust_remote_code cache collisions and a missing import_utils availability helper are unrelated." }, { - "left": "issue:45632", - "right": "issue:46234", + "left": "issue:45276", + "right": "issue:45584", "accept": false, - "reason": "trust_remote_code cache-path collision is unrelated to a missing import_utils availability symbol." + "reason": "Gemma4 embedding resizing and Whisper empty transcription handling are unrelated." }, { "left": "issue:42796", "right": "issue:44521", "accept": false, - "reason": "Tokenizer BPE-marker handling and multimodal assistant-mask generation are different code paths." + "reason": "Tokenizer BPE marker handling and multimodal assistant-mask generation are different subsystems." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Missing Qwen2.5-VLImageProcessor export is not the same as missing AutoImageProcessor from the package root." + "reason": "Both are import errors, but one is Qwen2.5-VL image processor export and the other is top-level AutoImageProcessor export; not the same bug." }, { - "left": "issue:45245", - "right": "issue:45584", + "left": "issue:36246", + "right": "issue:38617", "accept": false, - "reason": "Category-count runtime error is unrelated to Whisper's empty-transcription failure." + "reason": "Qwen image-processor import and configuration_utils layer_type_validation import are unrelated missing-symbol issues." }, { "left": "issue:46082", "right": "issue:46121", "accept": false, - "reason": "LlamaConfig head_dim validation and JSON-list rope-validation TypeError are distinct config bugs." + "reason": "Llama head_dim validation and rope-validation ignore_keys type handling are different config-validation paths." }, { - "left": "issue:36246", - "right": "issue:38617", + "left": "issue:45245", + "right": "issue:45584", "accept": false, - "reason": "Both are import errors, but for different missing symbols in different modules." + "reason": "Category-cardinality limits and Whisper empty transcription errors are unrelated." }, { "left": "issue:44246", "right": "issue:45003", "accept": false, - "reason": "Import performance complaints are unrelated to modeling_utils sys.modules access safety." + "reason": "Import slowness and unsafe sys.modules access are related only loosely; no evidence of the same underlying defect." }, { "left": "issue:42175", "right": "issue:44279", "accept": false, - "reason": "TensorFlow-backend packaging issue is too different from a generic dependency issue to treat as the same bug." - }, - { - "left": "issue:44683", - "right": "issue:45446", - "accept": false, - "reason": "Both flex_attention-related, but one is a torch>=2.9 compile failure and the other is a version-check/import guard bug." + "reason": "A missing TensorFlow backend in extras installation and a generic dependency issue are too broad and not the same concrete problem." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "AutoConfig missing fields with trust_remote_code is different from an unrecognized model lacking model_type." + "reason": "trust_remote_code field loss and unrecognized model_type/config.json loading failure are different config-loading bugs." }, { "left": "issue:44295", "right": "issue:46506", "accept": false, - "reason": "Registering position_ids as a buffer and CPU/CUDA output differences are unrelated symptoms." + "reason": "Position_ids buffer reading and CPU/CUDA output divergence are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both mention flex_attention, but one is a compile failure on torch>=2.9 and the other is a version-check/import bug; different code paths." }, { "left": "issue:44964", "right": "issue:45600", "accept": false, - "reason": "Phi-4 multimodal loading and stale auto_mappings references to Sam3LiteText are different issues." + "reason": "Model loading for Phi-4 multimodal and stale auto-mappings references are unrelated." }, { - "left": "issue:44186", - "right": "issue:44734", + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "A GraniteSpeech CI test issue and a Phi-4 model-loading issue are unrelated." + }, + { + "left": "issue:43854", + "right": "issue:44964", "accept": false, - "reason": "LayoutLMv2Tokenizer NER/padding crash is unrelated to serve KV-cache continuation indexing." + "reason": "Both are model-load failures, but they concern different models and distinct root causes." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Mistral3 weight-tying regression and Qwen3.5 save_pretrained corruption are different model serialization bugs." + "reason": "Mistral3 weight tying and Qwen3.5 save_pretrained checkpoint correctness are different regression paths." }, { - "left": "issue:45632", - "right": "issue:46421", + "left": "issue:41950", + "right": "issue:44016", "accept": false, - "reason": "trust_remote_code cache collisions are unrelated to spurious recompilation from lazy cache initialization." + "reason": "A pipeline/image-processor bug and a notebook syntax error are unrelated." }, { - "left": "issue:43854", - "right": "issue:44964", + "left": "issue:45632", + "right": "issue:46421", "accept": false, - "reason": "Both are model-loading issues, but for different models and distinct failure causes." + "reason": "trust_remote_code cache-path collision and lazy cache initialization recompilation are unrelated cache behaviors." } ] }, @@ -10947,7 +11162,6 @@ "issue:28282", "issue:32090", "issue:35141", - "issue:41950", "issue:42796", "issue:42907", "issue:43295", @@ -10965,15 +11179,15 @@ "issue:44016", "issue:44062", "issue:44075", - "issue:44112", + "issue:44186", "issue:44220", "issue:44279", "issue:44336", "issue:44368", "issue:44496", + "issue:44734", "issue:44792", "issue:44933", - "issue:44964", "issue:44987", "issue:45042", "issue:45083", @@ -10985,35 +11199,36 @@ "issue:45593", "issue:45600", "issue:45759", + "issue:46082", "issue:46123", "issue:46369", + "issue:46399", "issue:46424", "issue:46506", "issue:46515" ], "soft_pairs": [ - "issue:41950|issue:44016", - "issue:44112|issue:44964", - "issue:43673|issue:46424", + "issue:44186|issue:44734", "issue:27957|issue:35141", + "issue:43673|issue:46424", "issue:43650|issue:44220", "issue:44016|issue:44220", "issue:43479|issue:45375", + "issue:32090|issue:35141", "issue:45083|issue:46506", "issue:43531|issue:45600", - "issue:32090|issue:35141", "issue:28282|issue:35141", "issue:43976|issue:44987", "issue:43295|issue:45335", "issue:43526|issue:43901", "issue:44496|issue:45042", - "issue:42907|issue:43441", "issue:43673|issue:44336", + "issue:42907|issue:43441", "issue:43441|issue:45588", - "issue:46123|issue:46369", "issue:44016|issue:44062", - "issue:44792|issue:46424", + "issue:46123|issue:46369", "issue:43381|issue:45593", + "issue:44792|issue:46424", "issue:43976|issue:44279", "issue:43901|issue:44496", "issue:43824|issue:44933", @@ -11022,222 +11237,223 @@ "issue:44279|issue:46515", "issue:45335|issue:45584", "issue:44075|issue:44368", - "issue:42796|issue:43650" + "issue:42796|issue:43650", + "issue:46082|issue:46399" ], - "prepared_review_unit_hash": "3d5ee1f167c8eed076c91425bfaa2aabb627bbd705bc63a4db184b8fc4befad1", + "prepared_review_unit_hash": "e8902f540aa1699b7c0fe74f8e76a92b5c990a5b368fbaee090e0dad9c84b33b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36291, - "estimated_input_tokens": 9073, - "estimated_eval_tokens": 18402 + "serialized_chars": 36965, + "estimated_input_tokens": 9242, + "estimated_eval_tokens": 18740 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are a loose collection of distinct Transformers issues across unrelated models, pipelines, configs, and regressions. None of the soft-pair candidates look like the same underlying bug/change, so no duplicate merges are warranted.", - "confidence": 0.92, - "canonical_issue_reason": "issue:44964 is the strongest canonical candidate: it is a concrete, user-facing model-loading regression against a named model with clear impact and reproduction, while most other issues are narrower or orthogonal.", + "summary": "This cluster is a noisy false positive: the items span unrelated bugs across tokenizers, model configs, cache handling, training, docs, and import/export regressions. None of the soft-edge pairs look like the same underlying bug or a mergeable duplicate PR.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:44964 is the best representative issue for triage because it is specific enough to be actionable but broad enough to reflect a common class of latest-version loading regressions.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:41950", - "right": "issue:44016", + "left": "issue:44186", + "right": "issue:44734", "accept": false, - "reason": "Different bugs: a video-classification pipeline processor lookup vs a notebook syntax error." + "reason": "Different code paths and symptoms: LayoutLMv2 tokenizer NER/padding crash vs transformers serve KV-cache continuation indexing." }, { - "left": "issue:44112", - "right": "issue:44964", + "left": "issue:27957", + "right": "issue:35141", "accept": false, - "reason": "Unrelated failure modes: GraniteSpeech CI device override vs Phi-4 multimodal model loading." + "reason": "XLMRoberta FlashAttention-2 support is unrelated to embedding reinitialization after resize_token_embeddings." }, { "left": "issue:43673", "right": "issue:46424", "accept": false, - "reason": "Both mention cache, but they hit different code paths: GenerationMixin chunked prefill vs cache max length handling." - }, - { - "left": "issue:27957", - "right": "issue:35141", - "accept": false, - "reason": "Unrelated areas: Flash Attention 2 support for XLMRoberta vs embedding resize/reinit behavior." + "reason": "Both involve cache concepts, but one is a missing GenerationMixin cache in chunked_prefill and the other is max_cache_length only being honored for quantized cache." }, { "left": "issue:43650", "right": "issue:44220", "accept": false, - "reason": "No meaningful overlap; one is a placeholder/spam issue, the other is an audio feature extraction bug." + "reason": "No overlap in bug description or subsystem; one is an invalid title, the other is an audio feature extraction failure." }, { "left": "issue:44016", "right": "issue:44220", "accept": false, - "reason": "Notebook syntax error and fbank feature extraction are completely different issues." + "reason": "Notebook syntax error is unrelated to _torch_extract_fbank_features()." }, { "left": "issue:43479", "right": "issue:45375", "accept": false, - "reason": "Both are config-related, but they involve different models and different missing/default fields." + "reason": "Different multimodal config bugs: defaulting None values vs a missing deepstack_visual_indexes field." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer _gpu_broadcast_one NoneType crash is unrelated to output embedding reinitialization after resizing token embeddings." }, { "left": "issue:45083", "right": "issue:46506", "accept": false, - "reason": "Different subsystems: Qwen3 Omni feature-length helper vs CPU/CUDA output divergence in PPDocLayoutV3." + "reason": "Different model families and failure modes: helper length computation in qwen3_omni_moe vs CPU/CUDA divergence in PPDocLayoutV3." }, { "left": "issue:43531", "right": "issue:45600", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and Sam3LiteText auto-mapping CI breakage are unrelated." - }, - { - "left": "issue:32090", - "right": "issue:35141", - "accept": false, - "reason": "Trainer broadcast NoneType error is unrelated to token embedding resizing/post-init reinitialization." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to auto_mappings references to removed Sam3LiteText configs." }, { "left": "issue:28282", "right": "issue:35141", "accept": false, - "reason": "Missing PyTorch import and embedding reinitialization are different problems with no shared code path." + "reason": "Missing PyTorch dependency import error is not the same as embedding resize/post_init reinitialization." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Both are version regressions, but one is a Python support matrix problem and the other is a specific model-loading failure." + "reason": "Python version support regression and loading failure for a specific model are different issues." }, { "left": "issue:43295", "right": "issue:45335", "accept": false, - "reason": "Different regressions: processor/tokenizer image handling vs t5gemma decoder embedding resize propagation." + "reason": "Processor/tokenizer regression with images is unrelated to decoder.embed_tokens not updating after resize_token_embeddings." }, { "left": "issue:43526", "right": "issue:43901", "accept": false, - "reason": "A BeitImageProcessorFast label-processing bug is unrelated to stale pipeline documentation." + "reason": "BEiT image processor reduce_labels bug and TextClassificationPipeline docs mismatch are different subsystems." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Different domains: unrecognized model loading vs PIL image processors incorrectly requiring torchvision." - }, - { - "left": "issue:42907", - "right": "issue:43441", - "accept": false, - "reason": "Same model family but different concrete issues: save/dequantization failure vs FlashAttention load/runtime failure." + "reason": "Unrecognized model config loading failure is unrelated to PIL image processors incorrectly requiring torchvision." }, { "left": "issue:43673", "right": "issue:44336", "accept": false, - "reason": "Generation cache behavior and ANSI output formatting in loading_report are unrelated." + "reason": "Cache-missing regression has no relation to ANSI codes emitted by loading_report." }, { - "left": "issue:43441", - "right": "issue:45588", + "left": "issue:42907", + "right": "issue:43441", "accept": false, - "reason": "Both involve flash attention, but the failing code paths and symptoms are different enough that they are not the same bug." + "reason": "Both mention Ministral/FlashAttention, but one is a save/dequantization problem and the other is a model loading failure under FlashAttention in v5 RC." }, { - "left": "issue:46123", - "right": "issue:46369", + "left": "issue:43441", + "right": "issue:45588", "accept": false, - "reason": "MaskGenerationPipeline batching and sam3-litetext memory overflow are unrelated." + "reason": "Different failure modes: model-specific FlashAttention regression vs a flash_attention.py AttributeError on s_aux=None for sink-less models." }, { "left": "issue:44016", "right": "issue:44062", "accept": false, - "reason": "Notebook syntax error is not the same as the AddedToken keyword-argument TypeError." + "reason": "Notebook syntax error is unrelated to the AddedToken special keyword conflict." }, { - "left": "issue:44792", - "right": "issue:46424", + "left": "issue:46123", + "right": "issue:46369", "accept": false, - "reason": "A janus image-generation test failure and cache max_length behavior do not match." + "reason": "MaskGenerationPipeline partial-batch result loss and sam3-litetext memory overflow are unrelated." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss logic are different bugs." + "reason": "Gradient-checkpointing eval-mode restriction is unrelated to D-FINE auxiliary-loss behavior when denoising is off." + }, + { + "left": "issue:44792", + "right": "issue:46424", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to cache max_cache_length handling." }, { "left": "issue:43976", "right": "issue:44279", "accept": false, - "reason": "A Python/version compatibility issue is not the same as a generic dependency complaint." + "reason": "Python-version compatibility regression and a general dependency issue are not the same bug." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "Documentation mismatch for return_all_scores does not overlap with an unrecognized-model config error." + "reason": "Pipeline documentation mismatch is unrelated to model config loading failure." }, { "left": "issue:43824", "right": "issue:44933", "accept": false, - "reason": "Both are import errors, but they reference different symbols and modules with different root causes." + "reason": "Both are import/export regressions, but they concern different missing symbols from different modules and code paths." }, { "left": "issue:44496", "right": "issue:45759", "accept": false, - "reason": "These are different model/configuration failures: unknown model type vs not unwrapping text_config in composite multimodal configs." + "reason": "Unrecognized model config and AutoModelForCausalLM.from_config not unwrapping text_config are distinct loading bugs." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "Different regressions on different model families and code paths: loading failure vs save_pretrained checkpoint corruption." + "reason": "Qwen3VL MoE loading regression and Qwen3.5 save_pretrained checkpoint corruption are different code paths." }, { "left": "issue:44279", "right": "issue:46515", "accept": false, - "reason": "A vague dependency issue is not the same as a specific torch 2.4.1 incompatibility." + "reason": "Generic dependency issue is not the same as torch 2.4.1 incompatibility." }, { "left": "issue:45335", "right": "issue:45584", "accept": false, - "reason": "Token-embedding resize propagation and Whisper empty-transcription generation are unrelated." + "reason": "resize_token_embeddings not updating decoder.embed_tokens is unrelated to Whisper empty-transcription generation failure." }, { "left": "issue:44075", "right": "issue:44368", "accept": false, - "reason": "SGD argument handling and a tied-embeddings warning during Qwen fine-tuning are different issues." + "reason": "Unused SGD optimizer args and a tie_word_embeddings warning are unrelated." }, { "left": "issue:42796", "right": "issue:43650", "accept": false, - "reason": "A tokenizer BPE-marker bug and a placeholder issue are clearly unrelated." + "reason": "Tokenizer BPE marker handling bug has nothing to do with the placeholder 'ADD THE DATA' issue." + }, + { + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "LlamaConfig head_dim validation and Qwen3.5 use_kernels registration crash are different model/config bugs." } ] }, @@ -11307,21 +11523,19 @@ "issue:45907", "issue:46082", "issue:46209", - "issue:46399", "issue:46512" ], "soft_pairs": [ - "issue:46082|issue:46399", - "issue:41950|issue:45020", "issue:44279|issue:45405", + "issue:41950|issue:45020", "issue:43577|issue:45216", "issue:45103|issue:45372", "issue:44351|issue:45278", - "issue:43299|issue:45357", "issue:39692|issue:43901", + "issue:43295|issue:44779", + "issue:43299|issue:45357", "issue:44242|issue:45200", "issue:41950|issue:45823", - "issue:43295|issue:44779", "issue:44683|issue:45542", "issue:38175|issue:44617", "issue:43976|issue:45405", @@ -11332,203 +11546,197 @@ "issue:43479|issue:43525", "issue:44521|issue:45325", "issue:44466|issue:45276", - "issue:44016|issue:44291", "issue:44462|issue:45081", + "issue:44016|issue:44291", "issue:43653|issue:45593", "issue:43441|issue:45834", "issue:42994|issue:43317", "issue:45081|issue:45907" ], - "prepared_review_unit_hash": "db7dea7e414bbbd45d21926b5ec077b96fd38ee1b6707ab8d3c4be25a679a4c9", + "prepared_review_unit_hash": "00f174b81afb34d85958f0ec824c2f3920ed31abf532671b28a25ceb3c3b8675", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 35947, - "estimated_input_tokens": 8987, - "estimated_eval_tokens": 18230 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35038, + "estimated_input_tokens": 8760, + "estimated_eval_tokens": 17776 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is heterogeneous and does not contain any true duplicate pairs by the available evidence. Most soft matches are superficial (shared subsystem/model family) but involve different code paths, behaviors, or regression surfaces.", - "confidence": 0.82, - "canonical_issue_reason": "Issue 44521 is the strongest representative: it is open, well-scoped, and has the most follow-up activity/inbound references among the set, making it the clearest issue to anchor the cluster around.", + "summary": "This cluster is not a true duplicate set; it mixes many unrelated bugs across different models, tokenizer paths, docs, quantization, config, and loading/saving code. All soft edges look like false positives under strict duplicate triage.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue stands out because the cluster spans multiple independent problems rather than one underlying defect.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 is the best overall issue candidate because it is specific, reproducible, actively discussed, and still open for resolution.", + "best_issue_reason": "None of the issues is a good global representative for the whole cluster; the closest thing to a broad regression still only covers one narrow code path.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:46082", - "right": "issue:46399", + "left": "issue:44279", + "right": "issue:45405", "accept": false, - "reason": "Different bugs: LlamaConfig head_dim validation vs Qwen3.5 rotary/kernel registration crash." + "reason": "Both mention dependency/version problems, but one is a vague transformers dependency issue and the other is a specific PEFT version bump; different root causes." }, { "left": "issue:41950", "right": "issue:45020", "accept": false, - "reason": "Pipeline processor lookup bug vs a broad remote_code loading regression; not the same code path." - }, - { - "left": "issue:44279", - "right": "issue:45405", - "accept": false, - "reason": "Generic dependency complaint vs an unreleased PEFT version bump; different problem and fix." + "reason": "Video-classification processor lookup and remote_code breakage are unrelated code paths." }, { "left": "issue:43577", "right": "issue:45216", "accept": false, - "reason": "Blip2 dtype propagation on load vs Qwen3.5 save_pretrained serialization corruption." + "reason": "Both are regression-style model state issues, but one is about loaded dtype on Blip2 and the other about incorrect save_pretrained serialization for Qwen3.5." }, { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "auto_docstring future-annotations crash vs Gemma4 processor import failure; unrelated." + "reason": "Auto-docstring annotation handling and Gemma 4 processor import failure are unrelated." }, { "left": "issue:44351", "right": "issue:45278", "accept": false, - "reason": "Single HybridCache import error vs many import errors after upgrade; too broad and different." + "reason": "A missing HybridCache symbol is not the same as broad import failures after a version upgrade." }, { - "left": "issue:43299", - "right": "issue:45357", + "left": "issue:39692", + "right": "issue:43901", "accept": false, - "reason": "Qwen3VL load failure vs Qwen3.5 save_pretrained visual-key bug; distinct regressions." + "reason": "Both are docs-related, but they concern different API examples and different mistakes." }, { - "left": "issue:39692", - "right": "issue:43901", + "left": "issue:43295", + "right": "issue:44779", "accept": false, - "reason": "SigLIP2 doc example errors vs TextClassificationPipeline docs mismatch; documentation-only overlap is insufficient." + "reason": "Both involve tokenizer regressions, but they affect different behaviors and model families; not enough evidence of one shared bug." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3VL load failure and Qwen3.5 save_pretrained key serialization are different paths and symptoms." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "MoE load-balancing loss gating vs mm_token_type_ids defaulting; different model logic." + "reason": "MoE load-balancing loss logic and Gemma 4 text-only mm_token_type_ids handling are different multimodal/model concerns." }, { "left": "issue:41950", "right": "issue:45823", "accept": false, - "reason": "Video-classification processor lookup vs Gemma4 PLE device placement mismatch." - }, - { - "left": "issue:43295", - "right": "issue:44779", - "accept": false, - "reason": "Custom model processor.tokenizer regression vs Deepseek tokenizer output regression; different affected APIs." + "reason": "Video pipeline processor selection and Gemma4 device-mismatch during forward are unrelated." }, { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "flex_attention torch 2.9 compilation failure vs TensorBoard/TensorFlow backend dependency issue." + "reason": "Torch flex_attention compilation and tensorboard/TensorFlow backend setup have no common underlying bug." }, { "left": "issue:38175", "right": "issue:44617", "accept": false, - "reason": "SigLIP2 zero probabilities vs Sam3Video CUDA OOM; unrelated symptoms and models." + "reason": "SigLIP2 zero probabilities and Sam3Video CUDA OOM are unrelated failures." }, { "left": "issue:43976", "right": "issue:45405", "accept": false, - "reason": "Python version compatibility break vs PEFT minimum-version packaging issue." + "reason": "Python version support regression and PEFT dependency version bump are different issues." }, { "left": "issue:46082", "right": "issue:46512", "accept": false, - "reason": "LlamaConfig head_dim validation vs from_config dtype staleness; different configuration bugs." + "reason": "LlamaConfig head_dim validation and stale config.dtype after from_config(dtype=...) are separate configuration bugs." }, { "left": "issue:43650", "right": "issue:46209", "accept": false, - "reason": "Non-informative request vs FP8 inference failure on B200; no shared underlying bug." + "reason": "The first is a non-descriptive title; the second is an FP8 inference problem on B200. No shared underlying change is visible." }, { "left": "issue:44779", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different tokenizer logic." + "reason": "Both are tokenizer regressions, but they target different model families and likely different tokenizer code paths." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Qwen3VL loading regression vs GLM-4.7-Flash unit-test load issue; different models and failure modes." + "reason": "Qwen3VL loading and GLM-4.7-Flash unit-test loading are model-specific failures with different causes." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Phi4MultimodalConfig default init bug vs Llama4Config missing pad_token_id; separate config issues." + "reason": "Phi4MultimodalConfig default subconfig initialization and Llama4Config missing pad_token_id are different config bugs." }, { "left": "issue:44521", "right": "issue:45325", "accept": false, - "reason": "All-zero assistant masks vs Qwen2.5-VL rope/position-id scaling; different multimodal internals." + "reason": "Chat-template assistant masks and Qwen2.5-VL rope index scaling are separate multimodal bugs." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "Device-dependent lm_head serialization vs resize_token_embeddings not updating embeddings; different state-management bugs." + "reason": "Incorrect lm_head serialization and Gemma4 embedding resize propagation are unrelated state vs resize problems." }, { - "left": "issue:44016", - "right": "issue:44291", + "left": "issue:44462", + "right": "issue:45081", "accept": false, - "reason": "Notebook syntax error vs init_empty_weights argument mismatch; unrelated." + "reason": "AutoTokenizer ignoring tokenizer.json and Mistral regex patch crashing are different tokenizer-loading defects." }, { - "left": "issue:44462", - "right": "issue:45081", + "left": "issue:44016", + "right": "issue:44291", "accept": false, - "reason": "AutoTokenizer ignoring tokenizer.json vs Mistral regex patch crash; distinct tokenizer loading issues." + "reason": "Notebook syntax and init_empty_weights argument mismatch are unrelated." }, { "left": "issue:43653", "right": "issue:45593", "accept": false, - "reason": "BigBirdTokenizer special-token decode bug vs D-FINE auxiliary-loss behavior; unrelated." + "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss logic are entirely different subsystems." }, { "left": "issue:43441", "right": "issue:45834", "accept": false, - "reason": "FlashAttention failure for Ministral-3 vs Kosmos2.5 OCR index error; different code paths." + "reason": "FlashAttention failure for Ministral-3 and OCR index error in Kosmos2.5 are unrelated." }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Quantized model saving failure vs dequantized model loading/offload failure; related area but not same bug." + "reason": "Quantized model saving and dequantized model loading/offload are different stages and likely different fixes." }, { "left": "issue:45081", "right": "issue:45907", "accept": false, - "reason": "Tokenizer regex crash vs list_repo_files revision forwarding; unrelated utilities." + "reason": "Tokenizer regex patch crash and list_repo_files revision forwarding have no shared root cause." } ] }, @@ -11605,242 +11813,242 @@ "issue:44623|issue:44749", "issue:33290|issue:45656", "issue:44261|issue:44263", - "issue:45600|issue:45712", "issue:44623|issue:45397", + "issue:45600|issue:45712", "issue:44361|issue:45593", "issue:45542|issue:45800", - "issue:42175|issue:43976", "issue:44016|issue:45823", + "issue:42175|issue:43976", "issue:43441|issue:44683", - "issue:46291|issue:46399", "issue:43901|issue:43994", "issue:44671|issue:44779", - "issue:43673|issue:43761", "issue:43723|issue:44162", + "issue:43673|issue:43761", + "issue:46291|issue:46399", "issue:43525|issue:43881", "issue:43824|issue:43976", "issue:43901|issue:44246", - "issue:43525|issue:43643", "issue:44297|issue:45397", + "issue:43525|issue:43643", "issue:44188|issue:44336", "issue:44589|issue:45362", "issue:43784|issue:45542", - "issue:41628|issue:43976", "issue:44336|issue:44373", + "issue:41628|issue:43976", "issue:44246|issue:44496", "issue:45103|issue:46399", "issue:45561|issue:45563", "issue:44295|issue:45397", "issue:44242|issue:45593" ], - "prepared_review_unit_hash": "2b055198fa85b7a28ff5deb86b90de007fdcb7c545afdba1f1c8da9c4f3cb675", + "prepared_review_unit_hash": "c2fd70d2a36a5eb5ef78bf065a139abb004f5e9d4406cb726e7b8158a61ad77c", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36534, - "estimated_input_tokens": 9134, - "estimated_eval_tokens": 18524 + "serialized_chars": 36530, + "estimated_input_tokens": 9133, + "estimated_eval_tokens": 18522 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like a true duplicate; the set is heterogeneous across unrelated bugs, packaging issues, model regressions, and test/docs cleanup. I\u2019d treat this cluster as weakly related by repo/version, not by underlying defect.", - "confidence": 0.63, - "canonical_issue_reason": "No clear duplicate canonical emerges from these titles alone; issue 44246 is the broadest/highest-activity repo-wide bug report in the set, so it is the most reasonable representative issue.", + "summary": "This is a loose Transformers issue set with broad thematic overlap (loading/config/tokenizer/kernel regressions) but no true duplicate pairs. I would anchor the cluster on the generic loading regression in issue #43643, but all soft edges should be rejected.", + "confidence": 0.31, + "canonical_issue_reason": "Issue #43643 is the most broadly framed, concrete regression in the set: it describes a specific `AutoConfig.from_pretrained(..., trust_remote_code=True)` loading failure with missing fields, which is more reusable as an anchor than the model-specific or one-off reports.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44246 is the best single issue representative because it has the widest apparent scope and strongest discussion signal, even though the cluster is not actually a duplicate group.", + "best_issue_reason": "Issue #43643 is the best global representative because it is a clear API-level bug with a reproducible symptom and broad potential impact across remote-code models; it is closer to the cluster\u2019s loading/config theme than the other narrowly scoped reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "Different problems: missing saved processor files vs tokenizer/data-filtering slowdown after upgrade." + "reason": "Different bugs: missing files on `processor.save_pretrained` vs a slowdown/regression in filtering after upgrade." }, { "left": "issue:33290", "right": "issue:45656", "accept": false, - "reason": "Both mention DeepSpeed/optimizers, but one is OOM with Adafactor while the other is double optimizer stepping." + "reason": "Both mention DeepSpeed/optimizers, but one is OOM with Adafactor while the other is double optimizer stepping; not the same failure mode." }, { "left": "issue:44261", "right": "issue:44263", "accept": false, - "reason": "Unrelated components: MLA q_a_layernorm precision/config issue vs GlmMoeDsaIndexer torch.split behavior." + "reason": "Unrelated components and symptoms: MLA layernorm precision/config issue vs `torch.split` return-value behavior in a MoE indexer." }, { - "left": "issue:45600", - "right": "issue:45712", + "left": "issue:44623", + "right": "issue:45397", "accept": false, - "reason": "Both are repo-maintenance regressions, but they affect different files and symptoms: auto_mappings references vs leftover dummy classes leakage." + "reason": "`processor.save_pretrained` missing files and Gemma-4 ZeRO-3 `from_pretrained` loading are different code paths and failure causes." }, { - "left": "issue:44623", - "right": "issue:45397", + "left": "issue:45600", + "right": "issue:45712", "accept": false, - "reason": "Processor save_pretrained missing files is unrelated to Gemma4 Zero3 from_pretrained loading." + "reason": "Both are repo hygiene/auto-object issues, but one is stale auto-mapping references and the other is leftover dummy classes; not the same bug." }, { "left": "issue:44361", "right": "issue:45593", "accept": false, - "reason": "Different model/path: MLukeTokenizer task AttributeError vs D-FINE auxiliary-loss behavior when denoising is off." + "reason": "Different model/task bugs: MLuke tokenizer AttributeError vs D-FINE missing auxiliary losses when denoising is off." }, { "left": "issue:45542", "right": "issue:45800", "accept": false, - "reason": "Packaging/backend mismatch with TensorFlow vs a torch-transformers version incompatibility; not the same bug." - }, - { - "left": "issue:42175", - "right": "issue:43976", - "accept": false, - "reason": "One is missing TensorFlow in the torch extra; the other is Python version support regression." + "reason": "TensorFlow backend installation issue vs a torch/transformers version incompatibility; unrelated root causes." }, { "left": "issue:44016", "right": "issue:45823", "accept": false, - "reason": "Notebook syntax/docs issue vs runtime device-mismatch bug in Gemma4 forward." + "reason": "Notebook syntax error vs a Gemma4 device-mapping forward mismatch; no common underlying defect." }, { - "left": "issue:43441", - "right": "issue:44683", + "left": "issue:42175", + "right": "issue:43976", "accept": false, - "reason": "Both are attention-backend failures, but they hit different code paths and versions: FlashAttention in v5 RC vs compiled flex_attention on torch>=2.9." + "reason": "Backend packaging omission for `.[torch]` vs Python-version compatibility regression; not duplicates." }, { - "left": "issue:46291", - "right": "issue:46399", + "left": "issue:43441", + "right": "issue:44683", "accept": false, - "reason": "Both involve kernels, but one fails importing pipelines due to dependency requirements while the other fails registration with a missing decorator." + "reason": "Both involve attention kernels, but one is Ministral-3 FlashAttention failure and the other is compiled flex_attention on torch>=2.9; different breakages." }, { "left": "issue:43901", "right": "issue:43994", "accept": false, - "reason": "Docs mismatch about return_all_scores is not the same as incorrect model/pipeline outputs for SigLIP2." + "reason": "Docs drift for `return_all_scores` vs incorrect model/pipeline outputs for SigLIP2; unrelated issues." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "Both are v5 regressions, but one is CamemBERT MLM output quality and the other is Deepseek tokenizer correctness." + "reason": "CamemBERT masked-LM output regression and Deepseek tokenizer regression affect different model families and code paths." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "AutoTokenizer loading failure and ESM2 breakage are both model-loading problems, but they are distinct bugs with different symptoms." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Generation cache/chunked_prefill issue is unrelated to CLIPVisionModel hidden_states not being returned." + "reason": "Generation cache missing during chunked prefill vs CLIPVision hidden_states omission; different APIs and fixes." }, { - "left": "issue:43723", - "right": "issue:44162", + "left": "issue:46291", + "right": "issue:46399", "accept": false, - "reason": "Tokenizer loading failure and ESM2 model breakage are different components with different symptoms." + "reason": "Both mention kernels, but one is import-time kernel version metadata failure and the other is a missing decorator on rotary-pos-emb registration; not the same change." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Different model-specific load failures: missing pad_token_id on Llama4Config vs glm-4v-9b loading failure." + "reason": "`Llama4Config` missing `pad_token_id` and generic glm-4v-9b load failure are not concretely the same bug." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "ImportError for a model class is unrelated to Python 3.9+/3.10+ compatibility." + "reason": "ImportError for a missing model class vs Python 3.9/3.10 compatibility issue; unrelated." }, { "left": "issue:43901", "right": "issue:44246", "accept": false, - "reason": "Docs wording issue vs import-time performance problem; not the same underlying defect." + "reason": "Documentation mismatch and slow import performance are unrelated." }, { - "left": "issue:43525", - "right": "issue:43643", + "left": "issue:44297", + "right": "issue:45397", "accept": false, - "reason": "Both mention missing config fields, but one is a concrete model config attribute bug and the other is a trust_remote_code return-shape issue." + "reason": "Tokenizer save metadata mismatch vs Gemma-4 ZeRO-3 loading failure; different functionality." }, { - "left": "issue:44297", - "right": "issue:45397", + "left": "issue:43525", + "right": "issue:43643", "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch is unrelated to Gemma4 Zero3 from_pretrained loading." + "reason": "Both concern missing config fields, but one is model-specific `pad_token_id` and the other is a broader `trust_remote_code` loading regression; not clearly the same bug." }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "Attention-kernel divergence under torch.compile is unrelated to ANSI codes emitted by loading_report." + "reason": "Attention-kernel divergence under `torch.compile` vs ANSI codes in loading report; no overlap beyond being library bugs." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Float8 storage lookup failure and Qwen3.5 chat crash are different runtime failures." + "reason": "Storage-object lookup error and Qwen3.5 chat crash are different runtime failures with no shared concrete code path." }, { "left": "issue:43784", "right": "issue:45542", "accept": false, - "reason": "Sentence-transformers import NameError is unrelated to missing TensorFlow backend installation." + "reason": "Importing sentence-transformers breaks on missing `nn`, while the other is a packaging/backend dependency issue involving TensorFlow." }, { - "left": "issue:41628", - "right": "issue:43976", + "left": "issue:44336", + "right": "issue:44373", "accept": false, - "reason": "Importing AutoImageProcessor and Python version support are unrelated issues." + "reason": "Terminal ANSI output bug vs wrong docstring for `position_ids`; documentation/runtime mismatch but not the same issue." }, { - "left": "issue:44336", - "right": "issue:44373", + "left": "issue:41628", + "right": "issue:43976", "accept": false, - "reason": "Terminal ANSI output cleanup is unrelated to the position_ids docstring error." + "reason": "Missing `AutoImageProcessor` import and Python-version support regression are distinct issues." }, { "left": "issue:44246", "right": "issue:44496", "accept": false, - "reason": "Import slowness is unrelated to unrecognized model/config.json missing model_type." + "reason": "Slow import performance and unrecognized-model config errors are unrelated." }, { "left": "issue:45103", "right": "issue:46399", "accept": false, - "reason": "Auto-docstring crash under future annotations is unrelated to Qwen3.5 kernels registration failure." + "reason": "Auto-docstring crash with future annotations and Qwen3.5 kernel registration failure are different subsystems." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "Test-worker race on captured_info.txt is unrelated to a stale generation warning." + "reason": "A pytest-xdist race and a stale warning in paged generate are different test/runtime bugs." }, { "left": "issue:44295", "right": "issue:45397", "accept": false, - "reason": "position_ids buffer access error is unrelated to Gemma4 Zero3 from_pretrained behavior." + "reason": "Buffer registration access error vs Gemma-4 ZeRO-3 loading failure; not the same underlying defect." }, { "left": "issue:44242", "right": "issue:45593", "accept": false, - "reason": "Both are conditional loss omissions, but they affect different models and mechanisms: router logits vs denoising auxiliary losses." + "reason": "Load-balancing loss not added in MoE gating vs D-FINE auxiliary loss omission under denoising-off; similar-sounding loss issues but different models and logic." } ] }, @@ -11880,6 +12088,7 @@ "issue:43596", "issue:43673", "issue:43901", + "issue:44038", "issue:44162", "issue:44220", "issue:44242", @@ -11904,7 +12113,6 @@ "issue:45362", "issue:45375", "issue:45561", - "issue:45584", "issue:45759", "issue:45859", "issue:46132", @@ -11920,12 +12128,12 @@ "issue:41950|issue:44220", "issue:44336|issue:44485", "issue:44246|issue:44871", - "issue:43550|issue:46506", "issue:44336|issue:46424", + "issue:43550|issue:46506", "issue:36246|issue:42757", "issue:36246|issue:44351", - "issue:42886|issue:46132", "issue:44734|issue:46424", + "issue:42886|issue:46132", "issue:45003|issue:45561", "issue:36246|issue:43596", "issue:44779|issue:45292", @@ -11940,212 +12148,212 @@ "issue:44496|issue:45003", "issue:44162|issue:45200", "issue:45375|issue:45759", - "issue:45292|issue:45584", + "issue:43262|issue:44246", "issue:44877|issue:45375", - "issue:43262|issue:44246" + "issue:44038|issue:46506" ], - "prepared_review_unit_hash": "985a53ba8a5ca3b3c7373c589c356f50bed6f7ebdc772265566fe7871f386dc3", + "prepared_review_unit_hash": "85a7cc8694a95040016e97ea62965318d2e5b3ac42959b1ec389f7c629eda1db", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36315, - "estimated_input_tokens": 9079, - "estimated_eval_tokens": 18414 + "serialized_chars": 36274, + "estimated_input_tokens": 9069, + "estimated_eval_tokens": 18394 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft duplicate links are safe to accept here. The proposed pairs are mostly broad topic similarities (same model family, same subsystem, or both being regressions) rather than the same concrete bug, so this cluster should not be merged into a single canonical issue.", + "summary": "This cluster is a grab-bag of transformer regressions and import/config bugs, mostly centered on v5-era model-loading or pipeline issues. None of the soft pairs look like true duplicates; they share broad themes at most, not the same concrete bug.", "confidence": 0.89, - "canonical_issue_reason": "No clear canonical issue: the items split across unrelated bug classes (imports, cache/offline behavior, Qwen/MoE loading, pipelines, config validation, performance, etc.), and none of the soft pairs describe the same underlying defect.", + "canonical_issue_reason": "issue:44561 is the broadest representative of the cluster\u2019s recurring v5 regression theme and captures a user-facing breakage pattern better than the narrower model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "No issue is a good global representative for this set; the apparent similarity is superficial and the actual failures differ by model, API path, or root cause.", + "best_issue_reason": "issue:44561 is the best single issue anchor because it describes a clear, cross-cutting compatibility regression rather than a one-off model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42491", "right": "issue:43299", "accept": false, - "reason": "Both involve Qwen3/MoE loading, but one is a LoRA compatibility problem across hf4.x/hf5.x and the other is a v5.0 model-loading regression for Qwen3VL MoE models; different concrete failures." + "reason": "Both involve Qwen3-MoE loading regressions, but one is LoRA compatibility on hf4/hf5 and the other is Qwen3VL Moe loading on transformers 5.0; different failure modes." }, { "left": "issue:43531", "right": "issue:45362", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and Qwen3.5-35B chat crashes are different code paths and symptoms." + "reason": "Sliding-window behavior for Qwen3-MoE is unrelated to Qwen3.5-35B chat crashes; same family, different bug." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "BeitImageProcessorFast label reduction and unsafe sys.modules access in modeling_utils are unrelated bugs." + "reason": "BeitImageProcessorFast label reduction and unsafe sys.modules access are completely different code paths." }, { "left": "issue:41950", "right": "issue:44220", "accept": false, - "reason": "Video-classification image-processor lookup and `_torch_extract_fbank_features()` are different subsystems and failure modes." + "reason": "Video-classification pipeline image-processor lookup and audio fbank feature extraction are unrelated subsystems." }, { "left": "issue:44336", "right": "issue:44485", "accept": false, - "reason": "ANSI output in loading_report and GLM-5 RoPE implementation are unrelated issues." + "reason": "ANSI-code output in loading_report has nothing to do with GLM-5 RoPE implementation." }, { - "left": "issue:44242", + "left": "issue:44246", "right": "issue:44871", "accept": false, - "reason": "Missing load-balancing loss with `output_router_logits=False` is unrelated to the Gemma-3 eos_token_id config mismatch." + "reason": "Import-performance complaints and Gemma eos_token_id config mismatch are unrelated bugs." }, { - "left": "issue:43550", - "right": "issue:46506", + "left": "issue:44336", + "right": "issue:46424", "accept": false, - "reason": "Bamba-9B-v2 torch.compile+SDPA failure and PPDocLayoutV3 CPU/CUDA divergence are different model-specific bugs." + "reason": "Terminal-color handling in loading_report is unrelated to cache max length behavior." }, { - "left": "issue:44336", - "right": "issue:46424", + "left": "issue:43550", + "right": "issue:46506", "accept": false, - "reason": "Loading-report ANSI codes and cache `max_cache_length` semantics are not the same bug." + "reason": "torch.compile/SDPA failure in Bamba and CPU-vs-CUDA differences in PPDocLayoutV3 are distinct issues." }, { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols and different packages (`Qwen2_5_VLImageProcessor` vs `is_offline_mode`)." + "reason": "Both are import errors, but for different symbols and different dependency surfaces." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Different missing exports from different modules: Qwen2_5_VLImageProcessor vs HybridCache." + "reason": "Missing Qwen2_5_VLImageProcessor export and missing HybridCache export are separate import regressions." }, { - "left": "issue:42886", - "right": "issue:46132", + "left": "issue:44734", + "right": "issue:46424", "accept": false, - "reason": "Offline cache loading and `AttentionInterface.register` behavior changes are unrelated." + "reason": "KV-cache continuation tensor indexing in serve is not the same as cache max_cache_length handling." }, { - "left": "issue:44734", - "right": "issue:46424", + "left": "issue:42886", + "right": "issue:46132", "accept": false, - "reason": "KV-cache continuation crash in `transformers serve` is a separate bug from `max_cache_length` only applying to Quantized Cache." + "reason": "Offline tokenizer cache loading and AttentionInterface.register behavior change are unrelated." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "Unsafe `sys.modules` access and the pytest-xdist captured_info.txt race are unrelated." + "reason": "Unsafe sys.modules access and pytest-xdist file-race behavior are different problems." }, { "left": "issue:36246", "right": "issue:43596", "accept": false, - "reason": "Importing a missing processor symbol and a Zero3/BertModel index error are not the same defect." + "reason": "Qwen2_5_VLImageProcessor import failure and DeepSpeed ZeRO3 BertModel index error do not share a concrete bug." }, { "left": "issue:44779", "right": "issue:45292", "accept": false, - "reason": "DeepSeek tokenizer correctness and `resize_token_embeddings` not updating output embeddings are different bugs." + "reason": "DeepSeek tokenizer correctness and resize_token_embeddings not updating output embeddings are separate embedding/tokenizer bugs." }, { "left": "issue:45859", "right": "issue:46399", "accept": false, - "reason": "Both touch Qwen3.5/MoE, but one is CPU offload incompatibility in rotary embedding and the other is a missing decorator during kernel registration." + "reason": "Both involve Qwen3.5 rotary-related code, but one is CPU offload incompatibility and the other is missing decorator registration; not the same fix." }, { "left": "issue:44295", "right": "issue:44749", "accept": false, - "reason": "A buffer access error on `position_ids` is unrelated to a performance slowdown after upgrading Transformers." + "reason": "Registering position_ids as a buffer and version-upgrade filtering slowdown are unrelated." }, { "left": "issue:43475", "right": "issue:46369", "accept": false, - "reason": "SAM 3 Video missing an attribute and SAM 3 litetext memory overflow are different problems." + "reason": "SAM 3 Video missing an attribute and SAM3 litetext memory overflow are different defects." }, { - "left": "issue:44220", + "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "Audio feature extraction and an AMP crash in LwDetrImageLoss are unrelated." + "reason": "Missing load-balancing loss and AMP/CUDA crash in a different loss module are unrelated." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "Documentation mismatch for `return_all_scores` does not match the `sys.modules` access bug." + "reason": "Pipeline docs mismatch and sys.modules access bug are unrelated." }, { "left": "issue:43673", "right": "issue:44561", "accept": false, - "reason": "Both are v5 regressions, but chunked_prefill cache handling and `is_torch_fx_available` removal break different APIs and code paths." + "reason": "Both are v5 regressions, but chunked_prefill cache handling and trust_remote_code breakage are different concrete code paths." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "Nemotron lacking Flash Attention 2.0 support is not the same as the generic Flash Attention 2.0 issue." + "reason": "A model-specific Flash Attention 2.0 unsupported error is not the same as the generic flash-attention topic issue." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed and Zero3 deque underflow are different DeepSpeed failures." + "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed and empty-deque ZeRO3 failure are distinct DeepSpeed issues." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Missing `model_type` in config loading and unsafe `sys.modules` access are unrelated." + "reason": "Unrecognized model/config loading and unsafe sys.modules access are unrelated." }, { "left": "issue:44162", "right": "issue:45200", "accept": false, - "reason": "ESM2 breakage and Gemma 4 `mm_token_type_ids` defaults are unrelated model/config bugs." + "reason": "ESM2 being broken and Gemma 4 mm_token_type_ids defaults are unrelated model-specific bugs." }, { "left": "issue:45375", "right": "issue:45759", "accept": false, - "reason": "Both concern multimodal Qwen config handling, but one drops a vision config field under `@strict` while the other fails to unwrap `text_config` in `from_config`; not the same concrete bug." + "reason": "Both concern Qwen multimodal configs, but one drops a specific field under strict parsing while the other fails to unwrap text_config in from_config." }, { - "left": "issue:45292", - "right": "issue:45584", + "left": "issue:43262", + "right": "issue:44246", "accept": false, - "reason": "Embedding resize behavior and Whisper empty-transcription generation are unrelated." + "reason": "Audio chat-template sampling-rate defaults and import slowdown are unrelated." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Strict config rejection for granite_speech and missing Qwen3_5MoeVisionConfig field are different model/config issues." + "reason": "Different models and different config-handling bugs; only a loose strict-config similarity." }, { - "left": "issue:43262", - "right": "issue:44246", + "left": "issue:44038", + "right": "issue:46506", "accept": false, - "reason": "Audio chat-template sampling-rate handling and import-time slowness are unrelated." + "reason": "A Qwen3-VL-Moe v5 regression and CPU/CUDA result divergence in PPDocLayoutV3 are not the same bug." } ] }, @@ -12177,18 +12385,19 @@ "issue:43653", "issue:43906", "issue:43931", - "issue:44038", "issue:44060", "issue:44077", "issue:44162", "issue:44188", "issue:44297", + "issue:44361", "issue:44368", "issue:44466", "issue:44479", "issue:44483", "issue:44561", "issue:44623", + "issue:44749", "issue:44877", "issue:44977", "issue:44987", @@ -12215,20 +12424,19 @@ "issue:46202", "issue:46399", "issue:46421", - "issue:46431", - "issue:46506" + "issue:46431" ], "soft_pairs": [ - "issue:44038|issue:46506", - "issue:45292|issue:45701", + "issue:45292|issue:45584", "issue:44561|issue:46399", + "issue:45292|issue:45701", "issue:45412|issue:46123", "issue:45137|issue:45600", "issue:44466|issue:45335", "issue:44987|issue:45459", "issue:43475|issue:43540", - "issue:44995|issue:45200", "issue:45397|issue:45656", + "issue:44995|issue:45200", "issue:44060|issue:44077", "issue:44479|issue:46170", "issue:44368|issue:44977", @@ -12245,242 +12453,214 @@ "issue:44188|issue:46202", "issue:45987|issue:46082", "issue:44188|issue:46421", - "issue:43065|issue:46123", "issue:45584|issue:46170", - "issue:44297|issue:44623" + "issue:43065|issue:46123", + "issue:44297|issue:44623", + "issue:44361|issue:44749", + "issue:44749|issue:45656" ], - "prepared_review_unit_hash": "06463388f3a6cedd861ee8700954385697ec0fa536810184d22036ede0dc7f04", + "prepared_review_unit_hash": "f49a119a66ca0614963691cbba832aefa6a2428b9bdc01c777cd0ea811202673", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36337, - "estimated_input_tokens": 9085, - "estimated_eval_tokens": 18426 + "soft_pair_count": 31, + "serialized_chars": 37028, + "estimated_input_tokens": 9257, + "estimated_eval_tokens": 18770 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items do not form a single duplicate set; they span unrelated tokenizer, model-loading, deepspeed, video, and cache bugs. None of the soft edges look like the same underlying issue, so all should be rejected. If one issue must be chosen as the best representative, 44479 is the broadest and most concrete regression report.", - "confidence": 0.33, - "canonical_issue_reason": "issue:44479 is the strongest representative because it reports a specific v5.3.0 regression with clear scope and multiple affected Qwen variants, making it more canonical than the narrower or unrelated reports.", + "summary": "This set is highly heterogeneous, with many unrelated bugs across tokenizers, vision models, generation, and distributed/runtime behavior. I would not merge any of the soft edges; the strongest topical mini-cluster is around save_pretrained/serialization, but even there the paired reports are not the same concrete bug.", + "confidence": 0.74, + "canonical_issue_reason": "issue:44297 is the most concrete, self-contained bug report in the set and sits in the clearest mini-cluster (tokenizer/serialization/save_pretrained behavior).", "canonical_pr_reason": null, - "best_issue_reason": "issue:44479 is the most suitable global representative: concrete failure mode, broad impact, and well-scoped regression description.", + "best_issue_reason": "issue:44297 is the best representative issue to anchor the cluster summary because it is specific, reproducible, and better scoped than the broader or more speculative reports.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44038", - "right": "issue:46506", - "accept": false, - "reason": "Different models and failure modes: Qwen3-VL-Moe bug vs CPU/CUDA output divergence in PPDocLayoutV3." - }, { "left": "issue:45292", - "right": "issue:45701", + "right": "issue:45584", "accept": false, - "reason": "Tokenizer version-dependent behavior is unrelated to resize_token_embeddings/output_embeddings syncing." + "reason": "Both involve embeddings/tokenization-adjacent behavior, but one is resize_token_embeddings on output embeddings and the other is Whisper generation after align_special_tokens; different failures and code paths." }, { "left": "issue:44561", "right": "issue:46399", "accept": false, - "reason": "Both are v5 breakages, but one is trust_remote_code FX removal and the other is a Qwen3.5 kernel registration error." + "reason": "Unrelated problems: one is removal of is_torch_fx_available breaking trust_remote_code models, the other is a Qwen3.5 kernel registration/decorator error." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "resize_token_embeddings/output_embeddings is not the same bug as version-dependent tokenization changes; only a broad tokenization theme is shared." }, { "left": "issue:45412", "right": "issue:46123", "accept": false, - "reason": "Memory-release bug in RT-DETR is unrelated to a MaskGenerationPipeline batching bug." + "reason": "RT-DETR memory lifecycle leak and MaskGenerationPipeline partial-batch bug are unrelated code paths." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error and Sam3LiteText auto_mappings CI breakage are different code paths." + "reason": "DeepSpeed ZeRO3 deque failure and removed Sam3LiteText auto_mappings references are different CI/runtime issues." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "Both mention embeddings, but one is tied-weight serialization and the other is resize_token_embeddings not updating decoder.embed_tokens." + "reason": "Both mention resizing/tied embeddings, but one is inconsistent serialization of lm_head.weight and the other is resize_token_embeddings not updating decoder.embed_tokens; not the same defect." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "Loading physical-intelligence/fast and protobuf exception masking are separate tokenizer issues." + "reason": "Loading a model from hub and hiding tokenizer errors when protobuf is missing are distinct tokenizer/loading behaviors." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "Sam3VisionEncoderOutput missing field and Qwen3OmniMoe video ValueError are unrelated." + "reason": "Different model families and failures: missing fpn_position_embeddings in SAM3 video vs video input ValueError in Qwen3OmniMoe." }, { - "left": "issue:44995", - "right": "issue:45200", + "left": "issue:45397", + "right": "issue:45656", "accept": false, - "reason": "Stale indexer cache crash in GlmMoeDsa is unrelated to mm_token_type_ids defaults." + "reason": "Gemma-4 ZeRO3 from_pretrained and deepspeed double optimizer.step are separate training/distributed issues." }, { - "left": "issue:45397", - "right": "issue:45656", + "left": "issue:44995", + "right": "issue:45200", "accept": false, - "reason": "Both involve DeepSpeed, but one is from_pretrained loading and the other is duplicated optimizer stepping." + "reason": "Stale indexer cache on second forward pass is unrelated to mm_token_type_ids defaulting for text-only fine-tuning." }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "Tied-weights warning in Qwen3-Next and patchtsmixer post_init policy are not the same bug." + "reason": "Incorrect tied-weights warning in Qwen3-Next and patchtsmixer post_init allowance are unrelated configuration issues." }, { "left": "issue:44479", "right": "issue:46170", "accept": false, - "reason": "Video-input regression in Qwen models is unrelated to Windows/ROCm torch.distributed import failures." + "reason": "Video input regression in Qwen models and Windows ROCm import/type error are separate failures." }, { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "Tie-word-embeddings warning during LoRA fine-tuning is unrelated to flash-attention generation failure." + "reason": "A tied-embeddings warning during LoRA fine-tuning is not the same as flash-attention generation failure." }, { "left": "issue:45685", "right": "issue:46431", "accept": false, - "reason": "MPS histogram kernel limitation in MoE and missing chat template in a processor are unrelated." + "reason": "MPS histogram support for Int in MOE is unrelated to missing chat template validation in a processor." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "BigBirdTokenizer special-token registration is not the same as the broad ESM2 breakage report." + "reason": "BigBirdTokenizer special-token decode bug and ESM2 breakage are different tokenizer/model issues." }, { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "GlmMoeDsa second-pass crash and flash_attention s_aux=None AttributeError are different problems." + "reason": "Second-forward stale cache crash and flash_attention.py s_aux=None AttributeError are not the same code-path problem." }, { "left": "issue:43065", "right": "issue:43408", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder and sam3_video/sam3_tracker type warning are different SAM3 issues." + "reason": "Dummy Conv2d in Sam3PixelDecoder and sam3_video vs sam3_tracker type mismatch are separate SAM3 issues." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Offline tokenizer cache loading is unrelated to Gemma4 multi-GPU CUDA OOM." + "reason": "Offline cache loading for tokenizers and Gemma4 multi-GPU CUDA OOM are unrelated." }, { "left": "issue:43906", "right": "issue:46202", "accept": false, - "reason": "An isolated reproduction of an external issue is not the same as a Gemma pipeline roundtrip failure." + "reason": "An isolated reproduction of an older issue and a Gemma pipeline roundtrip failure on \u2581 are not clearly the same bug." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "Tie-word-embeddings warnings and paged generate stale num_return_sequences warning are unrelated." + "reason": "Tie-word-embeddings warning and stale num_return_sequences warning are different generate-time warnings." }, { "left": "issue:44877", "right": "issue:46170", "accept": false, - "reason": "Strict config loading for granite_speech is unrelated to the Windows/ROCm import error." + "reason": "Strict config rejection for granite_speech and a Windows ROCm import failure are unrelated." }, { "left": "issue:44483", "right": "issue:46202", "accept": false, - "reason": "Chat/completions request rejection is a service/API issue, not a tokenizer roundtrip bug." + "reason": "Chat/completions request rejection and Gemma \u2581 roundtrip failure are different user-facing bugs." }, { "left": "issue:45987", "right": "issue:46082", "accept": false, - "reason": "StaticCache return-type contract and LlamaConfig head_dim validation are different concerns." + "reason": "StaticCache return type contract and LlamaConfig head_dim validation are unrelated API/config issues." }, { "left": "issue:44188", "right": "issue:46421", "accept": false, - "reason": "torch.compile branching causing diverging kernels is not the same as cache lazy-init recompilation." - }, - { - "left": "issue:43065", - "right": "issue:46123", - "accept": false, - "reason": "SAM3 pixel decoder dummy layer and mask generation final-batch handling are unrelated." + "reason": "torch.compile kernel divergence from branching and cache lazy init causing recompilation are different performance/correctness problems." }, { "left": "issue:45584", "right": "issue:46170", "accept": false, - "reason": "Whisper empty transcription after align_special_tokens is unrelated to the Windows/ROCm dependency issue." - }, - { - "left": "issue:44297", - "right": "issue:44623", - "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch and processor save_pretrained missing files are different save-path bugs." - }, - { - "left": "issue:44188", - "right": "issue:46202", - "accept": false, - "reason": "Attention-kernel divergence under torch.compile is unrelated to Gemma pipeline roundtrip failure." - }, - { - "left": "issue:44995", - "right": "issue:45584", - "accept": false, - "reason": "Stale cache crash in GlmMoeDsa and Whisper empty transcription failure do not share a code path." - }, - { - "left": "issue:44297", - "right": "issue:45459", - "accept": false, - "reason": "Tokenizer_class mismatch on save_pretrained is separate from protobuf-related decode error masking." + "reason": "Whisper empty transcription failure and Windows ROCm torch.distributed import issues are unrelated." }, { - "left": "issue:45685", + "left": "issue:43065", "right": "issue:46123", "accept": false, - "reason": "MoE MPS histogram failure and MaskGenerationPipeline partial-batch bug are unrelated." + "reason": "SAM3 pixel decoder placeholder layer and mask generation batching behavior are not the same bug." }, { - "left": "issue:44561", - "right": "issue:44877", + "left": "issue:44297", + "right": "issue:44623", "accept": false, - "reason": "Both are v5 compatibility regressions, but they affect different loading mechanisms and configs." + "reason": "Both touch save_pretrained, but one is tokenizer_class mismatch and the other is missing processor files; similar area, different failure mode." }, { - "left": "issue:44038", - "right": "issue:44977", + "left": "issue:44361", + "right": "issue:44749", "accept": false, - "reason": "Qwen3-VL-Moe bug and Qwen3.5 flash-attention generation failure are different symptoms and code paths." + "reason": "MLukeTokenizer AttributeError and a performance slowdown after upgrade are unrelated." }, { - "left": "issue:45292", - "right": "issue:45335", + "left": "issue:44749", + "right": "issue:45656", "accept": false, - "reason": "Both concern resizing embeddings, but one is output_embeddings generally and the other is decoder.embed_tokens specifically." + "reason": "Tokenizer slowdown and double optimizer.step under deepspeed are different regressions." } ] }, @@ -12525,8 +12705,8 @@ "issue:44038", "issue:44162", "issue:44188", + "issue:44222", "issue:44351", - "issue:44361", "issue:44623", "issue:44749", "issue:44869", @@ -12554,9 +12734,7 @@ "issue:46399" ], "soft_pairs": [ - "issue:44361|issue:44749", "issue:43646|issue:44162", - "issue:44749|issue:45656", "issue:44623|issue:45656", "issue:44749|issue:45310", "issue:43232|issue:43673", @@ -12568,237 +12746,245 @@ "issue:42371|issue:45092", "issue:44038|issue:44623", "issue:28282|issue:30990", - "issue:43475|issue:45381", "issue:41628|issue:44933", + "issue:43475|issue:45381", "issue:45375|issue:45478", "issue:45310|issue:45375", "issue:42994|issue:43525", "issue:43784|issue:46399", "issue:43723|issue:46326", - "issue:45406|issue:46202", "issue:44964|issue:45137", + "issue:45406|issue:46202", "issue:43232|issue:44188", "issue:43873|issue:44945", "issue:42673|issue:43704", "issue:41628|issue:44351", "issue:44945|issue:45753", "issue:45081|issue:45103", - "issue:42994|issue:43824" + "issue:42994|issue:43824", + "issue:44222|issue:45310", + "issue:44038|issue:45310" ], - "prepared_review_unit_hash": "2250fc8983789a448027ff8f171e563aa4cd0c3d29a90bf60702ba52d4dbb644", + "prepared_review_unit_hash": "de54b21be21979bfb4a2fa22a1b752e56aa7d7c4e13ec6ab95c522f64722347f", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36483, - "estimated_input_tokens": 9121, - "estimated_eval_tokens": 18498 + "serialized_chars": 36453, + "estimated_input_tokens": 9114, + "estimated_eval_tokens": 18484 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of Transformers v5 / model-loading regression reports. A few items are true duplicates or near-duplicates, but most pairings are only thematically similar and should stay separate.", - "confidence": 0.66, - "canonical_issue_reason": "Issue 43646 is the broadest representative of the cluster\u2019s main theme: Transformers v5 breaking model initialization / loading paths. It anchors the wider regression family better than any single model-specific report.", + "summary": "This cluster is mostly heterogeneous transformer-related issues, with only a couple of clear duplicate pairs. The strongest exact duplicate is the Qwen3.5 MoE from_pretrained regression (45310/45478), and the Qwen3ForCausalLM VRAM leak reports (42673/43704) also look identical.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45310 is the clearest concrete regression report and has an exact duplicate in issue:45478; both describe the same Qwen3.5 MoE from_pretrained failure on transformers>=5.4.0.", "canonical_pr_reason": null, - "best_issue_reason": "43646 is the best overall issue candidate because it is broad, clearly central to the v5 regression theme, and usable as the cluster representative without being overly niche.", + "best_issue_reason": "issue:45310 is a strong cluster representative: it is specific, actionable, and clearly duplicated by issue:45478.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44361", - "right": "issue:44749", - "accept": false, - "reason": "Different failures in different subsystems; the titles are only superficially similar." - }, { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Both are broad breakage reports, but they describe different models/code paths and are not the same bug." - }, - { - "left": "issue:44749", - "right": "issue:45656", - "accept": false, - "reason": "Unrelated problems: data filtering slowdown vs DeepSpeed optimizer stepping." + "reason": "Different bugs: missing fields from trust_remote_code AutoConfig vs ESM2 breakage." }, { "left": "issue:44623", "right": "issue:45656", "accept": false, - "reason": "Processor save_pretrained/file output issue is unrelated to the DeepSpeed optimizer-step bug." + "reason": "Different code paths: processor.save_pretrained file emission vs DeepSpeed optimizer step duplication." }, { "left": "issue:44749", "right": "issue:45310", "accept": false, - "reason": "Different topics and different failure modes; not the same underlying issue." + "reason": "Different symptoms and models: general slowdown after upgrade vs Qwen3.5 MoE from_pretrained error." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Both involve generation internals, but one is a sync_gpus/update_kwargs bug and the other is a chunked_prefill cache regression." + "reason": "Both are generation-related, but one is sync_gpus/update_kwargs behavior and the other is chunked_prefill cache handling; not the same bug." }, { "left": "issue:44871", "right": "issue:46132", "accept": false, - "reason": "EOS configuration mismatch is unrelated to AttentionInterface.register behavior changes." + "reason": "Different functionality: eos_token_id config mismatch vs AttentionInterface.register behavior change." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "Generic bug report vs a specific Gemma-4 ZeRO3 from_pretrained issue; no evidence they are the same bug." + "reason": "Generic bug report vs gemma-4 zero3 from_pretrained issue; no evidence they are the same bug." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Saving quantized models and loading Qwen2.5-GGUF models are different operations and different failure modes." + "reason": "Quantized model saving failure vs Qwen2.5-GGUF loading failure are different operations." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Whisper timestamp decoding crash and Kimi tokenizer codec regression are unrelated." + "reason": "Whisper timestamp decode crash vs Kimi-K2.5 tokenizer codec regression; different code paths." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "Different model families and different root causes; only both are v5-related breakages." + "reason": "SAM 3 video attribute error vs AutoConfig trust_remote_code missing fields; unrelated model families and failures." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 settings warning is unrelated to remote-code checkpoint meta-init compatibility." + "reason": "TF32 settings warning vs meta-initialization incompatibility for old InternVL2 checkpoints; not the same issue." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Qwen3-VL-Moe loading bug and processor.save_pretrained missing files are separate problems." + "reason": "Qwen3-VL-Moe transformers bug vs processor.save_pretrained missing files; different subsystems." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "PyTorch-not-found import error is not the same as Sentence Transformers getting stuck while loading." + "reason": "AutoModel missing PyTorch ImportError vs Sentence Transformers loading hang; different failure modes." }, { - "left": "issue:43475", - "right": "issue:45381", + "left": "issue:41628", + "right": "issue:44933", "accept": false, - "reason": "SAM 3 vision encoder attribute error and Qwen2.5-VL video position-id issue are different model-specific bugs." + "reason": "Missing AutoImageProcessor export vs image_utils import failure; related area but not the same concrete bug." }, { - "left": "issue:41628", - "right": "issue:44933", + "left": "issue:43475", + "right": "issue:45381", "accept": false, - "reason": "Both are import-related, but they involve different symbols and likely different code paths." + "reason": "Different models and symptoms: SAM 3 vision encoder attribute error vs qwen2.5-vl video position ids issue." }, { "left": "issue:45375", "right": "issue:45478", - "accept": true, - "reason": "Same Qwen3.5-Moe from_pretrained regression on transformers>=5.4.0; 45375 appears to expose the underlying missing-field cause." + "accept": false, + "reason": "Different bugs: missing deepstack_visual_indexes field vs Qwen3.5 MoE from_pretrained error." }, { "left": "issue:45310", "right": "issue:45375", - "accept": true, - "reason": "Same Qwen3.5-Moe loading failure; 45375 looks like the more specific/root-cause report for the same bug." + "accept": false, + "reason": "One is a from_pretrained failure; the other is a config-field loss under @strict. Not the same bug." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized model saving and Llama4Config pad_token_id AttributeError are unrelated." + "reason": "Quantized save failure vs Llama4Config missing pad_token_id attribute; different model/config issues." }, { "left": "issue:43784", "right": "issue:46399", "accept": false, - "reason": "Sentence-transformers import NameError and Qwen3.5 kernel registration error are different bugs." + "reason": "Sentence-transformers import NameError vs Qwen3.5 kernel registration ValueError; unrelated." }, { "left": "issue:43723", "right": "issue:46326", "accept": false, - "reason": "Tokenizer loading failure in v5 and mistral_common continue_final_message behavior are not the same issue." + "reason": "Tokenizer loading error in v5 vs continue_final_message behavior for mistral_common tokenizers; not the same underlying bug." }, { - "left": "issue:45406", - "right": "issue:46202", + "left": "issue:44964", + "right": "issue:45137", "accept": false, - "reason": "Gemma4Processor missing _tokenizer and Gemma pipeline roundtrip on \u2581 are distinct failures." + "reason": "Model loading regression vs DeepSpeed ZeRO3 deque IndexError; unrelated failure classes." }, { - "left": "issue:44964", - "right": "issue:45137", + "left": "issue:45406", + "right": "issue:46202", "accept": false, - "reason": "Phi-4 multimodal loading issue and DeepSpeed ZeRO3 deque IndexError are unrelated." + "reason": "Gemma4Processor missing _tokenizer vs Gemma pipeline roundtrip on \u2581; different bugs." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation cache handling after sync_gpus is unrelated to attention-kernel divergence under torch.compile." + "reason": "Generation kwargs update issue vs attention kernel divergence under torch.compile; different code paths." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization offloading behavior and pipeline-parallel output correctness are different problems." + "reason": "Quantization offloading issue vs incorrect LLM output with pipeline parallelism; not the same problem." }, { "left": "issue:42673", "right": "issue:43704", "accept": true, - "reason": "Exact same title and same Qwen3ForCausalLM VRAM leak under multiple dataloader threads." + "reason": "Exact duplicate reports with the same title and same Qwen3ForCausalLM VRAM leak in multiple dataloader threads." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different missing symbols; no proof of a shared bug." + "reason": "Different missing exports: AutoImageProcessor vs HybridCache." }, { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Pipeline-parallel output corruption and Qwen3_5 image-specific infinite loop are unrelated." + "reason": "Pipeline parallelism output mismatch vs Qwen3_5 infinite loop on a specific image; different symptoms." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Tokenizer regex patch crash and auto_docstring annotation parsing crash are different subsystems." + "reason": "Tokenizer regex crash vs auto_docstring crash with future annotations; unrelated bugs." }, { "left": "issue:42994", "right": "issue:43824", "accept": false, - "reason": "Quantized model saving failed and missing Qwen2_5_VLForConditionalGeneration import are not the same underlying bug." + "reason": "Quantized saving failure vs missing Qwen2_5_VLForConditionalGeneration import; different issue types." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "FP8 save_pretrained moe vs Qwen3.5 MoE from_pretrained error; different operations." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Both mention Qwen/MoE, but the reported failures and model variants differ; not enough to treat as the same bug." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Same Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0; effectively identical duplicate reports." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: most items are indeed model-loading / v5-regression reports, and the chosen representative issue is broad enough for the cluster. The soft-edge verdicts are generally conservative, with the Qwen3.5-Moe trio being the only mildly speculative area but still plausibly the same bug family." + "feedback": "The summary is grounded in the packet and stays conservative on the soft-edge pairs. The two accepted duplicates (42673/43704 and 45310/45478) are well supported by matching titles and descriptions, and the rejected pairs are framed cautiously without overclaiming. No evidence of overstated duplicate/bundle claims beyond what the packet supports." }, "evaluator_used": true, "retried": false, @@ -12840,7 +13026,6 @@ "issue:43874", "issue:43976", "issue:44016", - "issue:44038", "issue:44075", "issue:44079", "issue:44222", @@ -12859,6 +13044,7 @@ "issue:45137", "issue:45200", "issue:45310", + "issue:45356", "issue:45561", "issue:45600", "issue:45656", @@ -12869,21 +13055,19 @@ "issue:46231" ], "soft_pairs": [ - "issue:44222|issue:45310", - "issue:44038|issue:45310", "issue:43232|issue:44792", "issue:36010|issue:41628", "issue:42831|issue:44222", - "issue:42175|issue:45712", "issue:42175|issue:44933", + "issue:42175|issue:45712", "issue:44496|issue:45561", "issue:43278|issue:45137", "issue:43504|issue:45561", "issue:45561|issue:45600", "issue:43976|issue:44933", "issue:46077|issue:46123", - "issue:30990|issue:35141", "issue:30333|issue:33290", + "issue:30990|issue:35141", "issue:43377|issue:44242", "issue:42898|issue:44568", "issue:44623|issue:45310", @@ -12898,216 +13082,211 @@ "issue:42898|issue:44779", "issue:44734|issue:45103", "issue:44295|issue:45656", - "issue:44075|issue:46231" + "issue:44075|issue:46231", + "issue:42898|issue:45356" ], - "prepared_review_unit_hash": "58f8727424de1e1afe8b1a0030bb03559b6f3bb36d65828d8077bd8a8086b800", + "prepared_review_unit_hash": "332e95d59ee7ec47214eecfc626fd653aa13e4d433122ed4df5c1017ee25bca5", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36726, - "estimated_input_tokens": 9182, - "estimated_eval_tokens": 18620 + "soft_pair_count": 29, + "serialized_chars": 36508, + "estimated_input_tokens": 9127, + "estimated_eval_tokens": 18510 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster does not contain any clear duplicate pairs; the soft matches are mostly superficial title similarity across unrelated models, tokenizer/config, generation, packaging, and training bugs. No PRs are present.", - "confidence": 0.95, - "canonical_issue_reason": "Issue 45561 is the most active and concrete standalone regression report in the set, with the strongest discussion and inbound reference trail, so it is the best representative if one issue must be chosen.", + "summary": "The candidate pairs are mostly superficial similarities across unrelated issues; I would reject all soft edges. The best issue anchor is a concrete tokenizer regression report rather than a broader or less specific packaging/import issue.", + "confidence": 0.8, + "canonical_issue_reason": "No single duplicate spans the set, but issue #44779 is the best anchor: it reports a concrete v4\u2192v5 tokenizer regression with a clear symptom and scope.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45561 has the clearest reproducible bug description and the most supporting activity, making it the strongest overall issue candidate here.", + "best_issue_reason": "Issue #44779 is the strongest standalone bug report here because it is specific, reproducible, and clearly describes the failing tokenizer path.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44222", - "right": "issue:45310", - "accept": false, - "reason": "Both mention FP8/MoE, but one is about save_pretrained and the other about from_pretrained loading; different code paths and failures." - }, - { - "left": "issue:44038", - "right": "issue:45310", - "accept": false, - "reason": "These are different Qwen model-family bugs with different symptoms and versions; shared wording is not enough to make them the same fix." - }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation sync/update logic is unrelated to the janus image-generation test failure." + "reason": "Generation kwargs/caching bug vs a Janus image-generation test failure; different code paths and symptoms." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but they concern different symbols and different modules; not the same underlying missing export." + "reason": "Two different missing imports (`GenerationMixin` vs `AutoImageProcessor`); same package, but not the same underlying defect." }, { "left": "issue:42831", "right": "issue:44222", "accept": false, - "reason": "One is an accuracy regression in FineGrainedFP8, the other is FP8 save_pretrained behavior for MoE; related theme but different bug." + "reason": "FineGrainedFP8 accuracy regression vs FP8 MoE save/load issue; different failure modes." }, { "left": "issue:42175", - "right": "issue:45712", + "right": "issue:44933", "accept": false, - "reason": "Packaging/backend dependency selection is unrelated to leftover dummy class cleanup." + "reason": "Packaging/backend dependency issue vs a nonexistent `image_utils` import; unrelated root causes." }, { "left": "issue:42175", - "right": "issue:44933", + "right": "issue:45712", "accept": false, - "reason": "One is about installed backends, the other about a non-existent image_utils import; different failure modes." + "reason": "Backend install metadata vs leftover dummy classes in `dummy_pt_objects.py`; not the same bug." }, { "left": "issue:44496", "right": "issue:45561", "accept": false, - "reason": "Model config validation for an OLMo checkpoint is unrelated to pytest-xdist file races in testing utils." + "reason": "Model config loading error vs pytest-xdist file race; unrelated areas." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "A dtype mismatch during evaluation is unrelated to the DeepSpeed ZeRO3 deque underflow." + "reason": "Dtype mismatch during evaluation vs DeepSpeed ZeRO3 deque underflow; different subsystems." }, { "left": "issue:43504", "right": "issue:45561", "accept": false, - "reason": "Legacy-field model loading and xdist test file races affect different subsystems and are not the same bug." + "reason": "Legacy-field model loading failure vs xdist capture-file race; unrelated." }, { "left": "issue:45561", "right": "issue:45600", "accept": false, - "reason": "Captured-info file races in tests are unrelated to stale auto_mappings entries for removed configs." + "reason": "Parallel-test file race vs stale auto_mappings references breaking CI; different causes." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "Python-version compatibility and missing image_utils import are distinct issues." + "reason": "Python-version compatibility issue vs nonexistent import; not the same defect." }, { "left": "issue:46077", "right": "issue:46123", "accept": false, - "reason": "Attention-mask construction for Gemma4 is unrelated to batch-finalization logic in MaskGenerationPipeline." + "reason": "Attention-mask construction bug vs partial-batch pipeline result loss; separate code paths." }, { - "left": "issue:30990", - "right": "issue:35141", + "left": "issue:30333", + "right": "issue:33290", "accept": false, - "reason": "Sentence Transformers loading hang and embedding reinitialization during post_init are unrelated bugs." + "reason": "MLFlow failure-reporting problem vs DeepSpeed Adafactor OOM; unrelated." }, { - "left": "issue:30333", - "right": "issue:33290", + "left": "issue:30990", + "right": "issue:35141", "accept": false, - "reason": "MLflow reporting on failed training jobs and Adafactor/deepspeed OOM are not the same code-path problem." + "reason": "SentenceTransformer loading stall vs token-embedding resize reinit bug; different functionality." }, { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "MIMI padding-mask inconsistency and router-logits load-balancing loss are separate multimodal/MoE issues." + "reason": "MIMI padding-mask batching bug vs MoE load-balancing loss omission; not mergeable as one fix." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Tokenizer cleanup-space behavior change and missing BOS/EOS addition are different tokenizer regressions." + "reason": "Both tokenizer-related, but one is `clean_up_tokenization_spaces` behavior and the other is `add_special_tokens` BOS/EOS handling; different bugs." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Processor save_pretrained missing files is unrelated to Qwen3.5 MoE from_pretrained loading errors." + "reason": "Missing files on `processor.save_pretrained` vs Qwen3.5 MoE `from_pretrained` error; unrelated." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "VitPose model breakage and Gemma4 mm_token_type_ids defaulting are unrelated model-specific bugs." + "reason": "Broken ViTPose models vs Gemma 4 multimodal token-type defaulting; different model families and issues." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Config default initialization and Whisper word timestamp decode crash are unrelated." + "reason": "Multimodal config default initialization vs Whisper word-timestamp decode crash; unrelated." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Default sub-config initialization has nothing to do with ModelOutput key assignment semantics." + "reason": "Config defaulting bug vs `ModelOutput` key assignment bug; distinct code paths." }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "A missing image-processor method and an InternVL2 meta-init incompatibility are different multimodal initialization issues." + "reason": "Missing image-processor method vs remote-code/meta-init incompatibility; both multimodal, but different root causes." }, { "left": "issue:44016", "right": "issue:46231", "accept": false, - "reason": "Notebook syntax error and documentation parameter-name mismatch are unrelated." + "reason": "Notebook syntax error vs documentation parameter-name mismatch; not duplicates." }, { "left": "issue:42898", "right": "issue:45812", "accept": false, - "reason": "These are both tokenizer regressions, but one is cleanup spacing and the other is Granite token ID mapping; different underlying changes." + "reason": "Generic tokenizer-space cleanup regression vs Granite token-ID regression; different tokenizer behaviors and fixes." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Importing sentence-transformers with latest transformers and missing Qwen2.5-VL export are separate import failures." + "reason": "Import-time `nn` NameError vs missing Qwen2.5-VL class export; separate import failures." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "AutoTokenizer loading failure in v5 and add_special_tokens BOS/EOS behavior are distinct tokenizer problems." + "reason": "`AutoTokenizer.from_pretrained` load issue in v5 vs mDeBERTa special-token regression; same area but not the same bug." }, { "left": "issue:42898", "right": "issue:44779", "accept": false, - "reason": "Deepseek tokenizer wrong outputs and clean_up_tokenization_spaces changes are different tokenizer regressions." + "reason": "Both are tokenizer regressions, but one is a global cleanup-behavior change and the other is DeepSeek-specific codec handling; not the same concrete defect." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "KV-cache continuation tensor indexing in serve is unrelated to auto_docstring crashes under future annotations." + "reason": "Serve/KV-cache tensor indexing crash vs auto-docstring annotation handling; unrelated." }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "position_ids buffer handling and duplicate optimizer steps under deepspeed are unrelated training/runtime bugs." + "reason": "Buffer registration error vs optimizer stepping twice under DeepSpeed; different code paths." }, { "left": "issue:44075", "right": "issue:46231", "accept": false, - "reason": "SGD arguments not being used is unrelated to a docs parameter-name mismatch." + "reason": "Optimizer argument wiring bug vs docs typo for training parameters; not a duplicate bug." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer cleanup behavior regression vs Kimi-K2.5 codec/warning regression; both tokenizers, but different underlying issues." } ] }, @@ -13168,7 +13347,6 @@ "issue:45125", "issue:45245", "issue:45310", - "issue:45356", "issue:45381", "issue:45405", "issue:45656", @@ -13181,22 +13359,21 @@ "issue:46431" ], "soft_pairs": [ - "issue:42898|issue:45356", "issue:42898|issue:43295", - "issue:45405|issue:45712", "issue:44060|issue:45125", + "issue:45405|issue:45712", "issue:44222|issue:44297", "issue:27957|issue:30333", - "issue:44933|issue:45712", "issue:30333|issue:34689", + "issue:44933|issue:45712", "issue:44297|issue:45656", "issue:30333|issue:36331", "issue:44222|issue:44623", "issue:42831|issue:44623", "issue:30333|issue:35141", "issue:43479|issue:43644", - "issue:44448|issue:46132", "issue:22355|issue:28282", + "issue:44448|issue:46132", "issue:42796|issue:43723", "issue:45245|issue:46132", "issue:41628|issue:42757", @@ -13207,226 +13384,220 @@ "issue:46209|issue:46431", "issue:45381|issue:46305", "issue:43122|issue:46209", + "issue:44373|issue:44485", "issue:43723|issue:44977", "issue:44222|issue:44484", - "issue:44373|issue:44485", "issue:44206|issue:45103", "issue:43232|issue:46424" ], - "prepared_review_unit_hash": "39fc6aa62106f4a84adfb027c7aaff87e62a9c9e6ab0d5ad5f5c7290109967b1", + "prepared_review_unit_hash": "b0a15ded78b8ed5e3b6e0bf11748342cc36ae62fa0ec4669c8f53d5c0eb3deea", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 36813, - "estimated_input_tokens": 9204, - "estimated_eval_tokens": 18664 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35894, + "estimated_input_tokens": 8974, + "estimated_eval_tokens": 18204 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is very heterogeneous; most items are unrelated regressions or feature-specific bugs that only share a broad Transformers-v5/tokenizer vibe. No pull requests are present. I\u2019d use a broad tokenizer-regression issue as the canonical anchor, but I would not merge any of the soft-edge pairs.", - "confidence": 0.92, - "canonical_issue_reason": "Issue 42898 is the broadest tokenization-regression report in the set and is the best single anchor for the cluster, even though the surrounding issues are mostly not true duplicates.", + "summary": "The items are largely unrelated Transformers issues spanning tokenization, multimodal models, FP8, config validation, and import/runtime regressions. None of the soft pairs look like the same underlying bug or mergeable change, so all soft edges are rejected.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43644 is the broadest regression-style report in the set, describing a Transformers 5.0.0-wide buffer initialization problem rather than a narrow model-specific edge case.", "canonical_pr_reason": null, - "best_issue_reason": "42898 is the most reusable representative because it targets a generic v5 tokenizer behavior change; the other issues are either narrower model-specific regressions or clearly unrelated subsystems.", + "best_issue_reason": "issue:43644 is a strong standalone bug report with a clear, version-scoped failure mode and broad user impact; it is the best representative issue here, though not a duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42898", - "right": "issue:45356", + "right": "issue:43295", "accept": false, - "reason": "Both involve tokenizer regressions, but one is a generic v5 cleanup-space change and the other is a Kimi tokenizer codec/BPE handling bug; different code paths." + "reason": "Both are v5 regressions, but one is tokenization cleanup behavior and the other is processor/tokenizer image-handling breakage; different code paths." }, { - "left": "issue:42898", - "right": "issue:43295", + "left": "issue:44060", + "right": "issue:45125", "accept": false, - "reason": "Both mention v5/tokenizer behavior, but 43295 is about processor.tokenizer/image passing regression, not the same issue as clean_up_tokenization_spaces." + "reason": "Different Qwen model-specific bugs: tied-weights warning versus missing tensor-parallel plan." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "Unrelated: one is a dependency version bump, the other is leftover dummy classes/import surface cleanup." - }, - { - "left": "issue:44060", - "right": "issue:45125", - "accept": false, - "reason": "Different Qwen3/MoE problems: tied-weights warning vs missing tensor-parallel plan." + "reason": "Unrelated: a dependency version bump issue versus leftover dummy class cleanup/import leakage." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "Both are save_pretrained-related, but one is FP8 MoE checkpoint saving and the other is tokenizer_class metadata mismatch." + "reason": "Both mention save_pretrained, but one is FP8 MoE checkpoint saving and the other is tokenizer_class metadata mismatch." }, { "left": "issue:27957", "right": "issue:30333", "accept": false, - "reason": "Completely different topics: Flash Attention support vs MLFlow reporting after failed training." + "reason": "No overlap in underlying problem: Flash Attention model support versus MLflow reporting after failed training." }, { - "left": "issue:44933", - "right": "issue:45712", + "left": "issue:30333", + "right": "issue:34689", "accept": false, - "reason": "Both touch imports, but one is a missing image_utils import and the other is dummy_pt_objects cleanup; not the same bug." + "reason": "Different failures: MLflow run-state reporting versus Llama 3.2 Vision model loading regression." }, { - "left": "issue:30333", - "right": "issue:34689", + "left": "issue:44933", + "right": "issue:45712", "accept": false, - "reason": "Different failures: MLFlow job tracking vs Llama 3.2 Vision model loading regression." + "reason": "Both touch imports/exports, but one is a missing image_utils import and the other is dummy class leakage in no-torch mode." }, { "left": "issue:44297", "right": "issue:45656", "accept": false, - "reason": "Tokenizer save_pretrained metadata issue is unrelated to a deepspeed optimizer-step double-call." + "reason": "Tokenizer save metadata bug is unrelated to DeepSpeed double optimizer stepping." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "Different training-stack failures: MLFlow reporting vs CustomTrainer compute_loss signature change." + "reason": "Training telemetry after failure is unrelated to the Trainer compute_loss signature change." }, { "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "Both concern save_pretrained, but model FP8 saving and processor file saving are different code paths." + "reason": "FP8 MoE save_pretrained and processor.save_pretrained missing files are different serialization paths." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "FP8 accuracy concerns are unrelated to processor.save_pretrained missing files." + "reason": "FP8 accuracy issue is unrelated to processor file saving." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "No overlap beyond general training/model behavior; these are different bugs." + "reason": "MLflow failure reporting is unrelated to embedding resize/post_init reinitialization." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Different areas: multimodal config default initialization vs non-persistent buffer fill behavior." + "reason": "Different problems: multimodal config defaults versus non-persistent buffer initialization." }, { - "left": "issue:44448", - "right": "issue:46132", + "left": "issue:22355", + "right": "issue:28282", "accept": false, - "reason": "PEGASUS output regression and AttentionInterface.register behavior change are unrelated." + "reason": "Both are import errors, but one is missing transformers.onnx and the other is missing PyTorch for AutoModel." }, { - "left": "issue:22355", - "right": "issue:28282", + "left": "issue:44448", + "right": "issue:46132", "accept": false, - "reason": "Both are import errors, but one is missing transformers.onnx and the other is missing PyTorch for AutoModel; not the same root cause." + "reason": "Different regressions: Pegasus output changes versus AttentionInterface.register behavior." }, { "left": "issue:42796", "right": "issue:43723", "accept": false, - "reason": "Both tokenizer-related, but one is BPE marker handling and the other is AutoTokenizer loading in v5; different bugs." + "reason": "Tokenizer BPE-marker handling is not the same as AutoTokenizer.from_pretrained failing in v5." }, { "left": "issue:45245", "right": "issue:46132", "accept": false, - "reason": "Category-count runtime error is unrelated to AttentionInterface.register behavior changes." + "reason": "No shared bug: category-count runtime error versus attention registration semantics." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Different missing-import failures from different packages/APIs." + "reason": "Different missing imports from different packages; not the same failure." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer_class metadata mismatch is not the same as the Qwen3.5 MoE from_pretrained failure." + "reason": "Tokenizer class metadata mismatch is unrelated to Qwen3.5 MoE from_pretrained errors." }, { "left": "issue:43475", "right": "issue:46305", "accept": false, - "reason": "SAM 3 Video missing attribute and Moonshine sliding-window off-by-one are unrelated model bugs." + "reason": "Different model components and symptoms: SAM 3 video attribute error versus Moonshine streaming mask off-by-one." }, { "left": "issue:45071", "right": "issue:46082", "accept": false, - "reason": "Both are config-related, but one is type checking in v5 and the other is Llama head_dim validation; different concrete issues." + "reason": "Both are config-related, but one is type-checking regression and the other is head_dim validation logic." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "Torch.split return-value handling in a MoE indexer is unrelated to remote-code checkpoint meta-initialization compatibility." + "reason": "torch.split indexer behavior is unrelated to remote-code checkpoint meta-init incompatibility." }, { "left": "issue:46209", "right": "issue:46431", "accept": false, - "reason": "FP8 inference on B200 and missing chat template in a processor are separate problems." + "reason": "FP8 inference on B200 and missing chat template are unrelated issues." }, { "left": "issue:45381", "right": "issue:46305", "accept": false, - "reason": "Both involve sequence/video masking logic, but one is vision_position_ids and the other is a sliding-window mask off-by-one." + "reason": "Both are sequence-position/window bugs, but in different models and code paths." }, { "left": "issue:43122", "right": "issue:46209", "accept": false, - "reason": "Tokenization output changes and FP8 inference on B200 are unrelated." + "reason": "Tokenization output drift across versions is unrelated to FP8 inference failure on a specific GPU." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A docstring mistake is not the same as a RoPE implementation discussion/bug." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "AutoTokenizer loading regression is unrelated to Qwen3.5 flash-attention generation issues." + "reason": "Tokenizer loading failure and flash-attention generation bug are different problems." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "FP8 MoE save_pretrained and max_shard_size default discussion are unrelated." - }, - { - "left": "issue:44373", - "right": "issue:44485", - "accept": false, - "reason": "A docstring mistake is not the same as a GLM-5 RoPE implementation discussion." + "reason": "Checkpoint saving in FP8 MoE is unrelated to the max_shard_size default question." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "Unsupported center argument in LasrFeatureExtractor is unrelated to auto_docstring / future-annotations crash." + "reason": "Feature extractor argument validation and auto-docstring annotation handling are unrelated." }, { "left": "issue:43232", "right": "issue:46424", "accept": false, - "reason": "Both are generation/cache-related, but one is sync_gpus kwargs update and the other is max_cache_length honoring only quantized cache; different code paths." + "reason": "Generation kwargs update after sync_gpus is unrelated to cache max length handling." } ] }, @@ -13504,10 +13675,10 @@ "issue:44295|issue:44623", "issue:43232|issue:45834", "issue:44190|issue:46082", - "issue:44279|issue:46328", "issue:43723|issue:44871", "issue:44263|issue:44360", "issue:44261|issue:45092", + "issue:44279|issue:46328", "issue:45092|issue:45357", "issue:42907|issue:46050", "issue:30990|issue:38175", @@ -13522,27 +13693,27 @@ "issue:43408|issue:46402", "issue:22355|issue:27957", "issue:42831|issue:45753", - "issue:22355|issue:28218", "issue:42757|issue:45278", + "issue:22355|issue:28218", "issue:39401|issue:44568", "issue:27957|issue:30990", "issue:42886|issue:43066", "issue:44368|issue:46305" ], - "prepared_review_unit_hash": "e81e640b3a41ad9f0ec434e7874b37c9394a5d9d679a3e629deb01871639a50e", + "prepared_review_unit_hash": "c803a54736ce0b8e2e15e582037e231037576805798f79a81987edd2dfca593f", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 35815, + "serialized_chars": 35814, "estimated_input_tokens": 8954, "estimated_eval_tokens": 18164 }, @@ -13551,174 +13722,180 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated bug reports, and the soft pairs look like false positives rather than true duplicates. The closest matches are still separated by different code paths, models, or failure modes, so I would not merge any pair.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 46050 is the most concrete, self-contained regression report in this set: it clearly describes a save/load round-trip failure for a quantized model, making it the best representative issue if one must be chosen.", + "summary": "This set is highly heterogeneous: the soft-similarity pairs share at most broad subsystem overlap, but not the same concrete bug or change. I rejected all soft edges as duplicates, and the best representative issue is the broad import-regression report rather than any narrow symptom report.", + "confidence": 0.84, + "canonical_issue_reason": "If a single representative is needed, issue 45278 is the broadest umbrella for the import-error regression subset; however, the overall cluster is not a true duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 46050 is the strongest standalone report here because it is specific, reproducible, and points to a clear serialization bug around quantized model persistence.", + "best_issue_reason": "Issue 45278 is the most general, user-facing regression report and best captures the broadest slice of the related import-breakage complaints in this mixed cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44734", "right": "issue:46421", "accept": false, - "reason": "Both mention cache/generation, but one is a serve-time KV indexing crash and the other is lazy cache initialization causing recompilation; different bugs." + "reason": "Both mention cache-related serving paths, but one is a KV cache continuation crash and the other is lazy cache init causing recompilation; different bugs." }, { "left": "issue:44295", "right": "issue:44623", "accept": false, - "reason": "Position_ids buffer access and processor.save_pretrained missing files are unrelated failure modes." + "reason": "One is about reading a bufferized position_ids tensor, the other about processor.save_pretrained missing files; unrelated failure modes." }, { "left": "issue:43232", "right": "issue:45834", "accept": false, - "reason": "Generation kwargs after sync_gpus and a long OCR input index error are different code paths and different symptoms." + "reason": "Generation kwargs after sync_gpus and a Kosmos2.5 OCR index error are different code paths and symptoms." }, { "left": "issue:44190", "right": "issue:46082", "accept": false, - "reason": "A local dataset loading script issue is not the same as LlamaConfig head_dim validation behavior." - }, - { - "left": "issue:44279", - "right": "issue:46328", - "accept": false, - "reason": "Generic dependency trouble and a padding performance bug have no concrete overlap." + "reason": "A local dataset loading script issue is not the same as a LlamaConfig head_dim validation error." }, { "left": "issue:43723", "right": "issue:44871", "accept": false, - "reason": "Tokenizer loading failure and Gemma eos_token_id mismatch are distinct tokenizer regressions." + "reason": "Tokenizer loading in v5 and Gemma-3 eos_token_id config mismatch are separate tokenizer/config bugs." }, { "left": "issue:44263", "right": "issue:44360", "accept": false, - "reason": "Both touch the DSA indexer, but one is about torch.split outputs and the other is about a missing ReLU; not the same bug." + "reason": "Both reference DSA indexer code, but one concerns torch.split outputs and the other a missing ReLU; not the same defect." }, { "left": "issue:44261", "right": "issue:45092", "accept": false, - "reason": "MLA precision error from missing rms_norm_eps is unrelated to old InternVL2 meta-initialization incompatibility." + "reason": "Missing rms_norm_eps in MLA q_a_layernorm and remote-code meta-init incompatibility are different multimodal issues." + }, + { + "left": "issue:44279", + "right": "issue:46328", + "accept": false, + "reason": "A generic dependency issue is not the same as a padding performance regression." }, { "left": "issue:45092", "right": "issue:45357", "accept": false, - "reason": "Checkpoint meta-init incompatibility and incorrect visual-encoder keys on save_pretrained are separate serialization/loading problems." + "reason": "One is an old InternVL2 meta-init compatibility problem, the other is Qwen3.5 save_pretrained writing wrong visual encoder keys." }, { "left": "issue:42907", "right": "issue:46050", "accept": false, - "reason": "Both concern quantization persistence, but one is about dequantized Ministrals/Devstrals saving and the other is a SINQ round-trip breakage; too different to treat as duplicates." + "reason": "Different quantization/save-load failures: dequantized Ministrals/Devstrals versus sinq quantized models." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Sentence Transformers loading stalls and SIGLIP2 zero probabilities are unrelated issues." + "reason": "Sentence Transformers loading hangs and SigLIP2 zero probabilities are unrelated." }, { "left": "issue:43644", "right": "issue:46305", "accept": false, - "reason": "Non-persistent buffer junk and a sliding-window off-by-one bug are different correctness problems." + "reason": "Non-persistent buffer junk and a streaming mask off-by-one are different bugs in different subsystems." }, { "left": "issue:42831", "right": "issue:45310", "accept": false, - "reason": "FP8 accuracy loss and a Qwen3.5-Moe from_pretrained error are unrelated." + "reason": "FineGrainedFP8 accuracy issues and Qwen3.5 Moe from_pretrained failures are not the same problem." }, { "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "A functional DSA indexer bug and a docstring issue are not duplicates." + "reason": "A runtime torch.split bug and a wrong docstring are not duplicates." }, { "left": "issue:44829", "right": "issue:46002", "accept": false, - "reason": "Degenerate training with flash_attention_3 and NaNs during checkout/init are both instability reports, but not the same concrete code-path bug." + "reason": "Training collapse with flash_attention_3 and NaNs on checkout/init point to different root causes." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "Chat/completions request rejection and Gemma4Processor missing _tokenizer are unrelated server-side failures." + "reason": "Chat-completions request rejection and Gemma4Processor missing _tokenizer are separate serving failures." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "A FineGrainedFP8 accuracy issue and a Qwen3-VL-Moe bug are different problems." + "reason": "A numerical accuracy issue and a Qwen3-VL-Moe bug are unrelated." }, { "left": "issue:41628", "right": "issue:43824", "accept": false, - "reason": "Missing AutoImageProcessor and missing Qwen2_5_VLForConditionalGeneration imports are different symbols and likely separate export regressions." + "reason": "Both are import errors, but for different symbols and model components; no evidence of the same missing export." }, { "left": "issue:43479", "right": "issue:44913", "accept": false, - "reason": "Default multimodal config initialization and GPTNeoX rotary_pct reload persistence are distinct config bugs." + "reason": "Default multimodal config initialization and GPTNeoX rotary_pct reload behavior are distinct config bugs." }, { "left": "issue:43408", "right": "issue:46402", "accept": false, - "reason": "A sam3_video/sam3_tracker type warning is not the same as text encoder weights not being loaded." + "reason": "A sam3 tracker/model-type warning and missing Sam3Model text encoder weights are different SAM3 issues." }, { "left": "issue:22355", "right": "issue:27957", "accept": false, - "reason": "Missing transformers.onnx and XLMRoberta Flash Attention 2 are unrelated issues." + "reason": "Missing transformers.onnx and XLMRoberta Flash Attention 2 are unrelated." }, { "left": "issue:42831", "right": "issue:45753", "accept": false, - "reason": "An FP8 accuracy issue and a Qwen3_5 infinite-loop bug are different failure modes." + "reason": "FineGrainedFP8 accuracy and a Qwen3_5 image-specific infinite loop are different defects." }, { - "left": "issue:22355", - "right": "issue:28218", + "left": "issue:42757", + "right": "issue:45278", "accept": false, - "reason": "Missing transformers.onnx and tokenizer spacing after added tokens are unrelated." + "reason": "The is_offline_mode import error may be one symptom in the broader upgrade report, but this pair does not clearly describe the same concrete bug." }, { - "left": "issue:42757", - "right": "issue:45278", + "left": "issue:22355", + "right": "issue:28218", "accept": false, - "reason": "The broad upgrade report mentions many import errors, but it is too umbrella-like to be treated as the same concrete bug as the specific is_offline_mode import failure." + "reason": "An import failure for transformers.onnx is unrelated to tokenizer spacing after added tokens." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Wrong offset_mapping and missing BOS/EOS insertion are distinct tokenizer behaviors on different models." + "reason": "Offset-mapping correctness and add_special_tokens BOS/EOS behavior are separate tokenizer bugs." + }, + { + "left": "issue:27957", + "right": "issue:30990", + "accept": false, + "reason": "Flash Attention 2 on XLMRoberta and Sentence Transformers loading hang are unrelated." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Offline cache loading and wrong decoder type in v5 are unrelated issues." + "reason": "Offline cache loading and wrong tokenizer decoder type are both tokenizer issues, but not the same underlying bug." }, { "left": "issue:44368", "right": "issue:46305", "accept": false, - "reason": "A tie_word_embeddings warning and a sliding-window mask off-by-one error are not the same bug." + "reason": "A tie_word_embeddings warning in Qwen3.5 LoRA and a streaming mask off-by-one are unrelated." } ] }, @@ -13797,247 +13974,247 @@ "issue:43408|issue:46369", "issue:42886|issue:42994", "issue:44188|issue:44373", - "issue:43646|issue:46002", - "issue:43550|issue:46132", "issue:43576|issue:43784", + "issue:43550|issue:46132", + "issue:43646|issue:46002", "issue:42371|issue:44263", - "issue:44295|issue:44297", "issue:44261|issue:44373", + "issue:44295|issue:44297", "issue:44117|issue:44704", "issue:42907|issue:44945", "issue:30990|issue:33357", "issue:43756|issue:44060", + "issue:41628|issue:45278", "issue:42907|issue:43937", "issue:43976|issue:44945", - "issue:41628|issue:45278", "issue:43784|issue:45600", "issue:39401|issue:44488", "issue:42907|issue:44964", "issue:44704|issue:45290", "issue:44360|issue:44485", "issue:39401|issue:44521", - "issue:43122|issue:44521", "issue:43122|issue:46326", + "issue:43122|issue:44521", "issue:44038|issue:44297", "issue:44625|issue:44829", "issue:43116|issue:45200", "issue:44038|issue:44295", "issue:43408|issue:43475" ], - "prepared_review_unit_hash": "b8c3571a6631d9252e3518e72bb151574231d8b69cbb3924fc74aff888642368", + "prepared_review_unit_hash": "5e7a24e52b5921dda04bd795d046baf80e207111352d7f44d7bd6db65aca3f2a", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36538, - "estimated_input_tokens": 9135, - "estimated_eval_tokens": 18526 + "serialized_chars": 36536, + "estimated_input_tokens": 9134, + "estimated_eval_tokens": 18524 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose semantic bucket of unrelated Transformers regressions rather than a true duplicate set. The broadest issue is the import-error umbrella (#45278), but none of the soft pairs look like the same concrete bug/change, so all should stay separate.", - "confidence": 0.86, - "canonical_issue_reason": "#45278 is the broadest user-facing regression report in the set (many import errors after upgrading), so it is the closest thing to an umbrella issue, though it is still too generic to represent the whole cluster cleanly.", + "summary": "This cluster is mostly a loose semantic collection of unrelated Transformers bug reports. None of the soft pairs look like true duplicates or mergeable PRs; they span different models, APIs, and failure modes. The broadest umbrella issue is 45278 (general import regressions after upgrading to v5), but it is only a loose representative, not a real duplicate anchor.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 45278 is the broadest, most general regression report in the set and best serves as an umbrella for the v5 import-breakage subset, though the overall cluster is heterogeneous.", "canonical_pr_reason": null, - "best_issue_reason": "#45278 is the best representative only because it is the widest-scoped issue among otherwise heterogeneous tokenizer/model/API breakages; it is not a true duplicate target for the rest.", + "best_issue_reason": "45278 is the most suitable global representative because it describes a wide import-regression class rather than a single model-specific failure, so it is the closest thing to a canonical issue here.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42796", "right": "issue:43122", "accept": false, - "reason": "Both involve tokenization, but one is a model-specific BPE marker bug and the other is a version-to-version tokenization difference; not the same defect." + "reason": "Different bugs: BPE marker handling for Ministral-3 vs tokenizer behavior changing across versions for another model." }, { "left": "issue:43232", "right": "issue:45593", "accept": false, - "reason": "Different code paths: generation kwargs after sync_gpus vs auxiliary losses in D-FINE." + "reason": "Unrelated code paths: generation kwargs after sync_gpus vs D-FINE auxiliary losses when denoising is off." }, { "left": "issue:43408", "right": "issue:46369", "accept": false, - "reason": "Same SAM3 family, but one is a model-type/config warning and the other is a memory overflow; distinct problems." + "reason": "Different SAM3 problems: a model-type warning during tracker init vs a memory overflow in sam3-litetext." }, { "left": "issue:42886", "right": "issue:42994", "accept": false, - "reason": "Offline tokenizer cache loading and quantized model saving are unrelated failure modes." + "reason": "Different failure modes: offline tokenizer cache loading vs quantized model saving." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "Attention-kernel branching under torch.compile is unrelated to a docstring issue for position_ids." + "reason": "torch.compile kernel divergence vs a docstring typo; not the same underlying bug." }, { - "left": "issue:43646", - "right": "issue:46002", + "left": "issue:43576", + "right": "issue:43784", "accept": false, - "reason": "Both mention initialization, but one is custom model init breakage and the other is NaNs after checkout/init; not the same bug." + "reason": "One is a broken CLI env command; the other is a NameError on importing sentence-transformers." }, { "left": "issue:43550", "right": "issue:46132", "accept": false, - "reason": "torch.compile + SDPA failure for Bamba is a different issue from AttentionInterface.register changing registered-function behavior." + "reason": "Bamba torch.compile/SDPA issue and AttentionInterface.register behavior change are separate bugs." }, { - "left": "issue:43576", - "right": "issue:43784", + "left": "issue:43646", + "right": "issue:46002", "accept": false, - "reason": "A broken CLI command and a NameError while importing sentence-transformers are separate regressions." + "reason": "Both are v5-era regressions, but one is custom model init breakage and the other is NaNs after checkout/init." }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "TF32 control settings and a torch.split return-value issue in a GLM indexer are unrelated." + "reason": "TF32 API warning and torch.split return values in a DSA indexer are unrelated." }, { - "left": "issue:44295", - "right": "issue:44297", + "left": "issue:44261", + "right": "issue:44373", "accept": false, - "reason": "Reading position_ids after buffer registration is unrelated to tokenizer_class mismatch on save_pretrained." + "reason": "Precision/config issue in MLA q_a_layernorm vs a wrong docstring; not the same bug." }, { - "left": "issue:44261", - "right": "issue:44373", + "left": "issue:44295", + "right": "issue:44297", "accept": false, - "reason": "Missing rms_norm_eps precision in MLA q_a_layernorm is not the same as a docstring mismatch." + "reason": "Reading a registered position_ids buffer and tokenizer_class mismatch on save_pretrained are different issues." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "Both touch auto-loading, but one is TOKENIZER_MAPPING_NAMES returning None and the other is cached_file kwargs propagation." + "reason": "TOKENIZER_MAPPING_NAMES None in from_pretrained vs AutoProcessor not forwarding kwargs to cached_file." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Quantized model saving/dequantization and incorrect pipeline-parallel outputs are different bugs." + "reason": "Saving dequantized Ministrals/Devstrals and incorrect pipeline-parallel outputs are unrelated." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Both are load-time failures, but Sentence Transformers loading stall and a MacOS bus error on CLIP are not the same underlying issue." + "reason": "Both mention loading, but one is Sentence Transformers hanging and the other is a MacOS bus error with a CLIP model." }, { "left": "issue:43756", "right": "issue:44060", "accept": false, - "reason": "RoPE-layer dropping in Smollm3 and a tied-weights warning in Qwen3-Next are unrelated model-architecture issues." + "reason": "Smollm3 RoPE layer count bug vs Qwen3-Next tied-weights warning; different model configs." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Related only at a broad v5-import level; missing AutoImageProcessor is not the same concrete bug as the wider import-error report." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Quantized model saving and invalid GenerationConfig for GLM-5 do not share a concrete code path." + "reason": "Quantized/dequantized save failure vs invalid GenerationConfig validation; unrelated." }, { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Python version compatibility in v5.1.0 and pipeline-parallel output corruption are unrelated." - }, - { - "left": "issue:41628", - "right": "issue:45278", - "accept": false, - "reason": "#45278 is a broad import-error umbrella, but this AutoImageProcessor import failure is only one symptom, not the same standalone bug." + "reason": "Python version compatibility breakage vs pipeline parallelism output corruption are distinct." }, { "left": "issue:43784", "right": "issue:45600", "accept": false, - "reason": "Import-time NameError in sentence-transformers is unrelated to stale auto_mappings references to removed Sam3LiteText configs." + "reason": "Import NameError in sentence-transformers vs stale auto_mappings references to removed Sam3LiteText configs." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Wrong offset_mapping for Qwen3 tokenization and failing to load a specific model are separate issues." + "reason": "Wrong offset_mapping for Qwen3 tokenizer vs inability to load cjvt/sleng-bert; different tokenizer/model problems." }, { "left": "issue:42907", "right": "issue:44964", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals is unrelated to loading Phi-4 multimodal instruct." + "reason": "Model saving failure vs model loading failure for Phi-4 multimodal instruct." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "Passing kwargs to cached_file in AutoProcessor is unrelated to apply_chat_template crashing on tool-call assistant messages." + "reason": "AutoProcessor kwargs forwarding and apply_chat_template crashes with tool calls are separate code paths." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "A missing ReLU in the DSA indexer is not the same as GLM-5 RoPE implementation issues." + "reason": "DSA indexer ReLU issue vs GLM-5 RoPE implementation; unrelated." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Tokenizer offset_mapping corruption and all-zero assistant masks in multimodal chat templating are different bugs." + "reason": "Tokenizer offset_mapping bug vs multimodal assistant_mask handling; not the same defect." }, { "left": "issue:43122", - "right": "issue:44521", + "right": "issue:46326", "accept": false, - "reason": "Different tokenization behavior across versions is not the same as assistant_mask generation for multimodal inputs." + "reason": "Version-to-version tokenization differences vs continue_final_message behavior for mistral_common tokenizers." }, { "left": "issue:43122", - "right": "issue:46326", + "right": "issue:44521", "accept": false, - "reason": "Version-dependent tokenization changes and continue_final_message handling for mistral_common are different tokenizer behaviors." + "reason": "Different symptoms and subsystems: tokenization drift vs apply_chat_template assistant mask generation." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Qwen3-VL-Moe model incompatibility and tokenizer_class mismatch on save are unrelated." + "reason": "Qwen3-VL-Moe bug and tokenizer_class mismatch on save_pretrained are unrelated." }, { "left": "issue:44625", "right": "issue:44829", "accept": false, - "reason": "num_labels propagation in Qwen3.5 is unrelated to flash_attention_3 training degeneration." + "reason": "num_labels propagation in Qwen3.5 vs flash_attention_3 causing degenerate training; different bugs." }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "Empty multi-label classification outputs in an example script and Gemma 4 mm_token_type_ids defaults are not the same issue." + "reason": "Empty multi-label classification results vs mm_token_type_ids defaulting to zeros are not the same underlying issue." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Qwen3-VL-Moe loading problems and position_ids buffer access errors are distinct." + "reason": "Qwen3-VL-Moe generation bug vs position_ids buffer access error; different failures." }, { "left": "issue:43408", "right": "issue:43475", "accept": false, - "reason": "Both are SAM3-related, but one is a model-type mismatch warning and the other is a missing output attribute; not the same concrete bug." + "reason": "SAM3 model-type warning and missing fpn_position_embeddings are related by family, but not the same concrete bug or fix." } ] }, @@ -14137,212 +14314,218 @@ "issue:44625|issue:46002", "issue:43976|issue:45446", "issue:44075|issue:44936", - "issue:42222|issue:43441", "issue:45357|issue:45478", + "issue:42222|issue:43441", "issue:43116|issue:43825" ], - "prepared_review_unit_hash": "f077e0e848ab2da9584bd602ce4011a914a9d8dcc089bd684994c9f9d58974ba", + "prepared_review_unit_hash": "3004ce0dbef9f429b44bf859bff8e2a5be75353cf12d3f552d15a2ec63d4ae72", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 35876, - "estimated_input_tokens": 8969, - "estimated_eval_tokens": 18194 + "serialized_chars": 35872, + "estimated_input_tokens": 8968, + "estimated_eval_tokens": 18192 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated regression reports across different models, APIs, and error classes. None of the soft pairs look like the same underlying bug/change, so I reject all soft edges. The cluster is heterogeneous; the broadest representative is the general import-regression report, but it is not a true duplicate hub.", - "confidence": 0.88, - "canonical_issue_reason": "issue:45278 is the broadest, most representative regression report in the set (many import errors after upgrading), but the cluster does not contain a clear duplicate set centered on it.", + "summary": "The set is broadly heterogeneous. A few issues share high-level themes like import failures, tokenization regressions, or Qwen3.5 model regressions, but none are safe duplicate matches under strict triage. If forced to pick an anchor, issue 45278 is the broadest umbrella for import-regression reports.", + "confidence": 0.84, + "canonical_issue_reason": "issue 45278 is the best umbrella anchor because it describes a broad post-upgrade import-failure regression and can plausibly subsume related import-error reports better than the more narrow model-specific tickets.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the best overall representative because it is the most general and could subsume related import-regression reports, though the evidence is still too mixed to treat as a true canonical duplicate.", + "best_issue_reason": "issue 45278 is the most generally representative issue in this set; it is broad enough to serve as a triage hub, even though the overall cluster is not a true duplicate group.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Different failure modes: saving dequantized Ministrals/Devstrals vs gguf_file loading via AutoTokenizer/AutoModelForCausalLM." + "reason": "Different failure modes: saving dequantized models vs.from_pretrained/gguf loading." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Both mention MoE/FP8, but one is a model support bug and the other is a save_pretrained serialization issue; not the same code-path." + "reason": "Both mention Qwen3.5/MoE, but one is a general 5.0 regression and the other is FP8 save_pretrained; different code paths." }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but they involve different symbols and different modules." + "reason": "Both are import errors, but they involve different missing symbols and different modules." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both concern Qwen3.5, but one is broken save_pretrained checkpoints and the other is a from_pretrained load error for MoE." + "reason": "Related to Qwen3.5 save/load, but one is save_pretrained checkpoint corruption and the other is a from_pretrained error." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions in v5, but they target different tokenizers and likely different underlying fixes." + "reason": "Both concern tokenizer behavior changes, but they refer to different models and different regressions." }, { "left": "issue:22355", "right": "issue:29159", "accept": false, - "reason": "No meaningful overlap: ONNX import failure vs slow/fast tokenizer inconsistency." + "reason": "No module import failure for transformers.onnx is unrelated to slow vs fast tokenizer inconsistency." }, { "left": "issue:42757", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but from different packages and unrelated missing names." + "reason": "Same broad import-error theme, but different missing symbols from different packages." }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Tensor-parallel training bug vs example-script multi-label classification output; unrelated issues." + "reason": "Tensor-parallel training bug and empty multi-label classification results are different problems." }, { "left": "issue:43278", "right": "issue:44945", "accept": false, - "reason": "Different symptoms and paths: dtype change in evaluate vs incorrect output under pipeline parallelism." + "reason": "Different regressions: dtype drift in evaluation vs incorrect output under pipeline parallelism." }, { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Packaging backend omission vs Python 3.14 load failure; unrelated." + "reason": "Backend packaging issue vs Python 3.14 load failure; not the same bug." }, { "left": "issue:43493", "right": "issue:45753", "accept": false, - "reason": "Different multimodal model problems: SigLIP2 implementation discrepancy vs Qwen3_5 infinite loop on a specific image." + "reason": "Distinct model-specific regressions affecting different behaviors and likely different code paths." }, { "left": "issue:22355", "right": "issue:29127", "accept": false, - "reason": "Transformers.onnx import error is unrelated to a LayoutLMv3 error-message clarity issue." + "reason": "Transformers.onnx import failure is unrelated to LayoutLMv3 box-validation messaging." }, { "left": "issue:44188", "right": "issue:46328", "accept": false, - "reason": "Attention-kernel divergence under torch.compile is unrelated to SequenceFeatureExtractor.pad performance." + "reason": "Attention-kernel divergence under torch.compile is unrelated to pad performance regression." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "Both involve GLM-5, but one is invalid GenerationConfig and the other is RoPE implementation; different bugs." + "reason": "Both are GLM-5-related, but one is GenerationConfig validation and the other is RoPE implementation." }, { "left": "issue:38617", "right": "issue:45278", "accept": false, - "reason": "Issue 38617 is one specific missing import; 45278 is a broad upgrade-related import regression, not the same bug." + "reason": "45278 is a broad import-error report, but 38617 is one specific missing import; insufficient evidence they are the same concrete bug." }, { "left": "issue:43066", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions in v5, but the reported model/tokenizer behavior differs and does not indicate one underlying defect." + "reason": "Wrong tokenizer decoder type in v5 and Deepseek tokenization drift are different tokenizer regressions." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM across dataloader threads is unrelated to dtype mismatch CI failures." + "reason": "VRAM leak in Qwen3ForCausalLM and dtype mismatch CI issue are unrelated." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Missing embed_tokens.weight on UMT5Encoder is a model-loading bug; AutoProcessor kwargs forwarding is a different loading path." + "reason": "Different from_pretrained plumbing issues: missing weights vs kwargs not passed to cached_file." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 settings guidance vs missing get_number_of_image_patches/AttributeError are unrelated." + "reason": "TF32 settings API change is unrelated to a missing image-patch method in GLM46V." }, { "left": "issue:43611", "right": "issue:46002", "accept": false, - "reason": "base_model_prefix loading regression and NaNs in classification heads are different failure modes." + "reason": "Base-model-prefix loading breakage and NaNs at init/checkout are different classes of failure." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Example-script multi-label classification output bug is unrelated to SigLIP2 model discrepancy." + "reason": "Classification script behavior and SigLIP2 implementation discrepancy are unrelated." }, { "left": "issue:44355", "right": "issue:46231", "accept": false, - "reason": "Compiled Python file errors and documentation parameter mismatch are not the same issue." + "reason": "Compiled Python load errors and a documentation parameter-name mismatch are not the same issue." }, { "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "Different Qwen3.5/MoE issues: one about Transformers 5.0 model support, the other about a from_pretrained error after 5.4.0." + "reason": "Both involve Qwen3.5/MoE, but they describe different API paths and symptoms." }, { "left": "issue:44625", "right": "issue:46002", "accept": false, - "reason": "Qwen3.5 num_labels propagation and classification-head NaNs are separate problems." + "reason": "Qwen3.5 num_labels propagation and NaNs in classification heads are different defects." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python version compatibility issue vs PyTorch version check bug in flex_attention; unrelated." + "reason": "Python version support failure and PyTorch version gate for AuxRequest are unrelated." }, { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "Optimizer SGD arguments not being used is unrelated to trainer.evaluate failing after trainer.train." + "reason": "Optimizer argument handling and trainer.evaluate() failing after train() are different training-stack bugs." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5 regressions, but one is save_pretrained visual-key corruption and the other is a from_pretrained error." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "Both are model-specific, but VitPose breakage and Ministral-3 FlashAttention failure are different code paths." + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are unrelated model-specific issues." }, { - "left": "issue:45357", - "right": "issue:45478", + "left": "issue:44355", + "right": "issue:46231", "accept": false, - "reason": "Both involve Qwen3.5, but one is save_pretrained visual-encoder keys and the other is a from_pretrained MoE error." + "reason": "Compiled-file loading errors are unrelated to a docs bug about set_training() parameters." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Multi-label example-script output bug is unrelated to the pipeline translation-task message regression." + "reason": "Multi-label classification empty output is unrelated to pipeline() translation-task messaging." } ] }, @@ -14372,6 +14555,7 @@ "issue:41628", "issue:42222", "issue:42371", + "issue:42490", "issue:42617", "issue:42673", "issue:42796", @@ -14393,7 +14577,6 @@ "issue:44466", "issue:44779", "issue:44829", - "issue:44877", "issue:44977", "issue:45071", "issue:45081", @@ -14410,7 +14593,6 @@ "issue:46002", "issue:46077", "issue:46178", - "issue:46231", "issue:46298", "issue:46328", "issue:46512" @@ -14429,214 +14611,226 @@ "issue:44291|issue:45071", "issue:39401|issue:44779", "issue:44466|issue:46002", - "issue:43450|issue:43986", "issue:43937|issue:44977", + "issue:43450|issue:43986", "issue:45405|issue:45446", "issue:44410|issue:45310", "issue:45341|issue:46328", - "issue:45341|issue:45561", "issue:38617|issue:41628", - "issue:44117|issue:45081", + "issue:45341|issue:45561", "issue:44829|issue:45216", + "issue:44117|issue:45081", "issue:45071|issue:46512", "issue:42222|issue:46178", "issue:43122|issue:46298", "issue:43408|issue:43742", "issue:42617|issue:45341", - "issue:45362|issue:46328", "issue:42371|issue:44261", - "issue:44877|issue:46231" + "issue:45362|issue:46328", + "issue:42490|issue:42673" ], - "prepared_review_unit_hash": "d73bb6c39ba926fd3904dbe765cfbf4e822d5e6009e5e40174f67d004e9371af", + "prepared_review_unit_hash": "35960bc6f31cab6c44437f9ee52f21720f84cbb4e037bec87adfbe24d9ec3a66", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36666, - "estimated_input_tokens": 9167, - "estimated_eval_tokens": 18590 + "serialized_chars": 36085, + "estimated_input_tokens": 9022, + "estimated_eval_tokens": 18300 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated regression reports across tokenizer, loading, config, attention, and training code paths. None of the suggested soft edges look like true duplicates, so the cluster does not have a strong duplicate center.", - "confidence": 0.93, - "canonical_issue_reason": "If one issue must serve as the cluster representative, issue 43122 is the best fit: it is a broad tokenizer regression across versions, has inbound references, and is closer to the common theme than the others.", + "summary": "This cluster is highly heterogeneous: most items are unrelated model-, tokenizer-, generation-, or testing/infrastructure-specific regressions. I did not find any soft edge pair that looks like the same underlying bug/change.", + "confidence": 0.61, + "canonical_issue_reason": "No clear canonical duplicate: the issues span many distinct code paths and product areas rather than one shared root cause.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43122 is the strongest single representative for triage because it describes a clear cross-version tokenizer behavior regression with broader impact than the more model-specific reports.", + "best_issue_reason": "issue:43122 is the broadest, clearest regression report here (tokenization changed from 4.57.3 to 5.0) and is the best representative of the recurring tokenizer-regression theme.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Different subsystems and failures: qwen3_omni_moe helper behavior vs a flex_attention import/version check." + "reason": "Different subsystems and failures: qwen3_omni_moe feature-length helper vs flex_attention version gating." }, { "left": "issue:42673", "right": "issue:42796", "accept": false, - "reason": "Both mention Qwen3/tokenizer-adjacent areas, but one is a VRAM leak in dataloader threads and the other is BPE marker handling." + "reason": "One is a VRAM leak in Qwen3ForCausalLM threading; the other is a tokenizer BPE-marker bug in Ministral." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Trainer/FSDP2 optimizer-state corruption is unrelated to GenerationMixin cache behavior during chunked prefill." + "reason": "Different code paths: trainer/FSDP2 optimizer-state mismatch vs generation cache missing in chunked_prefill." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "GLM-5 invalid GenerationConfig and a stale paged-generate warning are different code paths and symptoms." + "reason": "Both mention generation, but one is invalid GenerationConfig loading and the other is a stale warning in paged generate()." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Tokenizer loading regression and Qwen3 offset_mapping are both tokenizer issues, but not the same underlying bug." + "reason": "Both are tokenizer-related, but one is loading regression and the other is wrong offset_mapping for Qwen3." }, { "left": "issue:44380", "right": "issue:46077", "accept": false, - "reason": "Attention scaling ignored under backends is distinct from Gemma4 building causal masks when bidirectional attention is requested." + "reason": "Attention-related, but different models and bugs: GPT2 backend scaling vs Gemma4 bidirectional mask construction." }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs still referencing pipeline() is unrelated to the flex_attention PyTorch version check issue." + "reason": "Documentation cleanup for removed pipeline() APIs is unrelated to a PyTorch version check bug." }, { "left": "issue:44117", "right": "issue:45588", "accept": false, - "reason": "A None mapping in tokenizer lookup is not the same failure as flash_attention crashing on s_aux=None." + "reason": "Unrelated failures: tokenizer mapping None handling vs flash_attention crash on sink-less models." }, { "left": "issue:43673", "right": "issue:46328", "accept": false, - "reason": "Generation cache handling and SequenceFeatureExtractor.pad inefficiency are unrelated bugs." + "reason": "Generation cache bug and feature-extractor padding performance bug are unrelated." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "VRAM leakage in multi-threaded dataloading is not the same as an isolated reproduction of another issue." + "reason": "43906 is only an isolated reproduction reference; no evidence it is the same VRAM leak as 42673." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "Both are v5 regressions, but one is an init_empty_weights TypeError and the other is PretrainedConfig type checking." + "reason": "Different breakages: init_empty_weights argument handling vs PretrainedConfig type-checking in v5.4.0." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different tokenization symptoms." + "reason": "Both are tokenizer regressions, but they concern different models and symptoms (offset mapping vs general tokenization output)." }, { "left": "issue:44466", "right": "issue:46002", "accept": false, - "reason": "lm_head.weight serialization depends on device state; NaNs in classification heads after checkout/init is a different failure." + "reason": "Serialization of lm_head.weight is unrelated to NaNs in classification heads after checkout/init." }, { - "left": "issue:43450", - "right": "issue:43986", + "left": "issue:43937", + "right": "issue:44977", "accept": false, - "reason": "Incorrect batched video processor shapes are separate from a torchvision-missing crash in AutoProcessor." + "reason": "Different generation issues: invalid GenerationConfig vs flash-attention malfunction for Qwen3.5." }, { - "left": "issue:43937", - "right": "issue:44977", + "left": "issue:43450", + "right": "issue:43986", "accept": false, - "reason": "Invalid GenerationConfig for GLM-5 is not the same as Qwen3.5 flash-attention generation failure." + "reason": "Video-related but distinct: batched output shape bug vs missing torchvision crash in AutoProcessor." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "A bumped PEFT minimum version is unrelated to the flex_attention import version guard." + "reason": "A dependency version bump problem is unrelated to an auxiliary-request import version check." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Missing Qwen3Next projections and Qwen3.5 MoE from_pretrained errors are different model-loading bugs." + "reason": "Different Qwen model-family bugs: missing projections in qwen3next vs from_pretrained failure for Qwen3.5 MoE." }, { "left": "issue:45341", "right": "issue:46328", "accept": false, - "reason": "A testing_utils bug and a SequenceFeatureExtractor.pad performance issue do not share a concrete code-path." - }, - { - "left": "issue:45341", - "right": "issue:45561", - "accept": false, - "reason": "Both touch testing_utils, but one is a general bug report and the other is a pytest-xdist race on captured_info.txt." + "reason": "Different issues: a testing_utils bug vs a feature-extractor padding performance regression." }, { "left": "issue:38617", "right": "issue:41628", "accept": false, - "reason": "These are separate missing-import issues for different symbols and different APIs." + "reason": "Both are import errors, but they involve different missing symbols and different API surfaces." }, { - "left": "issue:44117", - "right": "issue:45081", + "left": "issue:45341", + "right": "issue:45561", "accept": false, - "reason": "Tokenizer mapping returning None and a Mistral regex patch crash are different loading-time failures." + "reason": "Shared file path is not enough; one is a generic testing_utils bug, the other is an xdist race on captured_info.txt." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "FlashAttention 3 causing degenerate training is unrelated to incorrect checkpoint serialization in save_pretrained." + "reason": "Different bugs: flash_attention_3 training degeneration vs incorrect save_pretrained checkpoint contents." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Both hit tokenizer loading, but one is a mapping lookup bug and the other is a Mistral regex patch crash." }, { "left": "issue:45071", "right": "issue:46512", "accept": false, - "reason": "PretrainedConfig type checking and stale config.dtype after from_config are different dtype/config lifecycle bugs." + "reason": "Type-checking breakage in v5.4.0 is unrelated to stale config.dtype after from_config(dtype=...)." }, { "left": "issue:42222", "right": "issue:46178", "accept": false, - "reason": "Broken vitpose models and a BACKENDS_MAPPING tensorflow_text error are unrelated." + "reason": "Different areas entirely: vitpose model breakage vs backend registration error for tensorflow_text." }, { "left": "issue:43122", "right": "issue:46298", "accept": false, - "reason": "Cross-version tokenizer regression is unrelated to a noisy logging message." + "reason": "Versioned tokenization regression is unrelated to a noisy logging message." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "A model-type warning between sam3 variants is not the same as a key error while loading MobileLLM-125M." + "reason": "Model-type warning for sam3_video/sam3_tracker is unrelated to a MobileLLM loading key error." }, { - "left": "issue:44877", - "right": "issue:46231", + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 API guidance and an MLA rms_norm_eps precision issue are not the same bug." + }, + { + "left": "issue:45362", + "right": "issue:46328", "accept": false, - "reason": "Strict granite_speech config loading and a documentation parameter-name mismatch are not the same bug." + "reason": "Qwen3.5 chat crash and feature-extractor padding slowdown are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:42673", + "accept": false, + "reason": "Path-with-slash import warning is unrelated to the Qwen3ForCausalLM VRAM leak." } ] }, @@ -14667,8 +14861,6 @@ "issue:42175", "issue:42222", "issue:42371", - "issue:42490", - "issue:42673", "issue:42831", "issue:43012", "issue:43066", @@ -14696,6 +14888,7 @@ "issue:44493", "issue:44496", "issue:44829", + "issue:44877", "issue:44933", "issue:45003", "issue:45071", @@ -14707,10 +14900,11 @@ "issue:45542", "issue:45656", "issue:46172", - "issue:46209" + "issue:46209", + "issue:46231" ], "soft_pairs": [ - "issue:42490|issue:42673", + "issue:44877|issue:46231", "issue:44484|issue:45310", "issue:39401|issue:46209", "issue:38617|issue:44220", @@ -14739,216 +14933,223 @@ "issue:43931|issue:44038", "issue:43097|issue:43452", "issue:43986|issue:45341", - "issue:45003|issue:45341" + "issue:45003|issue:45341", + "issue:43874|issue:44263" ], - "prepared_review_unit_hash": "2eea925bae67eb923b139ef95e250ed9d94bef59b3ad9d48cb17865a6f329fde", + "prepared_review_unit_hash": "035ae003ff72d14a837e1a11e27540011ab63ee23ec4ea56ac04d60a9c1f9fa5", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36096, - "estimated_input_tokens": 9024, - "estimated_eval_tokens": 18304 + "soft_pair_count": 31, + "serialized_chars": 36411, + "estimated_input_tokens": 9103, + "estimated_eval_tokens": 18462 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are mostly separate Transformers regressions or model-specific bugs. A few are thematically related to v5 upgrade/import/tokenization changes, but none are safe duplicate merges from the available evidence.", - "confidence": 0.66, - "canonical_issue_reason": "issue:45278 is the broadest, most umbrella-like report here: it explicitly covers many import errors after upgrading to 5.5.0, so it is the best anchor for a duplicate cluster around upgrade regressions.", + "summary": "The set is mostly a loose semantic cluster of Transformers regressions, but the pairs are not strong enough to treat as duplicates. The closest subgroup is import/config breakage in v5/v5.5, with 45278 being the broadest representative issue.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45278 is the most representative single issue in this set: it captures a broad, user-visible regression pattern (many import errors after upgrading) that best anchors the loose import/config breakage subgroup.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is also the strongest global representative because it is broad, high-signal, and likely to absorb related duplicate reports about v5 import regressions.", + "best_issue_reason": "It is the best overall issue to keep as the cluster anchor because it is broader than the more specific model- or feature-level reports and most closely matches the recurring upgrade/regression theme.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:42490", - "right": "issue:42673", + "left": "issue:44877", + "right": "issue:46231", "accept": false, - "reason": "Different problems: local-path slash warning vs VRAM leakage in multi-threaded Qwen3ForCausalLM usage." + "reason": "Strict config-loading regression vs a documentation parameter typo; unrelated symptoms and fixes." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "Unrelated: save_pretrained shard-size question versus a Qwen3.5-MoE from_pretrained loading error." + "reason": "Max shard-size behavior question vs Qwen3.5 MoE from_pretrained failure; different code paths." }, { "left": "issue:39401", "right": "issue:46209", "accept": false, - "reason": "Different code paths: tokenizer offset_mapping bug vs FP8 inference failure on B200 hardware." + "reason": "Tokenizer offset_mapping bug vs FP8 inference failure on B200; unrelated models and subsystems." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "Missing import symbol in configuration_utils is unrelated to the _torch_extract_fbank_features() issue." + "reason": "ImportError for a missing symbol vs audio feature-extraction issue; not the same bug." }, { "left": "issue:43937", "right": "issue:44496", "accept": false, - "reason": "GenerationConfig validation error and unrecognized-model/model_type error are distinct load-time failures." + "reason": "Both involve model/config loading, but one is invalid GenerationConfig and the other is an unrecognized model/config.json issue." }, { "left": "issue:44933", "right": "issue:45278", "accept": false, - "reason": "44933 is one specific missing import; 45278 is a broader umbrella for many import errors. Related theme, but not the same concrete bug." + "reason": "A specific missing import from image_utils is not clearly the same underlying bug as a broad \"many import errors\" report." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 API guidance warning is unrelated to the FineGrainedFP8 accuracy regression." + "reason": "TF32 API usage complaint vs FineGrainedFP8 accuracy regression; different topics." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Different failures: missing qwen3next projections vs Qwen3.5-Moe from_pretrained error." + "reason": "Missing Qwen3Next projections vs Qwen3.5 MoE from_pretrained error; different model bugs." }, { "left": "issue:42222", "right": "issue:45542", "accept": false, - "reason": "VitPose model breakage and missing-TensorFlow backend packaging are different issues." + "reason": "Broken vitpose models vs missing TensorFlow backend when only tensorboard is installed; unrelated." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "Unused SGD args and double optimizer.step under deepspeed are separate optimizer bugs." + "reason": "SGD args ignored vs optimizer.step called twice under deepspeed; different optimizer bugs." }, { "left": "issue:43066", "right": "issue:43122", "accept": false, - "reason": "Both concern tokenizer behavior in v5, but the concrete bugs differ: decoder type vs changed tokenization output." + "reason": "Tokenizer decoder-type regression vs changed tokenization output; related area but not the same concrete failure." }, { "left": "issue:42175", "right": "issue:42222", "accept": false, - "reason": "Backend packaging/install issue versus VitPose model loading failure; not the same bug." + "reason": "Package/backend install issue vs vitpose model breakage; no shared underlying bug." }, { "left": "issue:43874", "right": "issue:44261", "accept": false, - "reason": "Image-patch helper missing method is unrelated to the MLA q_a_layernorm eps/config precision issue." + "reason": "Missing image patch helper vs missing rms_norm_eps in MLA; separate model implementation issues." }, { "left": "issue:44263", "right": "issue:44485", "accept": false, - "reason": "Different GLM-related problems: torch.split handling vs RoPE implementation discussion." + "reason": "DsaIndexer torch.split return-value issue vs GLM-5 RoPE implementation bug; different code paths." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Removed config field and unexpected position-id keys are not the same concrete regression." + "reason": "Removed tie_embeddings API vs unexpected position-id key warnings; different regressions." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "Chunked-prefill cache issue is unrelated to a bug in testing_utils.py." + "reason": "Generation cache missing in chunked prefill vs a testing_utils bug; unrelated." }, { "left": "issue:43012", "right": "issue:43577", "accept": false, - "reason": "Compile-time float32 warning and BLIP2 dtype not propagating are distinct dtype-related issues." + "reason": "Warning during bfloat16 compilation vs Blip2 dtype not propagating to submodules; related dtype theme but not the same bug." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "Missing AutoImageProcessor export and PretrainedConfig type-checking breakage are separate API regressions." + "reason": "Missing AutoImageProcessor import vs PreTrainedConfig type-checking regression; unrelated." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "Type-checking breakage is unrelated to the TensorFlow backend detection/install issue." + "reason": "Config type-checking breakage vs undefined tf backend error; different failure modes." }, { "left": "issue:27957", "right": "issue:43352", "accept": false, - "reason": "Both mention Flash Attention 2, but they are model-specific unsupported-feature reports for different models." + "reason": "Both mention Flash Attention 2 support, but they concern different models and likely separate implementations." }, { "left": "issue:44483", "right": "issue:45464", "accept": false, - "reason": "General chat/completions request handling differs from a streaming inference failure on Qwen3.5-0.8B." + "reason": "Chat/completions request rejection vs streaming inference failure; same API family, but not a single concrete bug." }, { "left": "issue:44484", "right": "issue:45478", "accept": false, - "reason": "Save-pretrained shard sizing is unrelated to the Qwen3.5-MoE from_pretrained loading error." + "reason": "Different issues: one is save_pretrained shard-size behavior, the other is Qwen3.5 MoE loading failure." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Weight-shape mismatch when loading Qwen3-VL-30B is a different problem from missing qwen3next projections." + "reason": "Qwen3-VL weight-shape mismatch vs Qwen3Next missing projections; different model families and defects." }, { "left": "issue:44263", "right": "issue:46172", "accept": false, - "reason": "Torch.split indexer behavior and DeepSeek-v4 comb transpose are different code paths." + "reason": "DSA indexer split semantics vs DeepSeek-V4 transpose/dtype issue; unrelated." }, { "left": "issue:44360", "right": "issue:46172", "accept": false, - "reason": "Missing ReLU in the DSA indexer is unrelated to the DeepSeek-v4 transpose bug." + "reason": "Missing ReLU in DSA indexer vs DeepSeek-V4 tensor conversion/transposition bug; different changes." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "FlashAttention support failure and degenerate training under flash_attention_3 are different concrete issues." + "reason": "FlashAttention failure in Ministral-3 vs degenerate training with flash_attention_3; same broad feature, different concrete problems." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch and Qwen3-VL-Moe failure are separate model-loading bugs." + "reason": "Different Qwen3-VL issues with distinct symptoms: weight shapes mismatch vs MoE model bug." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "Removed tie_embeddings config field is unrelated to gguf_file breaking AutoTokenizer/AutoModel loading." + "reason": "API removal/change vs gguf_file loading breakage; not the same code path." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "AutoProcessor without torchvision and a testing_utils bug do not share the same failure path." + "reason": "AutoProcessor/torchvision dependency crash vs a testing utility bug; unrelated." }, { "left": "issue:45003", "right": "issue:45341", "accept": false, - "reason": "sys.modules access in modeling_utils is unrelated to the testing_utils issue." + "reason": "Unsafe sys.modules access in modeling_utils vs a testing utility bug; unrelated." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Image processor helper missing vs GLM-MoE split-return issue; different components." } ] }, @@ -14976,21 +15177,22 @@ "issue:42757", "issue:42890", "issue:42907", + "issue:43012", "issue:43064", "issue:43097", + "issue:43208", "issue:43502", "issue:43531", + "issue:43550", "issue:43673", - "issue:43824", "issue:43873", - "issue:43874", "issue:43986", "issue:43992", + "issue:44016", "issue:44038", "issue:44062", "issue:44117", "issue:44261", - "issue:44263", "issue:44315", "issue:44351", "issue:44380", @@ -15022,7 +15224,6 @@ ], "soft_pairs": [ "issue:45753|issue:46369", - "issue:43874|issue:44263", "issue:44933|issue:46231", "issue:43673|issue:46123", "issue:44261|issue:44485", @@ -15030,241 +15231,238 @@ "issue:44117|issue:44987", "issue:43531|issue:45341", "issue:45440|issue:46172", - "issue:43986|issue:45372", "issue:43064|issue:44315", + "issue:43986|issue:45372", "issue:38617|issue:42757", "issue:43097|issue:45588", "issue:43097|issue:46050", "issue:44410|issue:45375", "issue:44038|issue:44410", "issue:42890|issue:46298", - "issue:43502|issue:43992", "issue:38617|issue:44351", + "issue:43502|issue:43992", "issue:44062|issue:44117", "issue:43097|issue:43873", "issue:44936|issue:46365", "issue:44704|issue:45103", "issue:43986|issue:44117", "issue:44380|issue:46421", - "issue:43824|issue:43986", "issue:44938|issue:46178", + "issue:43986|issue:44016", "issue:42907|issue:44655", - "issue:43992|issue:44484" + "issue:43992|issue:44484", + "issue:43208|issue:43550", + "issue:43012|issue:43550", + "issue:42907|issue:45753" ], - "prepared_review_unit_hash": "a60fd596957dd545c6cffdcbb2b03a6a52fc34dc7d70ed6adfc37a4a215f7753", + "prepared_review_unit_hash": "9e522707287745b720183f2638dd1bd528cbc1384aae766a416ee6053310724f", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 35599, - "estimated_input_tokens": 8900, - "estimated_eval_tokens": 18056 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36684, + "estimated_input_tokens": 9171, + "estimated_eval_tokens": 18598 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is heterogeneous and the soft pairs mostly share only broad subsystem keywords. I rejected every soft edge because none looks like the same concrete bug/change with enough specificity to merge as duplicates.", - "confidence": 0.88, - "canonical_issue_reason": "If forced to pick a canonical issue, issue:43873 is the strongest anchor: it is open, broadly worded, and has the most discussion/references, so it best represents a user-facing bug cluster. That said, it is not a safe duplicate match for the other items.", + "summary": "The soft pairs are superficially related by broad themes (imports, loading, cache, quantization, model families), but each points to a different concrete failure mode or code path. I would not merge any of these pairs as duplicates, and there is no clear canonical issue in this heterogeneous set.", + "confidence": 0.92, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the most plausible cluster anchor because it is the broadest and most active bug report in the set; it is the best candidate to track centrally if any issue must be chosen.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45753", "right": "issue:46369", "accept": false, - "reason": "Different models and symptoms: Qwen3_5 infinite loop on one image vs SAM3-LiteText memory overflow." - }, - { - "left": "issue:43874", - "right": "issue:44263", - "accept": false, - "reason": "Both are GLM-related, but one is a missing image-processor method and the other is a torch.split indexing bug." + "reason": "Both are model-runtime failures, but one is an infinite loop on a specific image and the other is a memory overflow on sam3-litetext; different symptoms and likely different code paths." }, { "left": "issue:44933", "right": "issue:46231", "accept": false, - "reason": "Unrelated: image_utils import issue vs a documentation parameter-name mismatch." + "reason": "One is a missing import from image_utils; the other is a documentation/parameter-name mismatch. Not the same bug." }, { "left": "issue:43673", "right": "issue:46123", "accept": false, - "reason": "Different code paths: chunked_prefill cache handling vs MaskGenerationPipeline partial-batch finalization." + "reason": "Cache-related generation issue vs pipeline partial-batch result dropping. Different components and failure modes." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "Both mention GLM/MLA, but one is an rms_norm_eps precision issue and the other is RoPE implementation." + "reason": "MLA RMS norm epsilon precision problem vs GLM-5 RoPE implementation discussion. Same general area only at a very broad level." }, { "left": "issue:43986", "right": "issue:44484", "accept": false, - "reason": "Video-model loading without torchvision is unrelated to save_pretrained shard sizing." + "reason": "AutoProcessor video-model import crash without torchvision vs save_pretrained shard-size default question. Unrelated bugs." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Different failures: tokenizer-mapping None handling vs loading a specific model repo in v5.1.0." + "reason": "Tokenizer mapping None handling vs a model-loading failure on physical-intelligence/fast. Similar surface area, different root cause." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding-window bug is unrelated to a testing_utils bug." + "reason": "Qwen3-MoE sliding-window bug vs a testing utility bug. Not the same underlying issue." }, { "left": "issue:45440", "right": "issue:46172", "accept": false, - "reason": "Both are DeepSeek-family, but they describe different model versions and different tensor-path bugs." + "reason": "Different DeepSeek model-family correctness issues: remote-implementation divergence vs a tensor dtype/transpose bug." }, { - "left": "issue:43986", - "right": "issue:45372", + "left": "issue:43064", + "right": "issue:44315", "accept": false, - "reason": "Unrelated processor-loading crash vs missing ReasoningEffort import from mistral_common." + "reason": "Wrong optimizer states under FSDP2/PEFT/cpu_ram_efficient_loading vs Liger Kernel not applied with model_init. Different training paths." }, { - "left": "issue:43064", - "right": "issue:44315", + "left": "issue:43986", + "right": "issue:45372", "accept": false, - "reason": "Trainer/FSDP optimizer-state corruption is a different bug from Liger Kernel not applying with model_init." + "reason": "Both affect processor/model loading, but one needs torchvision and the other fails on a missing mistral_common import. Distinct causes." }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols from different packages and likely different fixes." + "reason": "Both are import errors, but for different symbols from different modules; not the same missing export." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "Tie-embeddings API removal and flash-attention sink handling are unrelated failure modes." + "reason": "Tie-embeddings API removal vs flash_attention crash on s_aux=None. Separate regressions." }, { "left": "issue:43097", "right": "issue:46050", "accept": false, - "reason": "One is an API removal complaint; the other is quantized model save/load breakage." + "reason": "API removal vs quantized save/load breakage. No shared code-path problem." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Same broad Qwen family, but one is missing attention projections and the other is a config field dropped by strict validation." + "reason": "Different Qwen configuration/model bugs: missing projections vs missing config field." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "Too vague to merge: generic Qwen3-VL-Moe bug report vs a specific qwen3next layer-0 projection issue." + "reason": "Both mention Qwen variants, but one is a broad transformers/Qwen3-VL-MoE bug and the other is a specific missing-projection issue." }, { "left": "issue:42890", "right": "issue:46298", "accept": false, - "reason": "Integration test seeding problem is unrelated to a noisy logging message." + "reason": "Integration-test flakiness from missing set_seed vs a noisy logging message. Not duplicates." }, { - "left": "issue:43502", - "right": "issue:43992", + "left": "issue:38617", + "right": "issue:44351", "accept": false, - "reason": "Offline/local-files behavior is unrelated to a missing embed_tokens.weight in UMT5Encoder loading." + "reason": "Both are import-name failures, but they involve different public symbols and different dependency layers." }, { - "left": "issue:38617", - "right": "issue:44351", + "left": "issue:43502", + "right": "issue:43992", "accept": false, - "reason": "Different missing imports from different namespaces; same broad symptom, not the same bug." + "reason": "local_files_only still making API calls vs missing embed_tokens.weight in UMT5Encoder loading. Different bugs." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "Tokenizer AddedToken kwarg duplication is unrelated to TOKENIZER_MAPPING_NAMES returning None." + "reason": "AddedToken constructor argument conflict vs tokenizer-mapping None handling. Similar tokenizer area, but unrelated failures." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "Different areas entirely: removed tie-embedding API vs quantization/offloading behavior." + "reason": "Tie_embeddings removal and quantization offloading are separate features with separate failure modes." }, { "left": "issue:44936", "right": "issue:46365", "accept": false, - "reason": "trainer.evaluate() after train() is unrelated to failing to import Trainer because datasets is installed." + "reason": "Evaluation-after-training failure vs importing Trainer blocked by datasets. Different lifecycle stages and causes." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "AutoProcessor kwargs forwarding is unrelated to auto_docstring breaking under future annotations." + "reason": "AutoProcessor kwargs propagation bug vs auto_docstring crash under future annotations. Different subsystems." }, { "left": "issue:43986", - "right": "issue:44117", + "right": "issue:44016", "accept": false, - "reason": "Processor-loading with missing torchvision is not the same bug as tokenizer mapping returning None." + "reason": "Processor loading crash vs notebook syntax error. Not the same issue." }, { "left": "issue:44380", "right": "issue:46421", "accept": false, - "reason": "GPT2 attention scaling under SDPA/FlashAttention is unrelated to lazy cache initialization causing recompilation." - }, - { - "left": "issue:43824", - "right": "issue:43986", - "accept": false, - "reason": "Different missing dependency/import paths: Qwen2.5-VL class import vs AutoProcessor video-model loading." + "reason": "GPT2 attention scaling ignored under SDPA/FlashAttention vs chunked-prefill cache lazy initialization causing recompilation. Different mechanisms." }, { "left": "issue:44938", "right": "issue:46178", "accept": false, - "reason": "Python 3.14 import failure is unrelated to an unsupported tensorflow_text backend entry." + "reason": "Python 3.14 import/load failure vs BACKENDS_MAPPING validation error for tensorflow_text. Unrelated." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals is a different bug from saving Pipeline objects." + "reason": "Saving dequantized models vs saving Pipeline objects. Different save_pretrained code paths." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing weights on from_pretrained is unrelated to the default max_shard_size in save_pretrained." + "reason": "Missing encoder weights on load vs shard-size default behavior on save. Not duplicates." }, { - "left": "issue:44933", - "right": "issue:44987", + "left": "issue:43208", + "right": "issue:43550", "accept": false, - "reason": "Image-utils import cleanup and loading a specific model repo are not the same underlying issue." + "reason": "xLSTM training blockers vs Bamba-9B-v2 torch.compile/SDPA failure. Different models and bugs." }, { - "left": "issue:43873", - "right": "issue:46050", + "left": "issue:43012", + "right": "issue:43550", "accept": false, - "reason": "Both involve quantization, but one is an offloading bug and the other is a save/load round-trip breakage." + "reason": "PyTorch bfloat16 precision warning vs compile-time SDPA failure. No shared root cause." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Quantized save/load failure vs image-specific infinite loop. Completely different problems." } ] }, @@ -15293,7 +15491,6 @@ "issue:42175", "issue:42490", "issue:42757", - "issue:42907", "issue:42994", "issue:43012", "issue:43064", @@ -15302,11 +15499,12 @@ "issue:43208", "issue:43299", "issue:43352", + "issue:43475", "issue:43502", "issue:43526", - "issue:43550", "issue:43582", "issue:43618", + "issue:43756", "issue:43761", "issue:43824", "issue:43825", @@ -15321,6 +15519,7 @@ "issue:44368", "issue:44380", "issue:44483", + "issue:44485", "issue:44514", "issue:44655", "issue:44683", @@ -15331,22 +15530,18 @@ "issue:45600", "issue:45698", "issue:45712", - "issue:45753", "issue:46132", "issue:46172", "issue:46298", "issue:46365" ], "soft_pairs": [ - "issue:43208|issue:43550", - "issue:43012|issue:43550", - "issue:42907|issue:45753", "issue:43012|issue:44368", "issue:43825|issue:44655", "issue:43066|issue:43906", "issue:45712|issue:46298", - "issue:43352|issue:44380", "issue:43097|issue:46298", + "issue:43352|issue:44380", "issue:43352|issue:43976", "issue:44162|issue:46172", "issue:45584|issue:46365", @@ -15358,246 +15553,242 @@ "issue:43064|issue:44380", "issue:42175|issue:46365", "issue:42490|issue:43012", - "issue:43012|issue:44683", "issue:43208|issue:43526", + "issue:43012|issue:44683", "issue:42757|issue:43824", "issue:38617|issue:43502", - "issue:44483|issue:44514", "issue:45698|issue:46172", + "issue:44483|issue:44514", "issue:45600|issue:46298", "issue:44038|issue:44230", "issue:43618|issue:44380", "issue:43618|issue:43761", "issue:43299|issue:43867", "issue:42994|issue:44655", - "issue:44483|issue:45290" + "issue:44483|issue:45290", + "issue:36683|issue:43475", + "issue:43756|issue:44485" ], - "prepared_review_unit_hash": "b91eb800d6ad4be3289632d7b86e64c26f8f6d0762fc922d60f2a9fdc27f57eb", + "prepared_review_unit_hash": "8fed415b2ae35c992507d2e6c89eb7dd1d529d56df608ab9d927a57e2712fef4", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 33, - "serialized_chars": 37280, - "estimated_input_tokens": 9320, - "estimated_eval_tokens": 18896 + "soft_pair_count": 32, + "serialized_chars": 37061, + "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 mostly a loose collection of Transformers v5 regressions, import errors, and model-specific bugs rather than a set of duplicates. No soft edges look like the same concrete underlying issue with enough confidence to merge.", - "confidence": 0.84, - "canonical_issue_reason": "issue:43097 is the best representative of the cluster\u2019s common theme: a broad v5 backward-compatibility/API regression with clear user impact and broader relevance than the more model-specific reports.", + "summary": "The set spans many unrelated regressions across model loading, generation, packaging, attention backends, and serialization. None of the soft pairs looks like the same concrete bug/change closely enough to merge as duplicates.", + "confidence": 0.73, + "canonical_issue_reason": "issue:45698 is the strongest representative because it describes a concrete, reproducible save/load bug in a core path, is still open, and has the most discussion/activity among the candidates.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43097 is the strongest single issue to anchor the cluster because it captures the recurring v5 breakage pattern better than the narrower model- or platform-specific bugs.", + "best_issue_reason": "issue:45698 is the best standalone issue to keep as the cluster representative: it is specific, actionable, and broad enough to subsume nearby custom-module serialization failures.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43208", - "right": "issue:43550", - "accept": false, - "reason": "Different models and failure modes: xLSTM training bugs vs Bamba torch.compile/SDPA failure." - }, - { - "left": "issue:43012", - "right": "issue:43550", - "accept": false, - "reason": "A PyTorch warning about bfloat16 precision is not the same as a model-specific compile failure under SDPA." - }, - { - "left": "issue:42907", - "right": "issue:45753", - "accept": false, - "reason": "Quantized/dequantized save failure vs infinite-loop image generation are unrelated bugs." - }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "Both mention warnings, but they concern different code paths and different config fields." + "reason": "Both mention warnings, but they are different warnings in different code paths: bfloat16 compilation vs tie_word_embeddings guidance during LoRA fine-tuning." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Pipeline task support messaging and saving Pipeline objects are separate problems." + "reason": "One is about pipeline task support/error messaging, the other about saving Pipeline objects; related subsystem, but different underlying bugs." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type is not the same as an isolated reproduction of another issue." + "reason": "The first is a tokenizer decoder-type regression; the second is only an isolated reproduction reference and does not clearly identify the same bug." }, { "left": "issue:45712", "right": "issue:46298", "accept": false, - "reason": "Left is a repo hygiene/dummy-class cleanup issue; right is just noisy logging." + "reason": "Dummy PT classes leaking into dir()/check_repo and a noisy logging message are unrelated maintenance issues." }, { - "left": "issue:43352", - "right": "issue:44380", + "left": "issue:43097", + "right": "issue:46298", "accept": false, - "reason": "Nemotron FA2 support error and GPT2 attention scaling under SDPA/FlashAttention are different bugs." + "reason": "Removal of tie_embeddings_and_encoder_decoder and a noisy logging message are unrelated changes." }, { - "left": "issue:43097", - "right": "issue:46298", + "left": "issue:43352", + "right": "issue:44380", "accept": false, - "reason": "A breaking API removal is unrelated to a noisy logging message." + "reason": "Both involve attention backends, but one is a model-level FlashAttention 2 support error and the other is GPT-2 scaling being ignored; not the same bug." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "Unsupported Flash Attention on one model is unrelated to a Python version compatibility problem." + "reason": "Nemotron FlashAttention support and a Python-version incompatibility are different failures with different causes." }, { "left": "issue:44162", "right": "issue:46172", "accept": false, - "reason": "ESM2 breakage and DeepSeek dtype/transpose behavior are different model-specific issues." + "reason": "ESM2 breakage and a DeepSeek tensor transpose issue are distinct model-specific bugs." }, { "left": "issue:45584", "right": "issue:46365", "accept": false, - "reason": "Whisper empty-transcription generation and importing Trainer are unrelated." + "reason": "Whisper empty transcription handling and Trainer import failure due to datasets are unrelated." }, { "left": "issue:44380", "right": "issue:46132", "accept": false, - "reason": "GPT2 attention scaling and AttentionInterface.register behavior are different code paths." + "reason": "The first is about GPT2 attention scaling being ignored, while the second is about AttentionInterface.register changing function behavior; not the same defect." }, { "left": "issue:44077", "right": "issue:46365", "accept": false, - "reason": "PatchTSMixer post_init validation and Trainer import failure are unrelated." + "reason": "A patchtsmixer post_init validation issue is unrelated to a Trainer import conflict." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "Attention scaling ignored and zero-loss classification are different functional bugs." + "reason": "Attention backend scaling ignored and degenerate zero-loss sequence classification are different model behaviors." }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "FP8 save_pretrained failure and Qwen3.5 MoE from_pretrained loading failure are not the same bug." + "reason": "Both mention MoE/FP8/Qwen3.5, but one is about save_pretrained and the other about from_pretrained loading; not enough to be the same concrete bug." }, { "left": "issue:36683", "right": "issue:43582", "accept": false, - "reason": "Gemma3Config missing vocab_size and Apple Silicon caching_allocator_warmup TypeError are unrelated." + "reason": "Gemma3Config missing vocab_size and Apple Silicon TypeError in caching_allocator_warmup are unrelated." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP2 optimizer-state corruption is not the same as GPT2 attention scaling being ignored." + "reason": "Distributed optimizer-state corruption and GPT2 attention scaling are different code paths." }, { "left": "issue:42175", "right": "issue:46365", "accept": false, - "reason": "Backend-install dependency selection and importing Trainer are unrelated import/package issues." + "reason": "One is backend selection after installing extras, the other is a Trainer import conflict with datasets; different packaging issues." }, { "left": "issue:42490", "right": "issue:43012", "accept": false, - "reason": "Local path slash warning and bfloat16 compile warning are separate issues." + "reason": "Local-path import warning and bfloat16 compile precision warning do not share the same bug." }, { - "left": "issue:43012", - "right": "issue:44683", + "left": "issue:43208", + "right": "issue:43526", "accept": false, - "reason": "A warning about precision is not the same as a torch>=2.9 flex_attention compile failure." + "reason": "xLSTM training blockers and BeitImageProcessorFast label reduction are unrelated model/component bugs." }, { - "left": "issue:43208", - "right": "issue:43526", + "left": "issue:43012", + "right": "issue:44683", "accept": false, - "reason": "xLSTM training bugs and BeitImageProcessorFast reduce_labels behavior are unrelated." + "reason": "A PyTorch precision warning and a compiled flex_attention failure on torch>=2.9 are different regressions." }, { "left": "issue:42757", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols and different root causes." + "reason": "Both are import errors, but for different symbols from different packages and likely different version mismatches." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "Missing layer_type_validation import and local_files_only still making API requests are unrelated." + "reason": "Missing layer_type_validation import and local_files_only still making API requests are unrelated failures." }, { - "left": "issue:44483", - "right": "issue:44514", + "left": "issue:45698", + "right": "issue:46172", "accept": false, - "reason": "Chat/completions request rejection and batched apply_chat_template crash are different failures." + "reason": "Wrong custom module after save_pretrained and a dtype/transpose bug in deepseek-v4 are different underlying issues." }, { - "left": "issue:45698", - "right": "issue:46172", + "left": "issue:44483", + "right": "issue:44514", "accept": false, - "reason": "Wrong custom module loaded after save_pretrained is unrelated to the DeepSeek dtype/transpose bug." + "reason": "A /v1/chat/completions API acceptance problem and a batched apply_chat_template crash are different layers of the stack." }, { "left": "issue:45600", "right": "issue:46298", "accept": false, - "reason": "Removed Sam3LiteText configs breaking CI is not the same as a noisy logging message." + "reason": "Broken auto_mappings references and a noisy logging message are unrelated." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Qwen3-VL-Moe loading breakage and fp8 inference support are different concerns." + "reason": "Both involve Qwen3-VL-Moe, but one is a general loading bug and the other is FP8 inference support; not the same change." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "CLIP attentions missing and GPT2 attention scaling are unrelated model regressions." + "reason": "CLIP attentions missing and GPT2 attention scaling ignored are different model regressions." }, { "left": "issue:43618", "right": "issue:43761", "accept": false, - "reason": "Both are CLIP regressions, but they target different missing outputs and do not clearly share one concrete bug." + "reason": "Both are CLIP regressions, but one affects attentions on CLIPOutput while the other drops hidden_states on CLIPVisionModel; not clearly the same bug." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Qwen3VL MoE loading failure and state_dict sorting load error are different loading bugs." + "reason": "Qwen3VL-Moe loading regression and a sorted-state_dict load error are not the same defect." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving failure and Pipeline save_pretrained failure are separate save-path issues." + "reason": "Quantized model saving failure and pipeline save_pretrained failure are different serialization paths." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "API request rejection and tool-call assistant-message chat-template crash are different problems." + "reason": "Both are chat-related, but one is API request acceptance and the other is tool-call message handling in apply_chat_template." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Gemma3Config missing vocab_size and Sam3VisionEncoderOutput missing fpn_position_embeddings are different attribute errors in different models." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping and GLM-5 RoPE implementation discussion are related only at a very broad topic level, not the same concrete bug." } ] }, @@ -15621,6 +15812,8 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-464", "nodes": [ + "issue:28282", + "issue:33453", "issue:36010", "issue:36296", "issue:36683", @@ -15632,12 +15825,10 @@ "issue:43064", "issue:43066", "issue:43352", - "issue:43475", "issue:43525", "issue:43575", "issue:43611", "issue:43618", - "issue:43756", "issue:43867", "issue:43937", "issue:43976", @@ -15646,11 +15837,11 @@ "issue:44077", "issue:44246", "issue:44355", - "issue:44485", "issue:44492", "issue:44625", "issue:44655", "issue:44855", + "issue:44857", "issue:44938", "issue:45071", "issue:45216", @@ -15667,16 +15858,15 @@ "issue:46077", "issue:46082", "issue:46202", + "issue:46285", "issue:46365" ], "soft_pairs": [ - "issue:36683|issue:43475", - "issue:43756|issue:44485", "issue:44625|issue:46365", "issue:42898|issue:43066", - "issue:43352|issue:43937", - "issue:36010|issue:42757", "issue:45276|issue:46077", + "issue:36010|issue:42757", + "issue:43352|issue:43937", "issue:43611|issue:44655", "issue:44655|issue:46050", "issue:36683|issue:44077", @@ -15703,254 +15893,266 @@ "issue:43867|issue:45357", "issue:43992|issue:45478", "issue:36296|issue:43575", + "issue:43618|issue:44077", "issue:43010|issue:43976", - "issue:43618|issue:44077" + "issue:28282|issue:33453", + "issue:43618|issue:46285", + "issue:43618|issue:44857" ], - "prepared_review_unit_hash": "3bbd1dc815d4cd9b09192d657500d580fcc1788d30a09729ba55e816710a5398", + "prepared_review_unit_hash": "4d202e58ff1d8ef6cd37920d3f039ad89e559f81a71ff69bfdf5ed8e5c54272b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 35, - "serialized_chars": 37121, - "estimated_input_tokens": 9281, - "estimated_eval_tokens": 18818 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38018, + "estimated_input_tokens": 9505, + "estimated_eval_tokens": 19266 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are mostly isolated regressions across different models, config fields, save/load paths, and environment compatibility issues. None of the soft-similarity pairs look like the same underlying bug/change closely enough to merge.", - "confidence": 0.97, - "canonical_issue_reason": "No single canonical issue stands out: the set is thematically similar in places, but the actual failures are diverse and model- or API-specific.", + "summary": "Mostly a heterogeneous mix of Transformers regression issues (imports, save/load, model configs, training, and tests). The only strong duplicate-like pair is the import-time slowdown reports.", + "confidence": 0.78, + "canonical_issue_reason": "issue:41669 is the clearest representative: it gives a concrete import-performance regression with a likely root cause (eager `import *` in models) and a measurable impact.", "canonical_pr_reason": null, - "best_issue_reason": "No good global representative issue: the cluster is too heterogeneous to safely collapse into one canonical issue.", + "best_issue_reason": "issue:41669 is the best global issue candidate because it is specific, actionable, and directly matches the shared import-slowdown theme better than the broader symptom-only report.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:36683", - "right": "issue:43475", - "accept": false, - "reason": "Both are AttributeErrors, but they hit different model classes and different missing fields; not the same bug." - }, - { - "left": "issue:43756", - "right": "issue:44485", - "accept": false, - "reason": "Both mention RoPE, but one is about dropped layers in Smollm3 and the other is a GLM-5 implementation discussion; different root causes." - }, { "left": "issue:44625", "right": "issue:46365", "accept": false, - "reason": "Qwen3.5 config propagation bug vs import failure involving datasets/Trainer; unrelated failures." + "reason": "Different failures: Qwen3.5 `num_labels` propagation vs `Trainer` import being affected by `datasets`." }, { "left": "issue:42898", "right": "issue:43066", "accept": false, - "reason": "Both are v5 tokenizer-related, but one is a behavior change flag and the other is a decoder-type regression; not the same issue." + "reason": "Both concern tokenization/decoder behavior, but they describe different regressions and different code paths." }, { - "left": "issue:43352", - "right": "issue:43937", + "left": "issue:45276", + "right": "issue:46077", "accept": false, - "reason": "Different model-specific errors: Flash Attention support vs invalid GenerationConfig." + "reason": "Same model family (Gemma4) but unrelated bugs: embedding resize propagation vs attention-mask construction." }, { "left": "issue:36010", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but from different packages and symbols; no shared underlying defect." + "reason": "Both are import errors, but for different symbols and different dependencies; no same underlying bug is evident." }, { - "left": "issue:45276", - "right": "issue:46077", + "left": "issue:43352", + "right": "issue:43937", "accept": false, - "reason": "Same model family, but one is embedding resize behavior and the other is attention mask construction; distinct code paths." + "reason": "Unrelated model-specific errors: Flash Attention support gating vs invalid `GenerationConfig`." }, { "left": "issue:43611", "right": "issue:44655", "accept": false, - "reason": "Model-loading breakage vs saving Pipeline objects; different APIs and failure modes." + "reason": "One is loading via `base_model_prefix`, the other is saving `Pipeline` objects; different serialization paths." }, { "left": "issue:44655", "right": "issue:46050", "accept": false, - "reason": "Both involve save_pretrained, but one is Pipeline serialization and the other is quantized model roundtrip; not mergeable as one fix." + "reason": "Both mention `save_pretrained`, but one is pipeline saving and the other is quantized model roundtrip; not the same bug." }, { "left": "issue:36683", "right": "issue:44077", "accept": false, - "reason": "Different model/config bugs with no shared code path." + "reason": "Different models and failures: missing `vocab_size` in Gemma3Config vs `patchtsmixer` post-init policy." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "Both are checkpoint/save-load regressions, but they affect different models and different serialization logic." + "reason": "Both affect model loading/saving, but one is a generic sorted-state_dict load failure and the other is Qwen3.5 checkpoint regression." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Both are Python-version compatibility complaints, but they target different release/runtime combinations." + "reason": "Both are Python-version compatibility reports, but they target different versions and likely different breakages." }, { "left": "issue:43352", "right": "issue:45593", "accept": false, - "reason": "Completely different subsystems: attention backend support vs training loss usage." + "reason": "Flash Attention support gating and missing auxiliary losses are unrelated model-behavior issues." }, { "left": "issue:44625", "right": "issue:46077", "accept": false, - "reason": "Unrelated model features: label propagation vs attention masking." + "reason": "Different Qwen3.5 vs Gemma4 bugs; no shared concrete code-path problem." }, { "left": "issue:46077", "right": "issue:46082", "accept": false, - "reason": "Gemma4 attention-mask bug and LlamaConfig head_dim validation are unrelated." + "reason": "Gemma4 attention-mask bug vs LlamaConfig validation bug are separate configuration/runtime issues." }, { "left": "issue:36683", "right": "issue:43525", "accept": false, - "reason": "Both are missing-config-attribute errors, but on different model configs and fields." + "reason": "Different missing-config-attribute failures in different models (`vocab_size` vs `pad_token_id`)." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "Different regressions: CLIPOutput attentions assignment vs Qwen3.5 checkpoint saving." + "reason": "CLIP output field removal and Qwen3.5 save_pretrained regression are unrelated." }, { "left": "issue:36683", "right": "issue:44625", "accept": false, - "reason": "Both concern config fields, but the models and missing properties differ; too broad to treat as one bug." + "reason": "Different missing-attribute / propagation bugs in different model configs." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "UMT5 encoder weight loading vs Qwen3.5 MoE from_pretrained error; distinct failures." + "reason": "Both involve Qwen3.5-related loading, but one is missing weights and the other is a from_pretrained error; too little evidence they are the same bug." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Training/optimizer state bug vs a typo in cache strategies; unrelated." + "reason": "Distributed training optimizer-state bug vs a typo in cache strategies; unrelated." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "no_grad decorator on cache updates vs generic import errors; different code paths." + "reason": "Decorator semantics for cache updates vs broad import-error regressions; not the same underlying issue." }, { "left": "issue:44075", "right": "issue:46365", "accept": false, - "reason": "Optimizer arguments not used vs importing Trainer blocked by datasets; unrelated." + "reason": "Optimizer argument handling and an import failure tied to `datasets` are unrelated." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Both are Python/import-time problems, but one is compiled-file execution and the other is a torch.jit parsing issue." + "reason": "Both are import-time failures, but one is about compiled Python files and the other is a TorchScript parse/indentation issue." }, { "left": "issue:43867", "right": "issue:45375", "accept": false, - "reason": "Both are serialization/config issues, but they affect different models and different missing fields." + "reason": "Generic load-model-state_dict issue vs a missing Qwen3_5MoeVisionConfig field; not a duplicate." }, { "left": "issue:43867", "right": "issue:46202", "accept": false, - "reason": "Checkpoint/state_dict load failure vs Gemma pipeline roundtrip on a tokenization character; unrelated." + "reason": "Both are model roundtrip/load failures, but they concern different models and different failure modes." }, { "left": "issue:44938", "right": "issue:45071", "accept": false, - "reason": "Different compatibility regressions: Python 3.14 load failure vs PretrainedConfig type-checking breakage in v5.4.0." + "reason": "Python 3.14 import failure and PretrainedConfig type-checking regression are different compatibility problems." }, { "left": "issue:36683", "right": "issue:43064", "accept": false, - "reason": "Config attribute error vs distributed optimizer-state bug; no common underlying bug." + "reason": "Config attribute bug vs distributed training optimizer-state corruption; unrelated." }, { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "Both are test flakiness issues, but one is missing set_seed and the other is an xdist file race; different causes." + "reason": "Missing `set_seed()` in tests and xdist file-race in captured_info are different test-infrastructure issues." }, { "left": "issue:43010", "right": "issue:45850", "accept": false, - "reason": "Specific no_grad cache bug vs a broad complaint that v5 breaks many models; not the same defect." + "reason": "Specific cache-update decorator change vs a broad report that v5 breaks too many models; not the same bug." }, { "left": "issue:41669", "right": "issue:44246", - "accept": false, - "reason": "Both touch import performance, but one identifies a concrete import-* cause while the other is a vague timing complaint." + "accept": true, + "reason": "Both point to the same import-time slowdown/regression in `transformers` import, with 41669 giving the concrete likely cause (`import *` in models)." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "Both are save/load regressions, but on different models and key mappings." + "reason": "Generic load failure vs Qwen3.5 visual-encoder key regression; distinct serialization bugs." }, { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Different Qwen3.5/UMT5 loading failures; no shared code path." + "reason": "Both reference Qwen3.5-ish loading problems, but one is missing embed weights and the other is a `from_pretrained` error; not enough to treat as the same bug." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Tensor-parallel training bug vs TP OOM while loading a large model; related area but not the same issue." + "reason": "Tensor-parallel training bug vs TP OOM on Qwen2-57B; different symptoms and likely different causes." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIP output field regression and `patchtsmixer` post_init policy are unrelated." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "Cache update decorator bug vs Python-version support issue; unrelated." + "reason": "Cache update no-grad semantics and Python-version support are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "PyTorch missing ImportError for AutoModel vs tokenizer-loading regression; different layers and failures." }, { "left": "issue:43618", - "right": "issue:44077", + "right": "issue:46285", + "accept": false, + "reason": "CLIP attentions regression vs CLIP text-model structural inconsistency; related area, but not the same concrete bug." + }, + { + "left": "issue:43618", + "right": "issue:44857", "accept": false, - "reason": "CLIP output regression vs patchtsmixer post_init restriction; different subsystems." + "reason": "CLIP attentions missing and an AMP/CUDA loss crash are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall and conservative. The canonical/best choice around 41669 vs 44246 is well supported by the packet: both are import-time slowdown regressions, with 41669 providing the concrete cause. The other soft-edge rejections mostly stay within the evidence and do not overclaim duplication." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -15968,7 +16170,6 @@ "cluster_id": "cluster-10000-464", "nodes": [ "issue:27957", - "issue:28282", "issue:30333", "issue:30990", "issue:31795", @@ -15986,7 +16187,6 @@ "issue:43299", "issue:43502", "issue:43576", - "issue:43618", "issue:43723", "issue:43824", "issue:43827", @@ -15994,6 +16194,7 @@ "issue:44062", "issue:44112", "issue:44222", + "issue:44230", "issue:44315", "issue:44351", "issue:44355", @@ -16003,7 +16204,8 @@ "issue:44556", "issue:44655", "issue:44749", - "issue:44857", + "issue:44908", + "issue:44938", "issue:44961", "issue:44964", "issue:45071", @@ -16013,298 +16215,298 @@ "issue:45468", "issue:45478", "issue:45600", - "issue:46132", - "issue:46285" + "issue:46132" ], "soft_pairs": [ - "issue:28282|issue:33453", - "issue:43618|issue:46285", - "issue:43618|issue:44857", "issue:44360|issue:45440", "issue:42890|issue:44964", "issue:43502|issue:44062", "issue:42890|issue:44112", "issue:42890|issue:45600", "issue:44387|issue:44556", - "issue:43723|issue:44556", "issue:44222|issue:45125", + "issue:43723|issue:44556", "issue:43010|issue:43065", "issue:43576|issue:46132", "issue:43502|issue:44485", "issue:45125|issue:45478", "issue:44355|issue:44749", - "issue:30333|issue:33453", "issue:42548|issue:44351", + "issue:30333|issue:33453", "issue:45125|issue:45375", "issue:43010|issue:43232", - "issue:30990|issue:31795", "issue:43502|issue:44222", + "issue:30990|issue:31795", "issue:43502|issue:44961", "issue:42831|issue:43874", "issue:43299|issue:44556", - "issue:30333|issue:36296", "issue:43827|issue:44961", + "issue:30333|issue:36296", "issue:44355|issue:45071", "issue:44355|issue:44655", "issue:33453|issue:34689", "issue:27957|issue:33453", - "issue:44315|issue:44360", "issue:42757|issue:43576", "issue:43576|issue:45071", + "issue:44315|issue:44360", "issue:43576|issue:43824", "issue:31795|issue:38617", - "issue:44961|issue:45468" + "issue:44961|issue:45468", + "issue:44485|issue:44961", + "issue:43576|issue:44938", + "issue:44908|issue:44961", + "issue:44222|issue:44230" ], - "prepared_review_unit_hash": "fcd4f9a9916aa930ab8b1a029c7731836d8c372ae22ae535f8ce6300c8db944c", + "prepared_review_unit_hash": "9a0e9bfae9427cd7e1fe3a1f31da0897fa70cc79d711f2731da8b8497c45cb88", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 37, - "serialized_chars": 38305, - "estimated_input_tokens": 9577, - "estimated_eval_tokens": 19410 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 38, + "serialized_chars": 37943, + "estimated_input_tokens": 9486, + "estimated_eval_tokens": 19228 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly semantically similar Transformers regressions, but they split across distinct code paths and symptoms (imports, tokenizer/model loading, CLI, docs, quantization, tensor-parallelism, and model-specific bugs). None of the soft pairs are safe duplicate merges.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43576 is the most broadly representative user-facing regression in this set: it is a concrete v5 breakage with a clear repro surface, and several other near-matches are also v5-era loading/compatibility regressions.", + "summary": "This cluster is mostly heterogeneous Transformers issues with a few loose v5/loading-regression similarities, but no pair is a safe duplicate match. I would not merge any of the soft edges.", + "confidence": 0.83, + "canonical_issue_reason": "issue:44556 is the broadest representative of the loading/regression theme seen across the set, especially version-upgrade breakages; however, the overall cluster is still too mixed to treat as one true duplicate group.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43576 is the best single issue to anchor triage because it is specific, actionable, and closest to the recurring v5 regression theme among the candidates.", + "best_issue_reason": "issue:44556 is the best umbrella issue for the repeated v5 reload/loading-breakage reports, but it is not a strong canonical duplicate for the whole cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:28282", - "right": "issue:33453", - "accept": false, - "reason": "Both are loader-related failures, but one is a PyTorch dependency ImportError and the other is a tokenizer-loading regression; different root causes and fixes." - }, - { - "left": "issue:43618", - "right": "issue:46285", - "accept": false, - "reason": "Both mention CLIP outputs/structure, but one is missing attentions assignment and the other is an API/attribute inconsistency after v5; not the same bug." - }, - { - "left": "issue:43618", - "right": "issue:44857", - "accept": false, - "reason": "One is a CLIPOutput field regression, the other is a float16/CUDA loss crash in LwDetrImageLoss; unrelated code paths." - }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "Different architectures and problems: a DSA/ReLU discussion versus DeepSeekV3MoE divergence from remote implementation." + "reason": "Different model families and bugs: DSA indexer ReLU vs DeepSeekV3 implementation divergence." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "One is a flaky test due to missing set_seed, the other is a model loading failure for Phi-4 multimodal; unrelated." + "reason": "CI test flakiness from missing set_seed is unrelated to Phi-4 multimodal loading failure." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "One is an offline/local-files behavior bug, the other is a tokenizer AddedToken argument conflict; different failure modes." + "reason": "local_files_only network violation and AddedToken keyword collision are different code paths." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Both are CI/test issues, but one is seeding-related flakiness and the other is a stale device override test; not the same underlying defect." + "reason": "Different test failures in different models/subsystems." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "One is test flakiness, the other is auto_mappings referencing removed configs and breaking CI; different fixes." + "reason": "Seeded test instability vs stale auto_mappings reference; not the same bug." }, { "left": "issue:44387", "right": "issue:44556", "accept": false, - "reason": "One is int4 quantization memory growth/OOM, the other is checkpoint reload incompatibility after upgrading to v5; different root causes." + "reason": "Both are version-regression reports, but one is int4 OOM memory behavior and the other is checkpoint reload failure." }, { - "left": "issue:43723", - "right": "issue:44556", + "left": "issue:44222", + "right": "issue:45125", "accept": false, - "reason": "Both involve v5 loading, but one is AutoTokenizer.from_pretrained and the other is checkpoint reload compatibility; not one concrete bug." + "reason": "FP8 save_pretrained bug and missing _tp_plan/tensor-parallel support are related only broadly, not the same defect." }, { - "left": "issue:44222", - "right": "issue:45125", + "left": "issue:43723", + "right": "issue:44556", "accept": false, - "reason": "FP8 save_pretrained for moe versus missing _tp_plan for tensor parallelism are separate model-export and parallelism issues." + "reason": "Tokenizer loading regression and general checkpoint reload regression are not the same concrete failure." }, { "left": "issue:43010", "right": "issue:43065", "accept": false, - "reason": "A cache/layer no_grad decoration request is not the same as a dummy Conv2d in Sam3PixelDecoder." + "reason": "Cache no_grad decoration and Sam3PixelDecoder dummy Conv2d are unrelated." }, { "left": "issue:43576", "right": "issue:46132", "accept": false, - "reason": "One is a broken CLI command in v5; the other is an AttentionInterface registration behavior change. Same broad area, different bug." + "reason": "Broken env CLI and AttentionInterface.register behavior change are different features." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "local_files_only network leakage is unrelated to GLM-5 RoPE implementation details." + "reason": "Network requests with local_files_only and GLM-5 RoPE implementation are unrelated." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "Both concern Qwen3.5 MoE, but one is tensor-parallel metadata and the other is from_pretrained loading error; not the same concrete fix." + "reason": "Same model family, but one is tensor-parallel metadata and the other is a from_pretrained loading error; not enough to merge." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Both mention compiled/loading/runtime problems, but one is compiled Python file errors and the other is data filtering slowdown after upgrade; not duplicates." + "reason": "Compiled Python file errors and slowed filtering after upgrade are different issues." }, { - "left": "issue:30333", - "right": "issue:33453", + "left": "issue:42548", + "right": "issue:44351", "accept": false, - "reason": "MLflow reporting on failed jobs is unrelated to tokenizer-loading regression." + "reason": "Different missing exports: PreTrainedModel vs HybridCache." }, { - "left": "issue:42548", - "right": "issue:44351", + "left": "issue:30333", + "right": "issue:33453", "accept": false, - "reason": "Both are import errors from transformers, but one is missing PreTrainedModel and the other is missing HybridCache; separate symbols and likely separate causes." + "reason": "MLFlow reporting and tokenizer loading are unrelated." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Different missing fields/APIs: tensor-parallel plan versus deepstack_visual_indexes being dropped by strict config parsing." + "reason": "Different Qwen3.5 config/model bugs: missing _tp_plan vs dropped config field." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "One is a no_grad decoration request for cache/layer updates; the other is generation kwargs update after sync_gpus. Not the same bug." + "reason": "Cache/layer no_grad change and generation kwargs sync bug are unrelated." }, { - "left": "issue:30990", - "right": "issue:31795", + "left": "issue:43502", + "right": "issue:44222", "accept": false, - "reason": "Both are about loading/docs around models, but one is Sentence Transformers hanging and the other is confusing forward() docs; unrelated." + "reason": "Local-files-only network leakage and FP8 save_pretrained are different failures." }, { - "left": "issue:43502", - "right": "issue:44222", + "left": "issue:30990", + "right": "issue:31795", "accept": false, - "reason": "local_files_only network calls and FP8 moe save_pretrained are different subsystems and failure modes." + "reason": "Loading hang and documentation confusion do not match." }, { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "The 'racoon' issue is not meaningfully related to local_files_only behavior." + "reason": "Network access bug and unrelated issue title." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "FineGrainedFP8 accuracy and a missing image-patch-count method in GLM46V are different bugs." + "reason": "FP8 accuracy issue and GLM46V patch-count AttributeError are different bugs." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Both are v5-related model loading regressions, but one is Qwen3VL MoE loading and the other is general checkpoint reload compatibility." + "reason": "Both involve loading regressions, but different model/version scenarios and likely different root causes." }, { - "left": "issue:30333", - "right": "issue:36296", + "left": "issue:43827", + "right": "issue:44961", "accept": false, - "reason": "MLflow reporting on failure and tensor-parallel training are unrelated." + "reason": "Docs still referencing pipeline() is unrelated." }, { - "left": "issue:43827", - "right": "issue:44961", + "left": "issue:30333", + "right": "issue:36296", "accept": false, - "reason": "Documentation references to removed pipeline() are unrelated to the 'racoon' issue." + "reason": "MLFlow reporting and tensor-parallel training bug are unrelated." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Compiled Python file errors are unrelated to PretrainedConfig type-checking regressions." + "reason": "Compiled-file load errors and PretrainedConfig type-checking regression are different." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Compiled-file execution errors and inability to save Pipeline objects are different issues." + "reason": "Compiled-file errors and save_pretrained pipeline serialization are unrelated." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Both are loading regressions, but tokenizer loading and Llama 3.2 vision model loading are distinct breakages." + "reason": "Different loading regressions affecting different components/models." }, { "left": "issue:27957", "right": "issue:33453", "accept": false, - "reason": "XLMRoberta Flash Attention 2 support and tokenizer-loading regression are unrelated." - }, - { - "left": "issue:44315", - "right": "issue:44360", - "accept": false, - "reason": "Liger Kernel model_init behavior and a DSA indexer/ReLU discussion are different problems." + "reason": "Flash Attention 2 support and tokenizer loading regression are unrelated." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "Missing is_offline_mode import and broken transformers env command are both v5-era breakages, but not the same code path." + "reason": "Import error from huggingface_hub and env CLI breakage are different issues." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "A broken CLI command and a PretrainedConfig type-checking issue do not share the same underlying defect." + "reason": "Broken env command and PretrainedConfig typing regression are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel model_init behavior and DSA indexer ReLU are unrelated." }, { "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "CLI breakage in v5 is unrelated to missing Qwen2_5_VLForConditionalGeneration import." + "reason": "env CLI bug and Qwen2.5-VL import error are different failure modes." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "A documentation clarification request is not a duplicate of an import error for layer_type_validation." + "reason": "Documentation clarification and missing symbol import are unrelated." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "The 'racoon' issue is unrelated to Gemma-4 audio relative positional encoding." + "reason": "Unrelated issues with no shared code path." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation and unrelated placeholder title." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug and unrelated title." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "Both mention FP8/MoE, but one is save_pretrained and the other is inference support; too broad to merge." } ] }, @@ -16344,7 +16546,6 @@ "issue:43901", "issue:43976", "issue:44038", - "issue:44222", "issue:44230", "issue:44246", "issue:44273", @@ -16353,21 +16554,20 @@ "issue:44315", "issue:44351", "issue:44393", - "issue:44485", "issue:44492", "issue:44556", "issue:44623", "issue:44704", "issue:44908", - "issue:44938", - "issue:44961", "issue:45003", "issue:45071", "issue:45092", "issue:45125", "issue:45230", "issue:45278", + "issue:45341", "issue:45362", + "issue:45478", "issue:45542", "issue:45753", "issue:45850", @@ -16375,13 +16575,10 @@ "issue:46143", "issue:46144", "issue:46209", - "issue:46218" + "issue:46218", + "issue:46328" ], "soft_pairs": [ - "issue:43576|issue:44938", - "issue:44485|issue:44961", - "issue:44908|issue:44961", - "issue:44222|issue:44230", "issue:46143|issue:46144", "issue:36296|issue:43901", "issue:41628|issue:42548", @@ -16408,271 +16605,288 @@ "issue:42890|issue:44908", "issue:43408|issue:44315", "issue:43352|issue:44393", - "issue:44246|issue:44273", "issue:43576|issue:44908", + "issue:44246|issue:44273", "issue:44295|issue:45230", "issue:44297|issue:45230", "issue:43976|issue:44393", "issue:44908|issue:45071", "issue:44704|issue:44908", - "issue:41669|issue:44351" + "issue:41669|issue:44351", + "issue:44492|issue:46328", + "issue:41669|issue:45341", + "issue:43976|issue:45850", + "issue:44393|issue:45478", + "issue:43976|issue:46218" ], - "prepared_review_unit_hash": "246092d427eb44a612ca7797bbeaa231c223e7b0d23ecd2d54dff35d9889190e", + "prepared_review_unit_hash": "0b13650cd6f70556cb3a167d36056aba6e43bca1437fbc790d07f276578801c2", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 38, - "serialized_chars": 38290, - "estimated_input_tokens": 9573, - "estimated_eval_tokens": 19402 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 39, + "serialized_chars": 38154, + "estimated_input_tokens": 9539, + "estimated_eval_tokens": 19334 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of unrelated Transformers regression reports centered on v5-era breakages, import/export issues, and model-specific bugs. I did not find any soft pair that clearly describes the same underlying bug/change, so I rejected all candidate merges.", - "confidence": 0.58, - "canonical_issue_reason": "issue:45850 is the broadest representative of the cluster\u2019s dominant theme: multiple v5 regressions causing breakage across models and APIs. It is still quite generic, but it best matches the overall cluster context.", - "canonical_pr_reason": null, - "best_issue_reason": "issue:45850 is the most cluster-like umbrella issue; it is broad enough to stand in for the repeated post-v5 breakage reports, even though the set is not a clean duplicate cluster.", - "best_pr_reason": null, - "soft_edge_verdicts": [ - { - "left": "issue:43576", - "right": "issue:44938", - "accept": false, - "reason": "Different failures: the env subcommand is broken in v5, while the other is a Python 3.14 import/load failure." - }, - { - "left": "issue:44485", - "right": "issue:44961", - "accept": false, - "reason": "Unrelated topics; GLM-5 RoPE discussion is not the same as a placeholder issue titled racoon." - }, - { - "left": "issue:44908", - "right": "issue:44961", - "accept": false, - "reason": "Scheduler kwargs handling is unrelated to the racoon issue." - }, - { - "left": "issue:44222", - "right": "issue:44230", - "accept": false, - "reason": "Both mention FP8/MoE, but one is save_pretrained serialization and the other is inference support; different code paths." - }, + "summary": "This cluster is a loose mix of Transformers v5 regression reports, import failures, and model-specific compatibility bugs. None of the soft pairs look like the same concrete defect, so all soft edges are rejected.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45278 is the broadest and most representative report in the set: it captures multiple import regressions after upgrading to v5.5.0, which matches the dominant theme across the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best global issue representative because it is concrete enough to summarize the upgrade/regression class while still broad enough to cover several nearby reports in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ { "left": "issue:46143", "right": "issue:46144", "accept": false, - "reason": "Related RoFormer area, but one is kwargs not forwarded and the other is attention interface usage; not the same bug." + "reason": "Related RoFormer area, but one is a kwargs passthrough bug and the other is an attention-interface implementation issue; not the same fix." }, { "left": "issue:36296", "right": "issue:43901", "accept": false, - "reason": "Tensor parallel training bug and text-classification docs mismatch are unrelated." + "reason": "Tensor-parallel training bug and TextClassificationPipeline docs mismatch are unrelated." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "Both are import errors, but for different symbols and likely different export paths." + "reason": "Both are import errors, but they fail on different symbols and likely different missing exports/code paths." }, { "left": "issue:36296", "right": "issue:43526", "accept": false, - "reason": "Tensor parallel training and BeitImageProcessorFast reduce_labels are unrelated." + "reason": "Tensor parallel training and BeitImageProcessorFast label reduction are unrelated subsystems." }, { "left": "issue:44315", "right": "issue:46209", "accept": false, - "reason": "Liger kernel with model_init and FP8 inference on B200 are separate concerns." + "reason": "Liger-kernel/model_init behavior and FP8 inference on B200 are different runtime problems." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Chunked prefill cache bug and a cache-strategies typo are not the same problem." + "reason": "Chunked-prefill cache regression and a cache-strategy typo are not the same underlying bug." }, { "left": "issue:45362", "right": "issue:45850", "accept": false, - "reason": "Qwen3.5 chat crash is a specific model bug; the other is a broad v5 breakage report." + "reason": "The Qwen3.5 chat crash may be one example of broader breakage, but the reports are not specific enough to treat as the same concrete bug." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "processor.save_pretrained missing files is unrelated to a generic bug report." + "reason": "Processor save_pretrained missing files is a specific serialization bug; the other issue is too generic to match it." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading malfunction and tensorboard/TensorFlow backend error are different issues." + "reason": "Lazy loading failure and TensorBoard/TF backend initialization error are unrelated." }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "Both involve Qwen/MoE, but one is a model bug and the other is missing _tp_plan for tensor parallelism." + "reason": "Both mention Qwen3.5/Moe, but one is a general model bug and the other is specifically missing tensor-parallel plan support." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Qwen3-VL bounding box errors and PretrainedConfig type-checking breakage are unrelated." + "reason": "A Qwen3-VL bounding-box generation issue and a PretrainedConfig type-checking regression are unrelated." }, { "left": "issue:44273", "right": "issue:45003", "accept": false, - "reason": "Both touch import mechanics, but lazy loading and unsafe sys.modules access are distinct bugs." + "reason": "Lazy loading problems and unsafe sys.modules access are not the same defect." }, { "left": "issue:44393", "right": "issue:45753", "accept": false, - "reason": "Same model family, but bbox-output errors and infinite looping on a specific image are different symptoms/root causes." + "reason": "Both involve Qwen3/VL outputs, but one is bounding-box hallucination/error and the other is an infinite loop on a specific image." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "Broken env command and checkpoint reload incompatibility after upgrade are separate failures." + "reason": "Both are v5 compatibility complaints, but one is about the env command and the other about loading old checkpoints; different code paths." }, { "left": "issue:44230", "right": "issue:46095", "accept": false, - "reason": "FP8 inference support is unrelated to asking for a weight conversion script." + "reason": "FP8 inference support and a DeepSeekV4 weight-conversion script request are different issues." }, { "left": "issue:43525", "right": "issue:45850", "accept": false, - "reason": "Missing pad_token_id in Llama4Config is a specific config bug, not the same as broad v5 breakage." + "reason": "Missing pad_token_id in Llama4Config is a specific config regression, not the same as a generic v5 breakage report." }, { "left": "issue:41669", "right": "issue:43673", "accept": false, - "reason": "Import-star performance regression and missing GenerationMixin cache are different issues." + "reason": "Import-star cleanup/performance and GenerationMixin cache missing in chunked_prefill are different underlying changes." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both are v5 compatibility problems, but they concern different models and different breakage mechanisms." + "reason": "Both concern v5 reload compatibility, but they affect different model families and likely different failure points." }, { "left": "issue:43572", "right": "issue:45850", "accept": false, - "reason": "StableLmConfig pad_token_idx regression is specific and not the same as the broad breakage report." + "reason": "StableLmConfig pad_token_idx regression is a specific config bug, not the same as a broad model-breakage umbrella." }, { "left": "issue:44492", "right": "issue:45850", "accept": false, - "reason": "A typo in cache strategies is not the same as a broad v5 regression report." + "reason": "A typo in cache-strategy text is unrelated to general v5 breakage." }, { "left": "issue:45278", "right": "issue:45850", "accept": false, - "reason": "Both are broad post-v5 breakage reports, but one is specifically about import errors after upgrade; not the same concrete bug." + "reason": "Both are broad v5 complaints, but one is specifically about import errors after upgrade and the other is a generic 'many models break' report; not the same defect." }, { "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "TF32 API settings guidance and a cache-strategies typo are unrelated." + "reason": "TF32 API settings and a cache-strategy typo are unrelated." }, { "left": "issue:41669", "right": "issue:46218", "accept": false, - "reason": "Slow import from import-* usage and a missing Qwen3_5 symbol are both import-related, but not the same defect." + "reason": "Both touch imports, but one is import-time slowdown from star imports and the other is a module import failure for a specific model." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "Missing set_seed in tests and scheduler kwargs handling are unrelated." + "reason": "Test seeding flakiness and scheduler kwargs handling are unrelated." }, { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "sam3_video/sam3_tracker config mismatch is unrelated to Liger Kernel application with model_init." + "reason": "Model-type warning for sam3 tracker and Liger kernel application are unrelated." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Flash Attention support for NemotronH and Qwen3-VL bounding box errors are different model-specific issues." + "reason": "Flash-Attn support for Nemotron and Qwen3-VL bounding-box errors are different model-specific bugs." }, { - "left": "issue:44246", - "right": "issue:44273", + "left": "issue:43576", + "right": "issue:44908", "accept": false, - "reason": "Import slowness and lazy-loading malfunction are related thematically but not the same concrete bug." + "reason": "Env command breakage and inverse_sqrt scheduler kwargs omission are unrelated." }, { - "left": "issue:43576", - "right": "issue:44908", + "left": "issue:44246", + "right": "issue:44273", "accept": false, - "reason": "Broken env command and inverse_sqrt scheduler kwargs are unrelated." + "reason": "Slow import and lazy-loading malfunction overlap superficially, but the reports do not identify the same concrete failure mode." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "Position_ids buffer read error is specific; the other is just a generic bug report." + "reason": "Position_ids buffer access error and a generic bug report are not the same issue." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "Tokenizer save_pretrained mismatch is unrelated to a generic bug report." + "reason": "Tokenizer save_pretrained metadata mismatch is a specific serialization bug, not the same as the generic issue." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python 3.9/3.10 compatibility breakage and Qwen3-VL bbox issues are unrelated." + "reason": "Python-version compatibility and Qwen3-VL output errors are unrelated." }, { - "left": "issue:44908", - "right": "issue:45071", + "left": "issue:45071", + "right": "issue:45092", "accept": false, - "reason": "Scheduler kwargs handling and PretrainedConfig type checking are different code paths." + "reason": "PretrainedConfig type-checking regression and old InternVL2 meta-init incompatibility are different bugs." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "AutoProcessor cached_file kwargs forwarding and inverse_sqrt scheduler kwargs are unrelated." + "reason": "AutoProcessor kwargs forwarding and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:44908", + "accept": false, + "reason": "Unsafe sys.modules access and scheduler kwargs handling do not share the same code path." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Slow import from import-* usage and missing HybridCache export are different import/export problems." + "reason": "Import-star cleanup/performance and HybridCache import failure are different issues." + }, + { + "left": "issue:44492", + "right": "issue:46328", + "accept": false, + "reason": "Cache-strategy typo and SequenceFeatureExtractor.pad inefficiency are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Model import cleanup and a testing_utils bug are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45850", + "accept": false, + "reason": "Python version support breakage and generic model breakage are not the same concrete defect." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL bounding-box hallucination and Qwen3.5-Moe from_pretrained failure are different symptoms and likely different fixes." + }, + { + "left": "issue:43976", + "right": "issue:46218", + "accept": false, + "reason": "Python-version compatibility and a specific module import failure are not the same bug." } ] }, @@ -16696,10 +16910,13 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-464", "nodes": [ + "issue:30333", "issue:31515", "issue:31795", "issue:33453", "issue:34689", + "issue:36296", + "issue:38617", "issue:41669", "issue:42371", "issue:42913", @@ -16709,20 +16926,20 @@ "issue:43502", "issue:43519", "issue:43525", + "issue:43572", "issue:43576", "issue:43606", "issue:43618", "issue:43704", "issue:43827", - "issue:43976", "issue:44077", + "issue:44230", "issue:44261", "issue:44263", "issue:44273", "issue:44315", "issue:44351", "issue:44360", - "issue:44393", "issue:44485", "issue:44492", "issue:44623", @@ -16735,22 +16952,14 @@ "issue:45230", "issue:45290", "issue:45310", - "issue:45341", "issue:45468", - "issue:45478", "issue:45850", "issue:46018", "issue:46095", "issue:46172", - "issue:46218", - "issue:46328" + "issue:46218" ], "soft_pairs": [ - "issue:44492|issue:46328", - "issue:41669|issue:45341", - "issue:43976|issue:45850", - "issue:44393|issue:45478", - "issue:43976|issue:46218", "issue:44077|issue:44861", "issue:43827|issue:44908", "issue:31795|issue:33453", @@ -16787,276 +16996,314 @@ "issue:44998|issue:45850", "issue:44938|issue:44998", "issue:44861|issue:45290", - "issue:43525|issue:44861" + "issue:43525|issue:44861", + "issue:31515|issue:38617", + "issue:43572|issue:44861", + "issue:44230|issue:44623", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:43298|issue:44230" ], - "prepared_review_unit_hash": "36a8c2b863051856dabc6ed408df7f34c63e78622fafa77a6ee4a2e6e3479ebe", + "prepared_review_unit_hash": "fd05a2d2cf11d669c06bdadac264010b3c06d4f8f400e9c7cdc93b16d8d6e2ec", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 42, - "serialized_chars": 39435, - "estimated_input_tokens": 9859, - "estimated_eval_tokens": 19974 + "soft_pair_count": 45, + "serialized_chars": 40347, + "estimated_input_tokens": 10087, + "estimated_eval_tokens": 20430 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a loose loading/import/tokenizer-regression cluster, but the soft pairs are mostly superficial title matches across different bugs; none look like true duplicates.", - "confidence": 0.79, - "canonical_issue_reason": "issue:43502 is the best anchor for the cluster because it is a concrete, broadly relevant loading-path bug (`local_files_only=True` still triggers requests) and fits the repeated load/regression theme better than the more niche or documentation-only reports.", + "summary": "This cluster is noisy and mostly heterogeneous. A few items touch v5 loading/tokenizer regressions, but the soft pairs are only superficially similar and none should be merged as duplicates.", + "confidence": 0.34, + "canonical_issue_reason": "issue:33453 is the clearest, concrete representative of the small loading/regression subset; it is more specific and actionable than the broader/vaguer v5 complaints.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43502 is the strongest representative issue: it is actionable, clearly a real bug, and broad enough to summarize the cluster\u2019s main failure mode without being as model-specific as several others.", + "best_issue_reason": "issue:33453 best fits as the cluster anchor because it describes a real, generic regression class without being overly broad or model-specific.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44492", - "right": "issue:46328", - "accept": false, - "reason": "Typo in cache strategy naming vs a SequenceFeatureExtractor padding performance issue; different code paths and symptoms." - }, - { - "left": "issue:41669", - "right": "issue:45341", - "accept": false, - "reason": "Import-star cleanup in model code vs a testing_utils bug; unrelated changes." - }, - { - "left": "issue:43976", - "right": "issue:45850", - "accept": false, - "reason": "Python version compatibility failure vs a vague general claim that many models break; not the same concrete bug." - }, - { - "left": "issue:44393", - "right": "issue:45478", - "accept": false, - "reason": "Qwen3-VL 2D box output problem vs Qwen3.5 MoE from_pretrained failure; different model families and failure modes." - }, - { - "left": "issue:43976", - "right": "issue:46218", - "accept": false, - "reason": "Python 3.9/3.10 compatibility issue vs an import failure for Qwen3_5ForCausalLM; not the same root cause." - }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Optional post_init policy for patchtsmixer vs an AttributeError in _get_tied_weight_keys; different utilities and code paths." + "reason": "Optional post_init policy and an AttributeError in tied-weight key handling are different code paths and failure modes." }, { "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Documentation still mentioning pipeline() vs a scheduler keyword-argument bug; docs cleanup is not the same as runtime behavior." + "reason": "Docs still mentioning pipeline() is unrelated to inverse_sqrt scheduler kwargs being ignored." }, { "left": "issue:31795", "right": "issue:33453", "accept": false, - "reason": "Docs confusion around model.forward arguments vs an actual tokenizer loading regression; one is documentation, the other is behavior." + "reason": "Forward-doc confusion about input_ids/past_key_values is not the same bug as tokenizer loading regression." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Liger Kernel not applied in model_init vs degenerate training with flash_attention_3; different subsystems." + "reason": "Liger Kernel not applying under model_init and flash_attention_3 causing degenerate training are different training issues." }, { "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Broken `transformers env` command vs lazy loading not working; unrelated command-line and loading bugs." + "reason": "A broken env command and lazy loading failures are unrelated symptoms." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer activation discussion vs Gemma4Audio positional encoding issue; no shared code path." + "reason": "DSA indexer ReLU discussion and Gemma-4 audio positional encoding are unrelated model internals." }, { "left": "issue:44655", "right": "issue:44861", "accept": false, - "reason": "Pipeline save_pretrained failure vs tied-weight-key handling crash; different saving code paths." + "reason": "Pipeline save_pretrained serialization and tied-weight key AttributeError are different subsystems." }, { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "Generic bug report vs a specific Qwen3.5 MoE from_pretrained error; too little evidence of the same underlying issue." + "reason": "A generic bug report is not enough to treat this as the same as a specific Qwen3.5 MoE from_pretrained error." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy loading bug vs inverse_sqrt scheduler kwargs being ignored; unrelated functionality." + "reason": "Lazy loading behavior and scheduler kwarg handling are unrelated." }, { "left": "issue:43576", "right": "issue:46218", "accept": false, - "reason": "Broken env command vs inability to import Qwen3_5ForCausalLM; different surfaces and causes." + "reason": "Broken env command and missing Qwen3_5 import are not the same failure." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Qwen3-VL timestamp calculation bug vs GLM-5 RoPE implementation discussion; separate model logic." + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are different processor/model bugs." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Timestamp computation in a processor vs scheduler kwarg handling; no meaningful overlap." + "reason": "Timestamp math in a processor is unrelated to scheduler kwargs being dropped." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Removing import-star usage vs a VRAM leak in multi-threaded inference; unrelated problems." + "reason": "Import-star performance cleanup and a VRAM leak in multithreaded inference are unrelated." }, { "left": "issue:46018", "right": "issue:46172", "accept": false, - "reason": "DeepSeek-V4 shared expert gating issue vs a `comb.to(...).transpose(...)` bug; different model internals." + "reason": "DeepSeek shared-expert gating and a comb.to(...).transpose bug do not match the same code path." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Documentation ambiguity in `forward` vs a model-loading regression for Llama 3.2; different categories." + "reason": "Forward API documentation confusion is not the same as a model-loading breakage for Llama 3.2 Vision." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Slow checkpoint loading vs confusing `forward` docs; one is performance, the other documentation." + "reason": "Slow checkpoint loading and forward docs confusion are unrelated." }, { "left": "issue:43704", "right": "issue:44485", "accept": false, - "reason": "VRAM leak with multiple dataloader threads vs GLM-5 RoPE implementation; unrelated issues." + "reason": "VRAM leakage and RoPE implementation issues are different problems." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "GLM-5 RoPE implementation discussion vs Qwen3.5 MoE from_pretrained error; different models and fixes." + "reason": "GLM-5 RoPE and Gemma-4 audio positional encoding are model-specific and unrelated." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 API settings migration vs VRAM leak in dataloader threads; not the same bug." + "reason": "TF32 API migration guidance has nothing to do with a VRAM leak." }, { "left": "issue:45850", "right": "issue:46218", "accept": false, - "reason": "Generic complaint that many models break vs a specific import failure for Qwen3_5ForCausalLM; too broad to be the same issue." + "reason": "A vague claim that v5 breaks many models is too broad to merge with a specific Qwen3_5 import failure." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "Network requests despite `local_files_only=True` vs incorrect timestamp calculation in Qwen3VL Processor; different behavior and components." + "reason": "local_files_only network requests and timestamp calculation are different bugs." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Processor timestamp bug vs missing files on `processor.save_pretrained(...)`; separate processor-related failures." + "reason": "Processor timestamp math is unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Continuous batching not supporting audio models and deepseek-v4 weight conversion are unrelated." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Processor timestamp bug vs VRAM leak in dataloader threads; unrelated symptoms." + "reason": "Timestamp calculation and VRAM leakage are unrelated." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "`torch.split()` return-value issue in GlmMoeDsaIndexer vs Qwen3.5 MoE from_pretrained error; different code paths." + "reason": "torch.split return handling and Gemma-4 audio positional encoding are unrelated." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch in CI vs Qwen3.5 MoE loading error; not the same underlying problem." + "reason": "CPU offload device mismatch and Gemma-4 audio positional encoding are unrelated." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "MLA layer norm epsilon precision discussion vs Qwen3.5 MoE from_pretrained error; unrelated implementation details." + "reason": "Missing rms_norm_eps in MLA q_a_layernorm is not the same as Gemma-4 audio positional encoding." }, { "left": "issue:44351", "right": "issue:46218", "accept": false, - "reason": "Cannot import HybridCache vs cannot import Qwen3_5ForCausalLM; both are import errors, but they concern different symbols and likely different fixes." + "reason": "Both are import errors, but they are for different symbols and likely different root causes." }, { "left": "issue:43352", "right": "issue:46218", "accept": false, - "reason": "Flash Attention 2 unsupported for NemotronHForCausalLM vs an import failure for Qwen3_5ForCausalLM; different failure modes." + "reason": "Flash Attention 2 support limitation and Qwen3_5 import failure are different issues." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "General tokenizer behavior change from v4 to v5 vs wrong tokenizer decoder type; related area, but not the same concrete bug." + "reason": "Both mention tokenizer/v5, but one is a behavior change and the other is a wrong decoder type; not enough to call them duplicates." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs bug vs an obviously unrelated placeholder title; no duplicate signal." + "reason": "Scheduler kwargs handling is unrelated to the placeholder 'Unemployment' issue." }, { "left": "issue:44492", "right": "issue:44998", "accept": false, - "reason": "Cache strategy typo vs unrelated placeholder title; no shared bug." + "reason": "A typo in cache strategies is unrelated to a placeholder bug report." }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions missing vs unrelated placeholder title; not a match." + "reason": "CLIPOutput attentions assignment and a placeholder issue are unrelated." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "A placeholder title does not establish the same underlying bug as a broad v5-breakage complaint." }, { "left": "issue:44938", "right": "issue:44998", "accept": false, - "reason": "Python 3.14 loading failure vs unrelated placeholder title; not a duplicate." + "reason": "Python 3.14 load failure and a placeholder issue are unrelated." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "Tied-weight-key AttributeError vs chat template crash on tool-call assistant messages; different model formatting/saving paths." + "reason": "Tied-weight key crashes and chat-template tool-call crashes are different code paths." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Missing `pad_token_id` on Llama4Config vs tied-weight-key crash; both are attribute errors but on different objects and code paths." + "reason": "Missing pad_token_id in Llama4Config and tied-weight key AttributeError are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow from_pretrained checkpoint loading and an ImportError for layer_type_validation are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig and tied-weight key handling are unrelated." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 support for Qwen3-VL/Qwen3.5 MoE and missing processor save files are different issues." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLFlow reporting on failed training jobs and slow checkpoint loading are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Slow checkpoint loading and a tensor-parallel training bug are not the same failure." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 inference support and a from_pretrained error for Qwen3.5 MoE are different symptoms." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Checkpoint load slowness is not the same as tokenizer-loading regression." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Audio continuous batching support and FP8 Qwen3-VL/Qwen3.5 support are different feature requests/bugs." } ] }, @@ -17080,36 +17327,17 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-464", "nodes": [ - "issue:30333", - "issue:31515", - "issue:33453", - "issue:36296", "issue:37934", - "issue:38617", "issue:42936", "issue:43240", - "issue:43298", - "issue:43572", "issue:43668", "issue:43717", "issue:43813", "issue:44031", - "issue:44230", - "issue:44623", - "issue:44861", - "issue:45310", "issue:46018", "issue:46144" ], "soft_pairs": [ - "issue:31515|issue:38617", - "issue:43572|issue:44861", - "issue:44230|issue:44623", - "issue:30333|issue:31515", - "issue:31515|issue:36296", - "issue:44230|issue:45310", - "issue:31515|issue:33453", - "issue:43298|issue:44230", "issue:43717|issue:46144", "issue:42936|issue:46018", "issue:43668|issue:44031", @@ -17123,154 +17351,106 @@ "issue:37934|issue:43668", "issue:42936|issue:43668" ], - "prepared_review_unit_hash": "954e428da656b98897903cb4714216bd2ff57db10c9b762a835a6e535af0cdd6", + "prepared_review_unit_hash": "ee3df0b014496603d4a02edf8497a317d8ceaecb8d65830888ae6640943b8e8b", "component_budget": { "node_count": 464, "item_count": 464, "soft_pair_count": 1705, - "serialized_chars": 884263, - "estimated_input_tokens": 221066, - "estimated_eval_tokens": 442388 + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 }, "budget": { - "node_count": 20, - "item_count": 20, - "soft_pair_count": 20, - "serialized_chars": 17139, - "estimated_input_tokens": 4285, - "estimated_eval_tokens": 8826 + "node_count": 9, + "item_count": 9, + "soft_pair_count": 12, + "serialized_chars": 8610, + "estimated_input_tokens": 2153, + "estimated_eval_tokens": 4562 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is a loose collection of unrelated Transformers issues across model loading, config typing, tokenizer warnings, training/runtime bugs, and inference features. None of the soft pairs look like the same underlying defect or change.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45310 is the clearest, most concrete bug report in the set: a version-specific from_pretrained failure for Qwen3.5 MoE with a well-scoped runtime symptom.", + "summary": "This cluster is heterogeneous: the issues cover unrelated model, tokenizer, config, and typo problems. None of the soft pairs look like the same underlying bug, so all should be rejected.", + "confidence": 0.91, + "canonical_issue_reason": "No true canonical duplicate issue emerges from this set; if one issue must anchor the cluster, issue 42936 is the most concrete functional bug report, but it is still distinct from the others.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the best representative issue because it is specific, actionable, and describes a concrete user-facing failure rather than a broad feature request or vague regression.", + "best_issue_reason": "Issue 42936 is the strongest representative only in the sense that it describes a specific runtime/model bug with clear behavior, but the cluster overall is not a duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:31515", - "right": "issue:38617", - "accept": false, - "reason": "One is a slow checkpoint load complaint; the other is an ImportError from a missing symbol. Different failure modes and code paths." - }, - { - "left": "issue:43572", - "right": "issue:44861", - "accept": false, - "reason": "StableLmConfig missing pad_token_idx and a tied-weights AttributeError are unrelated configuration vs weight-tie bugs." - }, - { - "left": "issue:44230", - "right": "issue:44623", - "accept": false, - "reason": "FP8 Qwen MoE inference support and processor.save_pretrained missing files are different feature/serialization problems." - }, - { - "left": "issue:30333", - "right": "issue:31515", - "accept": false, - "reason": "MLFlow not updating on failed training jobs is unrelated to slow from_pretrained checkpoint loading." - }, - { - "left": "issue:31515", - "right": "issue:36296", - "accept": false, - "reason": "Checkpoint load slowness and tensor-parallel training bugs do not share a concrete underlying bug." - }, - { - "left": "issue:44230", - "right": "issue:45310", - "accept": false, - "reason": "Both mention Qwen3.5 MoE, but one is FP8 inference support and the other is a from_pretrained error; not the same fix." - }, - { - "left": "issue:31515", - "right": "issue:33453", - "accept": false, - "reason": "Slow checkpoint loading and tokenizer loading regression are separate subsystems and symptoms." - }, - { - "left": "issue:43298", - "right": "issue:44230", - "accept": false, - "reason": "Continuous batching audio-model support is an inference pipeline feature gap, not FP8 Qwen MoE support." - }, { "left": "issue:43717", "right": "issue:46144", "accept": false, - "reason": "Mamba init_weights behavior and RoFormer attention-interface usage are distinct model implementation issues." + "reason": "Different model areas and failures: Mamba init-weights behavior vs RoFormer attention interface usage." }, { "left": "issue:42936", "right": "issue:46018", "accept": false, - "reason": "Mask2Former ignore_value handling and DeepSeek-V4 shared-expert gating are unrelated model-specific bugs." + "reason": "Mask2Former ignore_value handling and DeepSeek-V4 expert gating are unrelated bugs." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "ModernBERT norm_eps type hint and tokenizer regex warning after 4.57.3 are unrelated; one is typing, the other is warning behavior." + "reason": "One is a config type-hint issue; the other is a tokenizer regex warning regression." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "A Mask2Former ignore_value bug and kwargs not forwarded to loss calculation are different training-loss plumbing issues." + "reason": "Mask2Former ignore_value usage and loss-function kwargs propagation are separate code paths." }, { "left": "issue:43240", "right": "issue:46144", "accept": false, - "reason": "Loss kwargs propagation and RoFormer attention interface concerns are separate code paths." + "reason": "Loss kwargs forwarding and RoFormer attention interface usage do not describe the same defect." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Model ignore_value handling does not match tokenizer regex warning regressions." + "reason": "Different subsystems and symptoms: segmentation model ignore_value vs tokenizer warning behavior." }, { "left": "issue:42936", "right": "issue:46144", "accept": false, - "reason": "These are different model-specific implementation bugs with no clear shared root cause." + "reason": "No overlap in concrete bug or code path." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "Mask2Former ignore_value logic and a quantization-operation typo are unrelated." + "reason": "A Mask2Former behavior bug and a quantization-typo issue are unrelated." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Missing loss kwargs and an incorrect type hint are not the same defect." + "reason": "Loss kwargs forwarding and a config type annotation mistake are distinct issues." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "Type-hint correction and a string/attribute typo in quantization code are separate changes." + "reason": "These are different kinds of fixes: type-hint correction vs typo in a string/path reference." }, { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "A question about RMS norm semantics is not the same as a type-hint bug." + "reason": "Llama4 norm semantics and ModernBERT config typing are unrelated." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Mask2Former runtime behavior and ModernBERT type annotation are unrelated." + "reason": "Mask2Former ignore_value behavior and ModernBERT norm_eps typing are not the same bug." } ] }, @@ -17303,52 +17483,52 @@ "issue:10519|issue:11005", "issue:11005|issue:11307" ], - "prepared_review_unit_hash": "d3845436de1156660fe220685c9190fbf4b51c42210ecf95caa87840ff9205f7", + "prepared_review_unit_hash": "a83a8558adb7482c5210b7a13c9baeefd6d1a27062875064cc9edae94c81ee04", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2877, - "estimated_input_tokens": 720, - "estimated_eval_tokens": 1696 + "serialized_chars": 2876, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2877, - "estimated_input_tokens": 720, - "estimated_eval_tokens": 1696 + "serialized_chars": 2876, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These three issues are unrelated: token truncation behavior, ReduceLROnPlateau-like functionality, and Wav2Vec2 word time offsets. They should not be treated as duplicates.", - "confidence": 0.99, - "canonical_issue_reason": "issue #10519 is the most concrete and well-scoped, but the items are not duplicates of one another.", + "summary": "The three issues are not duplicates: one is about tokenizer truncation direction, one requests ReduceLROnPlateau-like scheduler behavior, and one asks for Wav2Vec2 word boundary offsets. The title overlap is superficial only.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue for the cluster: the topics are unrelated and do not describe the same underlying bug or feature request.", "canonical_pr_reason": null, - "best_issue_reason": "issue #10519 is the clearest, most actionable request; the other two are separate feature ideas in different areas.", + "best_issue_reason": "Issue 10519 is the clearest and most self-contained report, but it is not a cluster representative because the other issues are unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:10519", "right": "issue:11307", "accept": false, - "reason": "Truncation strategy for token sequences is unrelated to Wav2Vec2 word timestamp extraction." + "reason": "Different domains and requests: truncation behavior vs. audio alignment offsets." }, { "left": "issue:10519", "right": "issue:11005", "accept": false, - "reason": "Token truncation options and ReduceLROnPlateau-like scheduling are different features with no shared bug." + "reason": "Unrelated feature requests: tokenizer truncation direction vs. learning-rate scheduler behavior." }, { "left": "issue:11005", "right": "issue:11307", "accept": false, - "reason": "Learning-rate scheduling functionality is unrelated to Wav2Vec2 time-offset requests." + "reason": "No shared underlying bug or change; one is about schedulers, the other about Wav2Vec2 token timing offsets." } ] }, @@ -17380,7 +17560,7 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "458f54631d1a1979b628d16a54a65996639ad80f865fe99d9b92c16ef2e2fb60", + "prepared_review_unit_hash": "3973c443d761b27691eb5eb2cb0042fdd0640e2180cecbfd84698b65e918506f", "component_budget": { "node_count": 3, "item_count": 3, @@ -17402,24 +17582,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These issues are not a true duplicate cluster. Two are about model loading/download behavior, but one is about tokenizer parity for TAPAS and the other is about a private-model auth error message; neither matches the bert-base-uncased loading report closely enough to merge.", - "confidence": 0.78, - "canonical_issue_reason": "Issue 16618 is the broadest and most central of the three, since it concerns model loading itself, but it is still not a clean umbrella for the others.", + "summary": "These issues do not appear to describe the same underlying bug: one is about TAPAS tokenization differences, one about a misleading private-model download/auth error, and one about failing to load a specific public model.", + "confidence": 0.94, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "16618 is the best representative only in a loose sense because it is the most general user-facing failure in the set and has the most discussion.", + "best_issue_reason": "No clear canonical issue: the three reports concern different code paths and failure modes, so this cluster does not look like a true duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Different problems: TAPAS tokenization parity vs inability to load bert-base-uncased. They affect different code paths and would not be fixed by one change." + "reason": "Different problem domains: tokenization parity vs model loading failure for bert-base-uncased." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "15980 is about an incorrect error message when downloading a private model without login; 16618 is about failing to load a public model. Related area, but not the same underlying bug." + "reason": "One is an auth/download error for private models; the other is a generic inability to load a public model." } ] }, @@ -17449,7 +17629,7 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "ec5f5483f70c24a2f919f2dec45610de01631b6442e4ab33bd940578b9d1d5d4", + "prepared_review_unit_hash": "aa222752ae273a9c94064ddade8900882b303de330a3dc02d872ee6a5fc701a5", "component_budget": { "node_count": 2, "item_count": 2, @@ -17471,18 +17651,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are both broad community-contribution feature requests, but for different capabilities: Flash Attention 2 support vs GGUF support. They are related only at the template/category level, not the underlying change.", - "confidence": 0.98, - "canonical_issue_reason": "No canonical duplicate issue: the two issues request different features and would require unrelated implementation work.", + "summary": "These are two unrelated feature-request issues: one for Flash Attention 2 support on more architectures, the other for GGUF support on more architectures. They share a broad theme of architecture support but not the same underlying change or bug, so they should not be clustered as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "If a representative issue is needed, #26350 is slightly better as the more established and more discussed request, but it is not a duplicate of #33260.", + "best_issue_reason": "No single canonical issue fits because the requests target different features and code paths; if forced, issue 26350 is simply the older and more discussed thread, not a duplicate representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different feature requests for different support targets (Flash Attention 2 vs GGUF); not the same underlying bug/change." + "reason": "Different feature requests (Flash Attention 2 vs GGUF) with different implementation targets; same broad topic of architecture support is not enough to treat them as the same underlying change." } ] }, @@ -17541,187 +17721,187 @@ "issue:44370|issue:46226", "issue:46222|issue:46226" ], - "prepared_review_unit_hash": "340936a47e1a1f475a31e9c004d30759c8896a215f97b77fec21f5b19abc2746", + "prepared_review_unit_hash": "e513404bcbfc646d88d74b7586f9e02416545d838e79b9d66935df80043586a1", "component_budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 24, - "serialized_chars": 13526, - "estimated_input_tokens": 3382, - "estimated_eval_tokens": 7020 + "serialized_chars": 13529, + "estimated_input_tokens": 3383, + "estimated_eval_tokens": 7022 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 24, - "serialized_chars": 13526, - "estimated_input_tokens": 3382, - "estimated_eval_tokens": 7020 + "serialized_chars": 13529, + "estimated_input_tokens": 3383, + "estimated_eval_tokens": 7022 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Mostly locale-specific doc-translation requests, so the cluster is not one duplicate set; the only clear duplicate-looking pair is the identical Thai issue pair.", - "confidence": 0.87, - "canonical_issue_reason": "No single cluster-wide canonical: Turkish, Spanish, Romanian, and the placeholder-language issues are separate translation requests. If deduping the one exact duplicate pair, issue:46222 is the better canonical because it is earlier and still open.", + "summary": "The cluster is mostly repetitive i18n translation-template noise. Only the 46222/46226 pair has enough evidence to accept as a true duplicate; the other pairs are just soft textual matches without concrete shared targets, filenames, or diff evidence.", + "confidence": 0.72, + "canonical_issue_reason": "issue:46222 is the safest canonical anchor because it is part of the only exact duplicate pair in the packet and is the earlier of the two identical issues.", "canonical_pr_reason": null, - "best_issue_reason": "Issue:46222 is the best representative only for the exact duplicate pair, since it is the earlier open copy with the same Thai title/body as 46226.", + "best_issue_reason": "issue:46222 is the strongest issue representative here: it has an exact duplicate in 46226, while the rest of the cluster is only template-level similarity without stronger evidence.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Different translation requests; same template style does not make Turkish the same as another locale." + "reason": "Both are only textually similar translation-template issues; there is no concrete evidence they are the same work item." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "Different locale targets; these are separate doc-translation requests, not the same issue." + "reason": "High template similarity alone is not enough to collapse these into one issue." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Distinct language-specific translation request, not the same underlying task." + "reason": "The titles are too generic to prove duplication, and there is no supporting target or diff evidence." }, { "left": "issue:27088", "right": "issue:38435", "accept": false, - "reason": "Turkish vs Romanian are separate translation requests." + "reason": "The locale-tagged titles indicate related theme, but there is still no evidence they are the same issue." }, { "left": "issue:27088", "right": "issue:28936", "accept": false, - "reason": "Turkish and Spanish translation requests are different issues." + "reason": "These have different locale tags in the title and no evidence of the same underlying request." }, { "left": "issue:28936", "right": "issue:43834", "accept": false, - "reason": "Spanish request is separate from another locale-specific translation issue." + "reason": "The similarity is only at the template level; no concrete shared issue target is shown." }, { "left": "issue:28936", "right": "issue:44365", "accept": false, - "reason": "Different language targets; not the same request." + "reason": "No evidence beyond repetitive wording suggests these are duplicates." }, { "left": "issue:28936", "right": "issue:38435", "accept": false, - "reason": "Spanish vs Romanian translation requests are distinct." + "reason": "Different locale-tagged translation requests; nothing in the packet ties them to the same work item." }, { "left": "issue:28936", "right": "issue:44370", "accept": false, - "reason": "Separate locale-specific translation requests, not one bug/change." + "reason": "Soft similarity alone is insufficient to merge these as duplicates." }, { "left": "issue:38435", "right": "issue:44370", "accept": false, - "reason": "Romanian request and the placeholder-language issue are not clearly the same translation task." + "reason": "Near-identical template text is not enough without shared targets, files, or other concrete linkage." }, { "left": "issue:38435", "right": "issue:44365", "accept": false, - "reason": "Different locale requests; same broad subsystem is not enough." + "reason": "The packet only shows repeated template wording, not the same underlying issue." }, { "left": "issue:38435", "right": "issue:43834", "accept": false, - "reason": "These are separate language translation requests, not duplicates." + "reason": "These are too similar textually, but there is no concrete evidence they are the same issue." }, { "left": "issue:38435", "right": "issue:46222", "accept": false, - "reason": "Romanian translation request and the Thai request are different issues." + "reason": "The titles/body lengths are similar, but there is no evidence that they refer to the same request." }, { "left": "issue:38435", "right": "issue:46226", "accept": false, - "reason": "Romanian and Thai translation requests are distinct." + "reason": "Template similarity alone does not justify merging these as duplicates." }, { "left": "issue:43834", "right": "issue:44370", "accept": false, - "reason": "Placeholder titles likely correspond to different languages; not enough evidence of the same request." + "reason": "Both are generic placeholder-style issues, but the packet does not prove they are the same underlying item." }, { "left": "issue:43834", "right": "issue:44365", "accept": false, - "reason": "These appear to be different locale translation requests." + "reason": "Very similar wording, but no concrete evidence of a shared issue target or duplicate report." }, { "left": "issue:43834", "right": "issue:46222", "accept": false, - "reason": "No evidence these are the same language or the same request." + "reason": "The titles are generic and similar, but that is not enough to confirm duplication." }, { "left": "issue:43834", "right": "issue:46226", "accept": false, - "reason": "Different translation requests; do not merge on template similarity alone." + "reason": "No evidence beyond soft textual similarity supports treating these as the same issue." }, { "left": "issue:44365", "right": "issue:44370", "accept": false, - "reason": "Likely different locale requests despite identical template wording." + "reason": "Near-identical template text is suggestive, but still not enough to accept as a duplicate." }, { "left": "issue:44365", "right": "issue:46222", "accept": false, - "reason": "Separate translation tasks for different languages." + "reason": "The packet gives only similarity scores, not concrete shared issue evidence." }, { "left": "issue:44365", "right": "issue:46226", "accept": false, - "reason": "Separate translation tasks for different languages." + "reason": "Soft similarity alone is insufficient to collapse these into one issue." }, { "left": "issue:44370", "right": "issue:46222", "accept": false, - "reason": "Different locale requests; not the same underlying issue." + "reason": "There is no concrete link beyond repetitive template language." }, { "left": "issue:44370", "right": "issue:46226", "accept": false, - "reason": "Different locale requests; not the same underlying issue." + "reason": "The available data does not substantiate that these are the same underlying request." }, { "left": "issue:46222", "right": "issue:46226", "accept": true, - "reason": "Same Thai title/body and near-duplicate timestamps; this looks like the same translation request duplicated." + "reason": "Same title and identical body length with the strongest similarity score in the packet make this the only grounded duplicate pair." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative: it treats the Thai pair (46222/46226) as the only clear duplicate-looking pair, and it avoids overclaiming duplicate status for the locale/template translation requests where the packet only shows soft similarity. The canonical choice of 46222 as the earlier open copy is supported by the timestamps and states." + "accept": false, + "feedback": "The 46222/46226 duplicate call is grounded, but the rest of the analysis is not conservative enough. The packet only gives titles, lengths, dates, and soft-similarity scores; it does not substantiate claims that the placeholder-title issues are separate Turkish/Spanish/Romanian locale-specific work items. Several of the false-pair rationales overstate evidence by inferring locale separation from data that isn\u2019t present. The canonical choice of 43834 is also not strongly supported over the other nearly identical placeholder issues." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -17794,7 +17974,7 @@ "issue:43837|issue:45177", "issue:45322|issue:46133" ], - "prepared_review_unit_hash": "83a51dcb254da1e846308f2432935626f63b3e1a7098b73eb3d5681fe67e9129", + "prepared_review_unit_hash": "e5a3cb9e75d9cb920d5d5f3cea96c458fb40e835d5eef274e93ac8517472ad65", "component_budget": { "node_count": 16, "item_count": 16, @@ -17816,240 +17996,234 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a loose collection of model-request issues across different model families and modalities (text, audio, vision). They are related by topic, but they are not duplicates of the same underlying change.", + "summary": "These are all separate model-request issues across different model families and modalities (text, vision, audio, OCR, detection). The similarities are mostly generic 'add model/support' phrasing, not the same underlying change, so none of the soft edges should be merged.", "confidence": 0.97, - "canonical_issue_reason": null, + "canonical_issue_reason": "No true canonical duplicate exists here; the issues are distinct requests. If a hub-like representative is needed, issue 43671 is the closest because it is an active speech-model request with the strongest thematic overlap to the other audio-related proposals.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Issue 43671 is the strongest cluster anchor by discussion/activity and recurring overlap with related audio-model requests, but it still should not subsume the other issues as duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "EuroBert config and DEIMv2 are different model additions with different architecture/config needs." + "reason": "Different model requests: EuroBert config vs DEIMv2. Shared 'add model' wording only." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert is a text encoder model request; Qwen3-TTS is an audio generation support request." + "reason": "EuroBert config and Qwen3-TTS are unrelated models and code paths." }, { "left": "issue:36817", "right": "issue:36879", "accept": false, - "reason": "EuroBert and RF-DETR are unrelated model families (BERT-style encoder vs detection model)." + "reason": "EuroBert vs RF-DETR are separate architectures and tasks; not the same change." }, { "left": "issue:36879", "right": "issue:41211", "accept": false, - "reason": "RF-DETR and DEIMv2 are distinct detection/vision model requests, not the same fix." + "reason": "RF-DETR and DEIMv2 are different detection/vision models, not one duplicate request." }, { "left": "issue:36879", "right": "issue:43909", "accept": false, - "reason": "RF-DETR is a vision detection model; LFM2.5 Audio is an audio model request." + "reason": "RF-DETR is a detection model; LFM2.5 Audio 1.5B is an audio model request." }, { "left": "issue:36879", "right": "issue:44121", "accept": false, - "reason": "RF-DETR and a weight-sparse transformer are different model types and implementation paths." + "reason": "RF-DETR and a weight-sparse transformer are unrelated model families." }, { "left": "issue:36879", "right": "issue:43837", "accept": false, - "reason": "RF-DETR is vision detection; Qwen3-ASR is speech recognition support." + "reason": "RF-DETR and Qwen3-ASR are different modalities and implementation paths." }, { "left": "issue:36879", "right": "issue:43671", "accept": false, - "reason": "RF-DETR and Qwen3-TTS target unrelated modalities and code paths." + "reason": "RF-DETR vision detection and Qwen3-TTS speech support are not the same underlying change." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are different model requests with different architecture assumptions." + "reason": "NeoBERT and Qwen3-ASR are distinct model additions, not duplicates." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT is a BERT-like text model; circuit-sparsity is a sparse-transformer request." + "reason": "NeoBERT and circuit-sparsity transformer support are different model/code-path requests." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT and LFM2.5 Audio are unrelated text vs audio model additions." + "reason": "NeoBERT is a text model; LFM2.5 Audio 1.5B is an audio model request." }, { "left": "issue:36879", "right": "issue:37015", "accept": false, - "reason": "RF-DETR and NeoBERT do not share the same underlying bug or implementation change." + "reason": "RF-DETR and NeoBERT are unrelated models with no shared concrete fix." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT and ModernVBERT are different model families; both are additions, not duplicates." + "reason": "NeoBERT and ModernVBERT models are different model requests, not one duplicate." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 architecture and Qwen3-ASR are separate model support requests in different domains." + "reason": "MiniCPM3 architecture support and Qwen3-ASR support target different models and tasks." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 support is not the same change as adding a weight-sparse transformer model." + "reason": "MiniCPM3 and the weight-sparse transformer request are separate changes." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 and ModernVBERT are different model architecture requests." + "reason": "MiniCPM3 and ModernVBERT are different model additions." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 and Voxtral TTS support are unrelated model additions (text/LLM vs audio TTS)." + "reason": "MiniCPM3 is a model architecture request; Voxtral-4B-TTS is a speech model support request." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 and Qwen3-TTS are different modalities and would require separate implementations." + "reason": "DEIMv2 and Qwen3-TTS are unrelated vision vs speech requests." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT models and Qwen3-ASR are different model-support requests." + "reason": "ModernVBERT and Qwen3-ASR are different model families and modalities." }, { "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT and Qwen3-TTS are unrelated text vs audio model requests." + "reason": "ModernVBERT and Qwen3-TTS are distinct model support requests." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT is a concrete model request; BERT-like models with RoPE is a broad architectural class, not the same issue." + "reason": "ModernVBERT is a concrete model request; BERT-like models with RoPE is a broad architectural theme." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "BERT-like RoPE models and Voxtral TTS concern different model types and code paths." + "reason": "Generic BERT-with-RoPE discussion and Voxtral TTS support are not the same change." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS and Qwen3-ASR are related by prefix but represent different tasks and fixes." + "reason": "Same Qwen3 family, but TTS vs ASR are different modalities and code paths." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Qwen3-TTS and Voxtral TTS are both audio-related but for different models and implementations." + "reason": "Qwen3-TTS and Voxtral TTS are different model integrations, not one duplicate fix." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities (speech vs OCR/vision)." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated speech vs OCR model requests." }, { "left": "issue:43671", "right": "issue:44121", "accept": false, - "reason": "Qwen3-TTS support is not the same as adding a sparse transformer model." + "reason": "Qwen3-TTS and the sparse transformer request target different model architectures." }, { "left": "issue:43671", "right": "issue:43909", "accept": false, - "reason": "Qwen3-TTS and LFM2.5 Audio are separate audio model support requests." + "reason": "Qwen3-TTS and LFM2.5 Audio 1.5B are separate audio-model support requests." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR and LFM2.5 Audio are distinct speech/audio model requests." + "reason": "Qwen3-ASR and LFM2.5 Audio 1.5B are different audio-related models, not the same change." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and EUPE are unrelated audio vs perception-encoder model requests." + "reason": "ASR support and EUPE vision-language encoder support are unrelated." }, { "left": "issue:43837", "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR support is not the same underlying change as sparse-transformer support." - }, - { - "left": "issue:43837", - "right": "issue:45027", - "accept": false, - "reason": "Qwen3-ASR and Voxtral TTS are different speech tasks and implementations." - }, - { - "left": "issue:43837", - "right": "issue:45177", - "accept": false, - "reason": "Qwen3-ASR and DeepSeek-OCR2 are different modalities and model classes." + "reason": "Qwen3-ASR and sparse transformer support are different concrete requests." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "LFM2.5 Audio and SAM3-LiteText are different model families; one is audio, the other is a text-related request." + "reason": "Audio model support and SAM3-LiteText are different modalities and model paths." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "A sparse transformer request and EUPE are not the same code-path change." + "reason": "Sparse transformer request and EUPE encoder request are unrelated architectures." }, { "left": "issue:43837", "right": "issue:45027", "accept": false, - "reason": "Qwen3-ASR and Voxtral TTS are separate audio model support requests, not one merged fix." + "reason": "Qwen3-ASR and Voxtral TTS are different speech models with separate integrations." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT and Voxtral TTS target unrelated text vs audio model implementations." + "reason": "NeoBERT and Voxtral TTS are distinct models in different domains." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "Qwen3-ASR and DeepSeek-OCR2 are different modalities and model support paths." + "reason": "ASR support and OCR2 support are different model types and fixes." }, { "left": "issue:45322", "right": "issue:46133", "accept": false, - "reason": "EUPE and TIPSv2 are distinct vision-language model requests from different authors and model families." + "reason": "EUPE and TIPSv2 are both vision-language related, but they are different models and not one mergeable duplicate change." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer support and EUPE encoder support are different underlying changes." } ] }, @@ -18079,7 +18253,7 @@ "soft_pairs": [ "issue:38740|issue:46443" ], - "prepared_review_unit_hash": "56a055dcff0371a6de165fa7aaede292ab68887bc94752be1a627460453569bf", + "prepared_review_unit_hash": "89a25fef005fa7647b044253b990e08e4d9aa793199d3a724bd951cf1a01d306", "component_budget": { "node_count": 2, "item_count": 2, @@ -18101,18 +18275,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are unrelated: one is a documentation request about adding `pruna` as an optimization framework, and the other is a vision RoPE refactor. There is no duplicate cluster here.", + "summary": "These two issues are not duplicates: one is a docs request about adding Pruna as an optimization framework, while the other is an open Vision RoPE refactor. They do not describe the same bug, feature, or code path.", "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Neither issue is a good canonical representative because they concern unrelated topics and should not be clustered together.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:38740", "right": "issue:46443", "accept": false, - "reason": "Different topics and artifact types of work: a docs add-on request versus a vision model refactor. They do not describe the same underlying bug or change." + "reason": "Reject: the titles and contexts are unrelated (documentation/framework listing vs. Vision RoPE refactor), so they are not the same underlying issue." } ] }, @@ -18142,12 +18316,12 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "2ddf0da63bbb46bb90d5ebb71a8124851078c345d93b12a4cc0d129ec9d6129c", + "prepared_review_unit_hash": "b95fddc452f4489288b4a793ed551a4aba374dea21a84097767a4a8a005d82fe", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1480, + "serialized_chars": 1479, "estimated_input_tokens": 370, "estimated_eval_tokens": 996 }, @@ -18155,7 +18329,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1480, + "serialized_chars": 1479, "estimated_input_tokens": 370, "estimated_eval_tokens": 996 }, @@ -18164,18 +18338,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are unrelated: one is about block decoding behavior, the other requests support for a specific model/version (DeepSeek v3.2).", - "confidence": 0.98, - "canonical_issue_reason": "Neither issue is a clear duplicate of the other; they describe different feature/bug areas with no shared targets or evidence of the same underlying problem.", + "summary": "These two issues are not duplicates: one is a vague request about block decoding, while the other is a specific model-support request for DeepSeek v3.2. They do not appear to describe the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": "issue #43037 is the more concrete, actionable report because it names a specific feature/support target (DeepSeek v3.2), whereas #41084 is too vague to serve as a clear canonical issue.", "canonical_pr_reason": null, - "best_issue_reason": "No single canonical issue is appropriate because the topics are distinct and do not appear to be duplicate reports.", + "best_issue_reason": "#43037 is the better representative issue for the cluster simply because it is specific and scoped; #41084 is too ambiguous to anchor triage.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different subjects: decoding implementation vs model support request. No shared filenames, targets, or concrete overlap indicating the same underlying bug/change." + "reason": "The titles and available metadata point to unrelated concerns: generic block decoding vs DeepSeek v3.2 support. No shared targets, files, or explicit linkage suggest the same underlying problem." } ] }, @@ -18205,12 +18379,12 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "5322125063dc67d6131217613c5a62c6cac0ea47742ab0661d20190d354531a0", + "prepared_review_unit_hash": "d48756b022ca5936c9106a0c03a3cf181e98a64603e3c6094c84d00d3db3cd3c", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1583, + "serialized_chars": 1584, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -18218,7 +18392,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1583, + "serialized_chars": 1584, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -18227,18 +18401,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both issues mention custom trainers, but they address different behaviors: one is about `remove_unused_column` customization in `Trainer`, and the other is about FLOPS estimation for custom trainers. They are related in theme but not duplicates.", - "confidence": 0.93, - "canonical_issue_reason": "Issue #43598 is the more concrete Trainer customization request and centers on a specific code path (`remove_unused_column`).", + "summary": "These are related only at a broad Trainer/customization level, but they describe different problems: one is about `remove_unused_column` behavior, the other about FLOPS estimation for custom trainers. They should not be merged as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 43600 is the more concrete, user-facing bug report: it targets FLOPS estimation behavior for custom trainers with a specific failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "#43598 is the best representative because it is narrowly scoped and directly tied to a specific Trainer behavior that can be changed independently.", + "best_issue_reason": "Issue 43598 is the broader customization concern around `remove_unused_column` and may be the better umbrella topic for Trainer configurability.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Shared mention of custom trainers is too broad; one issue is about column pruning in `Trainer`, the other about FLOPS accounting. They do not appear to describe the same underlying bug or change." + "reason": "They mention custom trainers, but the underlying issues are different: unused-column removal customizability vs FLOPS counting logic. Not the same bug or change." } ] }, @@ -18268,7 +18442,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "714e86239e7fb18a61941ab01cdc67d4be7e9839d1b00af54cdbba3a2ef6badf", + "prepared_review_unit_hash": "25bfc1bb83cc8cb1810e6fe7a863ff60be65759e705ff74aa6fcae11c244e7d9", "component_budget": { "node_count": 2, "item_count": 2, @@ -18290,18 +18464,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related only at a very broad level (Qwen/AutoModel support) but they target different models and different capabilities: Qwen3-Omni registration vs. Qwen3.5 sequence-classification support.", + "summary": "These are related model-support requests, but they target different integrations: one asks for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, while the other asks for AutoModelForSequenceClassification support for Qwen3.5. They should not be merged as duplicates.", "confidence": 0.94, - "canonical_issue_reason": "Issue 43684 is the broader, more general AutoModel/AutoModelForConditionalGeneration registration request and is the better representative of the cluster.", + "canonical_issue_reason": "Issue #43684 is the broader model-registration request and is more likely to represent the core automation gap, but it is not the same underlying request as #44405.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43684 is the best issue anchor because it describes a concrete registration gap with a broader scope than the narrower Qwen3.5 sequence-classification request.", + "best_issue_reason": "#43684 is the better canonical issue because it is the broader and more general API registration request; #44405 is a separate task for sequence classification support on a different model family/config.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different underlying requests: one is Qwen3-Omni model registration for generic generation auto classes, the other is Qwen3.5 AutoModelForSequenceClassification support. Same family, but not the same bug/change." + "reason": "Different target APIs and model families: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration versus Qwen3.5 sequence classification support. Related area, but not the same bug or change." } ] }, @@ -18338,76 +18512,79 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "b75c7276d26113350a0e04c0f9371cbab3e12cd947f1bd2afdf26183550e5530", + "prepared_review_unit_hash": "8dadb17f96972fb5e481efd11aa998c0b19e74adfd9ed77af096ca7ea126a8b2", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4265, - "estimated_input_tokens": 1067, - "estimated_eval_tokens": 2390 + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4265, - "estimated_input_tokens": 1067, - "estimated_eval_tokens": 2390 + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster appears to be a set of low-information issue duplicates with generic or near-identical titles, but there is not enough body/detail evidence to confidently merge any pair.", - "confidence": 0.42, - "canonical_issue_reason": "Issue #43845 is the earliest of the clearly overlapping 'huggingface' issues and is as representative as the others given the lack of substantive detail.", + "summary": "Only the three \"huggingface\" issues have plausible cluster affinity, and that affinity is title-level only. With no body text, filenames, targets, or references, it is too weak to assert they are true duplicates; 43845 is just the earliest plausible representative among them. \"Fett\" looks unrelated.", + "confidence": 0.58, + "canonical_issue_reason": "issue:43845 is the earliest of the three title-similar \"huggingface\" issues, so it is the best provisional representative if this cluster is later confirmed.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #43845 is the best anchor for the cluster because it is the earliest near-duplicate title and has no evidence making it less representative than #43846 or #43859.", + "best_issue_reason": "issue:43845 is the best global issue candidate because it is the earliest and cleanest title match in the small title-only \"huggingface\" group.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Titles are too generic and there is no body/detail evidence showing the same underlying bug or request." + "reason": "The titles are different and there is no concrete overlap beyond weak soft similarity, so this is not enough to treat them as the same issue." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Only soft title similarity is present; no concrete shared problem is described." + "reason": "Different titles with no shared body details, targets, or filenames; the evidence is too thin to call them duplicates." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "Insufficient information beyond vague title similarity to treat these as duplicates." + "reason": "\"Fett\" does not have enough visible overlap with the title-only \"huggingface\" reports to justify a same-issue merge." }, { "left": "issue:43845", "right": "issue:43846", "accept": false, - "reason": "Near-identical titles alone are not enough without a shared concrete issue description." + "reason": "Same short title, but no body text or other concrete matching evidence is available; title similarity alone is not enough to confirm a duplicate." }, { "left": "issue:43845", "right": "issue:43859", "accept": false, - "reason": "No substantive evidence beyond the repeated generic title." + "reason": "The titles match, but the packet provides no additional evidence of the same underlying bug, so this should stay a soft cluster only." }, { "left": "issue:43846", "right": "issue:43859", "accept": false, - "reason": "The items look similar superficially, but the available data does not confirm the same underlying issue." + "reason": "Identical-looking titles are suggestive, but without body details or references this is still insufficient to confidently merge as duplicates." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "The clustering is plausible, but the writeup is a bit too strong for the packet. The only shared evidence for the three \u201chuggingface\u201d issues is title-level soft similarity; all pair records have deterministic_accept=false and no body text, filenames, targets, or references. So calling them duplicates/same topic and accepting all three pairings is not fully conservative. The claim that 43845 is the earliest representative is grounded, but the duplicate conclusion should be softer." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -18442,88 +18619,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "39f34c273925ef0b589e1d5698109b8fa1ce3c4a430a90762e8c406257cd70f4", + "prepared_review_unit_hash": "eeb65e7f966c79e59dbe8f869b0ae66b57a4a0cd0119764e5a7cf23ab4afaaed", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5782, - "estimated_input_tokens": 1446, - "estimated_eval_tokens": 3148 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5782, - "estimated_input_tokens": 1446, - "estimated_eval_tokens": 3148 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster looks like unrelated spam/placeholder issues with no substantive body text, explicit targets, or shared concrete bug. None are good duplicate candidates; all soft links should be rejected.", - "confidence": 0.97, - "canonical_issue_reason": "No canonical issue: the items are generic spam/placeholder reports with no actionable problem statement to unify around.", + "summary": "All five items are low-signal and appear unrelated; no soft duplicate pairs are convincing. Issue 44418 is the only one that looks like a plausible substantive report, but it is still too vague to anchor a duplicate cluster with confidence.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 44418 is the least spam-like and the only title that reads like a real user request, so it is the best available issue anchor among otherwise noisy items.", "canonical_pr_reason": null, - "best_issue_reason": "No issue is a strong global best; each entry is too vague and non-technical to serve as a meaningful duplicate anchor.", + "best_issue_reason": "44418 is the best issue candidate because it appears to be an actual documentation enhancement request, whereas the others look like spam or unrelated text.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both titles are just '' and provide no concrete shared bug or feature request beyond generic spam labeling." + "reason": "Both are titled '' but there is no concrete bug, feature, or target described to indicate the same underlying issue." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Both are spam placeholders with no body detail; identical generic titles are not enough to infer the same underlying issue." + "reason": "Generic spam titles with no shared technical context; similarity is purely superficial." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "'' and 'Documentation Enhancement Suggestion' are too vague and describe different kinds of low-signal content, not one bug." + "reason": "One is spam and the other is a documentation enhancement suggestion; they do not describe the same problem." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "A spam placeholder is not the same underlying issue as 'Check out \"Google LLC\"'; no concrete shared problem is visible." + "reason": "The titles are unrelated and contain no shared actionable subject matter." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "A documentation-suggestion placeholder and a spam placeholder are not clearly the same defect/change request." + "reason": "A documentation request and a spam post are not the same underlying issue." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "Both are low-information, but the titles do not establish a single underlying issue or duplicate report." + "reason": "No evidence of a common bug or change request beyond both being low-detail closed issues." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "Identical generic spam titles are insufficient evidence of the same underlying bug or change request." + "reason": "Both are spam-labeled titles, but there is no indication they refer to the same report or problem." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "The titles point to different spam-like content with no shared concrete issue description." + "reason": "The titles are unrelated; one is placeholder spam and the other is a specific quoted phrase." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "There is no evidence these are the same underlying report; the titles are unrelated and too vague." + "reason": "No shared concrete issue is visible; the similarity signal is not enough to treat them as duplicates." } ] }, @@ -18553,7 +18730,7 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "f372737fbb3182c420045dcb0c01d2d40ead99e7068291acfd49f6cba76a9549", + "prepared_review_unit_hash": "1f59f6538556bf6eaae22229e8f3d890cf4a30d84edcfab3e97d2bf02abd6746", "component_budget": { "node_count": 2, "item_count": 2, @@ -18575,28 +18752,28 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues appear to be near-exact duplicates: same title, similar body size, and no distinguishing metadata. They should be clustered together as the same underlying issue.", - "confidence": 0.98, - "canonical_issue_reason": "Issue #45145 is a clean canonical choice because it is effectively identical to #45151 and was created slightly earlier.", + "summary": "The two issues have identical titles and similar lengths, but there is no body-level evidence that they are true duplicates. They look like a possible duplicate / likely same topic, not a confirmed duplicate edge.", + "confidence": 0.69, + "canonical_issue_reason": "#45145 is the earlier report and the best representative for this topic, but the evidence only supports a likely duplicate, not a confirmed one.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #45145 is the best representative since it appears to be the same report as #45151 and is the earlier submission.", + "best_issue_reason": "#45145 is the best issue candidate because it is the earliest instance and has the same title as #45151, making it the most suitable cluster representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", - "accept": true, - "reason": "Exact same title and very similar body length with no other differentiating evidence; this looks like a duplicate issue." + "accept": false, + "reason": "Identical titles and close body lengths suggest the same topic, but there is no content or metadata evidence proving the same underlying issue; this is only a soft similarity." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The conclusion is grounded in the packet: the two issues have identical titles, very similar body lengths, and no stronger contrary evidence in the provided metadata. The soft-edge acceptance is conservative enough for a duplicate-issue cluster." + "accept": false, + "feedback": "The packet supports only a soft similarity judgment: identical titles and close body lengths, but no body/content evidence showing the issues are truly duplicates. The analyst overstates the evidence by calling them the same underlying issue and saying the earlier issue \u2018fully represents\u2019 the cluster. A more conservative verdict would be \u2018possible duplicate / likely same topic\u2019 rather than an accepted duplicate edge." }, "evaluator_used": true, - "retried": false, - "accepted_nontrivial_soft_edge": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -18673,40 +18850,40 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "fa234b4a83d05be4e495ef1b9c2e4ca808730a805209d720a3d1db1a3f2bd4c5", + "prepared_review_unit_hash": "c2805542e07c879f9fe3946da965e2ba2d8808513c9a9eeae8328f21a48c7067", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1672, - "estimated_input_tokens": 418, - "estimated_eval_tokens": 1092 + "serialized_chars": 1673, + "estimated_input_tokens": 419, + "estimated_eval_tokens": 1094 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1672, - "estimated_input_tokens": 418, - "estimated_eval_tokens": 1092 + "serialized_chars": 1673, + "estimated_input_tokens": 419, + "estimated_eval_tokens": 1094 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate 4-bit inference bugs in nearby areas, but they do not look like duplicates: one is a torch.finfo/uint8 audio-module failure, the other is a BitsAndBytesConfig skip-modules regression that removes the default lm_head exclusion.", + "summary": "These are two separate 4-bit inference issues: one is a Gemma4 audio-module uint8/torch.finfo TypeError, the other is a BitsAndBytesConfig lm_head exclusion regression causing an AssertionError. They share the broad theme of quantized inference failures but not the same code path or fix.", "confidence": 0.94, - "canonical_issue_reason": "No single canonical duplicate: issue 45672 is a torch.finfo/uint8 audio-module failure, while 45674 is a BitsAndBytesConfig regression where llm_int8_skip_modules clears the default lm_head exclusion and triggers an AssertionError.", + "canonical_issue_reason": "issue #45674 is the more general, configuration-level regression and is likely the better representative of the cluster than the model-specific Gemma4 audio-module bug.", "canonical_pr_reason": null, - "best_issue_reason": "45674 is the cleaner representative issue because it states a concrete config regression and failure mode; 45672 is a separate, more niche audio/NF4 dtype path.", + "best_issue_reason": "issue #45674 is the strongest standalone issue because it describes a clear, reproducible 4-bit inference regression tied to a specific configuration change and has broader applicability.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Different underlying bugs and code paths: TypeError from torch.finfo on uint8 audio-module weights vs AssertionError caused by skip_modules clearing the default lm_head exclusion. Similar 4-bit context is not enough to merge them." + "reason": "Different underlying bugs: one is a torch.finfo uint8 failure in Gemma4 audio modules, the other is a BitsAndBytesConfig skip-modules/ lm_head exclusion assertion. Same general area, but not the same code-path or fix." } ] }, @@ -18736,40 +18913,40 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "4e99415bcf748447cfa39afea866b3e1dda62c6dcfc583653a76bd59877af0d5", + "prepared_review_unit_hash": "2485bd27610dfaf21844ae8ba07e8427c4fe689a03c6c3cf0ced418fc320c0f1", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two closed issues are not duplicates; they are unrelated low-signal items with different personal names/dates and no shared technical context.", - "confidence": 0.99, + "summary": "The two issues are just closed placeholder-like entries with different personal-name titles and different dates; there is no evidence they describe the same underlying bug or request, so they should not be clustered as duplicates.", + "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good canonical representative because there is no underlying shared bug or feature request; they appear unrelated.", + "best_issue_reason": "Neither issue is a good canonical representative because there is no duplicate relationship to resolve; the titles point to different entities and there is no body evidence tying them together.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "Same broad pattern only via superficial text similarity; the titles refer to different people and dates, with no shared problem description, code path, or explicit linkage." + "reason": "Different names and dates in the titles, with no body text, links, or shared evidence indicating the same issue." } ] }, @@ -18799,40 +18976,40 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "caad7aebb56acebfc3af9cb7ae445b74360c96ec8c1154e179fdc0573915bf59", + "prepared_review_unit_hash": "9ba07b8f4e5f12357788f874257cbf99b4a297bd7a65c4d0f64e643820a828ac", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1480, - "estimated_input_tokens": 370, - "estimated_eval_tokens": 996 + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1480, - "estimated_input_tokens": 370, - "estimated_eval_tokens": 996 + "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": "The two closed issues appear unrelated and contain no technical description or shared context; they look like separate low-information entries rather than duplicates of the same bug.", - "confidence": 0.94, - "canonical_issue_reason": null, + "summary": "The two issues are unrelated non-technical placeholders (personal names plus dates) with no shared problem signal beyond superficial similarity.", + "confidence": 0.98, + "canonical_issue_reason": "Neither issue is a meaningful canonical technical report; both look like unrelated personal-entry stubs rather than duplicates of the same bug.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "No issue in this cluster is a good representative for duplicate triage, since neither contains actionable problem details.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "Different personal-name/date titles with no body, filenames, targets, or technical details; no evidence they describe the same underlying issue." + "reason": "Different names and dates; no evidence they describe the same underlying bug or request." } ] }, @@ -18862,40 +19039,40 @@ "soft_pairs": [ "issue:45810|issue:46295" ], - "prepared_review_unit_hash": "dc59f1357440f1073fb223b727e55f7db60baaabe258bb801198f345327260a4", + "prepared_review_unit_hash": "cff6777a29e1aad5a880bc76cfa3ff7a7da57ebe0a78fec0abf4f92495d9137a", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1541, - "estimated_input_tokens": 386, - "estimated_eval_tokens": 1028 + "serialized_chars": 1540, + "estimated_input_tokens": 385, + "estimated_eval_tokens": 1026 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1541, - "estimated_input_tokens": 386, - "estimated_eval_tokens": 1028 + "serialized_chars": 1540, + "estimated_input_tokens": 385, + "estimated_eval_tokens": 1026 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are not duplicates: one requests a new Qwen3_5 token-classification/value-model class, while the other asks for FP8 training support. They affect different features and code paths.", - "confidence": 0.96, + "summary": "The two issues are unrelated feature requests; this is not a duplicate cluster.", + "confidence": 0.97, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "No single canonical issue fits the cluster: 45810 asks for a Qwen3_5 token-classification/value-model class, while 46295 requests FP8 training support.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45810", "right": "issue:46295", "accept": false, - "reason": "Different underlying requests: model-class support for Qwen3_5 token classification vs. training-time FP8 support. They do not describe the same bug or change." + "reason": "Different feature requests in different areas (model class support vs FP8 training support); they do not describe the same underlying bug or change." } ] }, @@ -18935,12 +19112,12 @@ "issue:45871|issue:45873", "issue:45872|issue:45873" ], - "prepared_review_unit_hash": "7f29f841d935d9bce397e6a31ba8d59082dd98da3a5904310e240daa3626ddf0", + "prepared_review_unit_hash": "70f6389fcf624b5fe6e08c4a76c5cb39ee0346bd354e9feeb4547458d073d85c", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5725, + "serialized_chars": 5728, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -18948,7 +19125,7 @@ "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5725, + "serialized_chars": 5728, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -18957,66 +19134,66 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The set splits into two exact duplicate pairs\u201445870/45871 and 45872/45873\u2014while 45869 is a separate dataset/sample request and the cross-pairs are only topical financial-data similarity, not the same issue.", - "confidence": 0.96, - "canonical_issue_reason": "45872 is the strongest anchor here: it is the most detailed, concrete report and 45873 is an exact duplicate, making 45872 the clearest representative issue in the set.", + "summary": "The items split into two clear duplicate pairs: 45870/45871 are the same example request, and 45872/45873 are the same DataForge scraping-service request. 45869 is a separate dataset promo/request and does not match the others.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue for the full cluster: there are two separate duplicate topics (45870/45871 and 45872/45873), while 45869 is unrelated.", "canonical_pr_reason": null, - "best_issue_reason": "45872 is the most actionable and specific issue overall; the other near-matches are either a separate duplicate pair or only loosely related.", + "best_issue_reason": "45871 and 45873 are the best representatives of their respective duplicate pairs; 45869 should not be treated as canonical for either topic.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45869", "right": "issue:45870", "accept": false, - "reason": "Different asks: dataset bundle/sample offer vs. a financial time-series prediction example." + "reason": "Different topics: a dataset bundle promo/request vs an example about financial time series prediction." }, { "left": "issue:45869", "right": "issue:45871", "accept": false, - "reason": "Different asks: dataset bundle/sample offer vs. a financial time-series prediction example." + "reason": "Different topics: a dataset bundle promo/request vs an example about financial time series prediction." }, { "left": "issue:45870", "right": "issue:45871", "accept": true, - "reason": "Same title and apparent body content; clear duplicate." + "reason": "Same title and apparently same example request; these look like exact duplicates." }, { "left": "issue:45870", "right": "issue:45872", "accept": false, - "reason": "Related financial-data theme, but different concrete request: prediction example vs. prediction plus scraping service." + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." }, { "left": "issue:45870", "right": "issue:45873", "accept": false, - "reason": "Related financial-data theme, but different concrete request: prediction example vs. prediction plus scraping service." + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." }, { "left": "issue:45871", "right": "issue:45872", "accept": false, - "reason": "Related financial-data theme, but different concrete request: prediction example vs. prediction plus scraping service." + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." }, { "left": "issue:45871", "right": "issue:45873", "accept": false, - "reason": "Related financial-data theme, but different concrete request: prediction example vs. prediction plus scraping service." + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." }, { "left": "issue:45872", "right": "issue:45873", "accept": true, - "reason": "Same title and apparent body content; clear duplicate." + "reason": "Same title and apparently same service/request; these look like exact duplicates." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative: it only marks the two exact-title/jaccard-1.0 pairs as duplicates, treats the cross-pairs as merely topical similarity, and does not overstate evidence. The canonical issue choice is also supportable from the titles and body-length differences." + "feedback": "The analysis is grounded in the packet: 45870/45871 and 45872/45873 have identical titles and body lengths, while 45869 has a clearly different title/topic. The conclusion that the cluster splits into two duplicate pairs plus one unrelated issue is conservative and supported by the available evidence." }, "evaluator_used": true, "retried": false, @@ -19043,40 +19220,40 @@ "soft_pairs": [ "issue:46439|issue:46441" ], - "prepared_review_unit_hash": "8df69b07e52e02690549bb67f61b41a4c0da862d2f24ff43c266df60b60499d0", + "prepared_review_unit_hash": "91b021c40e97535b87037ad4af00bbcfba461216da485858be3efdac0de954e4", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1637, - "estimated_input_tokens": 410, - "estimated_eval_tokens": 1076 + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1637, - "estimated_input_tokens": 410, - "estimated_eval_tokens": 1076 + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both issues involve static cache and linear-attention, but they describe different observed failures: one is shape corruption leading to a RuntimeError, the other is a Qwen3.5 KeyError for `linear_attention`. They may be related, but the evidence is not strong enough to treat them as the same underlying bug.", - "confidence": 0.69, - "canonical_issue_reason": "Issue 46439 is the better canonical issue because it states a more general mechanism (`early_initialization` corrupting linear-attention layers) rather than a model-specific symptom.", + "summary": "The two issues are in the same static-cache / linear-attention area, but they describe different failure modes: one is shape corruption during `early_initialization`, the other is a Qwen3.5-specific `KeyError: 'linear_attention'`. Related, but not the same bug.", + "confidence": 0.84, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 46439 is the best representative of the cluster since it frames the broader static-cache/linear-attention defect and likely covers more downstream reports.", + "best_issue_reason": "Issue 46439 is the better representative of the broader static-cache/linear-attention corruption problem because it states a concrete shape/runtime failure and a likely root-cause area.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46439", "right": "issue:46441", "accept": false, - "reason": "Do not merge: the titles point to different concrete failures (`RuntimeError` from wrong shapes vs `KeyError: 'linear_attention'`) and there is no explicit shared fix or code path shown." + "reason": "Different concrete failures and likely different code paths: wrong-shape `RuntimeError` in early initialization vs Qwen3.5 `KeyError` for missing `linear_attention` handling. Similar subsystem, but not the same bug." } ] }, @@ -19106,7 +19283,7 @@ "soft_pairs": [ "issue:46551|issue:46552" ], - "prepared_review_unit_hash": "b4bc4e80c22edc58d232e2eced654ed55789f55633e5af8bbc83efd41ba16c09", + "prepared_review_unit_hash": "849ab6d615658470f0ead7b6cc29bb1579e3c13aadbdfee6244274ee6d1951c5", "component_budget": { "node_count": 2, "item_count": 2, @@ -19128,24 +19305,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are effectively exact duplicates: same title, same body size, and matching soft-similarity evidence. The open issue is the better cluster representative for ongoing tracking.", + "summary": "Both artifacts are the same issue report: identical title and essentially identical content about the Wav2Vec2CTCTokenizer v5 word-delimiter bug. They should be clustered together as duplicates/reposts of the same underlying bug.", "confidence": 0.99, - "canonical_issue_reason": "Issue #46552 is the better canonical issue because it is open and can serve as the active tracking item for this exact duplicate report.", + "canonical_issue_reason": "Issue #46551 is a closed instance with a bit more discussion activity, making it the more natural anchor for the duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #46552 best represents the cluster since it is open and identical in content to #46551.", + "best_issue_reason": "#46551 is the best issue representative because it is the closed/original-looking report and already carries the discussion context for the bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:46551", "right": "issue:46552", "accept": true, - "reason": "Exact same title and effectively the same report; these are duplicates describing the same tokenizer bug." + "reason": "Identical title and matching bug description indicate the same underlying issue, not just a related problem." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster summary is grounded in the packet: the two issues have identical titles, identical body_length, and a soft_similarity link with jaccard 1.0. Choosing the open issue as the canonical tracker is conservative and reasonable." + "feedback": "The cluster judgment is well grounded: the two issues have identical titles, identical body length, and a soft-similarity score/jaccard of 1.0, so treating them as duplicates is conservative and supported by the packet. The representative/canonical choice is also reasonable since #46551 is closed and has slightly more discussion activity." }, "evaluator_used": true, "retried": false, @@ -19202,12 +19379,12 @@ "pull_request:43823|pull_request:45213", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "db01b834dc24e83ae3fa91e10850aae1f64554bc91b90e56a9e513c9e2401ed1", + "prepared_review_unit_hash": "ffdb47284d0535eda03297df83dd065aa26e82d555f6bdee4b839ab1b508607f", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 19, - "serialized_chars": 38088, + "serialized_chars": 38086, "estimated_input_tokens": 9522, "estimated_eval_tokens": 19300 }, @@ -19215,7 +19392,7 @@ "node_count": 14, "item_count": 14, "soft_pair_count": 19, - "serialized_chars": 38088, + "serialized_chars": 38086, "estimated_input_tokens": 9522, "estimated_eval_tokens": 19300 }, @@ -19224,132 +19401,132 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly unrelated PRs spanning new-model additions, docs, tokenizer/processor tweaks, and model refactors. The only true duplicate-like pair is the two `fixed_cross_entropy` changes that both add `weight` and `label_smoothing` support in the same file; those should be grouped together. All other soft links are too broad to merge as duplicates.", - "confidence": 0.95, + "summary": "This cluster is mostly heterogeneous PRs from unrelated model/features. The only true duplicate-like pair is the two fixed_cross_entropy changes (43251 and 43254), which address the same code path and can plausibly be merged.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "`pull_request:43251` is the best canonical PR: it makes the same concrete `fixed_cross_entropy` fix, has the explicit tracking issue link, and is the clearer, more complete representation of the change. `43254` is a near-duplicate variant of the same edit.", + "canonical_pr_reason": "PR 43251 is the clearest representative of the shared cross-entropy kwargs fix: it is issue-linked, concrete, and overlaps directly with 43254 in the same function and behavior change.", "best_issue_reason": null, - "best_pr_reason": "`pull_request:43251` is the strongest representative of the only true duplicate pair and is the most suitable global PR anchor because it has the issue reference and the clearest intent. The rest of the PRs are unrelated to this loss-function fix.", + "best_pr_reason": "PR 43251 is the best PR candidate because it most directly captures the underlying bug fix and has the explicit tracking issue, while 43254 is a near-duplicate variant.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Unrelated PRs: VideoPrism model addition vs. ExecuTorch export test; different code paths and goals." + "reason": "Unrelated PRs: Videoprism model/docs addition vs. executorch dynamic-shape export test." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer method addition vs. exportability test; no shared underlying bug or change." + "reason": "Different subsystems: Mistral tokenizer methods vs. executorch export testing." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Mistral tokenizer methods vs. adding MobileLLM; unrelated model areas." + "reason": "Unrelated changes: tokenizer API additions vs. adding MobileLLM model support." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor robustness changes vs. adding MobileLLM; same broad library, but not the same fix." + "reason": "Different processor robustness fixes vs. new MobileLLM model implementation." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Loss-function kwargs support vs. adding MobileLLM; completely different subsystems." + "reason": "Cross-entropy loss kwargs fix is unrelated to adding MobileLLM." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Loss utility change vs. Mistral4 test/model work; unrelated changes." + "reason": "Different areas: loss utility change vs. Mistral4 test/model updates." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "ExecuTorch dynamic-shape test vs. Mistral4 fixes; not the same concrete bug." + "reason": "Executorch dynamic-shape export test is unrelated to Mistral4 tests/model changes." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Dynamic export test vs. new MobileLLM model support; unrelated." + "reason": "Executorch exportability test vs. MobileLLM model addition are unrelated." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "Model addition vs. class config resolution fix; no common code path." + "reason": "New MobileLLM model support is unrelated to config resolution ordering fix." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM model addition vs. Falcon output-interface refactor; unrelated." + "reason": "MobileLLM model addition and Falcon output-interface refactor are different code paths." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon refactor vs. model-creation skill PR; not mergeable as one concrete fix." - }, - { - "left": "pull_request:43995", - "right": "pull_request:44827", - "accept": false, - "reason": "Falcon standardized-output refactor vs. Mistral4 test/model changes; different problems." + "reason": "Falcon refactor is unrelated to the model-creation skill/docs PR." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test/model work vs. `layrnorm` renaming fix; unrelated." + "reason": "Mistral4 test/model work is unrelated to the layrnorm->layernorm rename fix." }, { "left": "pull_request:44827", "right": "pull_request:46267", "accept": false, - "reason": "Mistral4 fix vs. chat parsing overhaul; different subsystems and purposes." + "reason": "Mistral4 testing/model changes do not match the chat parsing overhaul." }, { "left": "pull_request:44827", "right": "pull_request:46048", "accept": false, - "reason": "Mistral4 work vs. TDT loss kernel integration; unrelated change sets." + "reason": "Different features: Mistral4 fixes vs. adding a TDT loss kernel integration." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 test/model updates vs. config resolution order fix; not the same underlying bug." + "reason": "Mistral4 test/model updates are unrelated to config resolution ordering." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Model changes vs. repo skill/documentation PR; no shared bug or code path." + "reason": "Mistral4 fixes are unrelated to the model-creation skill and related scaffolding." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "MobileLLM model addition vs. model-creation skill/documentation; not a duplicate change." + "reason": "Adding MobileLLM is unrelated to the model-creation skill PR." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactor and model-creation skill changes are unrelated." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both patches modify `fixed_cross_entropy` in the same file to pass `weight` and `label_smoothing` through to `nn.functional.cross_entropy`; they address the same concrete bug and could be merged into one PR." + "reason": "Same underlying fix in the same function: both expose/pass supported cross_entropy kwargs (weight, label_smoothing) through fixed_cross_entropy, and could be merged into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative: the only mergeable duplicate pair is the `fixed_cross_entropy` change in `pull_request:43251` and `pull_request:43254`, which share the same file, the same issue target, and the same concrete fix. The other soft links are correctly treated as unrelated or too broad to merge into one bugfix PR." + "feedback": "The summary is grounded in the packet and stays appropriately conservative. The only accepted soft-edge pair, 43251 vs 43254, is well supported by the diffs: both modify the same fixed_cross_entropy helper in the same file, target the same issue, and implement the same supported-kwargs pass-through with only minor signature/implementation differences. The other pairwise rejections are also conservative and match the packet content." }, "evaluator_used": true, "retried": false, @@ -19376,7 +19553,7 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "7115251010116f4125f1b73727b3bb7460cd03b737a48b324e766c2c6233c426", + "prepared_review_unit_hash": "fef3ecf1a4e7cd878c7bc490d66dcb45480b9211ede3d94810480ee1b3f09b73", "component_budget": { "node_count": 2, "item_count": 2, @@ -19398,18 +19575,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are about VibeVoice, but they target different scopes: #40546 is the core VibeVoice implementation, while #42781 adds a separate realtime/acoustic-tokenizer extension. They overlap on shared model scaffolding, but they are not the same concrete change and would not naturally merge into one PR.", - "confidence": 0.86, + "summary": "These are related VibeVoice pull requests, but they implement different scopes: one adds the core model, the other adds a realtime variant with acoustic-tokenizer/generation changes. Shared files reflect the same package, not the same fix.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "#40546 is the better canonical PR because it appears to be the primary, self-contained implementation of VibeVoice. It is broader and more central than the draft realtime follow-up.", + "canonical_pr_reason": "PR #40546 is the cleaner representative of the core VibeVoice addition; #42781 is a separate realtime extension and draft, so it should not be treated as the same change.", "best_issue_reason": null, - "best_pr_reason": "#40546 is the best PR for the cluster because it represents the base VibeVoice codepath and is the clearest standalone implementation; #42781 is a draft extension focused on realtime support and an additional acoustic tokenizer.", + "best_pr_reason": "PR #40546 is the best single representative for this cluster because it targets the base VibeVoice implementation and is more complete/stable than the draft realtime follow-up.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Reject: shared VibeVoice files indicate related work, but #42781 adds realtime-specific functionality and an acoustic tokenizer on top of the base implementation. They do not look like the same underlying change or a single mergeable fix." + "reason": "Reject: they share VibeVoice scaffolding, but #40546 is the base implementation while #42781 adds a distinct realtime path with tokenizer/generation changes; not the same concrete bug or mergeable change." } ] }, @@ -19439,12 +19616,12 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "e54a7928c93e94591c20dd4cf2abce583b5f437f140b9eaef08b231143178c86", + "prepared_review_unit_hash": "ae3796b8d76991d426520b4b6ebcc31fb8bed1be482bf5597984c42a70e6b2d7", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4675, + "serialized_chars": 4674, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -19452,7 +19629,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4675, + "serialized_chars": 4674, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -19461,18 +19638,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are unrelated feature additions in different model families: TimesFM covariate forecasting vs DINOv3 image classification support. They should not be clustered as duplicates.", + "summary": "These two PRs are unrelated feature additions in different model families, so they should not be clustered as duplicates.", "confidence": 0.99, "canonical_issue_reason": null, - "canonical_pr_reason": "Neither PR is a better canonical duplicate target because they address completely different code paths and model areas.", + "canonical_pr_reason": "No canonical duplicate PR: #40755 adds TimesFM covariate forecasting, while #41224 adds DINOv3 image classification support; they target different code paths and products.", "best_issue_reason": null, - "best_pr_reason": "No single PR stands out as a canonical representative since the changes are unrelated and not duplicates.", + "best_pr_reason": "#41224 is the better representative only in the sense that it is a focused, review-mature feature PR; #40755 is a separate, larger TimesFM change and not a duplicate.", "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different models, different functionality, and different implementation surfaces. One adds covariate forecasting for TimesFM; the other adds image classification support for DINOv3. They are not the same underlying change and could not plausibly be merged into one PR." + "reason": "Different models and implementations: TimesFM covariate forecasting vs DINOv3 image classification. They do not fix the same bug or add the same change and would not be merged into one PR." } ] }, @@ -19567,12 +19744,12 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "215b3bc25c6a5d9931479c67bbce2af511b5d2885257d945a7fb85468299c268", + "prepared_review_unit_hash": "51efc0afe41677b7f85d339899d636edbc9d54f80df4de758442bed478fc0fa5", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10435, + "serialized_chars": 10436, "estimated_input_tokens": 2609, "estimated_eval_tokens": 5474 }, @@ -19580,7 +19757,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10435, + "serialized_chars": 10436, "estimated_input_tokens": 2609, "estimated_eval_tokens": 5474 }, @@ -19589,24 +19766,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three items are separate model-addition PRs (DeepSeek OCR, Isaac, Kimi2-6) with overlapping integration scaffolding but different model-specific code and docs. They should not be deduplicated together.", - "confidence": 0.97, + "summary": "All three items are separate model-addition PRs (DeepSeek OCR, Isaac, Kimi2-6) that share framework wiring patterns but target different model families and code paths, so they are not duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45630 is the best representative only in the sense that it is the most substantial and mature of the three (highest review activity, broader integration surface), but it still describes a distinct model addition rather than a duplicate of the others.", + "canonical_pr_reason": "PR 45630 is the strongest representative if one must be chosen: it is the most mature and broad model-addition PR, with the widest set of model/auto/processor updates and the highest review activity.", "best_issue_reason": null, - "best_pr_reason": "PR #45630 has the strongest evidence of completeness and engagement, with many review comments and a broad set of model/auto/doc/test updates; however, it is still not a duplicate of the other two because the concrete model and code paths are different.", + "best_pr_reason": "PR 45630 is the best cluster representative on completeness and review depth, but it is still unrelated to the other PRs at the model level.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Both are new-model PRs and share generic Transformers integration files, but they add different models (DeepSeek OCR vs Isaac) with distinct configs, processors, and tests; not the same underlying change." + "reason": "Different model additions: DeepSeek OCR vs Isaac. They share generic auto-registration/docs scaffolding, but not the same underlying model or code path." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "These are separate model onboarding PRs for Isaac and Kimi2-6. Shared auto-registry/docs plumbing is expected, but the model-specific code paths, filenames, and intended behavior are different." + "reason": "Different model additions: Isaac vs Kimi2-6. Similar broad framework files are touched, but the concrete model implementations, processors, and tests are unrelated." } ] }, @@ -19636,7 +19813,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "3a162837f631195d48b082264ff1c8e72233f36d588908268f064a7799406aa5", + "prepared_review_unit_hash": "da90dfc1d067b105136eb92141170731f884ce0197f215b83316707553d0cc36", "component_budget": { "node_count": 2, "item_count": 2, @@ -19658,18 +19835,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate maintenance PRs: one upgrades checkout/cache/upload-artifact and related workflow actions for Node 24 compatibility, while the other upgrades docker/build-push-action and a different set of workflow actions. They overlap in general theme but not in the same concrete change.", - "confidence": 0.96, + "summary": "Both PRs are workflow maintenance upgrades, but they change different action families for different reasons: #42975 focuses on Node 24 compatibility for several GitHub Actions, while #42976 upgrades Docker build actions and a few other actions to newer releases. They overlap in files touched, but not in the underlying change.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "#42975 is the more specific and narrowly scoped PR, centered on a clear Node 24 compatibility upgrade across workflows; #42976 is broader and less directly tied to that concrete compatibility goal.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the better representative of the cluster only because it is the broader GitHub Actions compatibility update, but it is not a duplicate of PR 42976.", + "best_pr_reason": "#42975 is the better representative PR because its intent is explicit and cohesive, whereas #42976 is a looser general version bump across workflow tooling.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Both are workflow dependency bumps, but they update different action families and address different maintenance changes. They do not look like the same underlying bug or a single PR that could plausibly subsume both without becoming a broader release-level upgrade." + "reason": "Reject: both touch GitHub Actions workflows, but they update different actions for different purposes and are not the same underlying fix/change." } ] }, @@ -19708,12 +19885,12 @@ "pull_request:43451|pull_request:45490", "pull_request:45490|pull_request:46562" ], - "prepared_review_unit_hash": "8ed0d095701462e11a92a606554e7d71f97c83f496e83f8d6e592148ee2faee3", + "prepared_review_unit_hash": "c6cecf8af4738a73a63ace929149104524046d4f90acac52501eda46df2263d0", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 7, - "serialized_chars": 17510, + "serialized_chars": 17511, "estimated_input_tokens": 4378, "estimated_eval_tokens": 9012 }, @@ -19721,7 +19898,7 @@ "node_count": 5, "item_count": 5, "soft_pair_count": 7, - "serialized_chars": 17510, + "serialized_chars": 17511, "estimated_input_tokens": 4378, "estimated_eval_tokens": 9012 }, @@ -19730,54 +19907,54 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are five unrelated model-addition PRs. They share boilerplate Hugging Face patterns like docs toctree and auto-registry edits, but each introduces a different model family and code path (ViT NEPA, Molmo, Molmo2, CTSM, Step3.7 Flash). None looks like a true duplicate of another.", - "confidence": 0.97, + "summary": "These are all model-addition PRs for different architectures and model families; the shared files are mostly boilerplate registry/docs updates, not evidence of the same underlying change. No soft edges should be merged.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "PR #43451 (Add Molmo2) is the most substantial representative here: it has the broadest integration surface and the highest review activity, so it best stands in for the cluster if one PR must be chosen.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "#43451 is the best representative PR for this cluster because it is the most complete and actively reviewed model-addition change, whereas the others are separate model introductions.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different models and implementations: ViT NEPA vs Molmo. Shared registry/docs changes are boilerplate, not the same fix or feature." + "reason": "Different models: ViT NEPA vs Molmo. They share generic docs/auto-registry updates, but not the same code-path fix or mergeable change set." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "Different model families: ViT NEPA and Molmo2. No shared concrete code-path problem; overlap is only generic model-registration work." + "reason": "Different model families: ViT NEPA vs Molmo2. Shared auto/docs boilerplate is incidental; these are separate model implementations." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "Unrelated additions for ViT NEPA and CTSM. Similar support files do not indicate the same underlying change." + "reason": "ViT NEPA and CTSM are unrelated architectures (vision vs time-series). Similar framework wiring does not indicate the same underlying change." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are distinct model releases with separate configs, processors, and tests; not mergeable as one PR." + "reason": "Molmo and Molmo2 are distinct model additions with different files and APIs; they are not the same concrete bug/change." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo vs CTSM are unrelated model additions. Similar auto/docs edits are insufficient to treat them as duplicates." + "reason": "Molmo vs CTSM are unrelated model additions. Overlap is limited to shared registry/documentation plumbing." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM address different models and code paths; they only coincide on generic integration scaffolding." + "reason": "Molmo2 and CTSM target different model families and code paths; the shared auto-registration files are too generic to treat them as duplicates." }, { "left": "pull_request:45490", "right": "pull_request:46562", "accept": false, - "reason": "CTSM and Step3.7 Flash are different model integrations in different domains; shared auto-mapping/docs changes are not enough to merge them." + "reason": "CTSM is a time-series model, while Step3p7 is a multimodal/vision model addition. They are separate implementations, not one fix." } ] }, @@ -19818,12 +19995,12 @@ "pull_request:44815|pull_request:46380", "pull_request:46380|pull_request:46539" ], - "prepared_review_unit_hash": "c30f14bc64f152e56686936a6e77366d7797a4dcfbbda8cdff43cf617514f6c9", + "prepared_review_unit_hash": "832bab5aa24d3d6a9b7b1784349fbf4e963626cb2ade4b07b95508b83d7f23d5", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 14126, + "serialized_chars": 14125, "estimated_input_tokens": 3532, "estimated_eval_tokens": 7320 }, @@ -19831,7 +20008,7 @@ "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 14126, + "serialized_chars": 14125, "estimated_input_tokens": 3532, "estimated_eval_tokens": 7320 }, @@ -19840,66 +20017,66 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly unrelated PRs that all have superficial title/file similarity. The only substantive code fixes are in different areas, and the test-only PRs are clearly not duplicates of each other or of the code fixes.", - "confidence": 0.69, + "summary": "These PRs are not true duplicates: they span unrelated areas (Whisper decoding, CLIP/ViT tests, FP8 dequant/loading, generation watermark validation, PhiMoE integration tests). The strongest standalone production fix is the FP8/loading change, but the cluster as a whole should not be merged together.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46380 is the strongest canonical anchor: it contains an actual production bug fix in generation config validation (plus related generation-path updates and tests), unlike the mostly test-only or unrelated PRs.", + "canonical_pr_reason": "PR 44815 is the most concrete, self-contained production fix in the cluster: it adjusts core loading and FP8 kernel behavior for a specific dequant/loading bug, unlike the test-only or broad/mixed-change PRs.", "best_issue_reason": null, - "best_pr_reason": "PR 46380 is the best overall representative because it fixes a concrete runtime bug in core generation behavior and is more substantive than the other candidates.", + "best_pr_reason": "PR 44815 is the best overall PR candidate because it addresses a clear code-path problem with a focused implementation and is less ambiguous than the test-heavy or multi-area patches.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Whisper tokenizer logic vs. CLIP/ViT test edits; only superficial test-related similarity, not the same bug or change." + "reason": "Different targets and bugs: Whisper tokenizer decoding vs. unrelated CLIP/ViT test edits. Shared wording like \"fix\" and test changes is not enough." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper decode fix and FP8 dequant fix are unrelated code paths in different subsystems." + "reason": "Whisper decode/tokenization logic is unrelated to FP8 dequant/core loading. No shared code path or mergeable single fix." }, { "left": "pull_request:43291", "right": "pull_request:46380", "accept": false, - "reason": "Whisper tokenizer change versus generation watermarking/config validation fix; different underlying problems." + "reason": "Both mention Whisper only incidentally, but the actual fixes are different: tokenizer decode behavior vs. generation watermark/config handling and assorted test updates." }, { "left": "pull_request:43665", "right": "pull_request:46539", "accept": false, - "reason": "Both are test changes, but they target different models and different failure modes; not mergeable as one fix." + "reason": "Both are test-oriented, but they affect different models and failure modes. These do not look like the same underlying bug." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "Dummy test assertions in CLIP/ViT are unrelated to the FP8 dequant bug fix." + "reason": "CLIP/ViT test assertions are not the same underlying issue as FP8 loading/dequant logic." }, { "left": "pull_request:43665", "right": "pull_request:46380", "accept": false, - "reason": "CLIP/ViT test noise is unrelated to the generation config/runtime bug fix." + "reason": "The overlap is only generic test edits. The substantive code paths differ completely, so they are not duplicates." }, { "left": "pull_request:44815", "right": "pull_request:46539", "accept": false, - "reason": "FP8 kernel/dequant changes are unrelated to PhiMoE integration-test setup." + "reason": "FP8/core loading fixes and PhiMoE integration-test setup are unrelated code paths and could not plausibly be one PR." }, { "left": "pull_request:44815", "right": "pull_request:46380", "accept": false, - "reason": "Both touch core/integration code, but they fix different concrete issues in different execution paths." + "reason": "Different concrete bugs: FP8 dequant/loading vs. generation watermark validation plus other model/test changes. Too broad to treat as one fix." }, { "left": "pull_request:46380", "right": "pull_request:46539", "accept": false, - "reason": "Generation watermarking/config fix is unrelated to PhiMoE test environment handling." + "reason": "Generation config/watermark handling and PhiMoE integration-test behavior are unrelated; they only share a generic \"fix\" title." } ] }, @@ -19929,12 +20106,12 @@ "soft_pairs": [ "pull_request:43488|pull_request:43532" ], - "prepared_review_unit_hash": "e7bc2edc6d27d509fc87eda48127b0dc7233c478914836dffa4ac1d79bf294fc", + "prepared_review_unit_hash": "6c7765ff877e740da4b78916d5aabcdf565ccb3b6dcf73e665139beaa48d28be", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4562, + "serialized_chars": 4563, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -19942,7 +20119,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4562, + "serialized_chars": 4563, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -19951,18 +20128,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open '[don't merge]' pull requests with unrelated maintenance/test changes: one bumps version/deps and tweaks a Vit test, the other rewrites several CI/workflow files and test utilities. They are not the same underlying fix.", - "confidence": 0.97, + "summary": "The two pull requests are not duplicates: one is a small packaging/version/checking tweak in setup.py and ViT tests, while the other rewires workflow/testing utilities and also touches multiple files. They share only a test filename, not the same underlying change.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43532 is the broader, more representative change set in this cluster; it touches the CI/workflow path rather than a narrow setup/version tweak.", + "canonical_pr_reason": "No clear canonical PR for duplication: these PRs target different code paths and different kinds of repository-bot/demo changes.", "best_issue_reason": null, - "best_pr_reason": "Neither PR is a strong canonical merge target because both are intentionally non-mergeable test/demonstration PRs, but #43532 is the more substantial artifact overall.", + "best_pr_reason": "PR #43532 is the better representative item in this cluster because it is the broader, more substantive change; #43488 is a narrower setup/version tweak with a different purpose.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "The PRs change different code paths: #43488 is a setup/version/dependency + Vit test tweak, while #43532 modifies CI workflows, testing utilities, and related tests. Shared test filename alone is not enough to treat them as one fix." + "reason": "Reject: they do not fix the same concrete bug/change. #43488 edits setup/version/dependency metadata and a ViT test, while #43532 changes CI/workflow and notification/testing utilities with a different scope." } ] }, @@ -19992,7 +20169,7 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "96cdaf53f16f0742f8b98184fd024e8516cdb941876f4a849858d1c91f944475", + "prepared_review_unit_hash": "cf7bcfa0cd2825b8797cf2e1759ebe655ea23ced3f7244339ad6c58d29d52861", "component_budget": { "node_count": 2, "item_count": 2, @@ -20014,18 +20191,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Distinct PRs for two different segmentation pipelines; they share docs/registry scaffolding but do not implement the same feature or fix the same code path.", - "confidence": 0.98, + "summary": "Two open PRs add different new pipelines: one for promptable concept segmentation and one for promptable visual segmentation. They share documentation and registry scaffolding, but they target different tasks, models, and test suites, so they are not duplicates.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: PR 43612 adds promptable concept segmentation, while PR 43613 adds promptable visual segmentation; they overlap in plumbing but target different pipelines/models.", + "canonical_pr_reason": "PR #43613 is the better representative because it is the broader visual-segmentation pipeline change and has more model/test coverage, making it the more substantial member of this cluster.", "best_issue_reason": null, - "best_pr_reason": "Neither is a strong global duplicate representative because they address different features rather than one concrete fix.", + "best_pr_reason": "PR #43613 is the strongest standalone candidate: it implements a concrete pipeline addition across multiple segmentation-related models and tests, so it is the best exemplar of this class of changes.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different pipeline additions (concept vs visual segmentation) with shared boilerplate only; not the same underlying change." + "reason": "Different underlying features and code paths: concept segmentation vs visual segmentation. They overlap in docs/registry plumbing, but the pipeline implementations, model coverage, and tests are distinct, so they should not be merged as duplicates." } ] }, @@ -20055,40 +20232,40 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "addfb69eb306a956d15c2168558c761bfa69df716c828e903e3e683fb21df5fd", + "prepared_review_unit_hash": "1ced6893ee561db094e8665a772b10c51a3bfabcd4038dc8eaf7693239852798", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4293, - "estimated_input_tokens": 1074, - "estimated_eval_tokens": 2404 + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4293, - "estimated_input_tokens": 1074, - "estimated_eval_tokens": 2404 + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs touch `src/transformers/trainer.py`, but they solve different trainer extensibility problems: one changes gradient-accumulation loss scaling, the other extracts signature-column filtering for overrideability. They are not duplicate fixes and should stay separate.", - "confidence": 0.96, + "summary": "These are two separate Trainer extensibility PRs in the same file: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering. They do not fix the same underlying bug or implement the same change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43651 is the better representative because it makes a concrete trainer behavior change around loss scaling and has more discussion activity than the refactor-only column-filtering PR.", + "canonical_pr_reason": "PR 43651 is the more focused behavior change around training loss scaling, but it is not a duplicate of PR 43663.", "best_issue_reason": null, - "best_pr_reason": "PR 43651 is the stronger standalone change in the cluster: it addresses a specific runtime behavior in `training_step`, whereas PR 43663 is a separate refactor to expose signature-column logic.", + "best_pr_reason": "No single PR should be treated as the duplicate representative here; 43651 is only slightly more self-contained, while 43663 addresses a different Trainer override point.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Same file, but different underlying concerns: gradient accumulation loss normalization vs. model signature column filtering. They would not be merged as one PR." + "reason": "Different code paths and goals: 43651 adds _loss_is_scaled_for_ga for gradient-accumulation loss handling, while 43663 adds _get_signature_columns for column filtering. Shared file alone is not enough to merge them as one change." } ] }, @@ -20118,12 +20295,12 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "e8f46244c29ffeacffb05661ce7d55e4067507fd71a2537822164d2c3294a98f", + "prepared_review_unit_hash": "719d2fdd11e1eb45ae7058e256b93e72573102321ce7adedc0ee60753c66f042", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3664, + "serialized_chars": 3663, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -20131,7 +20308,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3664, + "serialized_chars": 3663, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -20140,18 +20317,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs touch `src/transformers/cli/serve.py` and mention a TypeAdapter-related NameError, but they are not the same change: one broadens the file to postponed annotations, while the other narrowly fixes the pydantic-absent import/annotation path. I would not merge them as duplicates.", - "confidence": 0.86, + "summary": "Both PRs touch the same CLI file and reference a TypeAdapter NameError, but they appear to fix different failure modes: one is a broader annotation/runtime typing cleanup, the other is specifically about pydantic being absent. They are not safe duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43842 is the more targeted and concrete fix: it directly addresses the runtime `TypeAdapter` annotation issue when pydantic is unavailable, with a smaller and clearer change surface.", + "canonical_pr_reason": "No single canonical PR stands out as the same fix for both items; 43842 is the tighter, more explicit match to the TypeAdapter/pydantic NameError, while 43656 looks broader and likely addresses a different annotation-evaluation problem.", "best_issue_reason": null, - "best_pr_reason": "PR #43842 is the best representative because it cleanly fixes the specific import/annotation failure mode with minimal collateral changes.", + "best_pr_reason": "43842 is the better representative of the cluster because its title and diff directly target the TypeAdapter NameError when pydantic is not installed, with a narrowly scoped change in serve.py.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": false, - "reason": "Same file and related symptom, but different fix strategy and likely different bug scope; 43656 applies a broader postponed-annotations refactor, while 43842 specifically guards `TypeAdapter` for the no-pydantic case." + "reason": "Reject: they share a file and broad CLI context, but the fixes are not clearly the same bug. 43656 appears to address runtime annotation evaluation via __future__ annotations, while 43842 explicitly fixes missing TypeAdapter import handling when pydantic is absent; they also target different issues." } ] }, @@ -20283,8 +20460,8 @@ "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:44086|pull_request:44722", "pull_request:44018|pull_request:44019", "pull_request:44019|pull_request:44071", "pull_request:44068|pull_request:44116", @@ -20308,8 +20485,8 @@ "pull_request:44071|pull_request:44086", "pull_request:44017|pull_request:44071", "pull_request:44073|pull_request:44101", - "pull_request:44059|pull_request:44071", "pull_request:44073|pull_request:44161", + "pull_request:44059|pull_request:44071", "pull_request:44056|pull_request:44101", "pull_request:44018|pull_request:44059", "pull_request:44059|pull_request:44068", @@ -20325,12 +20502,12 @@ "pull_request:44018|pull_request:44161", "pull_request:43997|pull_request:44071", "pull_request:44010|pull_request:44071", - "pull_request:44018|pull_request:44129", "pull_request:44010|pull_request:44073", + "pull_request: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:44071|pull_request:44129", "pull_request:43996|pull_request:44161", "pull_request:43996|pull_request:44017", "pull_request:44010|pull_request:44044", @@ -20343,448 +20520,364 @@ "pull_request:44044|pull_request:44129", "pull_request:44044|pull_request:44154" ], - "prepared_review_unit_hash": "716a6449ab196e4df2985ca5175b13361c5726ed22386ff2330f94042d58622b", + "prepared_review_unit_hash": "6b8b64c85a0fc08dc9b10057510e0b4cfdfbe88d0911d5075a74264144ecb2c2", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84896, - "estimated_input_tokens": 21224, - "estimated_eval_tokens": 42704 + "serialized_chars": 84903, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 }, "budget": { "node_count": 27, "item_count": 27, "soft_pair_count": 72, - "serialized_chars": 75092, - "estimated_input_tokens": 18773, - "estimated_eval_tokens": 37802 + "serialized_chars": 75098, + "estimated_input_tokens": 18775, + "estimated_eval_tokens": 37806 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of independent model-specific refactors toward standardized output tracing. The only clear duplicate groups are the GPT-Neo pair and the GPT-J/CodeGen trio; the rest are shared-pattern changes across different model families, not the same bug/change.", - "confidence": 0.84, + "summary": "This cluster is mostly a broad set of model-specific PRs that all refactor output tracing to the standardized decorator-based interface. The only strong duplicate candidates are same-model pairs within GPT-J and GPT-Neo; cross-model pairs share the theme and tracking issue but not the same concrete code path.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "44066 is the best anchor PR: it explicitly targets #43979, has the clearest GPT-J/CodeGen output-tracing refactor, and serves as the most representative self-contained change in the duplicate set.", + "canonical_pr_reason": "44066 is the best canonical PR: it is explicitly tied to #43979 and contains a focused, concrete GPT-J/CodeGen output-tracing refactor that sits at the center of the genuine duplicate chain.", "best_issue_reason": null, - "best_pr_reason": "44066 is the strongest representative PR for this cluster because it is explicit about the tracking issue and covers the core GPT-J/CodeGen tracing migration more cleanly than the later near-duplicates.", + "best_pr_reason": "44722 is the strongest single PR candidate in the duplicate set: it targets the same GPT-J/CodeGen output-tracing change with the clearest same-file overlap to the other accepted GPT-J PRs.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families and code paths; only the output-tracing pattern is similar." + "reason": "Different model families/files; the shared issue target only indicates the same refactor theme, not the same underlying code path." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "FNet/CVT refactor vs DeBERTa refactor; same standardization theme, not the same change." + "reason": "Different models and implementations; both are output-tracing refactors, but not the same bug or change." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Same GPT-J output-tracing migration in overlapping files; this looks like a duplicate/iteration of the same concrete change." + "reason": "Both are GPT-J output-tracing refactors in the same file, so they look like the same underlying change/iteration." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "ResNet vs EfficientNet; different models and implementation details." + "reason": "ResNet and EfficientNet are different code paths; same issue target is not enough to make them duplicates." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "EfficientNet refactor vs GPT-J/CodeGen refactor; not the same code path." + "reason": "EfficientNet and GPT-J are unrelated model implementations, so these are not the same change." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen vs EfficientNet; shared tracing motif only." + "reason": "GPT-J/CodeGen and EfficientNet are different model families; no shared code path." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 vs DeBERTa; different model code paths." + "reason": "MobileNetV2 and DeBERTa v2 are unrelated implementations, despite the same refactor pattern." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen tracing vs MGP-STR tracing; same infrastructure idea, different models." + "reason": "GPT-J/CodeGen and MGP-STR do not share the same concrete fix, only the broad output-tracing theme." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen vs MPT; not the same underlying fix." + "reason": "GPT-J/CodeGen and MPT are different model code paths; not a duplicate pair." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both are GPT-Neo output-tracing refactors in the same module; same concrete change with minor iteration." + "reason": "Both touch GPT-Neo output tracing in the same file and look like duplicate iterations of the same refactor." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen refactor vs GPT-Neo refactor; different model families." - }, - { - "left": "pull_request:44018", - "right": "pull_request:44066", - "accept": false, - "reason": "GPT-Neo vs GPT-J/CodeGen; related style, different code path." - }, - { - "left": "pull_request:44019", - "right": "pull_request:44722", - "accept": false, - "reason": "ResNet refactor vs GPT-J/CodeGen refactor; different model implementations." - }, - { - "left": "pull_request:44086", - "right": "pull_request:44722", - "accept": false, - "reason": "MGP-STR vs GPT-J/CodeGen; not the same concrete bug/change." - }, - { - "left": "pull_request:44019", - "right": "pull_request:44085", - "accept": false, - "reason": "ResNet vs GPT-J-related tracing change; different code paths." + "reason": "GPT-J and GPT-Neo are separate model implementations, so these are not the same concrete change." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo vs ResNet; same output-tracing theme only." + "reason": "GPT-Neo and ResNet are unrelated code paths." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet vs MPT; different model families and edits." + "reason": "ResNet and MPT are different models; same refactor theme is too broad." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo vs Flaubert/XLM-style attention refactor; not the same change." + "reason": "GPT-Neo and Flaubert are different model implementations." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo refactor vs GPT-J/CodeGen refactor; different underlying code paths." + "reason": "GPT-Neo and GPT-J are different model code paths, so this is not the same underlying change." }, { "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo vs Flaubert; unrelated model-specific refactors." + "reason": "GPT-Neo and Flaubert are unrelated implementations." }, { "left": "pull_request:44071", - "right": "pull_request:44722", + "right": "pull_request:44086", "accept": false, - "reason": "MPT vs GPT-J/CodeGen; same pattern, different implementation." + "reason": "MPT and MGP-STR are different models and fix different code paths." }, { - "left": "pull_request:44018", + "left": "pull_request:44071", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo vs GPT-J; different model families." + "reason": "MPT and GPT-J are not the same underlying bug/change." }, { - "left": "pull_request:44018", + "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo vs MGP-STR; not the same concrete change." + "reason": "GPT2 and MGP-STR are different model implementations." }, { - "left": "pull_request:44068", + "left": "pull_request:44085", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo vs MGP-STR; only broad tracing similarity." + "reason": "GPT-J and MGP-STR are unrelated code paths." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT vs Flaubert/XLM; different model-specific refactors." + "reason": "ViLT and Flaubert are different model families; no same-code-path match." }, { "left": "pull_request:44059", - "right": "pull_request:44086", + "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 vs MGP-STR; unrelated model code paths." + "reason": "GPT2 and GPT-J are different model code paths, so this is not a duplicate change." }, { "left": "pull_request:44018", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo vs MPT; same theme, different model refactors." + "reason": "GPT-Neo and MPT are different models; same refactor theme is too broad." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "Segformer vs LongT5; different components and fixes." - }, - { - "left": "pull_request:44017", - "right": "pull_request:44073", - "accept": false, - "reason": "Segformer vs VisualBert; same output-tracing pattern, not the same change." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44068", - "accept": false, - "reason": "GPT-2 vs GPT-Neo; different model families and code paths." + "reason": "Segformer and LongT5 are unrelated implementations." }, { - "left": "pull_request:44017", - "right": "pull_request:44101", + "left": "pull_request:44073", + "right": "pull_request:44722", "accept": false, - "reason": "Segformer vs Flaubert/XLM; unrelated model refactors." + "reason": "VisualBert and GPT-J are different code paths." }, { "left": "pull_request:44073", "right": "pull_request:44161", "accept": false, - "reason": "VisualBert vs LongT5; different model implementations." - }, - { - "left": "pull_request:44056", - "right": "pull_request:44101", - "accept": false, - "reason": "MPNet vs Flaubert/XLM; not the same concrete bug/change." + "reason": "VisualBert and LongT5 are different models and fixes." }, { "left": "pull_request:44059", "right": "pull_request:44071", "accept": false, - "reason": "GPT-2 vs MPT; different model families." + "reason": "GPT2 and MPT are distinct model implementations." }, { - "left": "pull_request:44073", - "right": "pull_request:44101", + "left": "pull_request:44074", + "right": "pull_request:44073", "accept": false, - "reason": "VisualBert vs Flaubert/XLM; same style of refactor only." + "reason": "TextNet and VisualBert are different code paths." }, { "left": "pull_request:44056", - "right": "pull_request:44161", + "right": "pull_request:44101", "accept": false, - "reason": "MPNet vs LongT5; distinct model-specific changes." + "reason": "MPNet and Flaubert are unrelated model implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44018", + "right": "pull_request:44017", "accept": false, - "reason": "RegNet vs GPT-Neo; different model families and code paths." + "reason": "RegNet and Segformer are different models; the shared refactor theme is not enough." }, { "left": "pull_request:43997", - "right": "pull_request:44161", - "accept": false, - "reason": "RegNet vs LongT5; unrelated refactors." - }, - { - "left": "pull_request:44010", - "right": "pull_request:44071", - "accept": false, - "reason": "SqueezeBert vs MPT; not the same underlying change." - }, - { - "left": "pull_request:44010", - "right": "pull_request:44018", + "right": "pull_request:44056", "accept": false, - "reason": "SqueezeBert vs GPT-Neo; same standardization theme only." + "reason": "RegNet and MPNet are separate model code paths." }, { - "left": "pull_request:44010", - "right": "pull_request:44044", + "left": "pull_request:44071", + "right": "pull_request:44076", "accept": false, - "reason": "SqueezeBert vs DeBERTa; different model code paths." + "reason": "MPT and ImageGPT are different models and changes." }, { - "left": "pull_request:44044", - "right": "pull_request:44129", + "left": "pull_request:44074", + "right": "pull_request:44101", "accept": false, - "reason": "DeBERTa vs SpeechT5; different concrete refactors." + "reason": "TextNet and Flaubert are unrelated implementations." }, { - "left": "pull_request:44044", - "right": "pull_request:44154", + "left": "pull_request:44056", + "right": "pull_request:44161", "accept": false, - "reason": "DeBERTa vs Vits; separate model-specific changes." + "reason": "MPNet and LongT5 do not share the same concrete fix." }, { "left": "pull_request:43997", - "right": "pull_request:44017", + "right": "pull_request:44018", "accept": false, - "reason": "RegNet vs Segformer; same pattern, not the same fix." + "reason": "RegNet and GPT-Neo are different model code paths." }, { "left": "pull_request:43997", - "right": "pull_request:44056", - "accept": false, - "reason": "RegNet vs MPNet; different model implementations." - }, - { - "left": "pull_request:44018", - "right": "pull_request:44059", - "accept": false, - "reason": "GPT-Neo vs GPT-2; different model families." - }, - { - "left": "pull_request:44098", - "right": "pull_request:44073", + "right": "pull_request:44161", "accept": false, - "reason": "ViLT vs VisualBert; both multimodal, but not the same code-path change." + "reason": "RegNet and LongT5 are unrelated implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44074", + "right": "pull_request:44071", "accept": false, - "reason": "RegNet vs TextNet; unrelated model-specific refactors." + "reason": "RegNet and MPT are different models." }, { - "left": "pull_request:43997", + "left": "pull_request:44010", "right": "pull_request:44071", "accept": false, - "reason": "RegNet vs MPT; different code paths." + "reason": "SqueezeBert and MPT are unrelated model code paths." }, { - "left": "pull_request:44056", + "left": "pull_request:44010", "right": "pull_request:44073", "accept": false, - "reason": "MPNet vs VisualBert; not the same underlying change." - }, - { - "left": "pull_request:44071", - "right": "pull_request:44076", - "accept": false, - "reason": "MPT vs ImageGPT; separate model-family refactors." - }, - { - "left": "pull_request:44074", - "right": "pull_request:44101", - "accept": false, - "reason": "TextNet vs Flaubert/XLM; unrelated model-specific edits." - }, - { - "left": "pull_request:44056", - "right": "pull_request:44161", - "accept": false, - "reason": "MPNet vs LongT5; different models and code paths." + "reason": "SqueezeBert and VisualBert are different implementations." }, { - "left": "pull_request:43997", - "right": "pull_request:44018", + "left": "pull_request:44018", + "right": "pull_request:44129", "accept": false, - "reason": "RegNet vs GPT-Neo; same tracing pattern only." + "reason": "GPT-Neo and SpeechT5 are unrelated model code paths." }, { - "left": "pull_request:43997", - "right": "pull_request:44071", + "left": "pull_request:44018", + "right": "pull_request:44059", "accept": false, - "reason": "RegNet vs MPT; not the same fix." + "reason": "GPT-Neo and GPT2 are different model implementations." }, { "left": "pull_request:44010", - "right": "pull_request:44071", + "right": "pull_request:44018", "accept": false, - "reason": "SqueezeBert vs MPT; different model implementations." + "reason": "SqueezeBert and GPT-Neo are not the same underlying change." }, { - "left": "pull_request:44018", + "left": "pull_request:44071", "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo vs SpeechT5; different concrete refactors." + "reason": "MPT and SpeechT5 are different model code paths." }, { "left": "pull_request:43996", "right": "pull_request:44161", "accept": false, - "reason": "FNet/CVT vs LongT5; only the output-tracing theme is shared." + "reason": "CVT/FNet and LongT5 are unrelated implementations." }, { "left": "pull_request:43996", "right": "pull_request:44017", "accept": false, - "reason": "FNet/CVT vs Segformer; different model families." + "reason": "CVT/FNet and Segformer are different model families." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBert vs DeBERTa; distinct code paths." + "reason": "SqueezeBert and DeBERTa v2 are different code paths." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa vs TextNet; unrelated model-specific changes." + "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 migration with overlapping files; these are clearly duplicate iterations of the same change." + "reason": "Same GPT-J/CodeGen output-tracing refactor in the same files; 44722 looks like a direct follow-up or revision of 44066." }, { "left": "pull_request:44085", "right": "pull_request:44722", "accept": true, - "reason": "Both touch the same GPT-J tracing refactor; the file-level change matches the same concrete migration." + "reason": "Both are GPT-J output-tracing refactors and share the same underlying file-level change." }, { "left": "pull_request:44085", "right": "pull_request:44129", "accept": false, - "reason": "GPT-J tracing refactor vs SpeechT5 tracing refactor; different model code paths." + "reason": "GPT-J and SpeechT5 are different models, so this is not the same fix." }, { "left": "pull_request:43996", "right": "pull_request:44007", "accept": false, - "reason": "FNet/CVT vs ResNet/RT-DETR; different model-specific refactors." + "reason": "CVT/FNet and ResNet are different model implementations." }, { "left": "pull_request:43996", "right": "pull_request:44129", "accept": false, - "reason": "FNet/CVT vs SpeechT5; same standardization pattern only." + "reason": "CVT/FNet and SpeechT5 are unrelated code paths." }, { "left": "pull_request:44044", "right": "pull_request:44129", "accept": false, - "reason": "DeBERTa vs SpeechT5; unrelated concrete changes." + "reason": "DeBERTa v2 and SpeechT5 are different models and changes." }, { "left": "pull_request:44044", "right": "pull_request:44154", "accept": false, - "reason": "DeBERTa vs Vits; different model families and edits." + "reason": "DeBERTa v2 and VITS are unrelated implementations." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative about cross-family similarities. The only accept edges are within the same model family/code path and look mergeable as the same refactor/iteration, while the rejects avoid overstating shared tracing patterns as the same bugfix." + "feedback": "The summary stays grounded in the packet and the soft-edge calls are generally conservative. The accepted duplicate pairs all have a plausible same-file/same-model basis, especially the GPT-J and GPT-Neo iterations. The cross-model pairs are rejected appropriately. One naming oddity remains in 44085, but the verdict does not overstate that evidence." }, "evaluator_used": true, "retried": false, @@ -20822,20 +20915,20 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "057642c71e380cc6d0c0db959ae2f5fb61650484686782886718c90376664ddd", + "prepared_review_unit_hash": "daf4fec55da030acfa937262405a8b9b72405b90817768ac33748de47fe86dfd", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84896, - "estimated_input_tokens": 21224, - "estimated_eval_tokens": 42704 + "serialized_chars": 84903, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 6, - "serialized_chars": 17125, + "serialized_chars": 17126, "estimated_input_tokens": 4282, "estimated_eval_tokens": 8820 }, @@ -20844,48 +20937,48 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "All items are model-specific refactors for standardized output tracing/capture_outputs, but they target different architectures and code paths. They are thematically similar, not duplicates, and should not be merged together as one PR.", - "confidence": 0.95, + "summary": "These PRs are all model-specific refactors to standardized output tracing/capture, but they touch different architectures and code paths, so they are thematically similar rather than true duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44129 is the best representative anchor: it is tightly scoped, has an explicit issue target, and clearly implements the output-tracing refactor for a single model family.", + "canonical_pr_reason": "No exact duplicate PR stands out in this set; if a representative is needed, PR #44129 is the clearest because it has an explicit issue target and a well-scoped standardized output capture refactor.", "best_issue_reason": null, - "best_pr_reason": "PR 44129 is the strongest standalone candidate in the set because it is concrete, issue-linked, and narrowly focused on one code path rather than being a broad pattern-only refactor.", + "best_pr_reason": "PR #44129 is the best representative of the cluster because it is the most explicitly tied to a tracking issue and clearly centers on the same standardized output-capture migration pattern.", "soft_edge_verdicts": [ { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "Different models (FocalNet vs ImageGPT) and different forward paths; same refactor theme is not enough to treat them as one duplicate change." + "reason": "Both are output-tracing refactors, but FocalNet and ImageGPT are different model implementations with separate forward paths and capture hooks; they are not the same underlying change." }, { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "VisualBert and TextNet are unrelated model implementations with separate output-handling code; these are parallel refactors, not the same change." + "reason": "VisualBert and TextNet are unrelated model families; the PRs share a refactor theme but not a common concrete code-path bug or a mergeable single fix." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "MPNet vs VisualBert affects different architectures and internals; they share a standardized-output theme but not a single concrete bug or fix." + "reason": "MPNet and VisualBert each introduce model-specific output-capture plumbing in different files and layers; this is the same pattern, not the same change." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "MPNet and TextNet are distinct code paths with different tracing details, so they cannot plausibly be merged as one PR." + "reason": "These touch different model backends (MPNet vs TextNet) and different forward/output mechanics, so they should not be merged as duplicates." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are separate model families; this is the same migration pattern applied in different places, not duplicate work." + "reason": "RegNet and SqueezeBert are unrelated subsystems; both refactor output handling, but they are separate model-local changes with no shared bug fix." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and SpeechT5 touch different subsystems and forward contracts; they are not the same underlying bug or change." + "reason": "SqueezeBert attention tracing and SpeechT5 output capture are model-specific refactors in different architectures; they do not fix the same concrete problem." } ] }, @@ -20965,274 +21058,274 @@ "pull_request:44028|pull_request:44030", "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "ced6bd90eda615269873454fdad3f55f888d00e3e676fc8150700d8618b12bd2", + "prepared_review_unit_hash": "347d6bb2975d2e8b06e16f0f25354f6ac7c2544206b0f89190de89df8f51add7", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40177, - "estimated_input_tokens": 10045, - "estimated_eval_tokens": 20346 + "serialized_chars": 40172, + "estimated_input_tokens": 10043, + "estimated_eval_tokens": 20342 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40177, - "estimated_input_tokens": 10045, - "estimated_eval_tokens": 20346 + "serialized_chars": 40172, + "estimated_input_tokens": 10043, + "estimated_eval_tokens": 20342 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are thematically similar refactors around output tracing, but they target different model implementations and code paths. I do not see duplicate pairs here; they look like parallel, model-specific cleanup work rather than one underlying bug/change.", - "confidence": 0.91, + "summary": "This cluster is a batch of separate model-specific output-tracing refactors, not duplicate bug fixes. The PRs share a common implementation pattern (`can_return_tuple` / output-capture cleanup) but target different model families and code paths.", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44003 is the broadest and most infrastructure-heavy of the set, spanning both `mamba` and `falcon_mamba` and adding the shared output-capturing plumbing, so it is the best representative if one must be picked.", + "canonical_pr_reason": "PR 44003 is the most representative of the shared refactor theme: it spans two related model files (`falcon_mamba` and `mamba`) and adds the new output-capturing plumbing, making it a good cluster anchor.", "best_issue_reason": null, - "best_pr_reason": "PR 44003 is the most representative of the cluster\u2019s theme because it centralizes the output-tracing mechanism across closely related files, whereas the others are narrower per-model refactors.", + "best_pr_reason": "PR 44003 is the strongest global representative because it is broader than the single-file refactors and most clearly shows the common output-tracing change pattern.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Both are output-tracing refactors, but they affect different models (`timm_backbone` vs `upernet`) and different forward paths." + "reason": "Different model families (`timm_backbone` vs `upernet`); same refactor motif only, not the same concrete code path." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Same general refactor theme, but separate model codepaths (`timm_backbone` vs `vision_text_dual_encoder`)." + "reason": "`timm_backbone` and `vision_text_dual_encoder` are unrelated model paths; both refactor output handling but do different changes." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Both are model-specific output tracing changes, not the same bug or interchangeable change." + "reason": "Separate models (`timm_backbone` vs `univnet`); no shared underlying bug or mergeable fix." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Different target models and different refactor scope; one is a single-backbone cleanup, the other updates Mamba plumbing across two files." + "reason": "`timm_backbone` and `mamba/falcon_mamba` are distinct code paths; similarity is only the generic output-tracing cleanup." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "Shared wording only; the PRs touch unrelated model internals (`timm_backbone` vs `rwkv`)." + "reason": "Different subsystems (`timm_backbone` vs `rwkv`); same style of refactor, not the same change." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "Both refactor output handling, but for unrelated architectures and separate forward implementations." + "reason": "`mobilenet_v1` and `univnet` are unrelated model implementations; not a duplicate fix." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different models and different tracing/capture changes; not mergeable as one concrete fix." + "reason": "Different model families (`mobilenet_v1` vs `upernet`); same output-tracing pattern only." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "They are separate model families (`mobilenet_v1` vs `depth_anything`) with no shared codepath." + "reason": "`mobilenet_v1` and `depth_anything` touch different forward paths and model architectures." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "These are distinct model wrappers, not the same underlying issue." + "reason": "Separate encoder models with different call paths; similar refactor theme is too broad to treat as duplicate." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Same refactor motif, but different encoder-decoder wrapper and different concrete implementation." + "reason": "`mobilenet_v1` vs `speech_encoder_decoder` are different model stacks; not the same bug/change." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "Both change output tracing, but they touch different models and separate forward signatures." + "reason": "`vision_text_dual_encoder` and `univnet` are unrelated; they only share the refactor style." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Different model families and no evidence of a shared bug or shared codepath." + "reason": "Different models and code paths (`vision_text_dual_encoder` vs `upernet`); not a single underlying fix." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "`vision_text_dual_encoder` and `mamba/falcon_mamba` are unrelated implementations; too broad to be duplicates." + "reason": "`vision_text_dual_encoder` and `mamba/falcon_mamba` are separate architectures; same theme, different change." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "No shared implementation path; only the generic refactor naming overlaps." + "reason": "`vision_text_dual_encoder` vs `depth_anything`; both adjust output handling but are not the same concrete PR." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "Different models and refactor surfaces; not the same concrete change." + "reason": "`univnet` and `upernet` are unrelated model code paths; no duplicate defect or shared fix." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "These target unrelated model codepaths and cannot plausibly be one merged PR." + "reason": "`univnet` vs `mamba/falcon_mamba`; only the generic output-tracing pattern overlaps." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "Separate model families; similarity is only at the refactor pattern level." + "reason": "Different model families (`univnet` vs `depth_anything`); not mergeable as one underlying change." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "Different model implementations (`upernet` vs `rwkv`) and different output handling details." + "reason": "`upernet` and `rwkv` are different architectures and forward paths; same refactor motif only." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "No shared codepath; one is segmentation model refactoring, the other is Mamba output capture plumbing." + "reason": "`upernet` vs `mamba/falcon_mamba`; distinct code paths and different output plumbing." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "Even though both are output-tracing refactors, they affect unrelated models and different forward logic." + "reason": "`mamba/falcon_mamba` and `superpoint` are unrelated model implementations; similar tracing cleanup is not a duplicate." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "`mamba/falcon_mamba` and `codegen` are distinct codepaths; this is not a duplicate change." + "reason": "`mamba` and `codegen` affect different forward implementations and return conventions; not the same fix." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "Different models and different refactor scope." + "reason": "`timm_backbone` and `codegen` are different model families; the overlap is only high-level refactoring." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "No shared bug or exact codepath; just same refactor theme." + "reason": "`univnet` and `codegen` touch unrelated model code paths; not a duplicate change." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Separate model implementations, not the same underlying change." + "reason": "`upernet` vs `codegen`; same generic output-refactor theme, but different concrete models." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "Both are tracing-related, but they modify different models and are not mergeable into one concrete fix." + "reason": "`vision_text_dual_encoder` and `codegen` are separate implementations; no shared underlying bug." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Both are encoder-decoder related, but they target different model classes and different files." + "reason": "`depth_anything` and `vision_encoder_decoder` are different model families; not the same change." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Different model wrappers and distinct codepaths; too broad to call duplicates." + "reason": "`depth_anything` vs `speech_encoder_decoder`; distinct code paths, only similar refactor style." }, { "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 unrelated models; not a duplicate fix." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "These are separate model refactors with no shared bug or codepath." + "reason": "`depth_anything` and `rwkv` are separate architectures; same output-tracing theme only." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Different models and different forward behavior." + "reason": "`depth_anything` vs `dpr`; different model code paths and return types, not the same bug/change." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "Distinct model families; only the refactor pattern matches." + "reason": "`vision_encoder_decoder` and `rwkv` are unrelated implementations; overlap is only the refactor pattern." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "Both are encoder-decoder wrappers, but they are separate implementations with different code paths." + "reason": "`vision_encoder_decoder` vs `speech_encoder_decoder`; similar wrapper pattern, but different concrete code paths." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "Unrelated models and no concrete overlap beyond the tracing refactor theme." + "reason": "`vision_encoder_decoder` and `superpoint` are different model families; not a duplicate." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "Different model internals and no shared bug fix." + "reason": "`vision_encoder_decoder` and `dpr` affect different model wrappers and outputs; not the same fix." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "Separate model implementations; not the same underlying change." + "reason": "`speech_encoder_decoder` and `superpoint` are unrelated model paths; same style of cleanup only." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "Different model wrappers and forward paths." + "reason": "`speech_encoder_decoder` vs `rwkv`; distinct architectures and code paths." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "Different models and different output semantics." + "reason": "`speech_encoder_decoder` and `dpr` are separate model wrappers; not a duplicate change." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "`superpoint` and `rwkv` have unrelated codepaths." + "reason": "`superpoint` and `rwkv` are unrelated models; only the tracing/refactor motif matches." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "Different models and separate output handling changes." + "reason": "`superpoint` vs `dpr`; different code paths and return handling, not the same underlying change." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "Both are tracing-related, but they affect unrelated models and cannot be merged as one fix." + "reason": "`rwkv` and `dpr` are distinct models; same broad output-tracing cleanup, but not a shared fix." } ] }, @@ -21262,12 +21355,12 @@ "soft_pairs": [ "pull_request:44662|pull_request:45401" ], - "prepared_review_unit_hash": "0370260e56a6e78058beb04d61390b2c3e0bf22ecbc7df41d1e19e5b037f0c94", + "prepared_review_unit_hash": "64d4a613513c07698063e39b7479adf643d6664cf2194682d5c4d311e91287e0", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6551, + "serialized_chars": 6550, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -21275,7 +21368,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6551, + "serialized_chars": 6550, "estimated_input_tokens": 1638, "estimated_eval_tokens": 3532 }, @@ -21284,18 +21377,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs, but they are unrelated feature additions: one adds PenguinVL, the other adds Voxtral TTS. They share only generic model-registration and docs plumbing, so they should not be deduplicated.", - "confidence": 0.97, + "summary": "These are two unrelated model-addition PRs that only overlap on shared Transformers scaffolding (auto registration, docs toc, package init). They do not appear to fix the same code-path problem or be mergeable into one PR.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45401 is the better representative only because it has an explicit issue target and a tightly scoped feature addition, whereas 44662 is a separate model implementation with no issue linkage.", + "canonical_pr_reason": "If a PR must represent the cluster, #45401 is the cleaner candidate because it is narrower, has an explicit issue target, and is a self-contained Voxtral TTS implementation.", "best_issue_reason": null, - "best_pr_reason": "PR 45401 is the stronger standalone PR candidate: it has a clear issue reference, focused Voxtral TTS support, and a cohesive change set.", + "best_pr_reason": "#45401 is the better standalone PR in this pair: it has explicit tracking, a focused scope, and clearer feature intent. #44662 is a separate PenguinVL implementation and not the same change.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different models and different code paths; shared auto-registration/docs files are boilerplate, not evidence of the same underlying change." + "reason": "Reject: PenguinVL and Voxtral TTS are different model implementations with different codepaths, files, and purposes; the overlap is only generic framework registration/docs, not the same underlying change." } ] }, @@ -21325,12 +21418,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "a93ead2000335199ec99276a3cf7e796dd7fe2180873fe3df7b0d6fa953371fe", + "prepared_review_unit_hash": "0dc1e2e980dff3d8317e915f1eddb60e316c5daeeadc5da8b7a0eb4f1b8b6696", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4408, + "serialized_chars": 4407, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -21338,7 +21431,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4408, + "serialized_chars": 4407, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -21347,18 +21440,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are about fusion, but they target different layers: #44979 introduces a generic module-fusion API, while #45363 wires n-to-1 kernel fusion through KernelConfig and model-loading paths. They are related feature work, not duplicates.", - "confidence": 0.88, + "summary": "Both PRs are about model/module fusion, but they target different layers and concrete changes: PR #44979 introduces a general Module Fusion API and tests, while PR #45363 wires n-to-1 kernel fusion through KernelConfig across existing integration points. They are related feature work, not the same underlying change.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45363 is the broader, more user-facing integration PR, spanning config, loading, integrations, and an example, while #44979 is the lower-level API foundation.", + "canonical_pr_reason": "PR #45363 is the broader, more end-to-end feature implementation: it modifies core model loading/integration code, adds a user-facing example, and connects fusion through KernelConfig. That makes it the better representative for the cluster.", "best_issue_reason": null, - "best_pr_reason": "PR #45363 is the best representative of the cluster because it covers the main end-to-end KernelConfig fusion workflow and is the more complete feature change.", + "best_pr_reason": "PR #45363 is the strongest standalone representative because it is the more complete integration of the fusion feature and touches the actual runtime path, whereas PR #44979 is mainly the API foundation.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Related fusion features, but not the same concrete code-path problem. #44979 adds a generic module-fusion API; #45363 implements KernelConfig-based n-to-1 kernel fusion in model loading/integrations. They could be merged separately." + "reason": "They share the general fusion theme, but PR #44979 defines a generic Module Fusion API while PR #45363 implements n-to-1 kernel fusion via KernelConfig in different files and code paths. They could not plausibly be merged as one duplicate PR." } ] }, @@ -21388,7 +21481,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "763e796e48732fd82f65a4f36c702e4582e3a36888c2e8a694fb82bd6b9b030e", + "prepared_review_unit_hash": "9d90558a1d93abd68ad85f06781dba68e61481665578f9edd6d0bc22452d3ea1", "component_budget": { "node_count": 2, "item_count": 2, @@ -21410,18 +21503,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same AttributeError in `_patch_mistral_regex` by switching from `backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer`; 45317 is the more complete version because it also adds a test.", + "summary": "Both pull requests address the same Mistral regex AttributeError by changing the pre-tokenizer access from backend_tokenizer.pre_tokenizer to tokenizer.pre_tokenizer; 45317 is the fuller version with tests.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "45317 is the better canonical PR: it fixes the same code-path bug as 45086 and includes an additional test, making it the more complete representative of the change.", + "canonical_pr_reason": "45317 is the better canonical PR because it contains the same fix as 45086 plus test coverage and a clearer description/title.", "best_issue_reason": null, - "best_pr_reason": "45317 is the strongest standalone PR candidate because it directly targets the reported bug, touches the same fix site, and adds coverage for the regression.", + "best_pr_reason": "45317 is the strongest representative of the cluster: same underlying code change as 45086, but more complete and validated by tests.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same underlying fix for the same AttributeError in `_patch_mistral_regex`; the code changes are effectively identical and 45317 simply adds test coverage, so these are mergeable as one PR." + "reason": "Same concrete bug fix in the same file for the same issue target; 45317 is essentially an expanded version of 45086 with tests." } ] }, @@ -21455,52 +21548,52 @@ "pull_request:45569|pull_request:46565", "pull_request:46332|pull_request:46565" ], - "prepared_review_unit_hash": "46b6c061bc28d3b5a101e001a77ca3c96a5c29a8f6681fa8c4a329d389caf838", + "prepared_review_unit_hash": "2aa0f234b08791945081aeba8fac3acaadda999a2776cbb01c59cf5671f631b8", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 3, - "serialized_chars": 13473, - "estimated_input_tokens": 3369, - "estimated_eval_tokens": 6994 + "serialized_chars": 13472, + "estimated_input_tokens": 3368, + "estimated_eval_tokens": 6992 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 3, - "serialized_chars": 13473, - "estimated_input_tokens": 3369, - "estimated_eval_tokens": 6994 + "serialized_chars": 13472, + "estimated_input_tokens": 3368, + "estimated_eval_tokens": 6992 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is heterogeneous: two PRs are about NemotronH model refactors, and two are about Nemotron ASR streaming support. None are true duplicates; the closest pairings are adjacent follow-ups in the same area, not the same concrete change.", - "confidence": 0.91, + "summary": "These PRs are related by the Nemotron family and some shared auto/doc plumbing, but they target different concrete changes: NemotronH refactoring, NemotronH dense/sparse split, Nemotron ASR streaming, and Nemotron 3.5 ASR streaming. I do not see duplicate-level overlap.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "No single PR cleanly subsumes the whole cluster. If one representative is needed, 45569 is the broadest NemotronH refactor, but it does not cover the ASR work.", + "canonical_pr_reason": "PR #46332 is the clearest standalone feature implementation in the cluster: it adds Nemotron ASR streaming end-to-end (model, losses, masking, auto mappings, docs). The others are either a different model family/version or a separate architectural refactor, so they are not better canonical duplicates.", "best_issue_reason": null, - "best_pr_reason": "45569 is the most substantial umbrella artifact within the NemotronH subgroup, with broad model/docs/auto wiring changes, but it is not a global duplicate canonical for the whole set.", + "best_pr_reason": "PR #46332 is the strongest representative PR for this cluster because it has the most complete, coherent change set and the richest implementation surface. PR #46565 is a related follow-up for Nemotron 3.5 ASR, while #45115 and #45569 concern NemotronH architecture/model refactoring rather than ASR streaming.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "Related NemotronH refactors, but 45115 is a GraniteMoeHybrid inheritance/mapping change while 45569 splits NemotronH into dense/sparse variants and updates docs/auto wiring. Different concrete changes, not one mergeable duplicate." + "reason": "Both touch NemotronH code, but #45115 is a structural inheritance/weight-mapping refactor and #45569 is a broader model split into NemotronH dense/sparse with docs and auto updates. They are related, not the same concrete change." }, { "left": "pull_request:45569", "right": "pull_request:46565", "accept": false, - "reason": "Different model families and tasks: NemotronH architecture refactor vs Nemotron 3.5 ASR streaming. Shared auto/docs files are not enough to treat them as the same change." + "reason": "#45569 is about NemotronH dense/sparse model architecture, while #46565 adds Nemotron 3.5 ASR streaming. Shared auto/docs files do not make them the same underlying change." }, { "left": "pull_request:46332", "right": "pull_request:46565", "accept": false, - "reason": "Both are ASR streaming PRs, but they target different model versions/families (Nemotron ASR vs Nemotron 3.5 ASR) and different code paths, so they should stay separate." + "reason": "Both are ASR-related, but they target different model versions and code paths: Nemotron ASR versus Nemotron 3.5 ASR. This is a family overlap, not a duplicate PR." } ] }, @@ -21530,7 +21623,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "bf5f9468ae4a7c75cc053a4eb1bfc7270645dd91be1f4ab6b679497745a08c8a", + "prepared_review_unit_hash": "bf8cb3894f8d056a3256f4c6ae658bf796efc2501728ab8aae4bd1faa6b63a76", "component_budget": { "node_count": 2, "item_count": 2, @@ -21552,18 +21645,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two related but distinct PRs: one adds heterogeneous per-layer configuration support, the other adds heterogeneous modeling/runtime support. They share infrastructure files, but they are not the same underlying change set.", - "confidence": 0.93, + "summary": "These are related but distinct PRs: one introduces per-layer heterogeneous configuration, the other builds on that to add heterogeneous modeling/cache support. They are not duplicates and should not be merged as the same change.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the broader umbrella change: it introduces heterogeneous model support across configuration and modeling code paths, making it the best representative of the cluster.", + "canonical_pr_reason": "PR #45333 is the more foundational, self-contained change: it adds the per-layer configuration machinery that the modeling PR builds on.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the best single representative because it covers the larger end-to-end heterogeneous-model feature rather than only the config layer.", + "best_pr_reason": "PR #45333 is the best representative of the cluster because it scopes a single coherent feature and provides the config layer needed by the broader modeling work.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Related feature work, but not the same concrete change: #45333 is limited to per-layer config support, while #45332 focuses on modeling/runtime integration and broader heterogeneous model behavior. They would not plausibly collapse into one PR without blending distinct implementation scopes." + "reason": "Different layers of the feature: #45333 adds heterogeneous per-layer config, while #45332 adds model/cache/masking behavior that consumes that config. Related, but not the same underlying change and not something to merge as duplicates." } ] }, @@ -21593,7 +21686,7 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "01857eed64ab97d4b083389aea6e702ba5e8ba85581452f2e89d1a4e530154ed", + "prepared_review_unit_hash": "a34123231e7373f0c71addcb9c1e0fcb3da0eb84bd40ff0e2b878f99fcdc9581", "component_budget": { "node_count": 2, "item_count": 2, @@ -21615,18 +21708,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are about typing, but they address different changes: one is a broad type-checking sweep across many files, the other adds typing aliases/protocols and related checks. They are related, not duplicates.", - "confidence": 0.91, + "summary": "Both PRs are typing-related, but they are not the same change: #45415 is a broad type-checking sweep across many `src/transformers/*` files, while #45425 adds focused typing helpers/protocols in `_typing.py` and related modules. The overlap in shared files looks incidental, not like one concrete fix.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45425 is the more coherent representative: it introduces specific typing helpers/protocols and updates the corresponding code paths, whereas 45415 is a wider repo-level type-checking cleanup.", + "canonical_pr_reason": "#45415 is the broader umbrella typing PR, spanning many modules plus type-checking config, so it is the clearest representative of the cluster\u2019s theme.", "best_issue_reason": null, - "best_pr_reason": "PR 45425 is the best standalone PR in this cluster because it has a narrower, well-scoped typing refactor. PR 45415 is broader and more of a general type-checking sweep, so it is less suitable as the canonical representative.", + "best_pr_reason": "#45415 is the best representative PR because it is the larger, more central typing pass rather than a narrow helper-type cleanup.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Both are typing-related, but they are not the same concrete change: 45415 is a broad multi-file type-checking pass, while 45425 adds typing protocols/aliases and targeted updates. They would not plausibly be merged as one duplicate PR." + "reason": "Both are typing-related, but they address different scopes and code paths: a wide repo type-checking sweep versus targeted typing protocol/alias additions. Shared files alone are not enough to treat them as one mergeable PR." } ] }, @@ -21656,12 +21749,12 @@ "soft_pairs": [ "pull_request:45513|pull_request:46084" ], - "prepared_review_unit_hash": "44034a5d1bccf9434278ddd02dd44f0f5ca3597bf899469e7d20d48db39dc47e", + "prepared_review_unit_hash": "84e144de4093796624d772b2e4941e54e33f87e0ec7a4768f6e4248f66a79a4e", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4857, + "serialized_chars": 4858, "estimated_input_tokens": 1215, "estimated_eval_tokens": 2686 }, @@ -21669,7 +21762,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4857, + "serialized_chars": 4858, "estimated_input_tokens": 1215, "estimated_eval_tokens": 2686 }, @@ -21678,18 +21771,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two unrelated pull requests: one fixes multi-token cached forward in Qwen3.5/GDN linear attention across several model files, the other fixes Mamba2 `use_cache` behavior when `seq_len > 1`. They do not appear to be the same underlying bug or mergeable into one PR.", - "confidence": 0.97, + "summary": "These are not duplicates: one fixes multi-token cached forward in Qwen3.5/Qwen3-next/Olmo hybrid linear attention paths, while the other fixes Mamba2 cache handling when seq_len > 1. The underlying code paths and bugs are different.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46084 is the cleaner canonical PR: it has an explicit issue target, a narrowly defined cache/sequence-length bug, and a focused fix path in one model family.", + "canonical_pr_reason": "No issue artifacts are present. Among the two PRs, #46084 is the cleaner canonical representative because it has a single model focus, an explicit linked issue, and a narrowly scoped cache fix.", "best_issue_reason": null, - "best_pr_reason": "PR 46084 is the best representative of the cluster because it is the most self-contained and clearly described bugfix, with a direct issue reference and limited scope.", + "best_pr_reason": "#46084 is the best representative PR: it targets one concrete bug in Mamba2 use_cache behavior for seq_len > 1 and is tightly scoped to that fix.", "soft_edge_verdicts": [ { "left": "pull_request:45513", "right": "pull_request:46084", "accept": false, - "reason": "Reject: these PRs address different model families and different cache bugs. 45513 fixes multi-token cached forward in Qwen3.5/GDN linear attention; 46084 fixes Mamba2 `use_cache` when `seq_len > 1`. They are not the same code-path problem and would not plausibly be merged into one PR." + "reason": "Reject: they address different model families and different failure modes. #45513 changes cached multi-token forward logic for Qwen3.5/Qwen3-next/Olmo hybrid linear attention, while #46084 fixes Mamba2 cache state handling when seq_len > 1. They do not look mergeable into one PR for the same underlying bug." } ] }, @@ -21719,7 +21812,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "30ffa7a36fb5350a7b5adaaf44a7d53c562700d226026e4c1666dd1a33ae3bed", + "prepared_review_unit_hash": "35b5004541f5c489e237098d8312aa7c42d9761e810f5c8763ba3f44c22d1265", "component_budget": { "node_count": 2, "item_count": 2, @@ -21741,18 +21834,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two related EP/loading PRs, but they fix different concrete bugs: one adds config/`has_ep` plumbing for EP + ZeRO-3 loading, the other prevents expert weights from being overwritten during FSDP2 loading. They overlap in subsystem, not in underlying defect.", - "confidence": 0.91, + "summary": "Both PRs are about expert parallelism, but they fix different bugs on different code paths: 45548 targets EP + DeepSpeed ZeRO-3 loading/config serialization, while 45662 targets EP + FSDP2 where experts can be overwritten by rank-0 broadcast. Similar area, not the same change.", + "confidence": 0.96, "canonical_issue_reason": null, "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR 45662 is the better representative of the cluster because it fixes a direct state-corruption bug in the model loading path with a clear runtime effect on expert parameters.", + "best_pr_reason": "45662 is the stronger representative of the cluster because it addresses the more concrete EP correctness bug across loading/sharding paths; 45548 is a narrower loading/serialization fix and not a duplicate.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Both touch EP/loading, but the failures differ: 45548 is about preserving `distributed_config` and exposing `has_ep` for ZeRO-3/accelerate launch, while 45662 fixes expert tensors being overwritten by rank-0 broadcast during FSDP2 loading. Different code paths and root causes." + "reason": "Different underlying bugs and fixes: one is ZeRO-3/accelerate loading and config handling, the other is FSDP2 expert overwriting via broadcast. Shared EP context is too broad to treat as the same PR." } ] }, @@ -21782,12 +21875,12 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "a8d133df3eff031bc7f6128b2cdcc27defb6f2627ca55fe4fe815de7db371b1a", + "prepared_review_unit_hash": "5432cedf76c6798b21c4de4449d0cd579720fbe78f0d56a27103ebeef210721a", "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 }, @@ -21795,7 +21888,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4589, + "serialized_chars": 4590, "estimated_input_tokens": 1148, "estimated_eval_tokens": 2552 }, @@ -21804,18 +21897,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same underlying xdist collision in patched testing debug artifacts, with 45645 extending the same core fix to CI log collection. They are close enough to treat as duplicate/overlapping fixes, with 45645 the broader and more complete candidate.", - "confidence": 0.86, + "summary": "Both PRs address the same xdist-related captured_info logging collision in the testing debug path. PR 45639 fixes the log writer to use worker-specific output, while PR 45645 applies the same core fix and also updates CI log collection to read multiple captured_info files, so they look like alternative/overlapping implementations of one bugfix.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45645 is the better canonical PR because it covers the core xdist-safe artifact fix in `testing_utils.py` and also updates CI handling to read multiple captured_info files, making it the more complete superset of the change.", + "canonical_pr_reason": "PR 45645 is the better canonical PR because it contains the core xdist-safe output fix plus the CI/workflow adjustment needed to consume the split log files, making it the more complete end-to-end change for this bug.", "best_issue_reason": null, - "best_pr_reason": "PR 45645 is the strongest overall PR candidate: it fixes the same captured_info collision problem and includes the downstream workflow adjustments needed to preserve debug output in CI.", + "best_pr_reason": "PR 45645 is the strongest representative because it covers both the producer-side collision fix and the downstream log aggregation behavior; PR 45639 is a narrower subset of the same change.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Both PRs fix the same concrete captured_info/xdist collision path in `src/transformers/testing_utils.py`; 45645 is a broader superset that plausibly subsumes 45639, so they can be treated as overlapping duplicate fixes." + "reason": "Same underlying xdist captured_info collision bug; both modify the same testing debug log path, and 45645 extends the same fix with workflow handling for the split files. They could plausibly be merged into one PR." } ] }, @@ -21848,12 +21941,12 @@ "pull_request:45670|pull_request:46401", "pull_request:46320|pull_request:46401" ], - "prepared_review_unit_hash": "d34fe31d21edbc2d38841e4774ab94cdced9716bbe9293f977d5695705b44be0", + "prepared_review_unit_hash": "58eb74f1dd5f00f7b784df1817c9f8568ab94a3d6e538b607397bf28f4d1e2a9", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7579, + "serialized_chars": 7580, "estimated_input_tokens": 1895, "estimated_eval_tokens": 4046 }, @@ -21861,7 +21954,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7579, + "serialized_chars": 7580, "estimated_input_tokens": 1895, "estimated_eval_tokens": 4046 }, @@ -21870,30 +21963,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These three PRs are separate, model-specific changes. Two add attn_implementation dispatch support for different architectures, and one is a small AutoModel registration nit. None are true duplicates.", - "confidence": 0.97, + "summary": "This cluster mixes one unrelated AutoModel mapping nit with two separate PRs that both add attn_implementation dispatch support, but for different model families. The RecurrentGemma and BLIP-2/InstructBLIP/InstructBLIPVideo PRs are thematically similar yet not the same underlying change.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "PR #46401 is the best representative of the meaningful subgroup: it makes a concrete attn_implementation dispatch change across several closely related multimodal models and has the broader, more complete implementation surface. PR #46320 is similar in theme but targets a different model family, and #45670 is unrelated.", "best_issue_reason": null, - "best_pr_reason": "PR #46320 is the strongest representative only in a loose sense: it is a self-contained, concrete implementation change with model code, conversion mapping, docs, and tests. It is still not a duplicate of the others.", + "best_pr_reason": "PR #46401 is the strongest standalone PR in the set because it addresses a concrete code-path change, spans multiple affected models in one coherent change, and has substantial implementation/test coverage. It is more representative than the small AutoModel registration nit in #45670.", "soft_edge_verdicts": [ { "left": "pull_request:45670", "right": "pull_request:46320", "accept": false, - "reason": "Different underlying changes: AutoModel registry mapping vs RecurrentGemma attn_implementation dispatch support." + "reason": "These are different changes: one adds a GLMASR AutoModel registration entry, the other adds attn_implementation dispatch for RecurrentGemma. Same repo area only at a very high level, not the same bug or patch." }, { "left": "pull_request:45670", "right": "pull_request:46401", "accept": false, - "reason": "Different underlying changes: AutoModel registry mapping vs BLIP-2/InstructBLIP attn_implementation dispatch support." + "reason": "#45670 is a model-mapping nudge for GLMASR, while #46401 implements attn_implementation dispatch for BLIP-2/InstructBLIP variants. They do not fix the same concrete code-path problem." }, { "left": "pull_request:46320", "right": "pull_request:46401", "accept": false, - "reason": "Same broad feature theme, but different model families and code paths; they should remain separate PRs." + "reason": "Both are attn_implementation-related, but they target different model families and different model classes. This is shared feature work, not a duplicate or mergeable single fix for the same underlying bug." } ] }, @@ -21923,12 +22016,12 @@ "soft_pairs": [ "pull_request:45720|pull_request:46259" ], - "prepared_review_unit_hash": "f287e5e2be311c6ab6bf13bb7403630bf57cf6931f93b6d0ff59d60323b23e06", + "prepared_review_unit_hash": "ca26dbb2efe28df53d4c0eafe8fb21c6804340ecad70603550010173f78267f9", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7247, + "serialized_chars": 7248, "estimated_input_tokens": 1812, "estimated_eval_tokens": 3880 }, @@ -21936,7 +22029,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7247, + "serialized_chars": 7248, "estimated_input_tokens": 1812, "estimated_eval_tokens": 3880 }, @@ -21945,18 +22038,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are both transformer lint/chore changes, but they target different rules and different code paths: TRF003 return-tuple handling vs TRF018 weight-initialization modeling patterns. They are not duplicates.", - "confidence": 0.98, + "summary": "The pair is not a duplicate: PR 45720 implements TRF003 by adjusting forward return-tuple behavior across model forward methods, while PR 46259 implements TRF018 by refactoring model weight initialization. The overlap in filenames is incidental to broad model coverage, not the same underlying change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No canonical PR for a duplicate cluster here; the two PRs target different lint rules and different code paths, so neither should stand in for the other.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "If one PR must be highlighted, 45720 is the more self-contained change: it addresses a single forward-path behavior across models. 46259 is a separate, broader initialization-rule refactor.", "soft_edge_verdicts": [ { "left": "pull_request:45720", "right": "pull_request:46259", "accept": false, - "reason": "Different underlying changes: 45720 adds can_return_tuple and adjusts forward return_dict handling for TRF003, while 46259 changes _init_weights implementations across models for TRF018. They touch similar model files, but not the same concrete bug or mergeable PR." + "reason": "Reject: TRF003 (forward return handling) and TRF018 (_init_weights refactor) are distinct changes affecting different behaviors, so they should not be merged as duplicates." } ] }, @@ -22052,12 +22145,12 @@ "soft_pairs": [ "pull_request:45932|pull_request:46100" ], - "prepared_review_unit_hash": "7d28c22b7d1c3e6996b69c617e7b8cc2d0cd290bb11a3f6be42d3cc9b57513ac", + "prepared_review_unit_hash": "acc6dee16e01f25991922f6264e2167a2f9ef3e63e3d44d45c1b2b65c3ff98c9", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5082, + "serialized_chars": 5083, "estimated_input_tokens": 1271, "estimated_eval_tokens": 2798 }, @@ -22065,7 +22158,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5082, + "serialized_chars": 5083, "estimated_input_tokens": 1271, "estimated_eval_tokens": 2798 }, @@ -22074,18 +22167,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated open PRs, not duplicates: #45932 is a broad/internal testing and model-code change set (workflow secrets plus multiple model files), while #46100 is a ViT test migration adding a new Bonbon ViT test file and related test updates.", - "confidence": 0.98, + "summary": "Both items are unrelated open PRs that only look similar because they involve tests/infrastructure. PR 45932 is a CI/telemetry and model-code experiment, while PR 46100 is a ViT test migration; they should not be deduplicated.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No clear canonical PR: the PRs address different workstreams and code paths, so neither is a duplicate representative of the other.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "No PR is a good cluster-level representative because the cluster does not reflect a single underlying change or bug.", + "best_pr_reason": "PR 46100 is the cleaner, self-contained change focused on test migration. PR 45932 is an experimental 'DO NOT MERGE' PR that mixes CI secrets and unrelated model-file edits, so it is a poorer representative.", "soft_edge_verdicts": [ { "left": "pull_request:45932", "right": "pull_request:46100", "accept": false, - "reason": "Reject: the diffs are about different changes. #45932 touches CI secrets and multiple model implementation/config files, while #46100 adds a new ViT test migration file and related test updates. They do not fix the same concrete code-path problem and would not plausibly be merged as one PR." + "reason": "Reject: these are different changes. 45932 touches CI/workflow secrets and multiple model implementations; 46100 adds/migrates ViT tests. They do not fix the same concrete code-path problem and would not plausibly be merged into one PR." } ] }, @@ -22118,7 +22211,7 @@ "pull_request:46073|pull_request:46196", "pull_request:46154|pull_request:46196" ], - "prepared_review_unit_hash": "e6ddafe9876ba4ae24c073f1cf3d9ba956be99bf51b49da36a9b1e6f146c7490", + "prepared_review_unit_hash": "a77679dd8f7deeecaf1791f6636b377248ae20b11389faa8574673ac810f9d44", "component_budget": { "node_count": 3, "item_count": 3, @@ -22140,30 +22233,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are about `attn_implementation=sdpa` dispatch support, but they target different model families and code paths (CTRL, T5, and wav2vec2/audio stack). They are related in theme, not duplicates of the same underlying change.", + "summary": "All three PRs are about adding SDPA/attn_implementation dispatch, but they target different model families and code paths (CTRL, T5, wav2vec2-derived audio models). They are thematically similar, not the same underlying change, so they should not be clustered as duplicates.", "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46196 is the broadest and most representative implementation in the cluster, touching many related audio model files and clearly centered on SDPA dispatch support for one concrete subsystem.", + "canonical_pr_reason": "No single PR is a true canonical duplicate for the others; if one representative is needed, [T5] Support attn_implementation=sdpa dispatch is the cleanest and most self-contained example of the pattern.", "best_issue_reason": null, - "best_pr_reason": "PR #46196 is the best single representative for the cluster because it has the largest scope and most complete cross-model implementation among the candidates, while still being a concrete code change rather than just docs.", + "best_pr_reason": "PR 46154 is the best representative because it is narrowly scoped to one model family and clearly implements the SDPA dispatch change without the broader multi-model scope of 46196.", "soft_edge_verdicts": [ { "left": "pull_request:46073", "right": "pull_request:46154", "accept": false, - "reason": "Both are SDPA dispatch PRs, but one is for CTRL and the other for T5; these are separate model implementations and not the same code-path fix." + "reason": "Different models and implementation targets (CTRL vs T5). Same feature theme, but not the same bug/change and not plausibly one merged PR." }, { "left": "pull_request:46073", "right": "pull_request:46196", "accept": false, - "reason": "CTRL attention dispatch and wav2vec2/audio attention dispatch are different model families with different internals, so they should not be merged as duplicates." + "reason": "CTRL support vs wav2vec2/seamless audio models are separate code paths and separate model families; too broad to treat as duplicates." }, { "left": "pull_request:46154", "right": "pull_request:46196", "accept": false, - "reason": "T5 decoder attention dispatch and wav2vec2/audio dispatch are distinct changes in unrelated model code, despite the shared SDPA theme." + "reason": "Both add SDPA dispatch, but T5 and wav2vec2/audio model stacks are unrelated enough that they are not the same concrete change." } ] }, @@ -22193,7 +22286,7 @@ "soft_pairs": [ "pull_request:46126|pull_request:46269" ], - "prepared_review_unit_hash": "e6946278c14609f14ebd5d94ad0e5e3dd331c6a96d993b22a8b3653d35ab8904", + "prepared_review_unit_hash": "b05910510a2cfe418a40350d92a3bc77f07e16217919502bf7b9fb3451d0899b", "component_budget": { "node_count": 2, "item_count": 2, @@ -22215,18 +22308,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one PR adds a focused EP/Trainer routing feature, while the other is a broad distributed branch base touching many subsystems and CI. The shared distributed filenames are too generic to indicate the same underlying change.", - "confidence": 0.92, + "summary": "These are related distributed-transformers PRs, but they do not look like the same underlying change: one is a focused EP/Trainer integration PR, the other is a broad distributed branch base touching many unrelated areas.", + "confidence": 0.84, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46126 is the most coherent and specific change set: it targets EP routing and Trainer integration with a clear code-path focus. PR #46269 is a broad base branch with many unrelated distributed edits, so it is not a good canonical representative.", + "canonical_pr_reason": "No issue in the cluster. As a PR representative, #46269 is the broader umbrella change and better captures the wider distributed refactor set in this cluster.", "best_issue_reason": null, - "best_pr_reason": "PR #46126 is the best representative for the cluster because it is narrowly scoped and has a concrete feature goal, whereas PR #46269 is too broad and mixed to serve as a canonical change.", + "best_pr_reason": "#46269 is the best single representative because it is the larger distributed base PR and subsumes more of the shared infrastructure touched here, whereas #46126 is a narrower EP/Trainer feature integration.", "soft_edge_verdicts": [ { "left": "pull_request:46126", "right": "pull_request:46269", "accept": false, - "reason": "Reject: the overlap is only at the subsystem level (distributed/tensor_parallel/utils/trainer files). #46126 implements EP router masking and Trainer integration, while #46269 is a large distributed branch base touching many files and CI, so they do not look like the same fix or mergeable into one PR." + "reason": "Shared distributed files are not enough: #46126 is specifically EP + Trainer integration, while #46269 is a broad distributed branch-base PR spanning many modules. They do not appear mergeable as one concrete code-path fix." } ] }, @@ -22288,190 +22381,184 @@ "pull_request:46504|pull_request:46563", "pull_request:46432|pull_request:46563" ], - "prepared_review_unit_hash": "e169768104d45bc8534835e78975eb3dc6f2fea820537ed58ae4afdc78d6618a", + "prepared_review_unit_hash": "26bb94552e9796fbede1a2203edb3a5494b9bf6e8b8410a0c2188a2fba9c3d99", "component_budget": { "node_count": 9, "item_count": 9, "soft_pair_count": 26, - "serialized_chars": 23355, - "estimated_input_tokens": 5839, - "estimated_eval_tokens": 11934 + "serialized_chars": 23361, + "estimated_input_tokens": 5841, + "estimated_eval_tokens": 11938 }, "budget": { "node_count": 9, "item_count": 9, "soft_pair_count": 26, - "serialized_chars": 23355, - "estimated_input_tokens": 5839, - "estimated_eval_tokens": 11934 + "serialized_chars": 23361, + "estimated_input_tokens": 5841, + "estimated_eval_tokens": 11938 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Mostly unrelated workflow-maintenance PRs. The only tight overlap is the pr-ci-security-gate.yml trio, but those PRs change different parts of the workflow and should stay separate.", + "summary": "These are all workflow-maintenance PRs, but they mostly touch different files or make different changes within the same workflow. The only near-cluster is the pr-ci-security-gate.yml trio, and those edits address different concerns, so there are no true duplicates here.", "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46554 is the strongest central representative of the only meaningful subcluster: it adds the actual security-gate checks for workflow/CI-infra changes in pr-ci-security-gate.yml.", + "canonical_pr_reason": "PR 46554 is the most central representative of the only related subgroup: it makes the substantive security-gate behavior change in .github/workflows/pr-ci-security-gate.yml, while the other same-file PRs change different mechanics (permissions/approval flow and workflow reference notes).", "best_issue_reason": null, - "best_pr_reason": "PR #46554 best represents the cluster because it is the most substantive and concrete workflow-security change among otherwise unrelated workflow edits.", + "best_pr_reason": "PR 46554 is the best global representative because it is the clearest, most substantive workflow security change and has the strongest overlap with the other same-file PRs, even though they are not duplicates.", "soft_edge_verdicts": [ { "left": "pull_request:46138", "right": "pull_request:46554", "accept": false, - "reason": "Different workflows and unrelated changes: stdin handling in self-comment-ci vs security checks in pr-ci-security-gate." + "reason": "Different workflow files and different automation paths; only a generic similarity in CI-related maintenance." }, { "left": "pull_request:46138", "right": "pull_request:46558", "accept": false, - "reason": "These touch different workflow files and solve different problems; not the same underlying change." + "reason": "Unrelated workflow file change; no shared concrete bug or code path." }, { "left": "pull_request:46138", "right": "pull_request:46563", "accept": false, - "reason": "Unrelated workflow edits: command piping in self-comment-ci vs reusable-workflow commentary in pr-ci-security-gate." + "reason": "Unrelated workflow file change; same broad area but different workflow and purpose." }, { "left": "pull_request:46138", "right": "pull_request:46504", "accept": false, - "reason": "Different workflow and different fix; only superficial similarity as small YAML churn." + "reason": "Different workflow file and different behavior; not the same underlying change." }, { "left": "pull_request:46138", "right": "pull_request:46432", "accept": false, - "reason": "Different workflow files and different security/operational fixes; not duplicate." + "reason": "Different workflow files and distinct fixes; not a duplicate pair." }, { "left": "pull_request:46432", "right": "pull_request:46503", "accept": false, - "reason": "SSH runner hardening vs benchmark trigger change are unrelated changes." + "reason": "Different workflow files with separate changes; no same concrete bug/change." }, { "left": "pull_request:46432", "right": "pull_request:46501", "accept": false, - "reason": "Different workflows and unrelated adjustments; no shared code-path problem." + "reason": "Different workflow files and unrelated edits; only superficial similarity." }, { "left": "pull_request:46432", "right": "pull_request:46502", "accept": false, - "reason": "Memory-monitor/download hardening and reviewer-assignment checkout behavior are separate changes." + "reason": "Different workflow files; these are separate workflow tweaks, not duplicates." }, { "left": "pull_request:46432", "right": "pull_request:46554", "accept": false, - "reason": "Different workflows with unrelated security fixes; not the same underlying bug or change." + "reason": "One updates ssh-runner, the other pr-ci-security-gate; different workflow behavior and no shared bug." }, { "left": "pull_request:46432", "right": "pull_request:46558", "accept": false, - "reason": "SSH runner secrets handling vs pr-ci-security-gate permissions/approval gating are distinct." + "reason": "Different workflow files and distinct workflow logic; not the same change." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46563", + "accept": false, + "reason": "Different workflow files; no evidence they fix the same issue." }, { "left": "pull_request:46501", "right": "pull_request:46503", "accept": false, - "reason": "Both are workflow tweaks, but they change different jobs and triggers; not mergeable as one fix." + "reason": "Both are generic workflow updates, but they change different files and behaviors." }, { "left": "pull_request:46501", "right": "pull_request:46502", "accept": false, - "reason": "Anti-slop permissions hardening and reviewer checkout ref change are unrelated." + "reason": "Different workflows with different fixes; not the same underlying problem." }, { "left": "pull_request:46501", "right": "pull_request:46504", "accept": false, - "reason": "Different workflow files and different logic changes; only broad YAML similarity." + "reason": "Different workflow files and unrelated CI behavior changes." }, { "left": "pull_request:46501", "right": "pull_request:46554", "accept": false, - "reason": "Different workflows and different security concerns; not the same change." + "reason": "Different workflow files; no concrete overlap beyond general CI maintenance." }, { "left": "pull_request:46502", "right": "pull_request:46503", "accept": false, - "reason": "Reviewer assignment checkout and benchmark dispatch support are unrelated workflow updates." + "reason": "Different workflow files and separate behavior changes." }, { "left": "pull_request:46502", "right": "pull_request:46504", "accept": false, - "reason": "Different workflows and unrelated implementation details." + "reason": "Different workflow files; these are not the same fix." }, { "left": "pull_request:46502", "right": "pull_request:46554", "accept": false, - "reason": "Reviewer workflow checkout fix is unrelated to pr-ci-security-gate security checks." + "reason": "Different workflow files and different concerns; not a duplicate." }, { "left": "pull_request:46503", "right": "pull_request:46504", "accept": false, - "reason": "Benchmark trigger expansion and build-image message parsing fix are separate changes." + "reason": "Different workflow files and unrelated CI logic changes." }, { "left": "pull_request:46503", "right": "pull_request:46554", "accept": false, - "reason": "Different workflow files and different problems; no shared bug." + "reason": "Different workflow files; no shared bug or change target." }, { "left": "pull_request:46504", "right": "pull_request:46554", "accept": false, - "reason": "Build-image commit-message handling and pr-ci-security-gate checks are unrelated." + "reason": "Different workflow files and different operational behavior." }, { "left": "pull_request:46504", "right": "pull_request:46558", "accept": false, - "reason": "Different workflows and different fixes; not the same underlying issue." + "reason": "Different workflow files; only broad CI context overlaps." }, { "left": "pull_request:46554", "right": "pull_request:46558", "accept": false, - "reason": "Same workflow file, but one adds file-diff security checks while the other changes permissions and approval gating; related, but not the same concrete change." + "reason": "Same workflow file, but the changes are about different concerns: security checks versus permissions/approval gating." }, { "left": "pull_request:46554", "right": "pull_request:46563", "accept": false, - "reason": "Direct security checks vs reusable-workflow wrapper notes are different implementations and different changes." + "reason": "Same workflow file, but one adds active security checks while the other swaps in a reusable workflow and comments; different underlying change." }, { "left": "pull_request:46558", "right": "pull_request:46563", "accept": false, - "reason": "Both touch pr-ci-security-gate.yml, but they modify different parts of the workflow and are not duplicates." - }, - { - "left": "pull_request:46504", - "right": "pull_request:46563", - "accept": false, - "reason": "Build CI image trigger parsing and pr-ci-security-gate workflow notes are unrelated." - }, - { - "left": "pull_request:46432", - "right": "pull_request:46563", - "accept": false, - "reason": "SSH runner hardening and pr-ci-security-gate workflow commentary are unrelated." + "reason": "Same workflow file, but these edit different parts of the gate and do not fix the same concrete issue." } ] }, @@ -22501,12 +22588,12 @@ "soft_pairs": [ "pull_request:46152|pull_request:46207" ], - "prepared_review_unit_hash": "346e98c4183a539d4c682927d07c0f5e32f29a95fb8b4656b5f31b3544b9e3ea", + "prepared_review_unit_hash": "c8cea507182c64838bb0ce43a734c0ad32433aa8b33ebe6204e345eca569eb06", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4276, + "serialized_chars": 4275, "estimated_input_tokens": 1069, "estimated_eval_tokens": 2394 }, @@ -22514,7 +22601,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4276, + "serialized_chars": 4275, "estimated_input_tokens": 1069, "estimated_eval_tokens": 2394 }, @@ -22523,18 +22610,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are related cleanup changes to stop reading Python source files with plain open(), but they target different code paths and concrete behaviors, so they are not duplicates.", - "confidence": 0.87, + "summary": "These two PRs are related at a theme level (both replace direct file-open source inspection), but they change different code paths for different reasons and should not be treated as duplicates.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46207 is the more concrete and self-contained fix: it updates _can_set_attn_implementation in modeling_utils.py and includes test coverage for that specific behavior.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #46207 is the better representative of the cluster because it addresses a specific source-reading bug in a core runtime path, rather than a broader helper refactor.", + "best_pr_reason": "PR 46207 is the cleaner representative of the cluster\u2019s theme: it has a focused source-inspection refactor in one code path and includes tests, while PR 46152 is a separate package-resource traversal fix in import_utils.", "soft_edge_verdicts": [ { "left": "pull_request:46152", "right": "pull_request:46207", "accept": false, - "reason": "They share the general theme of avoiding open() for source inspection, but the affected code paths are different (import_utils helper vs. modeling_utils attention-implementation check) and would not naturally merge into one PR." + "reason": "Reject: they address different underlying behaviors in different modules (_resolve_traversable vs _can_set_attn_implementation). The shared idea of avoiding open() is too broad to imply the same bug or mergeable change." } ] }, @@ -22566,7 +22653,7 @@ "pull_request:46436|pull_request:46474", "pull_request:46173|pull_request:46474" ], - "prepared_review_unit_hash": "d5936dfdd7af6d2567fbe54328dea7ac9df856e55b25fc92d731fe346baf6070", + "prepared_review_unit_hash": "e2a420a32c3e8c964156e024e94d866058597753744bb3270ae39557a8fe7507", "component_budget": { "node_count": 3, "item_count": 3, @@ -22588,30 +22675,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two PRs are near-duplicate fixes for StaticLayer.get_seq_length type hints; a third PR is a separate static-cache max-length configuration fix. Canonicalize on the earlier, more established type-hint PR.", - "confidence": 0.97, + "summary": "Two PRs are the same StaticLayer type-hint change in cache_utils.py and should be clustered together; the generation cache max-len fix is a separate change in a different code path.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46173 is the earlier, original StaticLayer.get_seq_length type-annotation fix, with the same file-level change as 46474 and an explicit issue link/review activity.", + "canonical_pr_reason": "PR #46474 is the cleanest representative of the StaticLayer.get_seq_length type-hint change: it has the same patch and target as #46173, but a simpler title and identical scope. PR #46436 is unrelated in implementation and bug surface, so it should not be the canonical PR for this cluster.", "best_issue_reason": null, - "best_pr_reason": "PR 46173 best represents the duplicate pair because it is the first occurrence of the identical cache_utils typing change and already anchors the referenced issue.", + "best_pr_reason": "#46474 is the best single PR for the duplicated StaticLayer annotation fix because it directly matches the code change in cache_utils.py and the same tracking issue. #46173 is effectively the same change; #46436 addresses a different static-cache length handling bug in generation/utils.py.", "soft_edge_verdicts": [ { "left": "pull_request:46436", "right": "pull_request:46474", "accept": false, - "reason": "Different fixes: 46436 changes generation cache max-length handling, while 46474 only adjusts the StaticLayer get_seq_length type hint. Same subsystem, but not the same underlying change." + "reason": "Different fixes: #46436 changes static cache max length selection in generation/utils.py, while #46474 only updates the StaticLayer.get_seq_length type annotation in cache_utils.py. Same broad subsystem, but not the same underlying bug or mergeable into one PR." }, { "left": "pull_request:46173", "right": "pull_request:46474", "accept": true, - "reason": "These are effectively the same PR: same file, same type-hint change, same explicit issue target, and identical diff intent." + "reason": "Same underlying change: both PRs update StaticLayer.get_seq_length return typing in src/transformers/cache_utils.py and target the same issue #45987. The diffs are effectively identical, so they are duplicates." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: 46173 and 46474 share the same file, same issue target, and identical type-hint diff, so treating them as duplicates/canonicalizing on the earlier PR is conservative. The separation of 46436 is also supported because it changes generation cache max-length handling in a different file and is a distinct fix. No evidence is overstated." + "feedback": "Grounded and conservative: the packet supports that #46173 and #46474 are duplicate diffs on the same cache_utils.py type-hint change and same issue target, while #46436 is a separate generation/utils.py change with no shared files or target. The soft-edge decisions are appropriately cautious." }, "evaluator_used": true, "retried": false, @@ -22701,40 +22788,40 @@ "soft_pairs": [ "pull_request:46310|pull_request:46546" ], - "prepared_review_unit_hash": "6196817b4ae9aa1fd6ac2d01ab87999f3f450d1d6c87c29e31a2830ba78db7d3", + "prepared_review_unit_hash": "2e95f980edea99a22e3a77a3d0f8b8dfd14626ebc379ee5a88d84dcb3fb9d659", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4786, - "estimated_input_tokens": 1197, - "estimated_eval_tokens": 2650 + "serialized_chars": 4851, + "estimated_input_tokens": 1213, + "estimated_eval_tokens": 2682 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4786, - "estimated_input_tokens": 1197, - "estimated_eval_tokens": 2650 + "serialized_chars": 4851, + "estimated_input_tokens": 1213, + "estimated_eval_tokens": 2682 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs fix the same Moonshine streaming sliding-window boundary bug: the right side mask should include the full boundary (`<= right_window_size`) in the same code paths. They target the same issue and are duplicative at the change level; the later PR is the better representative because it also adds tests.", - "confidence": 0.98, + "summary": "Both PRs fix the same off-by-one bug in the Moonshine streaming sliding-window mask: the right window should include the boundary element (`<= right_window_size` instead of `<`). PR 46546 is the stronger canonical PR because it includes the Moonshine fix plus test coverage, while 46310 is the narrower earlier patch.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46546 is the better canonical PR because it makes the same boundary fix as 46310 but also includes a test file, making it the more complete representative of the change.", + "canonical_pr_reason": "PR 46546 is the better canonical PR because it contains the same Moonshine streaming boundary fix as 46310, plus test coverage and a broader related adjustment, making it the more complete change set.", "best_issue_reason": null, - "best_pr_reason": "PR 46546 is the best cluster representative: same exact bug fix in the same files, with added regression coverage.", + "best_pr_reason": "PR 46546 is the best overall PR candidate: it addresses the same concrete bug, is more complete, and adds validation. PR 46310 is essentially a narrower version of the same fix.", "soft_edge_verdicts": [ { "left": "pull_request:46310", "right": "pull_request:46546", "accept": true, - "reason": "Same underlying boundary bug in Moonshine streaming masking, same changed code paths, and the second PR is just the more complete version with tests." + "reason": "Same underlying Moonshine streaming right-window boundary bug; both change the same mask logic from `< right_window_size` to `<= right_window_size`, so they are duplicate fixes for the same code path." } ] }, @@ -22764,12 +22851,12 @@ "soft_pairs": [ "pull_request:46376|pull_request:46417" ], - "prepared_review_unit_hash": "b649e21459886c0074961b187fcc5cfe9d6c7ca6599a763267f1cdacb41faf99", + "prepared_review_unit_hash": "70e123cd3ad46e00f5ffcdb6e523b2bade9966e5bc87f351330ddaa9d14315ae", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5939, + "serialized_chars": 5938, "estimated_input_tokens": 1485, "estimated_eval_tokens": 3226 }, @@ -22777,7 +22864,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5939, + "serialized_chars": 5938, "estimated_input_tokens": 1485, "estimated_eval_tokens": 3226 }, @@ -22786,18 +22873,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are unrelated pull requests: one is a draft infra change for multimodal continuous batching, the other adds the HunYuan VL model implementation and docs. They do not look like duplicate fixes or mergeable variants of the same change.", - "confidence": 0.98, + "summary": "Two unrelated pull requests: one adds multimodal support to continuous batching infrastructure, the other adds the HunYuan VL model package, docs, and tests. They do not fix the same code-path problem and should not be deduplicated.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46417 is the more complete and self-contained change: it adds a specific model with code, tests, and docs, whereas PR 46376 is a draft infrastructure feature spanning a different subsystem.", + "canonical_pr_reason": "PR 46417 is the better canonical representative because it is a concrete, self-contained model addition with docs and tests, whereas PR 46376 is an infrastructure change for continuous batching.", "best_issue_reason": null, - "best_pr_reason": "PR 46417 is the best representative PR in this cluster because it is concrete, non-draft, and narrowly focused on one model addition; PR 46376 is broader, draft, and on a separate continuous-batching feature path.", + "best_pr_reason": "PR 46417 is the strongest standalone PR in the cluster: it has the clearest deliverable, broadest implementation scope, and the most complete supporting test/documentation set.", "soft_edge_verdicts": [ { "left": "pull_request:46376", "right": "pull_request:46417", "accept": false, - "reason": "Different underlying work: continuous-batching multimodal support versus HunYuan VL model addition. They touch different subsystems and cannot plausibly be merged into one PR as the same change." + "reason": "Different changes: continuous batching multimodal support vs. adding a specific HunYuan VL model. They touch different subsystems and are not the same underlying fix." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index db34b036026df400af75ffa71340981ed30995b2..a6bb4ee933ec4c57a970574899563bf1d60c6832 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260611t180022z", + "analysis_id": "hybrid-model-20260612t000041z", "archived_artifacts": { - "hybrid": "snapshots/20260611T180022Z/analysis-runs/hybrid-model-20260611t180022z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260611T180022Z/analysis-runs/hybrid-model-20260611t180022z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-06-11T18:10:55Z", + "published_at": "2026-06-12T00:09:12Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260611T180022Z", + "snapshot_id": "20260612T000041Z", "variant": "hybrid" } diff --git a/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.json b/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..d941dc8e550ec86dbfef7c3b8532fa16d019e344 --- /dev/null +++ b/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.json @@ -0,0 +1,2252 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260612T000041Z", + "generated_at": "2026-06-12T00:09:10Z", + "evidence_quality": "full", + "llm_enrichment": true, + "meta_bugs": [ + { + "cluster_id": "cluster-43979-11", + "summary": "Cluster of 1 issues and 10 PRs centered on issue #43979.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 43979, + "canonical_pr_number": 44007, + "issue_numbers": [ + 43979 + ], + "pr_numbers": [ + 43996, + 44007, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 43996, + "right_pr_number": 44007, + "code_similarity": 0.179, + "size_similarity": 0.576, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.429, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44013, + "code_similarity": 0.122, + "size_similarity": 0.318, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.392, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44044, + "code_similarity": 0.245, + "size_similarity": 0.864, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.479, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44066, + "code_similarity": 0.225, + "size_similarity": 0.818, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.408, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44072, + "code_similarity": 0.14, + "size_similarity": 0.303, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.528, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44085, + "code_similarity": 0.216, + "size_similarity": 0.783, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44129, + "code_similarity": 0.163, + "size_similarity": 0.643, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.229, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44154, + "code_similarity": 0.153, + "size_similarity": 0.535, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.31, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44722, + "code_similarity": 0.225, + "size_similarity": 0.848, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.368, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44013, + "code_similarity": 0.19, + "size_similarity": 0.553, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.531, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44044, + "code_similarity": 0.186, + "size_similarity": 0.667, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.354, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44066, + "code_similarity": 0.188, + "size_similarity": 0.704, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.315, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44072, + "code_similarity": 0.212, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.708, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44085, + "code_similarity": 0.195, + "size_similarity": 0.735, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.318, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44129, + "code_similarity": 0.103, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.191, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44154, + "code_similarity": 0.238, + "size_similarity": 0.93, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.344, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44722, + "code_similarity": 0.178, + "size_similarity": 0.679, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.28, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44044, + "code_similarity": 0.126, + "size_similarity": 0.368, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.351, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44066, + "code_similarity": 0.127, + "size_similarity": 0.389, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.325, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44072, + "code_similarity": 0.29, + "size_similarity": 0.952, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.667, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44085, + "code_similarity": 0.131, + "size_similarity": 0.406, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.329, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44129, + "code_similarity": 0.07, + "size_similarity": 0.205, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.192, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44154, + "code_similarity": 0.177, + "size_similarity": 0.594, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.389, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44722, + "code_similarity": 0.118, + "size_similarity": 0.375, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.287, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44066, + "code_similarity": 0.25, + "size_similarity": 0.947, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.404, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44072, + "code_similarity": 0.136, + "size_similarity": 0.351, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.442, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44085, + "code_similarity": 0.24, + "size_similarity": 0.906, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.394, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44129, + "code_similarity": 0.147, + "size_similarity": 0.555, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.243, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44154, + "code_similarity": 0.17, + "size_similarity": 0.62, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.306, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44722, + "code_similarity": 0.257, + "size_similarity": 0.982, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.402, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44072, + "code_similarity": 0.133, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.393, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44085, + "code_similarity": 0.763, + "size_similarity": 0.957, + "file_overlap": 0.5, + "area_overlap": 0.825, + "patch_similarity": 0.887, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ] + } + ] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44129, + "code_similarity": 0.145, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.263, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44154, + "code_similarity": 0.174, + "size_similarity": 0.654, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.286, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44722, + "code_similarity": 0.907, + "size_similarity": 0.964, + "file_overlap": 1.0, + "area_overlap": 0.808, + "patch_similarity": 0.874, + "shared_filenames": [ + "src/transformers/models/codegen/modeling_codegen.py", + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/codegen/modeling_codegen.py", + "left_ranges": [ + [ + 245, + 266 + ] + ], + "right_ranges": [ + [ + 228, + 234 + ] + ] + }, + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44085, + "code_similarity": 0.137, + "size_similarity": 0.387, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44129, + "code_similarity": 0.074, + "size_similarity": 0.195, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.231, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44154, + "code_similarity": 0.175, + "size_similarity": 0.566, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.414, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44722, + "code_similarity": 0.124, + "size_similarity": 0.357, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.347, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44129, + "code_similarity": 0.141, + "size_similarity": 0.503, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.272, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44154, + "code_similarity": 0.18, + "size_similarity": 0.684, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.289, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44722, + "code_similarity": 0.728, + "size_similarity": 0.923, + "file_overlap": 0.5, + "area_overlap": 0.791, + "patch_similarity": 0.78, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44154, + "code_similarity": 0.099, + "size_similarity": 0.344, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.199, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44722, + "code_similarity": 0.146, + "size_similarity": 0.545, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.247, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44154, + "right_pr_number": 44722, + "code_similarity": 0.164, + "size_similarity": 0.631, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.255, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, + { + "cluster_id": "cluster-41211-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41211.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41211, + "canonical_pr_number": 44339, + "issue_numbers": [ + 41211 + ], + "pr_numbers": [ + 41356, + 44339 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41356, + "right_pr_number": 44339, + "code_similarity": 0.155, + "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": 0.797, + "size_similarity": 0.885, + "file_overlap": 0.75, + "area_overlap": 0.831, + "patch_similarity": 0.693, + "shared_filenames": [ + "docs/source/en/_toctree.yml", + "docs/source/en/model_doc/minicpm3.md", + "src/transformers/models/auto/modeling_auto.py", + "src/transformers/models/minicpm3/__init__.py", + "src/transformers/models/minicpm3/configuration_minicpm3.py", + "src/transformers/models/minicpm3/modeling_minicpm3.py", + "src/transformers/models/minicpm3/modular_minicpm3.py", + "tests/models/minicpm3/__init__.py", + "tests/models/minicpm3/test_modeling_minicpm3.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/_toctree.yml", + "left_ranges": [ + [ + 713, + 720 + ] + ], + "right_ranges": [ + [ + 730, + 737 + ] + ] + }, + { + "filename": "docs/source/en/model_doc/minicpm3.md", + "left_ranges": [ + [ + 1, + 67 + ] + ], + "right_ranges": [ + [ + 1, + 45 + ] + ] + }, + { + "filename": "src/transformers/models/auto/modeling_auto.py", + "left_ranges": [ + [ + 285, + 291 + ], + [ + 700, + 706 + ], + [ + 1298, + 1304 + ] + ], + "right_ranges": [ + [ + 720, + 726 + ], + [ + 1331, + 1337 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/__init__.py", + "left_ranges": [ + [ + 1, + 29 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/configuration_minicpm3.py", + "left_ranges": [ + [ + 1, + 141 + ] + ], + "right_ranges": [ + [ + 1, + 126 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 544 + ] + ], + "right_ranges": [ + [ + 1, + 522 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modular_minicpm3.py", + "left_ranges": [ + [ + 1, + 444 + ] + ], + "right_ranges": [ + [ + 1, + 342 + ] + ] + }, + { + "filename": "tests/models/minicpm3/test_modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 109 + ] + ], + "right_ranges": [ + [ + 1, + 136 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45647-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45647.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45647, + "canonical_pr_number": 45738, + "issue_numbers": [ + 45647 + ], + "pr_numbers": [ + 45737, + 45738 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 45737, + "right_pr_number": 45738, + "code_similarity": 0.526, + "size_similarity": 0.11, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.024, + "shared_filenames": [ + "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py", + "left_ranges": [ + [ + 491, + 497 + ] + ], + "right_ranges": [ + [ + 491, + 497 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45987-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45987.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45987, + "canonical_pr_number": 46173, + "issue_numbers": [ + 45987 + ], + "pr_numbers": [ + 46173, + 46474 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46173, + "right_pr_number": 46474, + "code_similarity": 1.0, + "size_similarity": 1.0, + "file_overlap": 1.0, + "area_overlap": 1.0, + "patch_similarity": 1.0, + "shared_filenames": [ + "src/transformers/cache_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cache_utils.py", + "left_ranges": [ + [ + 69, + 75 + ], + [ + 375, + 381 + ] + ], + "right_ranges": [ + [ + 69, + 75 + ], + [ + 375, + 381 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46291-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46291.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 46291, + "canonical_pr_number": 46293, + "issue_numbers": [ + 46291 + ], + "pr_numbers": [ + 46293, + 46478 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 46293, + "right_pr_number": 46478, + "code_similarity": 0.483, + "size_similarity": 0.325, + "file_overlap": 0.333, + "area_overlap": 0.862, + "patch_similarity": 0.111, + "shared_filenames": [ + "src/transformers/integrations/hub_kernels.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/hub_kernels.py", + "left_ranges": [ + [ + 89, + 180 + ], + [ + 233, + 247 + ] + ], + "right_ranges": [ + [ + 89, + 181 + ], + [ + 234, + 248 + ], + [ + 263, + 269 + ], + [ + 360, + 366 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45561-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45561, + "canonical_pr_number": 45645, + "issue_numbers": [ + 45561 + ], + "pr_numbers": [ + 45639, + 45645 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45639, + "right_pr_number": 45645, + "code_similarity": 0.494, + "size_similarity": 0.619, + "file_overlap": 0.5, + "area_overlap": 0.434, + "patch_similarity": 0.451, + "shared_filenames": [ + "src/transformers/testing_utils.py", + "tests/utils/test_testing_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/testing_utils.py", + "left_ranges": [ + [ + 3595, + 3602 + ], + [ + 3819, + 3853 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 86 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46305-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46305.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 46305, + "canonical_pr_number": 46546, + "issue_numbers": [ + 46305 + ], + "pr_numbers": [ + 46310, + 46546 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46310, + "right_pr_number": 46546, + "code_similarity": 0.566, + "size_similarity": 0.444, + "file_overlap": 0.5, + "area_overlap": 0.867, + "patch_similarity": 0.16, + "shared_filenames": [ + "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/moonshine_streaming/modeling_moonshine_streaming.py", + "left_ranges": [ + [ + 356, + 362 + ] + ], + "right_ranges": [ + [ + 355, + 361 + ] + ] + }, + { + "filename": "src/transformers/models/moonshine_streaming/modular_moonshine_streaming.py", + "left_ranges": [ + [ + 269, + 275 + ] + ], + "right_ranges": [ + [ + 269, + 275 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46504-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 46504, + "issue_numbers": [], + "pr_numbers": [ + 46504, + 46563 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46504, + "right_pr_number": 46563, + "code_similarity": 0.185, + "size_similarity": 0.923, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.0, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, + { + "cluster_id": "cluster-44018-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44068, + "issue_numbers": [], + "pr_numbers": [ + 44018, + 44068 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44018, + "right_pr_number": 44068, + "code_similarity": 0.766, + "size_similarity": 0.939, + "file_overlap": 1.0, + "area_overlap": 0.425, + "patch_similarity": 0.866, + "shared_filenames": [ + "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "tests/models/gpt_neo/test_modeling_gpt_neo.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "left_ranges": [ + [ + 26, + 46 + ], + [ + 138, + 143 + ], + [ + 182, + 187 + ], + [ + 283, + 295 + ], + [ + 327, + 341 + ], + [ + 348, + 354 + ], + [ + 360, + 369 + ], + [ + 399, + 406 + ], + [ + 411, + 419 + ], + [ + 428, + 433 + ], + [ + 472, + 492 + ], + [ + 507, + 513 + ], + [ + 519, + 528 + ], + [ + 541, + 559 + ], + [ + 562, + 567 + ], + [ + 595, + 601 + ], + [ + 607, + 614 + ], + [ + 627, + 643 + ], + [ + 685, + 690 + ], + [ + 708, + 714 + ], + [ + 720, + 727 + ], + [ + 740, + 757 + ], + [ + 761, + 766 + ], + [ + 780, + 786 + ], + [ + 791, + 798 + ], + [ + 807, + 822 + ], + [ + 840, + 845 + ] + ], + "right_ranges": [ + [ + 26, + 31 + ], + [ + 34, + 44 + ], + [ + 136, + 141 + ], + [ + 180, + 185 + ], + [ + 281, + 293 + ], + [ + 325, + 339 + ], + [ + 346, + 352 + ], + [ + 358, + 367 + ], + [ + 397, + 404 + ], + [ + 409, + 417 + ], + [ + 426, + 434 + ], + [ + 463, + 483 + ], + [ + 498, + 504 + ], + [ + 510, + 519 + ], + [ + 532, + 550 + ], + [ + 553, + 564 + ], + [ + 586, + 592 + ], + [ + 598, + 605 + ], + [ + 618, + 634 + ], + [ + 676, + 688 + ], + [ + 699, + 705 + ], + [ + 711, + 718 + ], + [ + 731, + 748 + ], + [ + 752, + 762 + ], + [ + 771, + 777 + ], + [ + 782, + 789 + ], + [ + 798, + 813 + ], + [ + 831, + 836 + ] + ] + }, + { + "filename": "tests/models/gpt_neo/test_modeling_gpt_neo.py", + "left_ranges": [ + [ + 458, + 464 + ] + ], + "right_ranges": [ + [ + 458, + 464 + ] + ] + } + ] + } + ] + } + ], + "duplicate_issues": [], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 45613, + "duplicate_pr_numbers": [ + 41116 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-44018-2", + "canonical_pr_number": 44068, + "duplicate_pr_numbers": [ + 44018 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45561-3", + "canonical_pr_number": 45645, + "duplicate_pr_numbers": [ + 45639 + ], + "target_issue_number": 45561, + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45647-3", + "canonical_pr_number": 45738, + "duplicate_pr_numbers": [ + 45737 + ], + "target_issue_number": 45647, + "reason": "PRs in cluster-45647-3 are treated as duplicates because they converge on issue #45647 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-45987-3", + "canonical_pr_number": 46173, + "duplicate_pr_numbers": [ + 46474 + ], + "target_issue_number": 45987, + "reason": "PRs in cluster-45987-3 are treated as duplicates because they converge on issue #45987 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46291-3", + "canonical_pr_number": 46293, + "duplicate_pr_numbers": [ + 46478 + ], + "target_issue_number": 46291, + "reason": "PRs in cluster-46291-3 are treated as duplicates because they converge on issue #46291 with closing_reference evidence." + }, + { + "cluster_id": "cluster-46305-3", + "canonical_pr_number": 46546, + "duplicate_pr_numbers": [ + 46310 + ], + "target_issue_number": 46305, + "reason": "PRs in cluster-46305-3 are treated as duplicates because they converge on issue #46305 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-46504-2", + "canonical_pr_number": 46504, + "duplicate_pr_numbers": [ + 46563 + ], + "target_issue_number": null, + "reason": "PRs in cluster-46504-2 are treated as duplicates because they share soft_similarity evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-41211-3", + "issue_number": 41211, + "reason": "Issue #41211 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 6 discussion comments plus 6 inbound references.", + "score": 148.0 + }, + "best_pr": { + "cluster_id": "cluster-43979-11", + "pr_number": 44007, + "reason": "PR #44007 is the strongest global PR candidate because it is open, belongs to a cluster with 11 artifacts, links to 1 issues, and carries 2 review/discussion events.", + "score": 238.886 + } +} diff --git a/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..2af3de24bef931a29d20d7e87fde8b750a54b4fb --- /dev/null +++ b/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,22899 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260612T000041Z", + "generated_at": "2026-06-12T00:09:10Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43425", + "issue:43541", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43827", + "issue:43828", + "issue:43866", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44805", + "issue:44918", + "issue:45081", + "issue:45161", + "issue:45237", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45800", + "issue:45823", + "issue:45859", + "issue:45874", + "issue:45901", + "issue:45902", + "issue:46015", + "issue:46165", + "issue:46294", + "issue:46304", + "issue:46402", + "issue:46468", + "issue:46515" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:46468", + "issue:45237|issue:45507", + "issue:30277|issue:35545", + "issue:35545|issue:42915", + "issue:35545|issue:41720", + "issue:45901|issue:46015", + "issue:44509|issue:44512", + "issue:35545|issue:36010", + "issue:44918|issue:46165", + "issue:44805|issue:45823", + "issue:43638|issue:44805", + "issue:43638|issue:46165", + "issue:43596|issue:44805", + "issue:44805|issue:46294", + "issue:44805|issue:45161", + "issue:46165|issue:46468", + "issue:44560|issue:46165", + "issue:43596|issue:46165", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43039|issue:45161", + "issue:43975|issue:44457", + "issue:44292|issue:44457", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:45859|issue:45902", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:45874|issue:46294", + "issue:44292|issue:45237", + "issue:39290|issue:45081", + "issue:45800|issue:46515", + "issue:45823|issue:45874", + "issue:43425|issue:44292", + "issue:34928|issue:41720", + "issue:43828|issue:45161", + "issue:43828|issue:44292", + "issue:43541|issue:43716", + "issue:43716|issue:46468", + "issue:43716|issue:45685", + "issue:44918|issue:45161", + "issue:44292|issue:44918", + "issue:45507|issue:45685", + "issue:43716|issue:44292", + "issue:43039|issue:44918", + "issue:41762|issue:43638", + "issue:43827|issue:44512", + "issue:43828|issue:45685", + "issue:44918|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:34928|issue:36010", + "issue:43039|issue:44805", + "issue:45237|issue:46015", + "issue:43866|issue:46402", + "issue:45800|issue:46304" + ], + "prepared_review_unit_hash": "d369b264f3aba78e0b3b00db699ce6cbe44f074605200fc8d0404ea8ea9e3471", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 63, + "serialized_chars": 46534, + "estimated_input_tokens": 11634, + "estimated_eval_tokens": 23524 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a similarity-based grab bag of unrelated runtime, model-loading, backend, and docs issues. The soft pairs share broad keywords like DeepSpeed, dtype/device mismatch, or version incompatibility, but they do not look like the same underlying bug or change, so all soft edges are rejected.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45237 is the most central hub in this set of mostly runtime/hardware failures, so it is the best anchor for the cluster, even though it does not truly duplicate the others.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45237 is the broadest representative of the cluster\u2019s hardware/backend failure reports, making it the best single issue to keep as a reference point.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different failures: a config attribute error in Qwen3.5 versus a streaming chat/completions API crash." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different backends and models: AMD GPU failure in GPT-OSS versus an MPS histogram op limitation in MoE." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different model paths and root causes: Mistral-3 image dtype mismatch versus GPT-OSS AMD GPU failure." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Different bugs: Phi MoE autocast dtype mismatch versus GPT-OSS AMD GPU incompatibility." + }, + { + "left": "issue:45237", + "right": "issue:46468", + "accept": false, + "reason": "One is a GPT-OSS GPU/runtime bug; the other is a broken Nemotron example, so they are not the same issue." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different model architecture bug: AMD GPU failure versus GraniteMoEHybrid calling an invalid method." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch in Jamba is unrelated to a ModernBERT ONNX export error." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "Different models and code paths: ONNX export failure versus Qwen3Moe FP8 failure." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ONNX export error and Qwen3 auto-device-map cuda assert are unrelated failure modes." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "A table QA crash is not the same as outdated RAG documentation." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs cleanup, but they reference different removed topics and are not the same change." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "Model export to ONNX is unrelated to the GenerationMixin import error." + }, + { + "left": "issue:44918", + "right": "issue:46165", + "accept": false, + "reason": "TRL SFT embedding unpacking is a different training bug from resuming local model training with trust_remote_code." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask shape mismatch and Gemma4 device mismatch are different tensor-path failures." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both are IndexErrors, but one is ZeRO-3 Bert init and the other is a mask/tensor shape mismatch." + }, + { + "left": "issue:43638", + "right": "issue:46165", + "accept": false, + "reason": "ZeRO-3 Bert initialization failure is unrelated to resume-training with trust_remote_code." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Different root causes: ZeRO-3 Bert index error versus indexed tensor/mask shape mismatch." + }, + { + "left": "issue:44805", + "right": "issue:46294", + "accept": false, + "reason": "Mask/tensor shape mismatch is unrelated to Qwen3.5 video token/feature mismatch." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Mask shape mismatch does not match GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:46165", + "right": "issue:46468", + "accept": false, + "reason": "Resume-training/trust_remote_code is unrelated to a broken Nemotron example." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Video StopIteration in Qwen3.5 is a different bug from resume-training with local models." + }, + { + "left": "issue:43596", + "right": "issue:46165", + "accept": false, + "reason": "ZeRO-3 Bert index error and resume-training failure are unrelated." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice_preset and ONNX export are different subsystems and bugs." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but Jamba collective mismatch and Bert ZeRO-3 index error are different failures." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": false, + "reason": "Same broad symptom class, but different Bert variants and trigger paths make this not a duplicate." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross_entropy usage is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "DeepSeek detokenization and LoRA merge/reload output mismatch are different bugs." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen-3-8B NVFP4 runtime failure is unrelated to LoRA save/reload inconsistency." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "Activation-checkpointing/FSDP recomputation mismatch is a different distributed-training bug from a mask shape error." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Both mention DeepSpeed, but a 2-D weight runtime error is not the same as a collective mismatch." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": false, + "reason": "CPU offload incompatibility and uninitialized-memory after to_empty() are different Qwen3_5Moe rotary embedding bugs." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "Bert ZeRO-3 init failure is unrelated to GPT-OSS tensor-parallel issues." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "FSDP checkpointing mismatch is unrelated to the CustomTrainer num_items_in_batch API break." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "ZeRO-3 Bert index error and GPT-OSS tensor-parallel failure are different code paths." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Mistral-3 image dtype mismatch and GraniteMoEHybrid invalid method are unrelated model bugs." + }, + { + "left": "issue:45874", + "right": "issue:46294", + "accept": false, + "reason": "Gemma4 reverse-embedding lookup and Qwen3.5 video token mismatch are different multimodal issues." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Qwen-3-8B NVFP4 runtime failure and GPT-OSS AMD GPU failure are different models/backends." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3 vLLM config regression and Mistral tokenizer regex patch crash are unrelated." + }, + { + "left": "issue:45800", + "right": "issue:46515", + "accept": false, + "reason": "Both are compatibility complaints, but they involve different version pairs and different breakage surfaces." + }, + { + "left": "issue:45823", + "right": "issue:45874", + "accept": false, + "reason": "Both are Gemma4 issues, but device-map mismatch and expensive reverse embedding lookup are distinct bugs." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "A broad torch-version incompatibility is not the same as a Qwen-3-8B NVFP4 runtime failure." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP activation-checkpointing mismatch is unrelated to Qwen3 auto device mapping cuda assertions." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Phi MoE autocast dtype mismatch and GPT-OSS tensor-parallel failure are different issues." + }, + { + "left": "issue:43828", + "right": "issue:44292", + "accept": false, + "reason": "Different models and bugs: Phi MoE autocast mismatch versus Qwen-3-8B NVFP4 runtime failure." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Torch Dynamo grouped_mm tracing failure is unrelated to Mistral-3 image dtype mismatch." + }, + { + "left": "issue:43716", + "right": "issue:46468", + "accept": false, + "reason": "A Mistral-3 preprocessing dtype bug is not the same as a broken Nemotron example." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Mistral-3 dtype mismatch and an MPS histogram integer limitation are unrelated backend/model bugs." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "TRL SFT embedding unpacking is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:44292", + "right": "issue:44918", + "accept": false, + "reason": "Same broad family only; Qwen-3-8B NVFP4 runtime failure and Qwen3.5 TRL SFT unpacking are distinct bugs." + }, + { + "left": "issue:45507", + "right": "issue:45685", + "accept": false, + "reason": "GraniteMoEHybrid invalid method and MPS histogram not implemented are unrelated failures." + }, + { + "left": "issue:43716", + "right": "issue:44292", + "accept": false, + "reason": "Different models and code paths: Mistral-3 dtype mismatch versus Qwen-3-8B NVFP4 runtime failure." + }, + { + "left": "issue:43039", + "right": "issue:44918", + "accept": false, + "reason": "Liger Kernel cross_entropy behavior is unrelated to TRL SFT embedding unpacking." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": false, + "reason": "Both are ZeRO-3 index errors, but Gemma3 loading and non-pretrained Bert initialization are different triggers." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs issues, but they refer to different removed commands/tasks and should not be merged." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch in Phi MoE is unrelated to an MPS histogram integer limitation." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 SFT embedding unpacking is unrelated to the MPS histogram backend error." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Liger Kernel cross_entropy and Phi MoE autocast mismatch are different training-stack bugs." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and CustomTrainer kwarg incompatibility are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP activation-checkpointing mismatch is unrelated to the GenerationMixin import failure." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross_entropy and a mask shape mismatch are not the same bug." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "GPU/runtime failure in GPT-OSS is unrelated to outdated RAG documentation." + }, + { + "left": "issue:43866", + "right": "issue:46402", + "accept": false, + "reason": "A corrupted Ovis2 checkpoint and missing Sam3Model text encoder weights are different loading failures." + }, + { + "left": "issue:45800", + "right": "issue:46304", + "accept": false, + "reason": "Both are version-compatibility regressions, but they involve different releases and different distributed-config behavior." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 2, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36010", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:43039", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43792", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44387", + "issue:44512", + "issue:44560", + "issue:44589", + "issue:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45161", + "issue:45237", + "issue:45362", + "issue:45464", + "issue:45685", + "issue:45800", + "issue:45812", + "issue:45823", + "issue:45901", + "issue:45920", + "issue:46015", + "issue:46294", + "issue:46304" + ], + "soft_pairs": [ + "issue:43296|issue:46304", + "issue:41762|issue:43596", + "issue:39290|issue:41720", + "issue:44589|issue:46304", + "issue:43901|issue:44512", + "issue:45823|issue:46294", + "issue:45237|issue:45901", + "issue:44841|issue:45823", + "issue:43039|issue:44560", + "issue:43975|issue:46015", + "issue:44863|issue:46015", + "issue:45020|issue:46304", + "issue:45812|issue:45920", + "issue:43866|issue:44863", + "issue:43872|issue:46304", + "issue:44292|issue:46015", + "issue:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43792|issue:45901", + "issue:43854|issue:43866", + "issue:43425|issue:45070", + "issue:43596|issue:45800", + "issue:43975|issue:45901", + "issue:43792|issue:45070", + "issue:45362|issue:45464", + "issue:44292|issue:45901", + "issue:39290|issue:45823", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:36010|issue:39290", + "issue:44841|issue:45084", + "issue:43404|issue:43866", + "issue:45800|issue:45823", + "issue:41093|issue:41762", + "issue:43296|issue:43366", + "issue:43572|issue:45070", + "issue:44805|issue:44918", + "issue:44387|issue:45005", + "issue:43541|issue:45161", + "issue:43828|issue:45464" + ], + "prepared_review_unit_hash": "e5e190a0247b02a27048f4e1e7fc8955f5e980b354e048cf91aca413edebcfa0", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 41, + "serialized_chars": 39729, + "estimated_input_tokens": 9933, + "estimated_eval_tokens": 20122 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are only superficially similar (shared error words or broad subsystem names) and do not look like the same bug. The only strong duplicate family here is the AutoTokenizer token-ID regression, where the broader v5 issue subsumes the Granite-specific report.", + "confidence": 0.87, + "canonical_issue_reason": "issue:45920 is the broadest and most central duplicate-family candidate: it describes the same AutoTokenizer v5 token-ID regression across multiple model families, while issue:45812 is a narrower Granite-only manifestation.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the best global issue representative because it has the widest reported impact, is still open, and cleanly captures the shared tokenizer regression across several models.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:46304", + "accept": false, + "reason": "Different failure domains: vLLM/PaddleOCR-VL load error vs PyTorch distributed build/support regression." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Both are IndexError-on-ZeRO-3 reports, but the models and triggering paths differ too much to treat as one bug without stronger evidence." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Gemma3 config attribute missing in vLLM is unrelated to the Qwen3 auto-device-map CUDA assert." + }, + { + "left": "issue:44589", + "right": "issue:46304", + "accept": false, + "reason": "Float8 storage type lookup failure is unrelated to the USE_DISTRIBUTED=0 support issue." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs issues, but they cover different documentation changes and different underlying behavior." + }, + { + "left": "issue:45823", + "right": "issue:46294", + "accept": false, + "reason": "Gemma4 device-mapping mismatch and Qwen3.5 video chat-template token mismatch are different model-path problems." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "AMD/GPT-OSS backend failure is unrelated to the table-question-answering crash." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Processor failure for Voxtral is a different code path from Gemma4 PLE device mismatch." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross_entropy usage and Qwen3-vl video StopIteration are not the same underlying bug." + }, + { + "left": "issue:43975", + "right": "issue:46015", + "accept": false, + "reason": "Tokenizer detokenization regression is unrelated to outdated RAG docs." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint-loading failure is unrelated to documentation examples." + }, + { + "left": "issue:45020", + "right": "issue:46304", + "accept": false, + "reason": "Remote_code breakage and USE_DISTRIBUTED=0 support regression are different compatibility problems." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same underlying v5 AutoTokenizer regression: both report wrong token IDs across model families, with 45920 being the broader umbrella." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Corrupted Ovis2 checkpoint and NemotronH loading failure are distinct model-specific problems." + }, + { + "left": "issue:43872", + "right": "issue:46304", + "accept": false, + "reason": "bitsandbytes Int8Params constructor incompatibility is unrelated to the PyTorch distributed toggle regression." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "Qwen-3 NVFP4 runtime failure is unrelated to RAG docs updates." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention sliding_window, but they affect different model families and appear to be different bugs." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Mask-shape mismatch during generation and ZeRO-3 BertModel zero-size tensor error are not the same issue." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Whisper cannot run and table-question-answering crashes are unrelated." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "GLM-4.7-Flash unit-test load failure and Ovis2 checkpoint corruption are different failures." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch version incompatibility and a pydantic PretrainedConfig regression are separate compatibility issues." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "ZeRO-3 BertModel index error and torch 2.4.1/transformers 5.8.0 incompatibility are not the same bug." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "DeepSeek detokenization regression is unrelated to table-question-answering crashes." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper load/run failure and pydantic model regression are unrelated." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Both concern Qwen3.5 chat, but one is a model crash and the other is a streaming chat/completions API failure; not enough to merge." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "Qwen-3 NVFP4 runtime failure is unrelated to table-question-answering." + }, + { + "left": "issue:39290", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3 sliding_window_pattern error and Gemma4 PLE device mismatch are different model-path issues." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 streaming failure is unrelated to the MoE MPS histogram op limitation." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Grouped_mm tracing failure and MPS histogram_mps integer support error are distinct MoE issues." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "MoE tracing/grouped_mm failure and autocast dtype mismatch are different runtime problems." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "Importing GenerationMixin and a Gemma3 config attribute error are unrelated." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Both are IndexError-ish, but one is a mask/tensor shape mismatch and the other is TRL SFT unpacking failure; different code paths." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Int4 OOM regression and tied-weights translation-model regression are different v5 issues." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Grouped_mm tracing error in Mixtral and GPT-OSS tensor-parallel failure are different MoE bugs." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Autocast dtype mismatch on Phi-tiny-MoE and Qwen3.5 streaming API failure are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the only accepted duplicate family\u2014issue:45812 with issue:45920\u2014is well supported by the titles. The remaining soft-edge verdicts stay conservative and do not overclaim equivalence from superficial similarity alone." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 3, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36331", + "issue:39186", + "issue:39692", + "issue:40990", + "issue:41093", + "issue:41762", + "issue:42915", + "issue:43054", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43541", + "issue:43596", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43873", + "issue:43950", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44360", + "issue:44479", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45183", + "issue:45399", + "issue:45464", + "issue:45507", + "issue:45632", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45902", + "issue:46294", + "issue:46463", + "issue:46489", + "issue:46491" + ], + "soft_pairs": [ + "issue:44291|issue:45070", + "issue:43366|issue:45464", + "issue:43531|issue:45070", + "issue:45020|issue:45800", + "issue:43716|issue:46294", + "issue:43541|issue:44589", + "issue:44479|issue:46294", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45759|issue:45902", + "issue:39186|issue:44918", + "issue:43716|issue:44918", + "issue:43645|issue:45399", + "issue:40990|issue:44960", + "issue:45632|issue:45698", + "issue:43404|issue:45005", + "issue:42915|issue:45399", + "issue:44530|issue:45005", + "issue:43873|issue:45005", + "issue:44360|issue:44512", + "issue:44863|issue:45005", + "issue:45183|issue:46463", + "issue:46489|issue:46491", + "issue:44918|issue:45507", + "issue:44492|issue:44512", + "issue:43975|issue:44292", + "issue:44292|issue:45507", + "issue:43950|issue:44534", + "issue:39692|issue:43054", + "issue:43644|issue:43950", + "issue:36331|issue:41762" + ], + "prepared_review_unit_hash": "4a273d5888c6e9d01ad4d9fb4214ea7ad322b355dc824eaacd5a0caa5ac296c0", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36939, + "estimated_input_tokens": 9235, + "estimated_eval_tokens": 18726 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose grouping of unrelated Transformers bug reports. The only clear duplicate chain is the non-persistent-buffer regression in v5.x, where 43644 and 44534 appear to be the same underlying bug as 43950. Most other pairs are only superficially similar and should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "43950 is the clearest, most specific report of the duplicated v5 non-persistent-buffer corruption bug (`from_pretrained` + `register_buffer(persistent=False)`), and it matches the duplicate reports 43644 and 44534 best.", + "canonical_pr_reason": null, + "best_issue_reason": "43950 is the strongest canonical issue candidate because it describes the root regression precisely and has the most actionable reproduction context among the matching reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Different bugs: empty-init helper argument handling vs a pydantic `PretrainedConfig` field regression." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "Different functionality: GGUF architecture support vs a streaming chat/completions failure." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior is unrelated to the pydantic field regression." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Both are version-compatibility complaints, but they target different failure modes and code paths." + }, + { + "left": "issue:43716", + "right": "issue:46294", + "accept": false, + "reason": "Different multimodal bugs: image preprocessor dtype mismatch vs video feature/token mismatch." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Distinct runtime errors in different paths: torch dynamo grouped_mm tracing vs float8 storage lookup." + }, + { + "left": "issue:44479", + "right": "issue:46294", + "accept": false, + "reason": "Both mention video input, but one is a broad v5.3 regression and the other is a chat-template mismatch for one model." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are index errors, but the mask-shape mismatch and ZeRO-3 zero-size tensor issue are different bugs." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is an embedding-quality complaint and the other is an AutoModel/pipeline correctness issue." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Different underlying problems: config unwrapping vs uninitialized rotary embedding materialization." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP `'weight' must be 2-D` is unrelated to the TRL SFT embedding-unpacking failure." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Different subsystems and symptoms: multimodal preprocessing vs trainer/input unpacking." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Notebook model-definition/init regression is unrelated to flash-attn fallback gating." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "Unrelated reports: perplexity on gpt-oss-20b vs GLM5." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": false, + "reason": "Both involve custom code loading, but one is a cache-path collision and the other is loading the wrong saved module." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Tied-weights problems, but on different model families and likely different root causes." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Fine-grained FP8 model failure is unrelated to flash-attn fallback behavior." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash is not the same as the tied-weights translation-model regression." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior is a different bug from tied-weight handling." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "A DSA indexer bug and a documentation typo are not duplicates." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint-loading failure is unrelated to translation tied-weights behavior." + }, + { + "left": "issue:45183", + "right": "issue:46463", + "accept": false, + "reason": "Both are input-validation error-message issues, but one is a broad XOR-logic bug and the other is T5-specific; insufficient evidence of the same root cause." + }, + { + "left": "issue:46489", + "right": "issue:46491", + "accept": false, + "reason": "Different tokenizer regressions affecting different models and different round-trip symptoms." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Unpacking Qwen3.5 embeddings in TRL is unrelated to GraniteMoEHybrid calling an invalid method." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "A cache-strategy typo and a removed-doc-command issue are unrelated documentation bugs." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Different problems: DeepSeek-Coder detokenization regression vs Qwen NVFP4 runtime failure." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Unrelated model/runtime failures." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Same underlying v5 non-persistent-buffer corruption regression; the titles describe the same bug in different words." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "Docs example errors and a text-embedding quality regression are not the same issue." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Both report the same non-persistent-buffer junk/corruption regression in Transformers v5." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall. The only clearly supported duplicate chain is the v5 non-persistent-buffer corruption regression (43644 / 44534 / 43950), and the remaining soft-edge rejections are conservative based on the titles. Minor caveat: phrases like \"most actionable reproduction context\" are not directly supported by the packet, so keep that wording a bit softer in future summaries." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 4, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:24643", + "issue:28530", + "issue:38175", + "issue:41553", + "issue:41762", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43493", + "issue:43643", + "issue:43742", + "issue:43782", + "issue:43824", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44368", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44589", + "issue:44661", + "issue:44910", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45356", + "issue:45399", + "issue:45520", + "issue:45663", + "issue:45685", + "issue:45701", + "issue:45715", + "issue:45823", + "issue:45859", + "issue:45902", + "issue:45920", + "issue:46315", + "issue:46489" + ], + "soft_pairs": [ + "issue:43828|issue:44928", + "issue:43872|issue:44589", + "issue:45920|issue:46489", + "issue:43866|issue:44403", + "issue:43054|issue:45701", + "issue:43950|issue:45902", + "issue:38175|issue:43054", + "issue:41553|issue:43883", + "issue:44910|issue:45859", + "issue:42947|issue:44387", + "issue:45663|issue:45902", + "issue:43054|issue:43493", + "issue:41553|issue:43881", + "issue:44403|issue:44451", + "issue:24643|issue:28530", + "issue:43975|issue:45520", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:43643|issue:45020", + "issue:41762|issue:42915", + "issue:45685|issue:45823", + "issue:44991|issue:45399", + "issue:43782|issue:44960", + "issue:43824|issue:43872", + "issue:43881|issue:43883", + "issue:45020|issue:45823", + "issue:44661|issue:45020", + "issue:45081|issue:45399", + "issue:43828|issue:45823", + "issue:42491|issue:44368", + "issue:45715|issue:46315", + "issue:43742|issue:45399", + "issue:45020|issue:45356" + ], + "prepared_review_unit_hash": "84d287f55ee3beb1e96fc6ab23d327641f9abaea3682f0b32f9bf76726acac7b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37920, + "estimated_input_tokens": 9480, + "estimated_eval_tokens": 19216 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose collection of mostly unrelated Transformers bug reports (tokenizers, remote code, MoE/FSDP/precision, model-loading, and performance regressions). The soft pairs are thematically similar but do not look like the same underlying bug or change, so I would not merge any of them as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": "Issue 45020 is the broadest umbrella-style report in the set, centered on recent Transformers version regressions affecting remote-code model loading, which makes it the best single issue to anchor the cluster even though the cluster is not truly one duplicate group.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45020 is the most representative/high-level issue in this mixed set; it is broader and more central than the narrowly scoped model-specific regressions, so it is the best issue to keep as the cluster anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Both mention Qwen3.5/position_ids and dtype/attention failures, but one is a Flash Attention illegal-access crash and the other is an RLHF NaN/SDPA-BF16 regression; different failure modes and code paths." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Both are quantization/storage compatibility errors, but they involve different concrete breakages: bitsandbytes Int8Params constructor mismatch vs missing Float8 storage lookup." + }, + { + "left": "issue:45920", + "right": "issue:46489", + "accept": false, + "reason": "Both concern tokenizer regressions, but one is a broad wrong-token-IDs issue across many models while the other is a DeepSeek-Coder v1-specific output bug tied to a particular prior PR gap." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "One is a corrupted checkpoint/model artifact problem; the other is a generic logging/noise issue when loading transformers. They are unrelated." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "Both discuss tokenization/model quality changes across versions, but one is a SigLIP2 text-embedding quality complaint and the other is a tokenizer-version behavior change; not the same bug." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "One is a from_pretrained buffer persistence corruption regression; the other is uninitialized memory in a specific MoE rotary embedding after meta-to-empty materialization. Different mechanisms." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Both are about SigLIP/SigLIP2 outputs, but one reports zero probabilities while the other reports worse text embeddings and a JAX/HF discrepancy; not the same underlying defect." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Both are loading-time errors, but one is an AutoTokenizer error for Voxtral and the other is a missing attribute on MolmoForCausalLM. Different models and different faults." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Both involve Qwen3.5 and rotary/position handling, but one is a Flash Attention crash from 3D position_ids and the other is CPU offload incompatibility in a MoE rotary embedding. Not mergeable as one bug." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory/gradient problems with PEFT/LoRA or quantization, but one is gradient checkpointing not taking effect and the other is increased CUDA reserved memory leading to OOM. Different root causes." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Gemma4 FSDP2 KeyError from per-layer shared_kv_states rebuilding is unrelated to Qwen3_5Moe rotary embedding uninitialized memory." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "Both are about SigLIP2 discrepancies, but one is a subjective embedding quality regression and the other is a specific HF-vs-JAX implementation mismatch. Related area, not duplicate." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Voxtral AutoTokenizer loading and glm-4v-9b loading failures are separate model-specific issues with different error paths." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise versus failure to load a specific model (ScandiBERT). Same broad subsystem, not the same bug." + }, + { + "left": "issue:24643", + "right": "issue:28530", + "accept": false, + "reason": "DeepSpeed weight shape error during training is unrelated to the early-stopping metric-for-best-model warning/disablement issue." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "One is incorrect detokenization for a specific model; the other is a Python 3.13 import_utils KeyError for flash_attn. Different components." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Both involve loading/reporting issues, but one is generic load noise and the other is a specific tokenizer/model load failure for cjvt/sleng-bert." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Both are training/memory related, but one is checkpointing not working with LoRA and the other is inefficient memory usage in Qwen3 MoE training. Different fixes." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Both touch config/loading regressions, but one is missing fields from AutoConfig with trust_remote_code and the other is a broad remote_code compatibility issue. Too broad to call duplicate." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading IndexError and Qwen3Moe FineGrainedFP8Config failure are distinct model/config-specific problems." + }, + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "MPS histogram_int error in MoE and Gemma4 PLE device mismatch under device_map=auto are unrelated backend/device-placement bugs." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading failure for a specific model and fallback-to-flash-attn2 blocking are different code paths and symptoms." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only load error and a GLM5 issue are unrelated model-loading failures." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Importing Qwen2.5-VL class and bitsandbytes Int8Params constructor mismatch are different failures with no shared fix path." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Both are model-loading errors, but one is glm-4v-9b loading failure and the other is a missing all_tied_weights_keys attribute on MolmoForCausalLM." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "Remote-code regressions and Gemma4 PLE device-mismatch are not the same bug; they only share a recent-version symptom surface." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Tokenizer mapping failure when adding a new model-like entry is a narrow registry bug, not the broader remote_code loading regression." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "One is a tokenizer regex patch crash with a backend_tokenizer attribute error; the other is a flash-attn2 fallback gating issue. Different fixes." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Both involve model execution on specialized hardware paths, but one is an autocast dtype mismatch in Phi MoE and the other is a Gemma4 device-map placement issue." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "LoRA compatibility across hf4.x/hf5.x for Qwen3 MoE and a warning about tie_word_embeddings in Qwen3.5-27B are not the same underlying bug." + }, + { + "left": "issue:45715", + "right": "issue:46315", + "accept": false, + "reason": "Both are tokenizer-performance regressions, but one is convert_ids_to_tokens rebuilding all_special_ids per iteration and the other is slow convert_tokens_to_ids for slow tokenizers with many added tokens." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "Key error loading a specific model and flash-attn2 fallback blocked by other checks are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Both are version regressions around loading/tokenization, but one is a broad remote_code breakage and the other is a Kimi-K2.5 tokenizer codec/fix_mistral_regex regression." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 5, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43792", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43872", + "issue:43940", + "issue:43950", + "issue:44155", + "issue:44220", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44560", + "issue:44589", + "issue:44683", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44991", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45488", + "issue:45584", + "issue:45588", + "issue:45663", + "issue:45823", + "issue:45902", + "issue:45923", + "issue:46402", + "issue:46489", + "issue:46515" + ], + "soft_pairs": [ + "issue:45488|issue:46489", + "issue:43827|issue:44509", + "issue:43404|issue:43950", + "issue:43872|issue:44291", + "issue:42915|issue:43872", + "issue:43792|issue:44220", + "issue:36331|issue:41093", + "issue:42947|issue:45923", + "issue:43257|issue:45305", + "issue:41553|issue:44991", + "issue:43940|issue:45313", + "issue:43854|issue:44863", + "issue:44928|issue:45588", + "issue:41093|issue:44805", + "issue:44479|issue:44560", + "issue:41553|issue:44488", + "issue:24643|issue:30064", + "issue:44863|issue:46402", + "issue:43596|issue:46515", + "issue:44530|issue:44863", + "issue:43366|issue:43828", + "issue:43638|issue:46515", + "issue:45325|issue:45902", + "issue:43366|issue:43531", + "issue:45663|issue:45823", + "issue:43716|issue:44560", + "issue:44155|issue:45584", + "issue:43366|issue:44589", + "issue:44683|issue:46515" + ], + "prepared_review_unit_hash": "dd81da81ec84b85512966db22ccd9f8e38b566307ccfa1caa8eae0730922f49b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36752, + "estimated_input_tokens": 9188, + "estimated_eval_tokens": 18632 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly heterogeneous issue set. The only strongly supported duplicate pairs are the `_is_hf_initialized` / bitsandbytes loading regression and the repeated mask-shape IndexError. The other same-theme pairs look related at the title level but the packet does not establish a shared underlying defect strongly enough to merge them.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 44291 is the clearest representative bug report: it names the concrete `init_empty_weights` / `_is_hf_initialized` incompatibility and describes the loading failure directly.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44291 is the most actionable and self-contained issue in the cluster, and it best captures the core loading-regression theme.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45488", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer-related, but they target different model families and the packet does not show the same root cause." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": false, + "reason": "This is a plausible docs-theme match, but the packet only supports a broad documentation similarity, not a confirmed duplicate defect." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Different bugs: untied `lm_head` in Mistral3 versus a `from_pretrained()` non-persistent buffer regression." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Same `_is_hf_initialized` / `init_empty_weights` loading incompatibility, with matching failure mode and context." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Both involve loading behavior, but the Qwen3MoE FP8 case is a different code path from the bitsandbytes `_is_hf_initialized` bug." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Whisper audio issues are related by theme, but the packet does not clearly establish the same underlying feature-extraction defect." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "A trainer-signature error and a mask/tensor shape mismatch are distinct failures." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is a LoRA behavior issue and the other is a missing flag on a trust_remote_code model." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Same DeepSpeed/Qwen ecosystem, but the defects differ: weight conversion/loading versus gradient averaging with GAS." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Both are model/tokenizer loading problems, but the packet does not show the same underlying cause." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 loading failures, but they affect different Qwen variants and likely different missing-module paths." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Different architectures and different load failures; no strong evidence of one bug." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "The Qwen3.5 RLHF/position-id issue and the flash-attention `s_aux=None` crash are unrelated code-path problems." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "Same mask-vs-tensor length mismatch IndexError, just reported with different concrete sizes." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "The titles suggest a common Qwen video regression, but the packet does not clearly prove the same defect." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Both concern loading behavior, but one is about a Voxtral error message and the other about a specific model load failure." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "DeepSpeed training shape error and void-segmentation-map processing are unrelated." + }, + { + "left": "issue:44863", + "right": "issue:46402", + "accept": false, + "reason": "Different model-loading failures on unrelated architectures." + }, + { + "left": "issue:43596", + "right": "issue:46515", + "accept": false, + "reason": "A BERT ZeRO-3 out-of-bounds error is not the same as a torch 2.4.1 incompatibility." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "PagedAttention `linear_attention` crash and NemotronH checkpoint loading are different defects." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "GGUF gpt-oss support and Phi-tiny-MoE autocast dtype mismatch are unrelated." + }, + { + "left": "issue:43638", + "right": "issue:46515", + "accept": false, + "reason": "Non-pretrained BERT ZeRO-3 out-of-bounds and torch-version incompatibility are separate regressions." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Both involve Qwen rotary/video-ish internals, but the packet does not establish the same failure mechanism." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "GGUF support for gpt-oss is unrelated to the Qwen3-MoE `sliding_window` issue." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Gemma4 FSDP2 shared-state rebuild and PLE device-map mismatch are different bugs." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Mistral-3 image-preprocessor dtype mismatch is not the same as the Qwen3-VL video `StopIteration` regression." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "AudioFlamingo cross-track leakage and Whisper empty-transcription generation are unrelated." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "GGUF gpt-oss support and Float8 storage lookup failure are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:46515", + "accept": false, + "reason": "Compiled flex_attention on torch >=2.9 is a separate torch-compatibility issue." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary is mostly grounded, but a few soft-edge accepts are stronger than the packet supports. In particular, the Whisper pair and the Qwen video pair look related by title, but the packet does not clearly establish the same underlying defect, so calling them duplicates is a bit too assertive. The `_is_hf_initialized` and mask-shape pairs are well supported; the rest should be kept more tentative." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 6, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43720", + "issue:43824", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44560", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:44991", + "issue:45030", + "issue:45276", + "issue:45313", + "issue:45357", + "issue:45507", + "issue:45685", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:46459", + "issue:46468", + "issue:46489", + "issue:46512", + "issue:46515" + ], + "soft_pairs": [ + "issue:43950|issue:45834", + "issue:44560|issue:45685", + "issue:43824|issue:43957", + "issue:42915|issue:43957", + "issue:43645|issue:43646", + "issue:44560|issue:44918", + "issue:43550|issue:44910", + "issue:42617|issue:43366", + "issue:45823|issue:46515", + "issue:36010|issue:42915", + "issue:45812|issue:46489", + "issue:44661|issue:45823", + "issue:44560|issue:44805", + "issue:38175|issue:43994", + "issue:43632|issue:45834", + "issue:44877|issue:45030", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:44661|issue:44805", + "issue:43317|issue:44756", + "issue:38175|issue:39692", + "issue:43644|issue:44534", + "issue:42491|issue:43720", + "issue:41720|issue:42915", + "issue:42915|issue:44756", + "issue:46459|issue:46512", + "issue:44451|issue:44488", + "issue:42491|issue:45276", + "issue:44960|issue:44977", + "issue:45313|issue:45357", + "issue:45507|issue:46468", + "issue:43644|issue:43645", + "issue:44488|issue:44991" + ], + "prepared_review_unit_hash": "172317dfc0020cba52e95b8a415baeaac44d8b2aadb6a3cfaccdc9b2afdaabee", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37432, + "estimated_input_tokens": 9358, + "estimated_eval_tokens": 18972 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are distinct regressions, with only a few clear duplicate pairs around custom model initialization, SigLIP2 output behavior, and non-persistent buffer handling.", + "confidence": 0.79, + "canonical_issue_reason": "issue:43644 is the strongest canonical anchor: it has a precise regression title and a confirmed near-exact duplicate (issue:44534) describing the same non-persistent buffer corruption bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43644 is the best overall issue candidate because it is narrowly scoped, clearly described, and directly represents a duplicated transformers v5 regression.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Different failures: buffer corruption during loading vs. Kosmos2.5 OCR index error." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Unrelated models and code paths: Qwen3-VL video embedding error vs. MoE MPS histogram issue." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Import/export failure for a missing class is not the same bug as meta-device loading breakage." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "FineGrainedFP8Config failure and meta-device loading failure are different root problems." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Both report Transformers 5.0 breaking custom model initialization; 43645 is the notebook-specific manifestation of the broader regression." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Different Qwen variants and different failure modes; no concrete shared bug is evident." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different models and distinct attention/compile code paths; not the same concrete bug." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py execution failure is unrelated to GGUF gpt-oss architecture support." + }, + { + "left": "issue:45823", + "right": "issue:46515", + "accept": false, + "reason": "Device-mismatch forward bug is unrelated to a torch version incompatibility report." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "ImportError for GenerationMixin is unrelated to Qwen3Moe FP8 loading." + }, + { + "left": "issue:45812", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer issues, but they affect different models and likely different loader paths." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "Model-list generation bug and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Video embedding StopIteration and mask-shape IndexError are unrelated failures." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Both describe SigLIP2 producing bad outputs; 43994 is the broader AutoModel/pipeline symptom of the same regression." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "_is_hf_initialized flag breakage is unrelated to Kosmos2.5 long-input indexing." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Both mention stricter config validation, but the model/config failures are not concrete duplicates." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Zero3 BertModel indexing failure is unrelated to tokenizer-mapping model addition logic." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Zero3 init BertModel issue is unrelated to tokenizer mapping failures." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping bug and tensor mask shape mismatch are different bugs." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Dequantized model offload failure and mmap OOM on Strix Halo are different loading problems." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "Same model family, but the docs/example issues are not the same concrete runtime bug." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "These are the same non-persistent buffer regression, described almost verbatim." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Qwen3 MoE LoRA compatibility and BitNet packed-weight loading are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Auto device mapping cuda assert and FineGrainedFP8Config failure are different bugs." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3Moe FP8 failure is unrelated to mmap/OOM behavior." + }, + { + "left": "issue:46459", + "right": "issue:46512", + "accept": false, + "reason": "Both involve dtype handling, but one is checkpoint loading and the other is from_config/config state; not the same bug." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Both fail to load different models, but there is no clear evidence of a shared concrete root cause." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "LoRA compatibility on qwen3_moe is unrelated to gemma4 token embedding resizing." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 and Qwen3.5 flash-attention failures are distinct model-specific issues." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "ZeRO-3 weight loading and save_pretrained visual-key regression are different paths and bugs." + }, + { + "left": "issue:45507", + "right": "issue:46468", + "accept": false, + "reason": "GraniteMoEHybrid invalid method call and Nemotron example breakage are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "They are related to Transformers 5 custom initialization, but 43645 is a notebook-specific symptom rather than the same exact bug." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-load failures, but on different models with no clear shared code-path evidence." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Overall grounded and conservative. The selected anchor issue and the accepted duplicate pairs are supported by the titles and similarity in the packet. Minor caution: phrases like \"confirmed near-exact duplicate\" are a bit stronger than the packet explicitly proves, but the inference is still reasonable from the provided evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 7, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42947", + "issue:43257", + "issue:43335", + "issue:43381", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43708", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44488", + "issue:44756", + "issue:44863", + "issue:44910", + "issue:44912", + "issue:44960", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45588", + "issue:45663", + "issue:45685", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:46050", + "issue:46468" + ], + "soft_pairs": [ + "issue:43708|issue:44075", + "issue:43856|issue:43873", + "issue:39692|issue:43994", + "issue:43854|issue:43931", + "issue:43856|issue:44387", + "issue:43257|issue:43854", + "issue:42491|issue:45292", + "issue:42491|issue:43931", + "issue:44960|issue:46050", + "issue:45488|issue:45812", + "issue:45362|issue:45406", + "issue:44164|issue:44291", + "issue:45276|issue:45335", + "issue:44910|issue:45588", + "issue:43425|issue:43957", + "issue:43404|issue:43454", + "issue:44402|issue:44488", + "issue:41720|issue:44155", + "issue:39692|issue:44960", + "issue:42947|issue:45704", + "issue:43873|issue:44387", + "issue:45663|issue:45759", + "issue:44291|issue:44756", + "issue:36010|issue:39404", + "issue:45663|issue:45859", + "issue:44387|issue:44756", + "issue:44292|issue:44912", + "issue:44155|issue:45823", + "issue:43957|issue:44756", + "issue:43335|issue:45663", + "issue:42947|issue:43381", + "issue:44863|issue:46468", + "issue:45685|issue:46468", + "issue:45276|issue:45292", + "issue:44164|issue:44756" + ], + "prepared_review_unit_hash": "e3f4480402b4098cfbb1ec97bfa3645b6e55eb299ce39449f38d4ff3e0c4fc05", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38254, + "estimated_input_tokens": 9564, + "estimated_eval_tokens": 19384 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated model- or subsystem-specific bug reports. The only plausible duplicate pair is the Gemma4/general `resize_token_embeddings` regression cluster; the rest should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 45292 is the broadest representative of the resize-token-embeddings regression: it states the core failure on `output_embeddings`, while the Gemma4 report is a narrower variant affecting additional tied embedding modules.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45292 is the best global issue candidate because it captures the underlying shared symptom in the most general form and is more likely to aggregate related reports than the model-specific variant.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Different training-time bugs: checkpoint resumption/max-steps calculation vs SGD optimizer args not being applied." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training memory usage and the other is quantization/offloading behavior; not the same code-path bug." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is a broken doc example and the other is wrong inference results from AutoModel/pipeline; different underlying failures." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different models and different load failures; no evidence they share the same root cause." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both are memory-related, but one is Qwen3 MoE training and the other is int4 quantization OOM/regression; too broad to merge." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Qwen3 MoE accelerate/deepspeed conversion problem vs GLM-4.7-Flash load issue; unrelated." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "LoRA qwen3_moe hf4.x vs hf5.x compatibility issue is not the same as resize_token_embeddings failing to update embeddings." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "Different model/version-loading problems with no shared concrete code path." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "Both concern model/quantization behavior, but one is a GLM5 issue and the other is save/load of sinq-quantized models; distinct bugs." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but one is a pre-tokenizer override issue for DeepSeek and the other is wrong token IDs for Granite; different failure modes." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Qwen3.5 chat crash and Gemma4Processor `_tokenizer` attribute error are unrelated." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Both involve model loading/saving internals, but extra_state handling and `_is_hf_initialized` arg errors are separate regressions." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Very similar symptom, but they affect different model implementations (`gemma4` vs `t5gemma`) and do not clearly share one concrete fix." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Both touch flash attention, but one is packed-sequence/position_ids misinterpretation and the other is `s_aux=None` attribute handling; different bugs." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch version incompatibility and meta-device model loading failure are separate issues." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both are untied `lm_head` weights, but in different model classes/code paths; not enough to merge as one bug." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Both relate to tokenizer/vocab inconsistencies, but the affected models and reported failures differ." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Qwen3 device-map CUDA assert vs AudioFlamingo3 track leakage are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Completely different models and failure types: SigLIP2 docs/load issues vs GLM5/quantization." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the T5 apex RMSNorm memory leak." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Both mention quantization/memory, but one is offloading not working and the other is reserved-memory growth leading to OOM; too generic to treat as duplicates." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Gemma4 FSDP2 attention state bug vs Qwen 3.5/3.6 composite config unwrapping bug are different code paths." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "`init_empty_weights` loader arg handling and Strix Halo mmap OOM are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "ImportError for `GenerationMixin` vs Whisper pipeline `return_language` regression are separate issues." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Gemma4 FSDP2 attention-state bug vs Qwen3_5Moe rotary embedding CPU-offload incompatibility are unrelated." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Different memory/OOM reports with no shared concrete mechanism." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Different quantization backends and models; not the same underlying bug." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "Audio token leakage between tracks and Gemma4 device-map mismatch are unrelated." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading regression is not the same as mmap causing OOM on Strix Halo." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "Sparse-layer creation bug in SwitchTransformers and Gemma4 FSDP2 attention bug are unrelated." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Gradient checkpointing in PEFT LoRA vs eval-mode checkpointing restriction are different issues." + }, + { + "left": "issue:44863", + "right": "issue:46468", + "accept": false, + "reason": "Both are Nemotron-related, but one is checkpoint loading and the other is an example-script breakage; too little evidence of a shared root cause." + }, + { + "left": "issue:45685", + "right": "issue:46468", + "accept": false, + "reason": "MOE MPS histogram dtype error and the Nemotron example issue are unrelated." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same underlying resize-token-embeddings regression: the Gemma4 report is a narrower instance of the broader `output_embeddings` failure." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "Extra_state save/load bug and Strix Halo mmap OOM are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The clustering is well-grounded in the packet. The 45276/45292 pair is the only clearly mergeable duplicate, and the rest of the soft-edge rejections stay conservative given the titles/body summaries. No overstatement of evidence stands out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 8, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36032", + "issue:39404", + "issue:42796", + "issue:43381", + "issue:43425", + "issue:43452", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43854", + "issue:43901", + "issue:43906", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44265", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44928", + "issue:45084", + "issue:45216", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45701", + "issue:45812", + "issue:45823", + "issue:46315", + "issue:46326", + "issue:46491" + ], + "soft_pairs": [ + "issue:44479|issue:44877", + "issue:43854|issue:45216", + "issue:45684|issue:45698", + "issue:43381|issue:44928", + "issue:45588|issue:45663", + "issue:42796|issue:46326", + "issue:43646|issue:43708", + "issue:43645|issue:44164", + "issue:45292|issue:45335", + "issue:43828|issue:43957", + "issue:43425|issue:43828", + "issue:44402|issue:44451", + "issue:44589|issue:44756", + "issue:45812|issue:46315", + "issue:46315|issue:46491", + "issue:44683|issue:45823", + "issue:39404|issue:45701", + "issue:43611|issue:43645", + "issue:43425|issue:45823", + "issue:43425|issue:43716", + "issue:44530|issue:45084", + "issue:43906|issue:44530", + "issue:44479|issue:44805", + "issue:44589|issue:45084", + "issue:44265|issue:44898", + "issue:43901|issue:44509", + "issue:43425|issue:44918", + "issue:39404|issue:44912", + "issue:43994|issue:44079", + "issue:36032|issue:43452" + ], + "prepared_review_unit_hash": "236d5d056b1c4b94170a89cc7e2733c2880728c12afa66db12cc461b0290163e", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36864, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a set of distinct Transformers v5 regressions with broad tokenizer/model-loading overlap, but only one clear duplicate pair emerged: the resize_token_embeddings embedding-update bug. The rest share subsystems or release context, not the same underlying defect.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45292 is the broader, model-agnostic report of the resize_token_embeddings regression; issue:45335 is just the t5gemma-specific manifestation of the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45292 best represents the duplicate cluster because it states the underlying embedding-resize failure in the most general form and would attract the widest set of affected reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Both involve model input behavior, but one is a video-regression report and the other is a generic mask-shape index error; different bugs." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Unrelated failures: GLM-4.7 unit-test loading issue versus a Qwen3.5 save_pretrained checkpoint regression." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Both touch save/from_pretrained for custom models, but one is about read-only source-file permissions and the other about loading the wrong custom module; not the same defect." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Eval-mode gradient checkpointing incompatibility is unrelated to the Qwen3.5 RLHF NaN caused by 3D position_ids/SDPA fallback." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Different code paths: flash_attention sink handling versus Gemma4 FSDP2 shared_kv_states rebuilding." + }, + { + "left": "issue:42796", + "right": "issue:46326", + "accept": false, + "reason": "Both involve tokenizer behavior, but one is BPE marker handling and the other is continue_final_message logic for mistral_common; not the same bug." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization in notebooks is unrelated to incorrect max_steps calculation when resuming Trainer with changed batch size." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Notebook custom-model initialization and extra_state save/load failures are separate issues." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same underlying bug: resize_token_embeddings fails to propagate updated embeddings; 45335 is the t5gemma-specific instance." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Both are dtype/loading-related, but one is an autocast mismatch in Phi-tiny-MoE and the other is a torch.device('meta') loading regression affecting different models." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch version incompatibility is not the same as the Phi-tiny-MoE autocast dtype mismatch." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer vocab-size inconsistency and inability to load ScandiBERT may be related in theme, but the reports do not point to the same concrete failure." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Float8 storage lookup failure is unrelated to the Strix Halo mmap/OOM workaround." + }, + { + "left": "issue:45812", + "right": "issue:46315", + "accept": false, + "reason": "Wrong token IDs for Granite models is a correctness regression, while convert_tokens_to_ids slowdown is a performance regression." + }, + { + "left": "issue:46315", + "right": "issue:46491", + "accept": false, + "reason": "Performance regression in convert_tokens_to_ids and CodeLlama leading-space round-trip bug are different tokenizer issues." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device_map auto mismatch are separate execution-path problems." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper pipeline return_language regression is unrelated to a general tokenization change report." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Base_model_prefix loading breakage and notebook custom-model initialization are both v5 custom-model concerns, but they are not the same failure mode." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "Torch 2.10 incompatibility is broader and distinct from the Gemma4 device mismatch during forward." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Global Torch compatibility issue does not match the Mistral-3 image preprocessor dtype mismatch." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and non-template-node compile error are unrelated compiler/cache failures." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "An isolated reproduction of an earlier issue is not the same as the Qwen3.5 PagedAttentionCache crash." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Video input regression and mask-shape mismatch are different symptoms with different triggers." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup failure is unrelated to compilation of non-template nodes." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export failure with torch_compilable_check is a different export-path bug than Perceiver non-default-resolution handling." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "One is a docs mismatch for return_all_scores; the other is broader docs still listing removed tasks." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch compatibility is unrelated to the Qwen3.5 embedding unpacking problem in trl SFTTrainer." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper return_language pipeline breakage and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "SigLIP2 nonsensical outputs and ModelOutput key assignment are distinct bugs." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5Tokenizer add_special_tokens conflict is unrelated to gguf_file breaking AutoTokenizer/AutoModel loading." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded in the packet. The one accepted duplicate pair, issue:45292 and issue:45335, is supported by the titles as the same resize_token_embeddings bug in a general vs t5gemma-specific form. The remaining soft-edge rejections are conservative and do not overclaim shared causality." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 9, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:32090", + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42175", + "issue:42796", + "issue:42915", + "issue:43299", + "issue:43404", + "issue:43421", + "issue:43541", + "issue:43582", + "issue:43596", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43646", + "issue:43716", + "issue:43856", + "issue:43906", + "issue:43931", + "issue:44164", + "issue:44451", + "issue:44462", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44661", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45759", + "issue:45859", + "issue:46050", + "issue:46178", + "issue:46285", + "issue:46326", + "issue:46402", + "issue:46489" + ], + "soft_pairs": [ + "issue:43299|issue:43931", + "issue:44671|issue:44991", + "issue:43596|issue:44683", + "issue:44164|issue:45759", + "issue:46050|issue:46326", + "issue:43906|issue:46326", + "issue:44991|issue:45081", + "issue:43856|issue:44756", + "issue:44164|issue:44479", + "issue:44488|issue:45081", + "issue:42796|issue:43404", + "issue:44661|issue:44683", + "issue:39404|issue:44671", + "issue:44488|issue:44987", + "issue:44462|issue:46489", + "issue:43611|issue:43644", + "issue:32090|issue:39404", + "issue:44737|issue:45859", + "issue:43716|issue:45084", + "issue:46285|issue:46402", + "issue:43582|issue:43632", + "issue:36032|issue:42915", + "issue:43643|issue:44530", + "issue:44451|issue:44987", + "issue:40990|issue:43421", + "issue:43541|issue:45084", + "issue:34567|issue:40444", + "issue:42175|issue:46178", + "issue:42915|issue:44530", + "issue:43644|issue:43646" + ], + "prepared_review_unit_hash": "8665c2fd074987e43281b044a1877835c3167cc7949a50567fb1290bda50dbd7", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36154, + "estimated_input_tokens": 9039, + "estimated_eval_tokens": 18334 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of Transformers v5 regression reports, with many model/tokenizer loading failures but no clear duplicate pair strong enough to merge. I found no PRs. The best representative issue is the broad model-loading regression around `base_model_prefix`.", + "confidence": 0.63, + "canonical_issue_reason": "Issue 43611 is the most representative broad v5 regression report here: it targets a core model-loading path and matches the cluster\u2019s general theme better than the more model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43611 is the strongest single issue to anchor the cluster because it is generic, clearly regression-focused, and covers a central load-path failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both are Qwen3-VL loading problems, but one is a generic MoE loading regression and the other is a specific weight-shape mismatch; not enough evidence they share the same concrete bug." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer/model regressions, but they affect different models and symptoms (CamemBERT predictions vs est-roberta tokenizer loading)." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Completely different failures: DeepSpeed/BertModel zero-sized index error vs compiled flex_attention on torch>=2.9." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "Different code paths: save/from_pretrained extra_state handling vs AutoModelForCausalLM.from_config text_config unwrapping." + }, + { + "left": "issue:46050", + "right": "issue:46326", + "accept": false, + "reason": "Unrelated issues: quantized save/load breakage vs mistral_common continue_final_message behavior." + }, + { + "left": "issue:43906", + "right": "issue:46326", + "accept": false, + "reason": "The first is a reproduction of an older issue; the second is a tokenizer prompt-format bug. No shared concrete bug." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Both involve tokenizer loading, but the failures are different models and different code paths (repo tokenizer load vs Mistral regex patch crash)." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Different memory issues in unrelated contexts: Qwen3 MoE training vs Strix Halo mmap/OOM." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "Different regressions: serialization extra_state vs video-input handling for Qwen VL models." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Different tokenizer/model loading bugs; one is loading a specific BERT repo, the other is a Mistral regex patch crash." + }, + { + "left": "issue:42796", + "right": "issue:43404", + "accept": false, + "reason": "Tokenizer BPE-marker handling and Mistral3 lm_head tying are unrelated bugs." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Unrelated areas: tokenizer mapping support vs flex_attention compilation." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline language return bug vs CamemBERT masked-LM regression; no shared code path." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Both are model-loading complaints, but the affected models and underlying causes differ." + }, + { + "left": "issue:44462", + "right": "issue:46489", + "accept": false, + "reason": "Both mention tokenizer issues, but one is repo tokenizer.json selection and the other is DeepSeek-Coder output correctness; not the same bug." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Related v5 regression theme, but one is base_model_prefix loading and the other is non-persistent buffer initialization; too different to merge." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast TypeError and Whisper pipeline return_language regression are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "Both are device/offload-related, but they hit different models and different mechanisms (XLNet CPU arange vs Qwen3.5 MoE rotary embedding offload)." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Different problem classes: image preprocessor dtype mismatch vs Jinja/template compilation error." + }, + { + "left": "issue:46285", + "right": "issue:46402", + "accept": false, + "reason": "Both are model-loading structure issues, but one is CLIP class inconsistency and the other is Sam3 text encoder weight loading." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Unrelated bugs: Apple Silicon allocator warmup TypeError vs `_is_hf_initialized` flag regression." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "Tokenizer method-name conflict and Qwen3Moe FP8 failure are unrelated." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "Different bugs: remote-code config field loss vs Qwen3.5 cache/group-type crash." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Both are load failures, but the concrete failures and model families differ." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "Perplexity regression on gpt-oss is unrelated to runtime special-token post-processor updates." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Torch Dynamo grouped_mm tracing error and template-node compilation error are unrelated." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "TrainerState token counting and multimodal IterableDataset finetuning failures are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:46178", + "accept": false, + "reason": "Both touch backend packaging, but one is missing tensorflow in a torch extra and the other is an unsupported tensorflow_text backend mapping entry." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Both involve Qwen models, but one is FP8 configuration and the other is PagedAttentionCache group-type handling." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "These are both v5 initialization regressions, but one is about non-persistent buffers and the other is a broader custom model-init breakage; not clearly the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 10, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28530", + "issue:34567", + "issue:36032", + "issue:40444", + "issue:42915", + "issue:43089", + "issue:43262", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43452", + "issue:43482", + "issue:43540", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43746", + "issue:43819", + "issue:43883", + "issue:43994", + "issue:44062", + "issue:44206", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45507", + "issue:45584", + "issue:45987", + "issue:46326" + ], + "soft_pairs": [ + "issue:44811|issue:45584", + "issue:44206|issue:44220", + "issue:43883|issue:44488", + "issue:40444|issue:46326", + "issue:36032|issue:43645", + "issue:43452|issue:43482", + "issue:43317|issue:43482", + "issue:45084|issue:45507", + "issue:44987|issue:44991", + "issue:44464|issue:45987", + "issue:43262|issue:43377", + "issue:43646|issue:44534", + "issue:43645|issue:44534", + "issue:43746|issue:46326", + "issue:43883|issue:44991", + "issue:44403|issue:45081", + "issue:42915|issue:45084", + "issue:43404|issue:44466", + "issue:44849|issue:45325", + "issue:43994|issue:44610", + "issue:28530|issue:34567", + "issue:43377|issue:43819", + "issue:43322|issue:43540", + "issue:44451|issue:45081", + "issue:44987|issue:45081", + "issue:42915|issue:44291", + "issue:43089|issue:45987", + "issue:44265|issue:44610", + "issue:44220|issue:44683", + "issue:36032|issue:43742", + "issue:44062|issue:44987", + "issue:42915|issue:44589", + "issue:45216|issue:45357" + ], + "prepared_review_unit_hash": "3446e2a41de625b731af3cdf0fed298e7f419e58530940d0e1d37a69611aeef6", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37690, + "estimated_input_tokens": 9423, + "estimated_eval_tokens": 19102 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous issue reports, not a tight duplicate cluster. The only strong duplicate-like pair is the Qwen3.5 save_pretrained regression (45216/45357); the rest share broad themes but look like different bugs or different code paths.", + "confidence": 0.41, + "canonical_issue_reason": "Issue 45216 is the broader, earlier report of the Qwen3.5 save_pretrained regression, and 45357 appears to be a later follow-up that the same visual-encoder-key problem still exists in 5.5.3.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45216 is the best canonical issue among the listed items because it most clearly captures the underlying Qwen3.5 save_pretrained regression and can absorb the later duplicate-like follow-up.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both involve Whisper/token handling, but one is about batch_decode skip_special_tokens and the other is an empty-transcription failure after align_special_tokens; different bug surface." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Both are audio feature-extraction regressions, but they hit different functions and failure modes (_center arg vs _torch_extract_fbank_features)." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Both are loading failures, but they concern different models and different missing-key/loader behavior." + }, + { + "left": "issue:40444", + "right": "issue:46326", + "accept": false, + "reason": "Both touch multimodal/chat templating, but one is about multi-image IterableDataset finetuning and the other is about continue_final_message with mistral_common tokenizers." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "One is a T5 tokenizer name conflict; the other is a Transformers 5 custom-model/Jupyter initialization break. Different issues." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": false, + "reason": "Both mention GGUF/loading, but one is a generic gguf_file from_pretrained regression and the other is a specific Qwen2.5-GGUF failure in v5." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "Different paths: device_map=auto dequantized model offload versus GGUF loading for Qwen2.5." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "These are unrelated model/template bugs; they do not look like the same underlying code-path problem." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are version-5 tokenizer/model-loading regressions, but for different models and likely different causes." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked generation/compiled forward inconsistency is a different bug from StaticCache.get_seq_length returning the wrong type." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio-related, but one is a default sampling-rate mismatch and the other is missing padding-mask support in MIMI encoder batching." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Custom model initialization break versus non-persistent buffers being filled with junk; different failures and code paths." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Jupyter/custom model initialization is unrelated to non-persistent buffer serialization behavior." + }, + { + "left": "issue:43746", + "right": "issue:46326", + "accept": false, + "reason": "Local checkpoint loading with PEFT adapters is unrelated to continue_final_message behavior for mistral_common tokenizers." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Missing all_tied_weights_keys on Molmo is unrelated to tokenizer loading for EMBEDDIA/est-roberta." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Noise when loading a transformer is a different issue from a Mistral tokenizer regex patch crash." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and template compilation failure are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Both mention lm_head/tied weights, but one is about a model defining untied weights and the other about device-dependent serialization inconsistency." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen model families and different bugs: hidden-states output versus rope-index temporal scaling." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Siglip2 bad outputs and OmDet-Turbo processor size mismatch are distinct vision bugs." + }, + { + "left": "issue:28530", + "right": "issue:34567", + "accept": false, + "reason": "Early stopping/eval metric handling is unrelated to TrainerState.num_input_tokens_seen not updating." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "MIMI batching/padding-mask support is a different bug from DAC.from_latents not matching forward pass without STE." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Both are multimodal loading/processing failures, but they affect different models and different error conditions." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "A specific model load failure is not the same as a tokenizer regex patch AttributeError." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Different models and different failure modes; no shared concrete code-path." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "FP8 model failure and init_empty_weights unexpected argument are separate loader/initialization bugs." + }, + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Generation performance overhead is not the same as the StaticCache return-type contract bug." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "torch.export exportability issue versus processor image-size mismatch are unrelated." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio fbank feature extraction and compiled flex_attention on torch>=2.9 are separate issues." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Tokenizer class-method conflict and MobileLLM key error are different loading regressions." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken special-argument duplication is a tokenizer construction bug, not the same as the model-loading regression in 44987." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Both involve float8-adjacent failures, but one is Qwen3Moe FP8 config and the other is a missing Float8 storage type; not the same bug." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Both report the same Qwen3.5 save_pretrained regression, with 45357 explicitly indicating the incorrect visual-encoder-key issue persists in a later release." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays appropriately cautious about the mostly heterogeneous set. The only accepted soft pair, 45216/45357, is well-supported by the titles as the same Qwen3.5 save_pretrained regression with a later follow-up report. The other soft-edge rejections are conservative and do not overclaim duplication." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 11, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42796", + "issue:43296", + "issue:43317", + "issue:43381", + "issue:43421", + "issue:43425", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43746", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:44079", + "issue:44291", + "issue:44403", + "issue:44462", + "issue:44479", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44756", + "issue:44811", + "issue:44829", + "issue:44877", + "issue:44991", + "issue:45325", + "issue:45479", + "issue:45538", + "issue:45542", + "issue:45663", + "issue:45812", + "issue:45923", + "issue:46315", + "issue:46402", + "issue:46491" + ], + "soft_pairs": [ + "issue:44479|issue:45325", + "issue:44991|issue:46315", + "issue:43531|issue:44877", + "issue:36032|issue:41720", + "issue:43425|issue:44403", + "issue:43611|issue:44534", + "issue:43296|issue:43531", + "issue:43632|issue:44756", + "issue:40444|issue:45325", + "issue:45812|issue:46491", + "issue:42796|issue:43746", + "issue:43792|issue:44589", + "issue:44403|issue:44991", + "issue:33666|issue:40444", + "issue:43632|issue:44534", + "issue:43381|issue:45923", + "issue:44079|issue:45663", + "issue:44291|issue:44589", + "issue:40990|issue:44811", + "issue:43482|issue:45538", + "issue:43482|issue:43525", + "issue:43421|issue:44568", + "issue:35707|issue:40444", + "issue:43317|issue:43856", + "issue:42175|issue:45542", + "issue:41553|issue:43582", + "issue:43531|issue:44291", + "issue:43818|issue:46402", + "issue:44079|issue:44534", + "issue:44829|issue:45479", + "issue:44462|issue:45812" + ], + "prepared_review_unit_hash": "f273a113f29cc95e06805f483cec77ed920f0a4e830b4f6bd755b7d8533c40f8", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36496, + "estimated_input_tokens": 9124, + "estimated_eval_tokens": 18504 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is a mix of several Transformer regression families, with a few plausible near-duplicates but not enough evidence to collapse it into one broad bucket. The clearest canonical issue is the Granite tokenizer regression; the TensorFlow packaging pair looks related enough to keep as a soft duplicate, while the Qwen2.5-VL video/rope-index pair is too weakly supported to merge.", + "confidence": 0.73, + "canonical_issue_reason": "Issue 45812 is the strongest representative: it is a concrete, model-wide Granite tokenizer regression, has downstream references, and best captures the cluster\u2019s most clearly supported duplicate-worthy bug.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45812 has the clearest user impact and the most specific evidence of a real regression affecting multiple Granite models, making it the best overall issue in the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Same model family and similar video-positioning terminology, but the packet only shows topical overlap; the underlying bug and fix scope are not clearly the same." + }, + { + "left": "issue:44991", + "right": "issue:46315", + "accept": false, + "reason": "Both concern tokenizers, but one is a load failure and the other is a performance regression in conversion; different defects." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Different model/configuration failures with only broad overlap around loading and config handling." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer attribute conflict vs auto device-mapping CUDA assert are unrelated code paths." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch version incompatibility and noisy loading behavior are not the same underlying bug." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 regressions, but they affect different internals and would not be fixed by one focused change." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "Different models and failure modes; too little evidence of a shared root cause." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "_is_hf_initialized breakage and mmap/OOM on Strix Halo are distinct issues." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen VL-related, but one is multi-image finetuning and the other is rope-index temporal scaling; not the same bug." + }, + { + "left": "issue:45812", + "right": "issue:46491", + "accept": false, + "reason": "Different tokenizer regressions on different model families and symptoms; too broad to merge." + }, + { + "left": "issue:42796", + "right": "issue:43746", + "accept": false, + "reason": "Tokenizer marker handling and GraniteSpeech PEFT checkpoint loading are unrelated." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper runtime failure and Float8 storage lookup are unrelated defects." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Generic loading noise is not the same as failing to load tokenizer files from the repository." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both mention Qwen VL, but they describe different training failures and likely different fixes." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Separate v5 regressions with different failure mechanisms." + }, + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "Gradient-checkpointing eval-mode limitation vs missing capability flag on a trust_remote_code model are different bugs." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key bookkeeping and Gemma4 FSDP2 shared_kv_states rebuilding are unrelated." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "init_empty_weights/_is_hf_initialized and Float8 storage lookup do not share the same root cause." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "GPT-OSS perplexity and Whisper batch_decode special-token handling are unrelated." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Different model-loading/tokenizer bugs with different symptoms and likely different fixes." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Distinct model-loading failures; no clear shared code-path problem." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime post-processor updates vs missing BOS/EOS insertion are different tokenizer behaviors." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL multi-image finetuning." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Device-map offload failure and Qwen3 MoE memory inefficiency are different issues." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": true, + "reason": "Both point to the same packaging/backend dependency problem: a torch-only install leaves TensorFlow backend support in a broken state." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad tokenizer error messaging and an Apple Silicon allocator TypeError are unrelated." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Sliding_window/Qwen3-MoE and init_empty_weights argument breakage are unrelated." + }, + { + "left": "issue:43818", + "right": "issue:46402", + "accept": false, + "reason": "Video-LLaVA tower initialization and Sam3Model text encoder loading are different model-specific bugs." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer filling are unrelated." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Both involve degenerate classification behavior, but one is flash_attention_3-specific and the other is a num_labels/problem_type edge case." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but one is about ignoring tokenizer.json while the other is Granite-specific wrong token IDs; too broad to merge confidently." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The canonical Granite issue is well supported by the packet, and the TensorFlow packaging pair looks plausibly related. However, the accepted Qwen2.5-VL pair (issue:44479 vs issue:45325) is not conservative enough: the packet only shows topical overlap in video/rope-index handling, not clear evidence they are the same underlying bug or mergeable into one fix. The summary also leans a bit too far by calling the cluster mostly distinct with only a couple of true duplicates, since that judgment isn\u2019t strongly evidenced from titles alone." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 12, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42796", + "issue:42915", + "issue:43262", + "issue:43296", + "issue:43377", + "issue:43454", + "issue:43482", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43701", + "issue:43746", + "issue:43749", + "issue:43792", + "issue:43818", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44877", + "issue:44898", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45491", + "issue:45584", + "issue:45701", + "issue:45800", + "issue:45823", + "issue:45834" + ], + "soft_pairs": [ + "issue:43377|issue:45491", + "issue:43818|issue:45357", + "issue:41720|issue:45081", + "issue:43644|issue:44079", + "issue:43611|issue:43646", + "issue:44265|issue:45072", + "issue:43749|issue:44079", + "issue:42175|issue:43296", + "issue:40444|issue:45381", + "issue:43792|issue:43957", + "issue:45325|issue:45381", + "issue:43994|issue:44898", + "issue:43454|issue:44466", + "issue:45325|issue:45834", + "issue:42915|issue:43482", + "issue:42796|issue:45381", + "issue:43632|issue:45072", + "issue:43701|issue:45325", + "issue:44291|issue:44877", + "issue:43296|issue:45800", + "issue:44479|issue:45823", + "issue:43873|issue:45701", + "issue:43606|issue:44265", + "issue:43262|issue:43746", + "issue:40990|issue:45584", + "issue:34634|issue:35707", + "issue:43881|issue:44488", + "issue:35707|issue:41720", + "issue:43454|issue:43550", + "issue:43606|issue:43819" + ], + "prepared_review_unit_hash": "23844196764736e8cff8391f6cb4637309825bf720a07c81fd99dd8a2b9e7378", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36745, + "estimated_input_tokens": 9187, + "estimated_eval_tokens": 18630 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous and spans many unrelated regressions. The only strong duplicate-like pair is the Qwen2.5-VL visual position-id regression; the rest look like separate bugs in different models/subsystems.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45325 is the most concrete root-cause report in the Qwen2.5-VL position-id regression area (`get_rope_index` / temporal scaling), so it\u2019s the best anchor for that narrow duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45325 is the most actionable issue in the cluster\u2019s one clear overlap area; the remaining issues are too diverse to make a single global canonical beyond that sub-bucket.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Both involve batching/masking behavior, but they are different models and different failure modes (MIMI padding mask vs Gemma3 sliding-window NaNs)." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Unrelated problems: Video-LLaVA tower architecture/weights vs Qwen3.5 save_pretrained key mapping." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Different subsystems entirely: device-map CUDA assert vs tokenizer regex patch crash." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "One is buffer initialization/serialization junk; the other is ModelOutput key assignment logic." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 loading regressions, but the concrete bugs differ (base_model_prefix loading vs custom model initialization)." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "Different code paths: torch.export with torch_compilable_check vs bf16 dtype mismatches in inference." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading is unrelated to ModelOutput key handling." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Packaging/backend dependency issue vs a model-loading failure in vLLM." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both mention multimodal Qwen models, but one is training with multiple images per prompt and the other is a video-position-id regression." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Whisper runtime failure and meta-device loading breakage are different bugs." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Both describe the same Qwen2.5-VL visual position-id / rope-index regression in transformers 5.3.0; one is the still-image path and the other the video-input path." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model failures: SigLIP2 nonsensical outputs vs Perceiver non-default-resolution interpolation failure." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Related to lm_head/tied weights, but one is a failed tie causing garbage generation while the other is inconsistent serialization depending on device." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Both are vision/multimodal issues, but Qwen2.5-VL position IDs and Kosmos2.5 OCR long-input indexing are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different regressions: FP8 config failure vs GGUF loading failure." + }, + { + "left": "issue:42796", + "right": "issue:45381", + "accept": false, + "reason": "Tokenizer BPE marker handling is unrelated to Qwen video position ids." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "Different regressions: `_is_hf_initialized` flag handling vs dtype mismatches in inference." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint resume key mismatch is not the same as Qwen2.5-VL rope-index/position-id scaling." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Both are loading/configuration issues, but one is an unexpected init argument and the other is strict config rejection." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "vLLM model-load failure and torch/transformers version incompatibility are separate problems." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Video regression in Qwen multimodal models vs Gemma4 PLE device mismatch; not the same code path." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Quantized offloading behavior and tokenizer-version changes are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU offload device mismatch for Bark is unrelated to torch.export / compilable-check failures." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio processor sample-rate defaults and GraniteSpeech checkpoint loading are different bugs." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on GPT-OSS is unrelated to Whisper empty-transcription generation failure." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "Bark voice_preset behavior and progressive generation with inputs_embeds/past_key_values are unrelated." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Different model-loading failures for different models; no shared concrete bug." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Progressive generation with embeddings/past KV is unrelated to a CUDA assert on auto device mapping." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "lm_head tying/serialization and Bamba torch.compile SDPA failures are different issues." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "CPU offload device mismatch for Bark is unrelated to DAC.from_latents / missing STE mismatch." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: most items are clearly heterogeneous, and the Qwen2.5-VL position-id/rope-index cluster is the only plausible duplicate sub-bucket. The soft-edge rejections are suitably conservative, and the one accepted pair is reasonably supported by the titles and shared model/version context." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 13, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:26413", + "issue:28530", + "issue:36331", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42796", + "issue:42915", + "issue:42947", + "issue:43065", + "issue:43257", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43454", + "issue:43526", + "issue:43572", + "issue:43720", + "issue:43749", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44265", + "issue:44387", + "issue:44466", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45357", + "issue:45704", + "issue:45759", + "issue:45823", + "issue:45920", + "issue:45923", + "issue:46050", + "issue:46209", + "issue:46305", + "issue:46315", + "issue:46402" + ], + "soft_pairs": [ + "issue:44186|issue:44265", + "issue:39692|issue:40444", + "issue:41720|issue:45823", + "issue:42796|issue:46209", + "issue:46305|issue:46315", + "issue:43377|issue:44610", + "issue:44568|issue:45920", + "issue:43454|issue:45072", + "issue:43257|issue:45923", + "issue:43317|issue:43873", + "issue:42796|issue:44977", + "issue:43377|issue:44898", + "issue:40990|issue:42947", + "issue:43454|issue:43720", + "issue:43572|issue:45759", + "issue:43322|issue:46050", + "issue:42617|issue:42915", + "issue:43749|issue:44534", + "issue:41720|issue:43322", + "issue:43526|issue:43883", + "issue:44466|issue:45357", + "issue:26413|issue:28530", + "issue:43881|issue:44991", + "issue:41720|issue:44977", + "issue:44387|issue:45704", + "issue:36331|issue:42617", + "issue:36331|issue:42915", + "issue:44877|issue:44991", + "issue:43065|issue:46402" + ], + "prepared_review_unit_hash": "e8833962c9c9cbd4e29c208844e24d3d521f4ac3fd25c8c0ab3759efc0ef32ce", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35971, + "estimated_input_tokens": 8993, + "estimated_eval_tokens": 18242 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: the soft pairs share broad themes (tokenizers, loading/offloading, vision models, quantization), but each pair points to a different concrete failure mode, so I reject all duplicate edges. There are no PRs in the cluster; if forced to anchor an issue, 43873 is the broadest umbrella-style report.", + "confidence": 0.65, + "canonical_issue_reason": "Issue 43873 is the best available anchor because it is broad, active, and closest to an umbrella loading/offloading/quantization complaint among the set.", + "canonical_pr_reason": null, + "best_issue_reason": "43873 is the most suitable global issue representative since it has the widest problem framing and enough discussion/inbound refs to serve as a cluster anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Different bugs: LayoutLMv2 tokenizer crash vs torch.export export compatibility." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "Different artifacts and failures: SigLIP2 doc example errors vs Qwen2.5-VL iterable-dataset multi-image finetuning." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both involve device mapping, but the symptoms and models differ: cudaErrorAssert on Qwen3 vs forward device mismatch on Gemma4." + }, + { + "left": "issue:42796", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer BPE-marker handling and FP8 inference on B200 are unrelated failures." + }, + { + "left": "issue:46305", + "right": "issue:46315", + "accept": false, + "reason": "Different code paths: sliding-window mask off-by-one vs slow convert_tokens_to_ids." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "Both are input-shape/processing issues, but one is MIMI padding-mask consistency and the other is OmDet-Turbo resolution mismatch." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "Tokenizer regressions, but not the same bug: missing BOS/EOS insertion vs wrong token IDs across multiple models." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Different concrete problems: tied lm_head weights vs bfloat16 dtype mismatches." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "Different model-loading concerns: Qwen3 MoE weight conversion vs missing supports_gradient_checkpointing flag." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Related offloading/quantization theme, but different failures and scopes; not the same underlying bug." + }, + { + "left": "issue:42796", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer marker handling vs flash-attention generation failure are unrelated." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Different bug classes: batched-vs-single consistency vs non-default-resolution image classification failure." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity on a specific model/task vs gradient-checkpointing ineffectiveness are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Different subsystems: weight tying in AyaVision vs packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "Different config problems: missing pad_token_idx vs from_config not unwrapping text_config." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Llava Next segmentation fault and sinq save/load breakage are different loading bugs." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Different failures: 3d_parallel.py runtime issue vs FineGrainedFP8Config failure." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "Both regressions touch loading, but one is FSDP CPU RAM efficient loading and the other is non-persistent buffer initialization." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Different models and symptoms: Qwen3 auto device mapping assert vs Llava Next loading segmentation fault." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "Unrelated bugs: BeitImageProcessorFast label reduction vs Molmo missing all_tied_weights_keys." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Both involve serialization/loading of weights, but the concrete bugs differ: lm_head serialization vs visual encoder keys." + }, + { + "left": "issue:26413", + "right": "issue:28530", + "accept": false, + "reason": "Resume-from-checkpoint epoch iterator failure and early-stopping metric lookup are different trainer issues." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "glm-4v-9b loading failure and tokenizer loading regression for EMBEDDIA/est-roberta are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Same model family name only; one is device-map assert, the other is flash-attention generation bug." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Different memory issues: int4 reserved-memory/OOM regression vs apex FusedRMSNorm leak." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "CustomTrainer compute_loss signature change and 3d_parallel.py failure are unrelated." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Different runtime errors: compute_loss kwarg mismatch vs Qwen3Moe FineGrainedFP8Config failure." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Strict config blocking granite_speech and tokenizer loading regression for est-roberta are different issues." + }, + { + "left": "issue:43065", + "right": "issue:46402", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and Sam3Model text-encoder weights not loaded 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": 21, + "review_unit_index": 14, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28530", + "issue:30064", + "issue:33357", + "issue:33666", + "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:43540", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43708", + "issue:43720", + "issue:43782", + "issue:43873", + "issue:43881", + "issue:44008", + "issue:44186", + "issue:44361", + "issue:44423", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44756", + "issue:44898", + "issue:44913", + "issue:45072", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45823", + "issue:46291" + ], + "soft_pairs": [ + "issue:43450|issue:45701", + "issue:44186|issue:45072", + "issue:40990|issue:45701", + "issue:43873|issue:44756", + "issue:43550|issue:43720", + "issue:41720|issue:42617", + "issue:43577|issue:43708", + "issue:43404|issue:45357", + "issue:44008|issue:45823", + "issue:43550|issue:43606", + "issue:44186|issue:44610", + "issue:44610|issue:44898", + "issue:43295|issue:45701", + "issue:42915|issue:43317", + "issue:44423|issue:45406", + "issue:45632|issue:45684", + "issue:36010|issue:42617", + "issue:41720|issue:45161", + "issue:43531|issue:43572", + "issue:42915|issue:46291", + "issue:28530|issue:33666", + "issue:44671|issue:44743", + "issue:36331|issue:41720", + "issue:44186|issue:44361", + "issue:43526|issue:43881", + "issue:42915|issue:43322", + "issue:43540|issue:43782", + "issue:43316|issue:44913", + "issue:28530|issue:30064", + "issue:33357|issue:35707" + ], + "prepared_review_unit_hash": "edae83e62f798135001a7e9ce26db20657462977266f13712edeb95607af5bed", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36677, + "estimated_input_tokens": 9170, + "estimated_eval_tokens": 18596 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the suggested pairs span unrelated bugs in tokenization, multimodal processors, quantization/offload, configs, training, and serve/runtime paths. I don\u2019t see any true duplicates among the soft edges.", + "confidence": 0.96, + "canonical_issue_reason": "No strong canonical issue exists here because the items are not a coherent duplicate set; if one must be picked, issue 45701 is the broadest and most active, but it is still not representative of the rest of the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45701 is the most general and externally active item in the list, but only as a weak anchor; the surrounding issues are too unrelated for it to be a meaningful canonical duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video processor batch-shape bug vs tokenization/version behavior; different component and failure mode." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "LayoutLMv2 tokenizer crash vs dtype mismatch in inference; unrelated subsystems." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity regression on a specific model is not the same as a generic tokenization-version complaint." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Quantization/offload behavior and mmap OOM are different root causes and code paths." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "torch.compile/SDPA bug for Bamba vs packed-weight loading bug for BitNet; not the same fix." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Qwen3 device-map crash vs failure to run a 3d_parallel script; no shared underlying bug." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "BLIP2 dtype propagation and trainer checkpoint step math are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Weight tying bug in Mistral3 vs incorrect save_pretrained keys in Qwen3.5; different models and issues." + }, + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Gemma3n forward variable collision vs Gemma4 device-map mismatch; unrelated code paths." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "SDPA/compile failure and CPU offload device mismatch are distinct problems." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer padding/truncation crash vs processor output size mismatch; different components." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both concern image resolution, but they are model-specific processor/mode mismatches with different fixes." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "processor.tokenizer regression and tokenization-version changes are not the same concrete bug." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "FP8 config failure vs dequantized-model offload loading failure; different quantization paths." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Continuous-batching multimodal crash and missing _tokenizer in Gemma4Processor are separate serve issues." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collision and read-only permission propagation are unrelated save/load filesystem issues." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "ImportError for GenerationMixin is unrelated to a 3d_parallel runtime failure." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Qwen3 device-map assert and GPT-OSS MoE tensor-parallel support are different distributed inference bugs." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and StableLm pad_token_idx config are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:46291", + "accept": false, + "reason": "FP8 model loading and kernels package version/revision import failure are separate issues." + }, + { + "left": "issue:28530", + "right": "issue:33666", + "accept": false, + "reason": "Trainer early-stopping metric selection vs Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT masked-LM regression and Qwen3.5 recurrent-state reset are different model bugs." + }, + { + "left": "issue:36331", + "right": "issue:41720", + "accept": false, + "reason": "CustomTrainer compute_loss signature mismatch vs Qwen3 device-map crash; no common code path." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both are tokenizer-related, but they affect different tokenizers and different failures." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug and glm-4v loading failure are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "FP8 config issue vs Llava Next segmentation fault; different models and root causes." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Qwen3OmniMoe video processing error and Qwen3VL weight_only loading error are not the same bug." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Gemma3TextConfig API mismatch and GPTNeoX rotary_pct reload issue are different config regressions." + }, + { + "left": "issue:28530", + "right": "issue:30064", + "accept": false, + "reason": "Early stopping metric lookup and void segmentation-map processing are unrelated." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are different failure modes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 15, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:29159", + "issue:30064", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36032", + "issue:39692", + "issue:41720", + "issue:42490", + "issue:43329", + "issue:43388", + "issue:43531", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43873", + "issue:44062", + "issue:44077", + "issue:44368", + "issue:44387", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44610", + "issue:44625", + "issue:44898", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45059", + "issue:45072", + "issue:45357", + "issue:45406", + "issue:45464", + "issue:45479", + "issue:45588", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45859", + "issue:46402" + ], + "soft_pairs": [ + "issue:29159|issue:36032", + "issue:30064|issue:34634", + "issue:43550|issue:43653", + "issue:43606|issue:43720", + "issue:43873|issue:45704", + "issue:43388|issue:45059", + "issue:44387|issue:45701", + "issue:44492|issue:44509", + "issue:41720|issue:43746", + "issue:43531|issue:45759", + "issue:45406|issue:45464", + "issue:45357|issue:46402", + "issue:43606|issue:43653", + "issue:42490|issue:43329", + "issue:34634|issue:41720", + "issue:34567|issue:35707", + "issue:43531|issue:44589", + "issue:44451|issue:44991", + "issue:45759|issue:45859", + "issue:44077|issue:44625", + "issue:44625|issue:45479", + "issue:44898|issue:45072", + "issue:45588|issue:45759", + "issue:44062|issue:44451", + "issue:44987|issue:45042", + "issue:44514|issue:44545", + "issue:44368|issue:45357", + "issue:44918|issue:44936", + "issue:39692|issue:44610" + ], + "prepared_review_unit_hash": "1da5fc515a78481ca246e56d1cd5c81ee3d152879d4d90753bff80e241f3c603", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36679, + "estimated_input_tokens": 9170, + "estimated_eval_tokens": 18596 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated Transformers issues that only look superficially similar by subsystem. The only clear duplicate is the Qwen2_5_VLProcessor batched apply_chat_template crash; the rest should stay separate.", + "confidence": 0.91, + "canonical_issue_reason": "issue:44514 is the earlier, more complete report of the same Qwen2_5_VLProcessor.apply_chat_template crash; issue:44545 is effectively the same bug with nearly identical title and should fold into it.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44514 is the best canonical issue because it is the earlier, direct statement of the bug and matches the duplicate report exactly.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Both are tokenizer-related, but one is slow-vs-fast inconsistency and the other is a T5 tokenizer class/method naming conflict; different failure modes and fixes." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Both are processor bugs, but one is about void segmentation maps in image processing and the other is Bark voice_preset handling; unrelated code paths." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Different models and different bugs: torch.compile/SDPA failure versus BigBirdTokenizer special-token decoding." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both involve loading/offload, but one is a device-mismatch CPU offload bug and the other is unpacking BitNet packed weights during accelerate loading; not the same change." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization offloading versus an apex FusedRMSNorm memory leak in T5 are distinct problems." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "One is gather_for_metrics truncating tuple labels; the other is SAM3 PCS behavior with text and boxes. No shared code path." + }, + { + "left": "issue:44387", + "right": "issue:45701", + "accept": false, + "reason": "Both mention memory/tokenization regressions, but one is int4 CUDA reserved memory/OOM and the other is tokenization changing across versions; too broad and not the same bug." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategies is not the same as docs still referencing removed pipeline tasks." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Both touch model loading, but one is Qwen3 auto device mapping cudaErrorAssert and the other is GraniteSpeech local checkpoint/PEFT adapter loading." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "sliding_window on Qwen3-MoE is a generation/runtime issue, while from_config not unwrapping text_config is a config-construction issue." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer is a processor initialization bug; chat/completions streaming on Qwen3.5-0.8B is an API/streaming failure." + }, + { + "left": "issue:45357", + "right": "issue:46402", + "accept": false, + "reason": "Both are Qwen3.5 multimodal loading-related, but one is save_pretrained key serialization and the other is Sam3Model text encoder weights loading; different artifacts and fixes." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to BigBirdTokenizer mask token registration/empty decode output." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Trailing-slash local path import warning is a path parsing issue; undefined video-patch helpers in multimodal token counting is a separate processor bug." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset failure and Qwen3 auto device-mapping cudaErrorAssert are different models and different runtime paths." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState.num_input_tokens_seen not updating is unrelated to progressive generation with inputs_embeds and past_key_values." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Qwen3-MoE sliding_window and Float8 storage lookup are unrelated inference/runtime failures." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Loading one specific model failing and tokenizer loading for another model are both loading symptoms, but there is no evidence they share the same underlying defect." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "One is from_config/text_config handling; the other is rotary embedding not compatible with CPU offload. Different layers, different fix." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "patchtsmixer optional post_init validation and Qwen3.5 num_labels propagation are unrelated config issues." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 is different from the zero-loss bug when problem_type=single_label_classification and num_labels=1." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Both are dtype/resolution mismatches, but one is Perceiver interpolation at non-default resolution and the other is bfloat16 inference dtype mismatches for SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash attention s_aux=None crash in sink-less models is unrelated to Qwen multimodal config unwrapping." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "AddedToken special-argument conflict and a specific model loading failure are not the same underlying tokenizer bug from the available evidence." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "A model loading failure for physical-intelligence/fast and PIL backend processors unnecessarily requiring torchvision are different problems." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same exact Qwen2_5_VLProcessor.apply_chat_template crash on batched input with padding=False; nearly identical titles and likely duplicate reports." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "tie_word_embeddings warning during LoRA fine-tuning is a config warning; save_pretrained saving wrong visual encoder keys is a serialization bug." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Unpacking Qwen3.5 input embeddings with TRL SFT trainer and trainer.evaluate() failing after trainer.train() are different trainer lifecycle bugs." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "SigLIP2 docs example errors and OmDet-Turbo processor/model input-size mismatch are different issues, even though both involve multimodal processors." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative. The one accepted duplicate pair (issue:44514 / issue:44545) is supported by near-identical titles and timing, and the rest of the soft-edge rejections are appropriately cautious rather than overstated. I did not see an unsupported merge or an overclaim that would warrant rejection." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 16, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:39692", + "issue:41720", + "issue:42886", + "issue:43295", + "issue:43381", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43596", + "issue:43618", + "issue:43643", + "issue:43645", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43819", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44488", + "issue:44661", + "issue:44671", + "issue:44843", + "issue:44898", + "issue:45072", + "issue:45081", + "issue:45356", + "issue:45381", + "issue:45663", + "issue:45704", + "issue:45834", + "issue:45859", + "issue:45907", + "issue:45923", + "issue:46032", + "issue:46050", + "issue:46294", + "issue:46463" + ], + "soft_pairs": [ + "issue:43381|issue:43844", + "issue:43421|issue:43927", + "issue:45704|issue:45923", + "issue:43746|issue:46050", + "issue:43531|issue:43645", + "issue:41720|issue:45704", + "issue:45704|issue:45859", + "issue:43782|issue:45381", + "issue:42886|issue:44843", + "issue:46032|issue:46463", + "issue:45663|issue:45923", + "issue:43421|issue:43653", + "issue:43994|issue:44671", + "issue:39692|issue:43873", + "issue:43746|issue:43873", + "issue:43421|issue:44186", + "issue:43596|issue:44220", + "issue:45081|issue:45356", + "issue:30064|issue:36010", + "issue:45356|issue:45907", + "issue:43720|issue:43819", + "issue:43653|issue:43720", + "issue:43540|issue:46294", + "issue:44220|issue:44661", + "issue:43295|issue:45834", + "issue:43618|issue:43697", + "issue:43550|issue:45072", + "issue:44062|issue:44488", + "issue:43550|issue:43819", + "issue:43540|issue:44008", + "issue:44186|issue:44898", + "issue:43531|issue:43643", + "issue:43653|issue:43819" + ], + "prepared_review_unit_hash": "afb4e08a5dc54c129a989ed7cd89b7f552ffccccaee638e66ce93f240434bf88", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37990, + "estimated_input_tokens": 9498, + "estimated_eval_tokens": 19252 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These issues are broadly heterogeneous: a few tokenizer/offline-loading reports are semantically adjacent, but most pairs are clearly different bugs across models, modalities, and distributed-training paths. I did not find any pair that is safe to merge as the same concrete fix.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43421 is the best representative of the tokenizer-special-token subtheme: it describes a concrete backend/post-processor update bug and is the most central issue among the tokenizer-related reports in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43421 is the strongest standalone issue in this group because it has a clear repro, a specific code-path target, and the broadest overlap with the few tokenizer-adjacent candidates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Gradient-checkpointing eval-mode behavior and ZeRO-3 gradient growth are different code paths and different failure modes." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Both involve token handling, but one is a runtime post-processor update bug and the other is save/load corruption of custom token IDs." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "Apex RMSNorm memory leak and missing supports_gradient_checkpointing are unrelated model/runtime issues." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "Both are loading/saving problems, but PEFT local-checkpoint loading and sinq quantized save/load breakage are distinct mechanisms." + }, + { + "left": "issue:43531", + "right": "issue:43645", + "accept": false, + "reason": "Qwen3-MoE sliding_window and Jupyter custom-model initialization are unrelated regressions." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "Auto device-mapping cuda assert and T5 apex RMSNorm leak are unrelated." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "Different problems: Apex RMSNorm leak versus CPU-offload incompatibility in Qwen3_5MoeTextRotaryEmbedding." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Different Qwen VL failures: weight_only loading error versus incorrect video vision_position_ids." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": false, + "reason": "Both are offline/tokenizer-loading themed, but 44843 is a specific Mistral-regex model_info bug while 42886 is a broader cache/offline failure; not enough to treat as the same fix." + }, + { + "left": "issue:46032", + "right": "issue:46463", + "accept": false, + "reason": "Mamba2 use_cache correctness and T5 decoder-input error messaging are unrelated." + }, + { + "left": "issue:45663", + "right": "issue:45923", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states and missing gradient-checkpointing support are different issues." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both touch tokenizer special-token handling, but one is runtime post-processor refresh and the other is BigBird mask-token registration; different code paths." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "SigLIP2 model/processor mismatch and add-new-model-like mapping failure are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "SigLIP2 documentation/model mismatch and quantization-offloading behavior are not the same bug." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Local checkpoint loading with PEFT and quantization offloading are different persistence/offload issues." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Both involve tokenizers, but the reported crashes are different: post-processor updates versus NER/padding/truncation failures." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "ZeRO-3/BertModel indexing and fbank feature extraction are unrelated." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer-regex related, but one is an AttributeError in fix_mistral_regex while the other is a Kimi codec regression with a warning issue; distinct fixes." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "ImageProcessor void segmentation maps and GenerationMixin import failure are unrelated." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Kimi tokenizer regression and list_repo_files revision forwarding are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "BitNet packed-weight loading and DAC from_latents/STE mismatch are different model bugs." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "BigBird special-token registration and BitNet accelerate-loading unpacking are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:46294", + "accept": false, + "reason": "Both are video-input issues, but they hit different stages: Qwen3OmniMoe video processing versus apply_chat_template token/feature mismatch in Qwen3.5-9B." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "fbank feature extraction and TOKENIZER_MAPPING_NAMES add-new-model-like behavior are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Custom model processor/tokenizer regression and Kosmos2.5 long-OCR index error are different failures." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIPOutput attentions regression and RTDetrV2 output drift are different model behavior changes." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Torch.compile/SDPA failure in Bamba and dtype mismatch in SwitchTransformers/TimmWrapperModel are distinct inference bugs." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "AddedToken special-argument conflict and sleng-bert loading failure are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Different model-path problems: Bamba compile/SDPA versus DAC latent reconstruction mismatch." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Qwen3OmniMoe video processing and Gemma 3n audio_mel_mask variable collision are unrelated multimodal bugs." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "LayoutLMv2 tokenizer padding/NER crash and Perceiver non-default-resolution failure are unrelated." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Qwen3-MoE sliding_window and trust_remote_code AutoConfig field loss are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "BigBird mask-token registration and DAC STE mismatch 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": 21, + "review_unit_index": 17, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43322", + "issue:43335", + "issue:43388", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43643", + "issue:43644", + "issue:43716", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43818", + "issue:43950", + "issue:44008", + "issue:44186", + "issue:44442", + "issue:44448", + "issue:44479", + "issue:44485", + "issue:44509", + "issue:44671", + "issue:44805", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45183", + "issue:45381", + "issue:45479", + "issue:45584", + "issue:45685", + "issue:45834", + "issue:45987", + "issue:46032", + "issue:46512" + ], + "soft_pairs": [ + "issue:43335|issue:43756", + "issue:43643|issue:44877", + "issue:30064|issue:36331", + "issue:44843|issue:45081", + "issue:41720|issue:44991", + "issue:43322|issue:45685", + "issue:44821|issue:45084", + "issue:45183|issue:45479", + "issue:43388|issue:46032", + "issue:43716|issue:44008", + "issue:39692|issue:45834", + "issue:43950|issue:45987", + "issue:44186|issue:44442", + "issue:44448|issue:44671", + "issue:45584|issue:45834", + "issue:43644|issue:43749", + "issue:43582|issue:46512", + "issue:44485|issue:44509", + "issue:43606|issue:45072", + "issue:34634|issue:36010", + "issue:44805|issue:44936", + "issue:44479|issue:45834", + "issue:43577|issue:43818", + "issue:43540|issue:45381", + "issue:43782|issue:44977" + ], + "prepared_review_unit_hash": "d1d30f8c6d01bd57472f16670edf8da2d88d9459dcfe9ccfb876d76ddbf3eb6e", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 25, + "serialized_chars": 35561, + "estimated_input_tokens": 8891, + "estimated_eval_tokens": 18038 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is not a true duplicate set: the soft pairs are mostly superficial topical matches across unrelated model, tokenizer, dtype, config, and training bugs. None of the candidate pairs look mergeable into a single fix.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the items span many distinct regressions and the pairwise similarities are only broad topic overlap, not the same underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No single issue cleanly represents the set; the cluster is too heterogeneous to justify a representative duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "Different bugs: sparse-layer creation in SwitchTransformers config vs Smollm3 RoPE-layer count mismatch." + }, + { + "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." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Unrelated subsystems: image processor segmentation-map handling vs CustomTrainer compute_loss signature change." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Both touch Mistral tokenizer patching, but one is an offline model_info call and the other is a backend_tokenizer AttributeError." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "Different areas entirely: Qwen3 device-mapping CUDA assert vs tokenizer loading for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Segfault when loading Llava Next is unrelated to the MPS histogram/int error in moe." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading and template-node compilation are separate failures with no shared code path." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "XOR validation for input_ids/inputs_embeds is unrelated to the single-label-classification zero-loss bug." + }, + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Different code paths: gather_for_metrics last-batch label trimming vs Mamba2Mixer use_cache incorrect results." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Both are model-specific regressions, but dtype mismatch in Mistral-3 is unrelated to Gemma3n's variable-name collision." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "Docs/example errors for SigLIP2 do not match Kosmos2.5 long-input OCR index failures." + }, + { + "left": "issue:43950", + "right": "issue:45987", + "accept": false, + "reason": "Both involve state/buffer semantics, but one is a from_pretrained non-persistent-buffer regression and the other is a StaticCache return-type contract bug." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer crashes in LayoutLMv2 and AutoTokenizer loading FastSpeech2ConformerTokenizer are distinct bugs." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Both are output regressions in v5, but Pegasus output drift and CamemBERT masked-LM prediction errors are different model-specific issues." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Whisper empty-transcription generation failure is unrelated to Kosmos2.5 long OCR input indexing." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Non-persistent-buffer corruption in from_pretrained is unrelated to FSDP_CPU_RAM_EFFICIENT_LOADING being broken." + }, + { + "left": "issue:43582", + "right": "issue:46512", + "accept": false, + "reason": "Apple Silicon caching_allocator_warmup TypeError and stale model.config.dtype after from_config are different bugs." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "RoPE implementation discussion is not the same as docs mentioning removed pipeline tasks." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "CPU-offload device mismatch for bark-small is unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "BarkProcessor voice_preset and GenerationMixin import failure are unrelated." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Mask/shape index errors are not the same as trainer.evaluate() failing after train()." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Qwen video-input regression and Kosmos2.5 OCR long-input index error are different model paths." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "BLIP2 dtype staying float32 is unrelated to Video-LLaVA's missing temporal attention/shared weights." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both mention vision/video inputs, but Qwen3OmniMoe video ValueError and qwen2.5-vl vision_position_ids mismatch are not the same bug." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation failure are separate load-time and runtime bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 18, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36331", + "issue:42175", + "issue:43262", + "issue:43421", + "issue:43526", + "issue:43577", + "issue:43582", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43994", + "issue:44060", + "issue:44291", + "issue:44368", + "issue:44561", + "issue:44589", + "issue:44610", + "issue:44671", + "issue:44811", + "issue:44857", + "issue:44912", + "issue:44991", + "issue:45003", + "issue:45020", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45406", + "issue:45538", + "issue:45701", + "issue:45987", + "issue:46192", + "issue:46285", + "issue:46291", + "issue:46294", + "issue:46463" + ], + "soft_pairs": [ + "issue:43825|issue:43994", + "issue:43421|issue:44811", + "issue:44991|issue:46291", + "issue:46294|issue:46463", + "issue:44561|issue:45020", + "issue:44610|issue:45072", + "issue:43720|issue:45072", + "issue:43577|issue:43582", + "issue:28218|issue:29159", + "issue:45987|issue:46192", + "issue:43825|issue:43827", + "issue:43749|issue:44291", + "issue:44671|issue:45701", + "issue:42175|issue:45081", + "issue:43782|issue:45325", + "issue:43582|issue:44857", + "issue:34634|issue:36331", + "issue:45003|issue:46285", + "issue:43262|issue:43819", + "issue:43653|issue:44912", + "issue:43526|issue:43697", + "issue:45081|issue:45406", + "issue:43742|issue:45020", + "issue:44060|issue:44368", + "issue:45084|issue:45538", + "issue:44589|issue:45042" + ], + "prepared_review_unit_hash": "d0747a64de03b9ddf0042a7f65d164ac778df1301b16c357947be2f31e744796", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35256, + "estimated_input_tokens": 8814, + "estimated_eval_tokens": 17884 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous bucket of unrelated issue reports (tokenization, model loading, processor behavior, dtype, CI/backends). None of the soft pairs look like true duplicates or mergeable same-bug reports, so this is not a strong duplicate cluster.", + "confidence": 0.41, + "canonical_issue_reason": "issue:45701 is the broadest umbrella-style report in the set and the closest thing to a cluster anchor, but it still does not actually duplicate the other issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45701 is the best representative issue because it is the most generic version-regression/tokenization report and has the broadest potential overlap, though the cluster is too mixed for a true canonical match.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "One is about a misleading pipeline error message for translation support; the other is about a specific model producing bad outputs with AutoModel/pipeline. Different bugs and code paths." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Both involve tokenization-related behavior, but one is about runtime special-token/post-processor updates and the other about Whisper batch_decode skipping special tokens. Not the same underlying bug." + }, + { + "left": "issue:44991", + "right": "issue:46291", + "accept": false, + "reason": "Tokenizer loading failure for one model versus a kernels dependency/version requirement error on importing pipelines. Unrelated failures." + }, + { + "left": "issue:46294", + "right": "issue:46463", + "accept": false, + "reason": "Video chat-template mismatch versus T5 decoder-input validation messaging. Different models and failure modes." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "Both mention remote_code breakage, but one is specifically about removing is_torch_fx_available and the other is a broader recent-version regression report. Not concrete enough to be one PR." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Processor input-size mismatch versus bf16 dtype mismatch. Different subsystems and fixes." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight unpacking during loading is unrelated to bfloat16 dtype mismatch in inference." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "Blip2 dtype propagation bug and Apple Silicon allocator TypeError are separate issues." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": false, + "reason": "Tokenization whitespace after added tokens may be one symptom, but the second report is a broader slow/fast tokenizer inconsistency. Not clearly the same defect." + }, + { + "left": "issue:45987", + "right": "issue:46192", + "accept": false, + "reason": "Cache sequence-length return-type bug and return_tensors being ignored are unrelated API issues." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "One is an incorrect runtime error message; the other is stale docs referencing removed pipeline behavior. Same theme, but different artifacts and fixes." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Both are model-loading related, but one is FSDP CPU RAM efficient loading and the other is an unexpected _is_hf_initialized argument with init_empty_weights. Different code paths." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "CamemBERT masked-LM regression and a generic tokenization-change report are not the same underlying bug." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Backend installation packaging problem versus a Mistral tokenizer regex crash. No shared code-path." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen model families and different bugs: weight_only loading error versus rope-index scaling." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "AppleSilicon TypeError in allocator warmup versus CUDA float16 AMP crash. Separate platform-specific problems." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "BarkProcessor voice_preset failure and CustomTrainer.compute_loss signature incompatibility are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:46285", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to CLIP text model structure inconsistency." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio processor chat-template sampling-rate default versus DAC.from_latents/STE mismatch. Different components and fixes." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "Label-reduction bug in an image processor and RTDetrV2 output differences are different regression classes." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Both involve processor/tokenizer internals, but they are model-specific attribute errors in different classes. Not mergeable as one fix." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "Specific MobileLLM loading KeyError versus generic remote_code regressions. Too broad to be the same bug." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Tied-weights warning in Qwen3-Next and LoRA tie_word_embeddings warning in Qwen3.5-27B are distinct configuration issues." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template compilation TypeError and CLIPTokenizer model_max_length default are unrelated." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Float8 storage lookup failure versus torchvision dependency requirement in PIL image processors. Separate issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 19, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:42175", + "issue:42490", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43638", + "issue:43643", + "issue:43645", + "issue:43650", + "issue:43696", + "issue:43720", + "issue:43746", + "issue:43819", + "issue:43854", + "issue:43906", + "issue:43927", + "issue:43931", + "issue:44387", + "issue:44479", + "issue:44554", + "issue:44610", + "issue:44617", + "issue:44811", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45042", + "issue:45372", + "issue:45459", + "issue:45600", + "issue:45759", + "issue:45803", + "issue:46209", + "issue:46291", + "issue:46326", + "issue:46421", + "issue:46424" + ], + "soft_pairs": [ + "issue:43906|issue:46209", + "issue:45372|issue:46326", + "issue:43526|issue:44554", + "issue:43329|issue:45600", + "issue:43638|issue:44936", + "issue:43452|issue:44991", + "issue:43550|issue:44912", + "issue:43819|issue:43927", + "issue:36010|issue:43645", + "issue:43257|issue:43329", + "issue:42175|issue:46291", + "issue:36010|issue:36331", + "issue:43531|issue:45042", + "issue:43746|issue:44811", + "issue:43720|issue:44610", + "issue:38175|issue:44610", + "issue:44387|issue:44617", + "issue:39692|issue:44912", + "issue:43854|issue:44977", + "issue:42490|issue:43257", + "issue:46421|issue:46424", + "issue:45459|issue:45803", + "issue:43650|issue:43906", + "issue:43643|issue:45759", + "issue:43262|issue:44811", + "issue:43257|issue:43696", + "issue:43329|issue:44479", + "issue:43329|issue:43450", + "issue:43257|issue:43931" + ], + "prepared_review_unit_hash": "1c7a3662f4e9f36dab0c51b84573aca9e4ff6d506120c2cae4ef44f3ff258bfa", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36053, + "estimated_input_tokens": 9014, + "estimated_eval_tokens": 18284 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a duplicate cluster: the items span unrelated bugs across imports, training, quantization, cache behavior, multimodal/video processing, and packaging. None of the soft-edge pairs look like the same underlying change or defect, so all should be rejected.", + "confidence": 0.74, + "canonical_issue_reason": "No strong canonical issue exists because the set is heterogeneous. If an anchor is required, issue 43329 is the broadest concrete code-path bug here (_get_num_multimodal_tokens video branch with undefined names) and is the most actionable representative.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43329 is the best single issue to anchor on because it describes a specific failing branch with a clear fix surface and broader relevance across multiple VLM processors.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43906", + "right": "issue:46209", + "accept": false, + "reason": "A reproduction of a prior issue is not the same bug as an FP8 inference OOM on B200; different symptoms and code paths." + }, + { + "left": "issue:45372", + "right": "issue:46326", + "accept": false, + "reason": "ImportError in Gemma 4 processor loading from mistral_common is unrelated to continue_final_message behavior in tokenizer chat handling." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "BEIT label reduction bug and MPS attention correctness issue are unrelated subsystems and failure modes." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Multimodal video token counting bug vs stale auto_mappings config references; different files and defects." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Zero3/non-pretrained Bert index error is a training initialization bug, not the same as evaluate() failing after train()." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "gguf_file loading breakage and tokenizer load failure for EMBEDDIA/est-roberta are separate model-loading problems." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "torch.compile+SDPA failure in Bamba and MXFP4 quantization fallback in git-oss are different model/backend bugs." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "DAC forward/from_latents mismatch and DiaConfig save/load token ID loss are different model families and bug types." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "GenerationMixin import error and Jupyter custom-model initialization breakage are both import-adjacent but not the same defect." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MoE weight conversion/loading with accelerate+deepspeed is unrelated to video token counting undefined names." + }, + { + "left": "issue:42175", + "right": "issue:46291", + "accept": false, + "reason": "Missing tensorflow in the torch extra is packaging-related; kernels v0.15 import failure is a separate dependency/API issue." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "GenerationMixin import failure and unexpected num_items_in_batch in CustomTrainer.compute_loss are unrelated bugs." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window bug and PIL backend torchvision dependency regression are different subsystems." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "GraniteSpeech PEFT local checkpoint loading and Whisper batch_decode skip_special_tokens handling do not share a concrete code path." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "BitNet packed-weight unpacking during accelerate loading and OmDet input-size mismatch are unrelated issues." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 is not the same bug as OmDet processor producing 640px inputs." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both are OOM-related, but one is an int4 reserved-memory regression and the other is a Sam3Video memory exhaustion report." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Documentation/example errors for SigLIP2 are not the same as MXFP4 quantization loading fallback for git-oss." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "GLM-4.7 Flash load failure in tests and Qwen3.5 flash-attention generation bug are different models and code paths." + }, + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "Trailing-slash local-model import warning and Qwen3 MoE conversion/loading failure are distinct loading edge cases." + }, + { + "left": "issue:46421", + "right": "issue:46424", + "accept": false, + "reason": "Both concern cache behavior, but lazy-init recompilation with chunked prefill is not the same as max_cache_length being ignored except for Quantized Cache." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Hidden tokenizer decode errors when protobuf is absent and bare-except swallowing interrupts/errors are different exception-handling bugs." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "A nonspecific 'ADD THE DATA' issue is unrelated to the isolated reproduction of issue 38071." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Missing fields from trust_remote_code AutoConfig and unwrapped text_config in composite multimodal configs are separate config-loading bugs." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio apply_chat_template sample-rate defaulting and Whisper batch_decode skip_special_tokens are unrelated processor behaviors." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Qwen3 MoE loading/conversion bug and GPT-oss-20b CUDA OOM are unrelated model/runtime failures." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Both mention video, but one is undefined helpers in token counting and the other is a broader v5.3.0 video input regression; not the same concrete bug." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Undefined video token-count helpers and batched video processor shape errors are different failures in the video stack." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3 MoE accelerate/deepspeed conversion issues and Qwen3-VL weight-shape mismatches are separate model-loading problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 20, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28282", + "issue:29127", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43317", + "issue:43404", + "issue:43425", + "issue:43452", + "issue:43454", + "issue:43525", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43824", + "issue:43927", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44560", + "issue:44568", + "issue:44756", + "issue:44821", + "issue:44849", + "issue:44912", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45127", + "issue:45137", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45405", + "issue:45538", + "issue:45701", + "issue:45859", + "issue:46291", + "issue:46512" + ], + "soft_pairs": [ + "issue:43749|issue:44756", + "issue:43454|issue:43577", + "issue:42915|issue:43742", + "issue:44291|issue:45042", + "issue:43606|issue:45859", + "issue:34567|issue:40990", + "issue:42915|issue:44821", + "issue:43525|issue:46291", + "issue:29159|issue:34567", + "issue:44568|issue:45356", + "issue:44987|issue:45538", + "issue:43404|issue:45127", + "issue:43927|issue:44912", + "issue:43697|issue:45357", + "issue:42915|issue:45538", + "issue:43425|issue:46291", + "issue:44849|issue:45701", + "issue:42175|issue:43824", + "issue:43577|issue:46512", + "issue:28282|issue:34689", + "issue:43317|issue:45538", + "issue:45137|issue:45405", + "issue:43452|issue:43582", + "issue:33357|issue:34567", + "issue:44062|issue:44991", + "issue:43742|issue:44291", + "issue:41720|issue:43540", + "issue:28282|issue:29127", + "issue:43540|issue:44560", + "issue:44479|issue:45381", + "issue:43540|issue:44479", + "issue:42915|issue:43749" + ], + "prepared_review_unit_hash": "e36f6e7da34d1029457985b4619dffc878a5daaa402ebdb98b5c3e9fb6a418a7", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37312, + "estimated_input_tokens": 9328, + "estimated_eval_tokens": 18912 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is highly heterogeneous: most items are unrelated model-loading, tokenizer, dtype, offload, or backend-specific bugs. None of the soft pairs look like the same underlying change or defect, so all soft edges are rejected. No PRs are present.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 45701 is the broadest, most central-sounding item in the set (generic tokenization regression with multiple inbound refs), but the overall cluster is too mixed for a truly strong canonical issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45701 is the best available representative by breadth and linkage activity, though it does not actually unify the heterogeneous bugs in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Both relate to memory/offload behavior, 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:43454", + "right": "issue:43577", + "accept": false, + "reason": "One is lm_head weight tying in Mistral3, the other is dtype propagation in BLIP2; unrelated bugs." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different models and symptoms: FineGrainedFP8Config failure vs a key error loading MobileLLM-125M." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "Init-empty-weights argument handling is unrelated to PIL backend torchvision requirements." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "Both mention CPU offload, but they involve different models and distinct runtime errors." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token counting is unrelated to perplexity on gpt-oss-20b." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "FineGrainedFP8Config loading failure is unrelated to AutoImageProcessor URL loading." + }, + { + "left": "issue:43525", + "right": "issue:46291", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config is unrelated to kernels version/revision import failure." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer consistency vs TrainerState token accounting are different subsystems and bugs." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different tokenizers and different concrete failures." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Model-loading failure for a specific model is unrelated to CLIPTokenizer model_max_length behavior." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both involve tied embeddings/lm_head, but one is a missing tie in a specific model and the other is LoRA merge collapse after vocab extension; not the same bug." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Custom token ID persistence on save/load is unrelated to MXFP4 quantization fallback." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "RTDetrV2 output drift and Qwen3.5 visual encoder key saving are unrelated issues." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "FP8 model load failure is unrelated to tokenizer max-length settings." + }, + { + "left": "issue:43452", + "right": "issue:46291", + "accept": false, + "reason": "gguf_file loading breakage is unrelated to kernels version/revision import requirements." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 hidden-state output behavior is unrelated to a broad tokenization-change report." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "TensorFlow backend packaging and missing Qwen2.5-VL class import are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:46512", + "accept": false, + "reason": "BLIP2 dtype propagation and config.dtype staleness are related only superficially; different affected code paths and models." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "PyTorch-not-found import error is unrelated to Llama 3.2 vision model loading regression." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Device-map/offload loading failure is unrelated to tokenizer max-length behavior." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque error is unrelated to a PEFT version pin issue." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf loading and Apple Silicon allocator warmup TypeError are unrelated bugs." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "MacOS bus error in CLIP loading is unrelated to TrainerState token counting." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "Different tokenizer errors with different root causes; not the same underlying bug." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Key error loading a model and init_empty_weights argument handling are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "CUDA assert under auto device mapping and video-input ValueError in Qwen3OmniMoe are different failures." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "Missing PyTorch import error is unrelated to LayoutLMv3 box-information error messaging." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Both involve video inputs, but the reported errors and likely code paths differ." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": false, + "reason": "Related family and version area, but one is a broad video-input regression across multiple models and the other is a specific position-id bug in qwen2.5-vl." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Qwen3OmniMoe video-processing failure and the multi-model v5.3.0 video regression are not concrete duplicates." + }, + { + "left": "issue:42915", + "right": "issue:43749", + "accept": false, + "reason": "FineGrainedFP8Config failure is unrelated to FSDP CPU RAM-efficient loading." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 21, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:24643", + "issue:27957", + "issue:28282", + "issue:29159", + "issue:34567", + "issue:36331", + "issue:38175", + "issue:42175", + "issue:42796", + "issue:43257", + "issue:43404", + "issue:43452", + "issue:43493", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43742", + "issue:43756", + "issue:43906", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44488", + "issue:44568", + "issue:44589", + "issue:44737", + "issue:44792", + "issue:44821", + "issue:44898", + "issue:44964", + "issue:44987", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45405", + "issue:45479", + "issue:45488", + "issue:45588", + "issue:45600", + "issue:45823", + "issue:45907", + "issue:45987", + "issue:46399", + "issue:46463", + "issue:46468", + "issue:46491" + ], + "soft_pairs": [ + "issue:44987|issue:45907", + "issue:24643|issue:29159", + "issue:42175|issue:44488", + "issue:44568|issue:45488", + "issue:38175|issue:44898", + "issue:44737|issue:46463", + "issue:43493|issue:43994", + "issue:44291|issue:44821", + "issue:27957|issue:28282", + "issue:34567|issue:36331", + "issue:43257|issue:43756", + "issue:43577|issue:45072", + "issue:44062|issue:45081", + "issue:44792|issue:44964", + "issue:43906|issue:46468", + "issue:45042|issue:45081", + "issue:44568|issue:46491", + "issue:43742|issue:44589", + "issue:43526|issue:45479", + "issue:44737|issue:45588", + "issue:43756|issue:46399", + "issue:43756|issue:45325", + "issue:43404|issue:45987", + "issue:45823|issue:46463", + "issue:43531|issue:44821", + "issue:43452|issue:44488", + "issue:45405|issue:45600", + "issue:42796|issue:43906" + ], + "prepared_review_unit_hash": "d72472684c4a4aaa5fe92e36aea7be14faf117a4066693658298310734650d9b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35854, + "estimated_input_tokens": 8964, + "estimated_eval_tokens": 18184 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: tokenizer regressions, model-loading failures, dtype/device issues, and trainer bugs across unrelated code paths. I only see one plausible duplicate pair in the SigLIP2 reports; the rest should stay separate.", + "confidence": 0.76, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:43994 is the best single issue to anchor on because it is a concrete, user-facing SigLIP2 regression with a clear repro, and it is the closest match to the only likely duplicate pair in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Different problems: model loading failure for physical-intelligence/fast vs list_repo_files not forwarding revision in CI." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "Unrelated subsystems: DeepSpeed training weight shape error vs slow/fast tokenizer inconsistency." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Different bugs: packaging/backend dependency issue vs model load failure for cjvt/sleng-bert." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Both are tokenizer regressions, but on different models and code paths: mdeberta special-token handling vs DeepSeek/Llama tokenizer pre-tokenizer override." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model failures: SigLIP2 zero probabilities vs Perceiver non-default resolution interpolation failure." + }, + { + "left": "issue:44737", + "right": "issue:46463", + "accept": false, + "reason": "Distinct model/code-path issues: XLNet CPU arange bug vs T5 decoder input validation message." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "Same model family and same underlying SigLIP2 wrong-output regression; the JAX-vs-HF discrepancy and nonsensical AutoModel/pipeline outputs plausibly stem from the same implementation bug." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Different loading failures: unexpected _is_hf_initialized argument vs AutoImageProcessor-from-URL behavior." + }, + { + "left": "issue:27957", + "right": "issue:28282", + "accept": false, + "reason": "Different issues: Flash Attention 2 with XLM-Roberta vs missing PyTorch import error." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "Trainer metrics bug vs custom compute_loss signature change; related area but not the same defect." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Both mention model internals, but one is MoE weight conversion under accelerate/deepspeed and the other is RoPE layer dropping in SmolLM3." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Separate dtype issues in different models and contexts: BLIP2 loading vs SwitchTransformers/TimmWrapperModel bfloat16 inference." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer bugs, but different failures: AddedToken special-arg duplication vs Mistral regex patch crash." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Different multimodal failures: Janus test failure vs Phi-4 multimodal load failure." + }, + { + "left": "issue:43906", + "right": "issue:46468", + "accept": false, + "reason": "No clear overlap; one is a reproduction of a prior issue, the other is a Nemotron example breakage." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Different subsystems: PIL image processor torchvision dependency vs Mistral tokenizer regex patching." + }, + { + "left": "issue:44568", + "right": "issue:46491", + "accept": false, + "reason": "Both tokenizer regressions, but different models and symptoms: missing BOS/EOS vs CodeLlama round-trip leading-space loss." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "Different load-time failures: MobileLLM key error vs missing Float8 storage object." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to sequence-classification zero-loss behavior across multiple models." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "XLNet CPU tensor allocation bug vs flash_attention crash on s_aux=None are unrelated." + }, + { + "left": "issue:43756", + "right": "issue:46399", + "accept": false, + "reason": "Different bugs: SmolLM3 RoPE-layer dropping vs Qwen3.5 kernel registration crash." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both touch RoPE-related code, but the model families and defect types differ (layer dropping vs still-image temporal scaling)." + }, + { + "left": "issue:43404", + "right": "issue:45987", + "accept": false, + "reason": "Mistral3 lm_head tying issue is unrelated to StaticCache.get_seq_length returning a tensor." + }, + { + "left": "issue:45823", + "right": "issue:46463", + "accept": false, + "reason": "Gemma4 device-map mismatch and T5 decoder input validation are different problems." + }, + { + "left": "issue:42796", + "right": "issue:43906", + "accept": false, + "reason": "Ministral tokenizer BPE-marker handling is unrelated to the isolated reproduction of a separate prior issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: it stays conservative about the cluster\u2019s heterogeneity, and the single accepted SigLIP2 pair is a reasonable soft match rather than an overclaim. I don\u2019t see any soft-edge verdicts that cross into unsupported certainty." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 22, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:32090", + "issue:33357", + "issue:38175", + "issue:41762", + "issue:42898", + "issue:43295", + "issue:43452", + "issue:43577", + "issue:43643", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43749", + "issue:43819", + "issue:44361", + "issue:44442", + "issue:44466", + "issue:44484", + "issue:44493", + "issue:44514", + "issue:44625", + "issue:44704", + "issue:44811", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44936", + "issue:45005", + "issue:45200", + "issue:45216", + "issue:45357", + "issue:45372", + "issue:45375", + "issue:45381", + "issue:45542", + "issue:45823", + "issue:45920", + "issue:46050", + "issue:46121", + "issue:46170", + "issue:46178", + "issue:46463", + "issue:46491" + ], + "soft_pairs": [ + "issue:45357|issue:45920", + "issue:44913|issue:46121", + "issue:44625|issue:44877", + "issue:43720|issue:44898", + "issue:43697|issue:45216", + "issue:43643|issue:45375", + "issue:24643|issue:29127", + "issue:41762|issue:44936", + "issue:45200|issue:46463", + "issue:28282|issue:32090", + "issue:44493|issue:46050", + "issue:44466|issue:45005", + "issue:28282|issue:33357", + "issue:43452|issue:44843", + "issue:43688|issue:43749", + "issue:43819|issue:44811", + "issue:44936|issue:45823", + "issue:44361|issue:44442", + "issue:38175|issue:43295", + "issue:43653|issue:44361", + "issue:45542|issue:46178", + "issue:45372|issue:46170", + "issue:42898|issue:46491", + "issue:44484|issue:44704", + "issue:44514|issue:45381", + "issue:43452|issue:43577" + ], + "prepared_review_unit_hash": "e1e2351b45af73adc07994a6ce9c4da5f591d572843b530604752ccd536d2c8d", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35930, + "estimated_input_tokens": 8983, + "estimated_eval_tokens": 18222 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are unrelated single-bug reports across different model families and code paths. The only plausible duplicate pair is the tied-weights/save_pretrained regression in v5 (issues 44466 and 45005), with 45005 as the broader umbrella.", + "confidence": 0.63, + "canonical_issue_reason": "Issue 45005 is the best umbrella representative for the tied-weights regression cluster: it is broader, explicitly about tied weights in v5, and can subsume the narrower device-dependent serialization symptom reported in 44466.", + "canonical_pr_reason": null, + "best_issue_reason": "45005 is the strongest issue candidate because it captures the shared underlying save_pretrained/tied-weights problem at a higher level than the narrower 44466 report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Different regressions: Qwen3.5 visual-encoder key serialization vs AutoTokenizer returning wrong token IDs across multiple models." + }, + { + "left": "issue:44913", + "right": "issue:46121", + "accept": false, + "reason": "Both touch config/rope-related code, but one is a silent reload defaulting bug and the other is a TypeError on JSON-loaded lists; different root cause." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-loading issues, but they affect different models and code paths: num_labels propagation vs strict config rejection." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "Unrelated model bugs: BitNet packed-weight loading vs Perceiver interpolation handling." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Different failure modes and models: RTDetrV2 output regression vs Qwen3.5 save_pretrained checkpoint corruption." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both involve config fields, but one is a generic AutoConfig trust_remote_code issue and the other is a model-specific strict-field omission." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "Completely different bugs: DeepSpeed weight shape error vs LayoutLMv3 box-information error message." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "Different code paths: Gemma3 loading with ZeRO-3 vs trainer.evaluate failing after train." + }, + { + "left": "issue:45200", + "right": "issue:46463", + "accept": false, + "reason": "One is a missing-default input requirement for Gemma 4 fine-tuning; the other is a T5 decoder-input validation message issue." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Different errors and subsystems: missing PyTorch import vs Trainer NoneType broadcast failure." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Different regressions: unexpected position-id keys vs quantized model save/load breakage." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": true, + "reason": "Both point to the same v5 tied-weights/save_pretrained serialization problem; 45005 is the broader umbrella and 44466 looks like a device-dependent manifestation." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Different environment failures: PyTorch import missing vs macOS bus error with a CLIP model." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "Both are loading-related, but one is gguf file handling and the other is an offline-mode mistral regex patch bug." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Different problems: auxiliary-loss normalization vs broken FSDP CPU RAM-efficient loading." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "Unrelated: DAC latent decoding mismatch vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Different runtime failures: trainer lifecycle issue vs Gemma4 device-map mismatch." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Both are tokenizer-related, but they involve different tokenizers and different failure modes." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Different regressions: SIGLIP2 zero probabilities vs processor.tokenizer/custom model regression." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizer bugs with different models and symptoms." + }, + { + "left": "issue:45542", + "right": "issue:46178", + "accept": false, + "reason": "Both concern backend availability, but they are different backend-mapping failures (tensorflow vs tensorflow_text) with different causes." + }, + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Different dependency/import failures: mistral_common ReasoningEffort vs Windows ROCm torch.distributed dependency." + }, + { + "left": "issue:42898", + "right": "issue:46491", + "accept": false, + "reason": "Both involve tokenization behavior, but one is a v5 API-default change and the other is a CodeLlama whitespace round-trip regression." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "Different save/load API issues: shard-size question vs kwargs not passed to cached_file." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Related model family, but distinct bugs: batched chat-template crash vs wrong vision_position_ids for video input." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "Different loading regressions: gguf handling vs dtype not propagating to Blip2 submodules." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster call is grounded in the packet. The only accepted soft pair, issue:44466 \u2194 issue:45005, is a conservative and plausible duplicate/umbrella match around v5 tied-weights save_pretrained serialization. The rest of the pair judgments also stay appropriately cautious and do not overclaim shared root causes. The summary is consistent with the evidence and does not overstate similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 23, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:29159", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:36010", + "issue:41720", + "issue:43295", + "issue:43504", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43606", + "issue:43643", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43927", + "issue:44361", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44514", + "issue:44554", + "issue:44561", + "issue:44568", + "issue:44610", + "issue:44625", + "issue:44743", + "issue:44821", + "issue:44843", + "issue:44977", + "issue:44991", + "issue:45290", + "issue:45356", + "issue:45440", + "issue:45491", + "issue:45701", + "issue:45704", + "issue:46002", + "issue:46050", + "issue:46209", + "issue:46365", + "issue:46402" + ], + "soft_pairs": [ + "issue:43295|issue:45356", + "issue:43531|issue:43742", + "issue:43577|issue:43697", + "issue:43653|issue:44568", + "issue:44464|issue:44743", + "issue:43646|issue:43749", + "issue:44843|issue:44991", + "issue:43688|issue:45704", + "issue:29159|issue:33357", + "issue:43756|issue:46402", + "issue:43653|issue:43927", + "issue:43688|issue:45440", + "issue:43550|issue:43927", + "issue:43688|issue:45701", + "issue:44479|issue:44625", + "issue:43550|issue:44361", + "issue:45491|issue:46002", + "issue:44464|issue:44977", + "issue:33666|issue:36010", + "issue:43742|issue:44821", + "issue:32090|issue:33666", + "issue:41720|issue:44464", + "issue:43643|issue:44561", + "issue:44466|issue:46050", + "issue:43504|issue:43550", + "issue:43504|issue:43720", + "issue:44514|issue:45290", + "issue:43577|issue:44554", + "issue:43504|issue:44610", + "issue:43720|issue:43927", + "issue:43606|issue:43927", + "issue:44554|issue:46365", + "issue:44977|issue:46209" + ], + "prepared_review_unit_hash": "6094adf0ec07a5ea6bf793b656004f269fb7494d5128a61f38a418b448f19154", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37676, + "estimated_input_tokens": 9419, + "estimated_eval_tokens": 19094 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is broadly about Transformers regressions and compatibility failures, with many tokenizer- and model-loading-related reports, but the soft pairs mostly describe different concrete bugs, not duplicates.", + "confidence": 0.42, + "canonical_issue_reason": "issue:45701 is the broadest umbrella report in the set (tokenization changes across Transformers versions) and best fits the shared regression theme; the other issues are narrower model- or code-path-specific failures.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45701 is the best global representative because it captures the general tokenizer/version-regression pattern seen across the cluster, even though the cluster is heterogeneous and not a tight duplicate group.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but one is about processor.tokenizer/images in custom code and the other is Kimi-K2.5 codec handling; different bug paths." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and a MobileLLM-125M load key error are unrelated failures." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Blip2 dtype propagation and RTDetrV2 output mismatch are different model-specific issues." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both involve special tokens, but BigBird mask-token registration and mdeberta BOS/EOS omission are distinct tokenizer bugs." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Chunked generation with compiled forward and Qwen3.5 recurrent-state reset are different generation-path bugs." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Custom model initialization breakage and FSDP CPU RAM efficient loading are separate startup/loading regressions." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Offline model_info patching and est-roberta tokenizer loading failures are both tokenizer-related but not the same underlying issue." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Aux-loss normalization and T5 apex RMSNorm memory leak are unrelated model implementation bugs." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency and a MacOS CLIP bus error are not the same failure mode." + }, + { + "left": "issue:43756", + "right": "issue:46402", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and Sam3 text-encoder weights not loading are different model config/loading problems." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "Missing BigBird special-token registration and DiaConfig losing custom token IDs after save/load are distinct tokenizer/state bugs." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "MoE auxiliary-loss normalization and DeepseekV3MoE divergence are separate algorithm/implementation issues." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "torch.compile with SDPA in Bamba and DiaConfig token-ID loss are unrelated." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "One is a MoE loss-scaling bug; the other is a broad tokenization regression report." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression and num_labels propagation are different Qwen3.5 problems." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Bamba SDPA/compile failure and MLukeTokenizer AttributeError have no shared code path." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Gemma3 sliding-window NaNs and classification-head NaNs on checkout/init are separate numerical issues." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Compiled-forward chunked generation and flash-attention generation failure are different generation bugs." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Multi-GPU Qwen2-VL training and GenerationMixin import error are unrelated." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "MobileLLM key error and AutoImageProcessor URL loading are different loader failures." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "NoneType gpu broadcast in Trainer and Qwen2-VL multi-GPU training are not the same issue." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert and chunked generation with compiled forward are different runtime paths." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Missing config fields with trust_remote_code and removal of is_torch_fx_available are separate trust_remote_code regressions." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "lm_head.weight serialization inconsistency and sinq quantization save/load breakage are different persistence bugs." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Beit legacy-field loading and Bamba SDPA compile failure are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy-field model loading and packed-weight unpacking during accelerate loading are different model-loading bugs." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both concern apply_chat_template, but one is batched padding=False on Qwen2.5-VL and the other is tool-call assistant messages with no content." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Blip2 dtype propagation and an MPS attention correctness issue are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Beit legacy-field loading and OmDet-Turbo processor image-size mismatch are different processor/model configuration issues." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "BitNet packed-weight loading and DiaConfig custom token-ID loss are distinct." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch for bark-small and DiaConfig token-ID loss are unrelated." + }, + { + "left": "issue:44554", + "right": "issue:46365", + "accept": false, + "reason": "Attention correctness on MPS and Trainer import affected by datasets are entirely different problems." + }, + { + "left": "issue:44977", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 flash-attention generation failure and FP8 inference on B200 are different hardware/backend issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 24, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28218", + "issue:33357", + "issue:33666", + "issue:36331", + "issue:39692", + "issue:43257", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43452", + "issue:43493", + "issue:43525", + "issue:43540", + "issue:43572", + "issue:43646", + "issue:43688", + "issue:43701", + "issue:43720", + "issue:43824", + "issue:43825", + "issue:43844", + "issue:43927", + "issue:43994", + "issue:44361", + "issue:44451", + "issue:44466", + "issue:44488", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44843", + "issue:44849", + "issue:44964", + "issue:44977", + "issue:45357", + "issue:45405", + "issue:45520", + "issue:45584", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:46050", + "issue:46192", + "issue:46209", + "issue:46489" + ], + "soft_pairs": [ + "issue:43262|issue:43295", + "issue:33666|issue:36331", + "issue:43525|issue:43572", + "issue:43452|issue:44451", + "issue:43278|issue:43688", + "issue:43720|issue:44361", + "issue:45823|issue:46050", + "issue:43994|issue:46192", + "issue:43493|issue:43825", + "issue:44964|issue:45405", + "issue:39692|issue:43493", + "issue:44849|issue:46209", + "issue:44625|issue:45584", + "issue:28218|issue:33357", + "issue:44849|issue:45357", + "issue:39692|issue:43295", + "issue:43278|issue:43844", + "issue:43493|issue:43844", + "issue:44488|issue:44843", + "issue:43646|issue:44466", + "issue:44568|issue:46489", + "issue:33357|issue:33666", + "issue:43540|issue:44977", + "issue:44589|issue:45520", + "issue:43257|issue:43299", + "issue:44568|issue:45812", + "issue:43334|issue:43572", + "issue:43927|issue:44361", + "issue:43824|issue:44589", + "issue:43278|issue:43701", + "issue:43334|issue:45759" + ], + "prepared_review_unit_hash": "9b812e7c7abcd1b6dde8cd674ac938e5f0d40354f8531f4c001de45172cb8327", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36515, + "estimated_input_tokens": 9129, + "estimated_eval_tokens": 18514 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous: tokenizer regressions, model-loading failures, config attribute issues, and unrelated training/runtime bugs. None of the soft pairs look like true duplicates under a strict triage standard. The clearest representative issue is the Granite tokenizer regression (#45812).", + "confidence": 0.82, + "canonical_issue_reason": "#45812 is the most concrete, high-signal, and broadly scoped regression in the set: it describes a silent tokenizer bug affecting all Granite models with a clear v4\u2192v5 change. It is a better canonical representative than the narrower or more model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "#45812 is the best global representative issue because it is specific enough to be actionable, yet broad enough to represent the main tokenizer-regression theme present in the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Different subsystems and failure modes: audio chat-template sampling-rate defaulting vs processor/tokenizer access and image-tokenizer regression." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Qwen2-VL multi-GPU training is unrelated to a CustomTrainer compute_loss signature mismatch." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing pad token fields, but they affect different models and different config attributes; not the same underlying bug." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "Both are loading failures, but one is a gguf/auto-loader regression and the other is a specific model load issue for ScandiBERT." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Training/eval dtype mismatch is unrelated to auxiliary-loss normalization." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet packed-weight loading and MLukeTokenizer AttributeError are unrelated code paths." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Device-map forward mismatch is not the same as sinq quantized model save/load corruption." + }, + { + "left": "issue:43994", + "right": "issue:46192", + "accept": false, + "reason": "SigLIP2 inference quality issues are unrelated to return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "A SigLIP2 implementation discrepancy is unrelated to a pipeline error message bug." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Model-loading failure and a dependency-version bump are not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Docs example errors are not the same underlying problem as the SigLIP2 implementation discrepancy." + }, + { + "left": "issue:44849", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 hidden-state behavior and FP8 inference on B200 are unrelated issues." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and Whisper empty-transcription generation are different model/component bugs." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to a MacOS bus error in CLIP loading." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Both involve Qwen3.5, but one is hidden-states output and the other is save_pretrained visual-encoder key serialization." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "A SigLIP2 docs example bug is unrelated to the processor.tokenizer/image-tokenizer regression." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype changes and gradient explosion with ZeRO-3 are different training pathologies." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation mismatch is unrelated to ZeRO-3 gradient growth." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "A specific model load failure is not the same as AutoTokenizer offline-mode breakage in _patch_mistral_regex." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Custom model initialization breaks broadly, but inconsistent lm_head serialization is a different save/load bug." + }, + { + "left": "issue:44568", + "right": "issue:46489", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different models and different failure mechanisms; too broad to merge." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "MacOS bus error in CLIP loading and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Video-input processing in Qwen3OmniMoe is unrelated to Qwen3.5 flash-attention generation behavior." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Float8 storage lookup failure is unrelated to a flash_attn import_utils KeyError on Python 3.13." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both are Qwen3-MoE-related, but one is weight conversion with accelerate/deepspeed and the other is multimodal loading in dev; not the same bug." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but one is mDeBERTa special-token handling while the other is Granite token IDs; different models and fixes." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Missing pad_token_id in Qwen3-VL text config is not the same as missing pad_token_idx in StableLmConfig." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "DiaConfig custom token IDs lost on save/load is unrelated to MLukeTokenizer task AttributeError." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "An import error for Qwen2.5-VL is unrelated to a Float8 storage type error." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype drift in evaluation is unrelated to a resume_from_checkpoint key mismatch." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "pad_token_id missing in Qwen3-VL config is unrelated to AutoModelForCausalLM.from_config not unwrapping text_config." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 25, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29159", + "issue:29942", + "issue:30333", + "issue:32090", + "issue:34567", + "issue:38175", + "issue:42490", + "issue:43065", + "issue:43493", + "issue:43504", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43653", + "issue:43696", + "issue:43701", + "issue:43844", + "issue:43873", + "issue:43994", + "issue:44112", + "issue:44291", + "issue:44423", + "issue:44464", + "issue:44493", + "issue:44554", + "issue:44734", + "issue:44792", + "issue:44964", + "issue:45072", + "issue:45084", + "issue:45198", + "issue:45200", + "issue:45229", + "issue:45405", + "issue:45406", + "issue:45446", + "issue:45520", + "issue:45584", + "issue:45588", + "issue:45823", + "issue:46050", + "issue:46121", + "issue:46285", + "issue:46369", + "issue:46424" + ], + "soft_pairs": [ + "issue:43504|issue:43606", + "issue:44464|issue:46424", + "issue:43575|issue:45405", + "issue:43572|issue:44291", + "issue:44792|issue:45405", + "issue:45200|issue:45229", + "issue:43606|issue:44112", + "issue:44493|issue:46285", + "issue:43504|issue:43653", + "issue:45229|issue:45823", + "issue:44464|issue:45584", + "issue:43994|issue:44493", + "issue:29942|issue:30333", + "issue:43873|issue:46050", + "issue:43844|issue:44964", + "issue:45588|issue:46050", + "issue:43065|issue:46369", + "issue:42490|issue:43696", + "issue:29159|issue:32090", + "issue:44423|issue:44734", + "issue:45084|issue:45520", + "issue:28218|issue:34567", + "issue:44554|issue:45446", + "issue:15354|issue:38175", + "issue:45406|issue:46121", + "issue:43493|issue:43701", + "issue:43540|issue:45200", + "issue:45072|issue:45198" + ], + "prepared_review_unit_hash": "2fe8a7f6565e003e322f74ae69b6bf73cbea4a6693acfdf74c7f0a46dd7b1d28", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36113, + "estimated_input_tokens": 9029, + "estimated_eval_tokens": 18314 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is over-broad: the items span unrelated tokenizer, model-loading, serving, quantization, training, and CI failures. I rejected all soft duplicate edges because they do not describe the same concrete bug/change.", + "confidence": 0.96, + "canonical_issue_reason": "No single issue is a reliable duplicate hub; the set is too heterogeneous to justify one canonical issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the closest representative of the few related quantization/offloading reports, but it still is not a true duplicate target for the rest.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy field loading failure vs CPU offload device mismatch; different code paths and symptoms." + }, + { + "left": "issue:44464", + "right": "issue:46424", + "accept": false, + "reason": "Compiled-forward generation inconsistency vs cache-length config behavior; related area, but not the same bug." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "Tensor-parallel OOM vs PEFT version pin; unrelated issues." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Missing config field vs init_empty_weights TypeError; different loading failures." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Janus image-generation test failure vs PEFT version bump; no overlap in underlying problem." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Missing multimodal token IDs for fine-tuning vs multi-GPU inference OOM; distinct bugs." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Both mention device behavior, but one is a model offload mismatch and the other is a stale CI test." + }, + { + "left": "issue:44493", + "right": "issue:46285", + "accept": false, + "reason": "Unexpected position-id key handling vs CLIP text-model structure change; not the same issue." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Pretrained loading field issue vs tokenizer special-token registration; unrelated." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "OOM during multi-GPU inference vs PLE device mismatch in Gemma4 forward; different failure modes." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Compiled forward generation inconsistency vs Whisper empty-transcription failure after token alignment; different models and code paths." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 bad outputs vs position-id key warnings; not the same underlying bug." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention CI tests vs MLflow failure-reporting bug; unrelated." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Both involve quantization/offloading, but one is offload behavior and the other is save/load after sinq quantization; not mergeable as one fix." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "ZeRO-3 gradient anomaly vs Phi-4 multimodal load failure; different problems." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "Flash-attention sink-less AttributeError vs quantized save/load breakage; unrelated code paths." + }, + { + "left": "issue:43065", + "right": "issue:46369", + "accept": false, + "reason": "Dummy Conv2d implementation vs memory overflow in sam3-litetext; one is structural, the other is a runtime OOM." + }, + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Import warning for local paths vs GPT-oss OOM; no relation." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer fast/slow inconsistency vs Trainer broadcast NoneType error; unrelated subsystems." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Serve crashes in different multimodal/KV-cache paths; same surface area, but not the same concrete bug." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Template-node compilation error vs flash_attn import KeyError on Python 3.13; unrelated." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Added-token spacing bug vs TrainerState token counter not updating; different functionality." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness issue vs PyTorch version check for AuxRequest import; unrelated." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp export failure vs SigLIP2 zero-probability outputs; different models and failures." + }, + { + "left": "issue:45406", + "right": "issue:46121", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer vs rope-parameter JSON list TypeError; unrelated." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 implementation discrepancy vs resume_from_checkpoint key mismatch; different bug classes." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video-input processing error vs Gemma4 text-only token defaults; not the same issue." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "Bfloat16 dtype mismatch in inference vs Wav2Vec2 save/tokenization failure; unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 26, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28218", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:43232", + "issue:43295", + "issue:43334", + "issue:43450", + "issue:43452", + "issue:43493", + "issue:43525", + "issue:43550", + "issue:43572", + "issue:43575", + "issue:43653", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43950", + "issue:43957", + "issue:44112", + "issue:44186", + "issue:44466", + "issue:44521", + "issue:44554", + "issue:44610", + "issue:44792", + "issue:44849", + "issue:44898", + "issue:44987", + "issue:45127", + "issue:45200", + "issue:45290", + "issue:45356", + "issue:45381", + "issue:45412", + "issue:45520", + "issue:45563", + "issue:45823", + "issue:46178", + "issue:46192", + "issue:46209", + "issue:46285", + "issue:46512" + ], + "soft_pairs": [ + "issue:37428|issue:43957", + "issue:45200|issue:45823", + "issue:43295|issue:43450", + "issue:44186|issue:46192", + "issue:43232|issue:44849", + "issue:45823|issue:46192", + "issue:43295|issue:44610", + "issue:43295|issue:43873", + "issue:44521|issue:45290", + "issue:43572|issue:44987", + "issue:45520|issue:46178", + "issue:43550|issue:43844", + "issue:39692|issue:46192", + "issue:44554|issue:45356", + "issue:43572|issue:43957", + "issue:37428|issue:43824", + "issue:43450|issue:43575", + "issue:45381|issue:46209", + "issue:44466|issue:45127", + "issue:43950|issue:46285", + "issue:42490|issue:43493", + "issue:44898|issue:45412", + "issue:43334|issue:43525", + "issue:44792|issue:45563", + "issue:42915|issue:43824", + "issue:38175|issue:43493", + "issue:43653|issue:44112", + "issue:43452|issue:46512", + "issue:28218|issue:35141" + ], + "prepared_review_unit_hash": "37621dac511fdb9db72157b42ba84ee71158b9080f7bd952b2e990a5a4f1f6c8", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36147, + "estimated_input_tokens": 9037, + "estimated_eval_tokens": 18330 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is heterogeneous: the soft-similarity pairs mostly share broad subsystem words (loading, tokenizer, multimodal, config) but not the same concrete bug. I reject all proposed duplicate links; the strongest standalone issue in the set is a core `from_pretrained()` regression.", + "confidence": 0.88, + "canonical_issue_reason": "issue:43950 is the strongest representative report here: it describes a concrete, high-impact `from_pretrained()` regression with a clear failure mode and broad loading impact, unlike the more model-specific or UI/doc-related reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43950 is the best issue to anchor the cluster because it is the most general and actionable library regression in the set, with the broadest downstream effect.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "Different root causes: missing flash-attn import symbol vs. meta-device loading breakage." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Gemma4 fine-tuning input defaulting vs. PLE device mismatch during forward; different code paths." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor/tokenizer image-handling regression vs. batched video-shape bug; not the same failure." + }, + { + "left": "issue:44186", + "right": "issue:46192", + "accept": false, + "reason": "NER padding/truncation crash vs. `return_tensors` being ignored when `text_kwargs` is passed." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs after `sync_gpus` vs. Qwen3.5 hidden-state output bug; unrelated." + }, + { + "left": "issue:45823", + "right": "issue:46192", + "accept": false, + "reason": "Device-map/PLE mismatch is unrelated to `return_tensors` handling." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Different multimodal processor bugs: tokenizer/image path vs. OmDet-Turbo size mismatch." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor/tokenizer regression is not the same as quantization offloading behavior." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both hit chat templates, but one is all-zero assistant masks and the other is a tool-call/no-content crash." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Missing `pad_token_idx` in StableLmConfig vs. model-loading failure for `physical-intelligence/fast`; different bugs." + }, + { + "left": "issue:45520", + "right": "issue:46178", + "accept": false, + "reason": "Import registry failure for `flash_attn` vs. unsupported `tensorflow_text` backend mapping entry." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "`torch.compile`+SDPA bug in Bamba vs. DeepSpeed ZeRO-3 gradient inflation; different runtime/training issues." + }, + { + "left": "issue:39692", + "right": "issue:46192", + "accept": false, + "reason": "Doc example/model-processor mismatch and quantization failure vs. tokenizer argument handling; unrelated." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness issue vs. Kimi-K2.5 tokenizer codec regression; different layers." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Config field regression vs. meta-device model loading failure; different code paths." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different missing symbols/modules and different models." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video processor batch-shape bug vs. tensor-parallel OOM while loading Qwen2; unrelated." + }, + { + "left": "issue:45381", + "right": "issue:46209", + "accept": false, + "reason": "Qwen2.5-VL video position IDs bug vs. FP8 inference failure on B200; no shared concrete fix." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Serialization inconsistency of tied weights vs. LoRA merge collapse after vocab extension; related domain, different bug." + }, + { + "left": "issue:43950", + "right": "issue:46285", + "accept": false, + "reason": "Non-persistent buffer corruption in `from_pretrained()` vs. CLIP class structure inconsistency; different symptoms and code paths." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "Trailing-slash import warning vs. SigLIP2 HF-vs-JAX discrepancy; unrelated." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver interpolation failure vs. RT-DETR memory not releasing; different subsystems." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Same symptom class (`pad_token_id` missing), but different model configs and likely separate model-specific fixes." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure vs. stale paged-generate warning; not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3Moe FP8 failure vs. Qwen2_5_VL import error; different model/runtime problems." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Both concern SigLIP2, but one reports zero probabilities and the other a broader HF-vs-JAX discrepancy; not enough to treat as one bug." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "BigBird tokenizer special-token registration bug vs. GraniteSpeech CI stale override test; unrelated." + }, + { + "left": "issue:43452", + "right": "issue:46512", + "accept": false, + "reason": "GGUF `from_pretrained()` breakage vs. `from_config(dtype=...)` leaving stale config dtype; different loading paths." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer added-space behavior vs. resized-embedding reinit bug; different components entirely." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 27, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:29942", + "issue:32090", + "issue:37428", + "issue:38175", + "issue:41720", + "issue:43262", + "issue:43278", + "issue:43335", + "issue:43388", + "issue:43425", + "issue:43450", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43819", + "issue:43824", + "issue:43827", + "issue:43856", + "issue:43901", + "issue:43975", + "issue:44112", + "issue:44206", + "issue:44442", + "issue:44514", + "issue:44521", + "issue:44561", + "issue:44610", + "issue:44857", + "issue:45083", + "issue:45198", + "issue:45200", + "issue:45325", + "issue:45412", + "issue:45520", + "issue:45584", + "issue:45593", + "issue:45759", + "issue:46170", + "issue:46192", + "issue:46209", + "issue:46234", + "issue:46285", + "issue:46402", + "issue:46431", + "issue:46512", + "issue:46566" + ], + "soft_pairs": [ + "issue:43278|issue:43856", + "issue:43824|issue:43975", + "issue:43388|issue:43526", + "issue:43526|issue:46192", + "issue:45200|issue:45325", + "issue:44442|issue:45198", + "issue:43262|issue:43450", + "issue:43819|issue:44112", + "issue:41720|issue:46209", + "issue:38175|issue:45412", + "issue:44561|issue:46234", + "issue:43827|issue:43901", + "issue:45759|issue:46285", + "issue:44610|issue:44857", + "issue:44206|issue:45584", + "issue:43819|issue:45083", + "issue:43335|issue:43550", + "issue:44857|issue:45593", + "issue:45520|issue:46170", + "issue:43606|issue:45083", + "issue:37428|issue:43425", + "issue:43611|issue:46402", + "issue:45759|issue:46512", + "issue:44514|issue:44521", + "issue:29942|issue:32090", + "issue:43335|issue:43653", + "issue:46431|issue:46566" + ], + "prepared_review_unit_hash": "3fe922a61254b95f4251432ec28d3efa2cb7413eed97730afdf0c9b2c384779b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36006, + "estimated_input_tokens": 9002, + "estimated_eval_tokens": 18260 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The pairs are only superficially similar by keywords; they span unrelated bugs across audio, multimodal chat templates, import/version breaks, training, and device/runtime issues. I would not merge any of the soft edges.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 44521 is the most active and well-scoped item in the set, with multiple inbound references and discussion, so it is the best representative issue even though it is not a duplicate of the others.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the strongest standalone issue to keep as the cluster representative because it is concrete, reproducible, and has the most external references/activity in this group.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Different bugs: dtype mismatch during evaluation vs MoE training memory usage." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "Unrelated failures: missing Qwen2.5-VL import vs incorrect detokenization in DeepSeek Coder." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Both concern labels, but one drops tuple labels in metrics and the other truncates BEiT image labels." + }, + { + "left": "issue:43526", + "right": "issue:46192", + "accept": false, + "reason": "Different processor issues: label reduction bug vs silently ignored return_tensors with text_kwargs." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Different multimodal Qwen bugs: missing token-type defaults vs rope index scaling error." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Both are tokenizer/speech related, but they affect different models and different failure modes." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Different modalities and code paths: audio chat-template sampling rate vs video batch output shape." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "Distinct problems: DAC latent reconstruction vs a GraniteSpeech CI device-override test." + }, + { + "left": "issue:41720", + "right": "issue:46209", + "accept": false, + "reason": "Both are hardware/runtime failures, but they involve different models, precisions, and root causes." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "Completely different issues: zero probabilities in SigLIP2 vs RT-DETR memory reclamation." + }, + { + "left": "issue:44561", + "right": "issue:46234", + "accept": false, + "reason": "Both are import-utils breakages, but they are for different missing helpers and different downstream failures." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs cleanup around v5 pipelines, but they target different docs and different deprecated APIs." + }, + { + "left": "issue:45759", + "right": "issue:46285", + "accept": false, + "reason": "Different model-configuration bugs: unwrapping text_config vs inconsistent CLIP text model structure." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Different detection-related bugs: wrong processor image size vs AMP/CUDA crash in a loss implementation." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Different audio bugs: unsupported feature-extractor arg vs Whisper empty-transcription generation failure." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "Both mention feature-extraction helpers, but they are separate code paths and separate model families." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Unrelated: SwitchTransformers config layer construction vs Bamba torch.compile/SDPA failure." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "Different detection model bugs: mixed-precision crash vs missing auxiliary losses when denoising is off." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Different platform/version issues: Python 3.13 flash_attn lookup vs Windows ROCm distributed import failure." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "Different subsystems and symptoms: CPU-offload device mismatch vs Qwen3 Omni feature-length helper behavior." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Version compatibility problems, but one is a missing flash-attention symbol and the other is Torch 2.10 incompatibility." + }, + { + "left": "issue:43611", + "right": "issue:46402", + "accept": false, + "reason": "Both are model-loading regressions, but they affect different models and different loading stages." + }, + { + "left": "issue:45759", + "right": "issue:46512", + "accept": false, + "reason": "Both involve from_config in multimodal setups, but one is about unwrapping text_config and the other about stale dtype metadata." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Related to apply_chat_template, but one is batched padding behavior and the other is assistant mask generation." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Completely different failures: flash-attention tests vs Trainer GPU broadcast TypeError." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Different tokenizer/config bugs with no shared code path." + }, + { + "left": "issue:46431", + "right": "issue:46566", + "accept": false, + "reason": "Unrelated: missing chat template vs weights stuck on meta with auto device mapping." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 28, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:26413", + "issue:28218", + "issue:29127", + "issue:31501", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:42490", + "issue:43278", + "issue:43335", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43720", + "issue:43819", + "issue:43825", + "issue:43854", + "issue:44112", + "issue:44206", + "issue:44351", + "issue:44360", + "issue:44448", + "issue:44493", + "issue:44561", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:45072", + "issue:45083", + "issue:45127", + "issue:45200", + "issue:45245", + "issue:45290", + "issue:45310", + "issue:45381", + "issue:45459", + "issue:45478", + "issue:45588", + "issue:45593", + "issue:45823", + "issue:45859", + "issue:46234", + "issue:46326" + ], + "soft_pairs": [ + "issue:45290|issue:46326", + "issue:43550|issue:44112", + "issue:44448|issue:44743", + "issue:44855|issue:45083", + "issue:43335|issue:43504", + "issue:37428|issue:44351", + "issue:45200|issue:45588", + "issue:45127|issue:45245", + "issue:43278|issue:43825", + "issue:29127|issue:32090", + "issue:43335|issue:45083", + "issue:43720|issue:44112", + "issue:45588|issue:45859", + "issue:39692|issue:43526", + "issue:43819|issue:44493", + "issue:44206|issue:45823", + "issue:43335|issue:44112", + "issue:26413|issue:31501", + "issue:43335|issue:45593", + "issue:45310|issue:45478", + "issue:45459|issue:46234", + "issue:34689|issue:35141", + "issue:42490|issue:43854", + "issue:39692|issue:44493", + "issue:43606|issue:44857", + "issue:44448|issue:44779", + "issue:44360|issue:44561", + "issue:28218|issue:32090", + "issue:44857|issue:45072", + "issue:45290|issue:45381", + "issue:38175|issue:44493" + ], + "prepared_review_unit_hash": "aa848f6ee382b2f9c94a5717810ed6e0ba3d62600f59760022a3ffe14f302b62", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37244, + "estimated_input_tokens": 9311, + "estimated_eval_tokens": 18878 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous and does not form one duplicate set. The only clear duplicate pair is the Qwen3.5 MoE from_pretrained regression: issue 45310 and 45478 describe the same failure.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 45310 is the best canonical issue for the only true duplicate pair here: it matches 45478 on title and reported failure, and it is the earlier/more detailed report.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45310 is the strongest representative artifact in this cluster because it is the clearest exact duplicate target among the items and has the most complete description for that specific bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45290", + "right": "issue:46326", + "accept": false, + "reason": "Both are tokenizer/chat-template bugs, but they concern different APIs and failure modes; not the same underlying issue." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Both mention CI/device-related failures, but they affect different models and code paths with different root causes." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Different regressions in different model implementations; no shared concrete bug path." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "One is a Python 3.13 import/parsing failure, the other is a helper output-length behavior issue; unrelated." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "SwitchTransformers sparse-layer config and BEiT pretrained loading are unrelated bugs." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols/modules and different causes." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Different models and different failure surfaces; not the same underlying bug." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge collapse and category-cardinality limits are unrelated problems." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Training/eval dtype drift is unrelated to the pipeline translation-task error message." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "Error-message clarity in LayoutLMv3 is unrelated to the Trainer NoneType broadcast failure." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "Different subsystems and different bug classes; no duplicate evidence." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "Both are CI failures, but one is packed-weight loading and the other is a stale device override test; not the same bug." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Both involve flash/offload-related code, but the concrete failure paths and affected components differ." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "Documentation/example failures and a BeitImageProcessorFast label bug are unrelated." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC latent/forward mismatch and unexpected position-id keys are different issues." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "Unsupported center argument in a feature extractor is unrelated to the Gemma4 PLE device mismatch." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "A configuration bug in SwitchTransformers is unrelated to a CI test failing due to stale device overrides." + }, + { + "left": "issue:26413", + "right": "issue:31501", + "accept": false, + "reason": "Both affect training startup, but the checkpoint resume iterator error and the FineWeb SLM startup failure are different root causes." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "Sparse-layer configuration and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Identical reported bug: both describe the same transformers>=5.4.0 Qwen3.5 MoE from_pretrained error, with matching title and failure context." + }, + { + "left": "issue:45459", + "right": "issue:46234", + "accept": false, + "reason": "Tokenizer protobuf-error masking and a missing import_utils availability helper are unrelated." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Model-loading breakage for Llama 3.2 Vision and embedding reinitialization after resize are different code-path problems." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path slash import warning and GLM-4.7 Flash test-loading failure are different issues." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 example errors and unexpected position-id keys are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch and float16 AMP CUDA crash are different failure modes." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus v4/v5 output regression and Deepseek tokenizer regression are unrelated." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer ReLU discussion and removal of is_torch_fx_available breaking trust_remote_code are unrelated." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer added-space behavior and Trainer broadcast NoneType error are unrelated." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "Different dtype-related crashes in different models/components; not the same underlying bug." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Tool-call chat-template crash and qwen2.5-vl video position-id bug are different issues." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Zero probabilities in SigLIP2 and unexpected position-id keys are not the same bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: 45310 and 45478 have identical titles and very similar failure descriptions, while the other soft pairs are conservatively rejected with issue-specific reasons. The canonical choice is also reasonable and not overstated." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 29, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:25251", + "issue:27957", + "issue:29942", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:40444", + "issue:41720", + "issue:42490", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43412", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43701", + "issue:43746", + "issue:43873", + "issue:43927", + "issue:43976", + "issue:44448", + "issue:44466", + "issue:44561", + "issue:44568", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:44987", + "issue:45083", + "issue:45229", + "issue:45440", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45698", + "issue:45823", + "issue:45920", + "issue:46192", + "issue:46209", + "issue:46399" + ], + "soft_pairs": [ + "issue:43701|issue:45083", + "issue:40444|issue:43526", + "issue:45920|issue:46192", + "issue:32090|issue:33357", + "issue:29942|issue:35141", + "issue:43550|issue:45083", + "issue:42490|issue:43575", + "issue:44448|issue:44977", + "issue:44855|issue:46399", + "issue:44977|issue:45229", + "issue:43927|issue:45698", + "issue:27957|issue:34689", + "issue:10000|issue:15354", + "issue:43746|issue:46209", + "issue:37428|issue:44987", + "issue:43278|issue:43329", + "issue:44857|issue:45588", + "issue:44561|issue:44779", + "issue:43278|issue:43381", + "issue:43873|issue:45229", + "issue:44977|issue:45563", + "issue:41720|issue:45229", + "issue:43976|issue:45440", + "issue:25251|issue:32090", + "issue:32090|issue:33290", + "issue:44568|issue:45491", + "issue:45588|issue:45823", + "issue:44448|issue:44466", + "issue:43412|issue:45588" + ], + "prepared_review_unit_hash": "94eeded5e9b831a9ead200450df1c1a22b988174a5bd547175f3366341106c2b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35960, + "estimated_input_tokens": 8990, + "estimated_eval_tokens": 18236 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster does not contain clear duplicates: the soft pairs span unrelated models, features, and failure modes. I would not merge any of the suggested edges; the most representative issue is the broad tokenizer regression report.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45920 is the strongest representative because it is an open, high-discussion regression affecting multiple tokenizers/models, making it broader and more central than the narrowly scoped bug reports elsewhere in the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the best cluster anchor for triage: it has the widest blast radius and the strongest sign of an actual platform-wide regression, so it is the most useful issue to keep visible.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint key mismatch and a Qwen3 Omni feature-length helper bug are unrelated code paths." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "One is a Qwen2.5-VL training/data-loading failure; the other is a BEiT image processor label-reduction bug." + }, + { + "left": "issue:45920", + "right": "issue:46192", + "accept": false, + "reason": "Both touch tokenization, but one is a broad v5 token-ID regression while the other is a specific return_tensors argument-handling bug." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer broadcast NoneType error and a MacOS CLIP bus error are different failures in different execution environments." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures and token-embedding reinitialization in post_init are unrelated issues." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "A torch.compile/SDPA failure in Bamba-9B-v2 does not match a Qwen3 Omni feature-length helper bug." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "A trailing-slash import warning and a tensor-parallel OOM on Qwen2-57B are different problems." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus v4/v5 output drift and Qwen3.5 flash-attention generation failure are different model-specific regressions." + }, + { + "left": "issue:44855", + "right": "issue:46399", + "accept": false, + "reason": "Both involve import/decorator parsing, but they affect different models and different Torch/JIT failure modes." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Flash-attention generation corruption and a multi-GPU OOM are not the same underlying bug." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "A saved config losing custom token IDs is distinct from loading the wrong custom module after save_pretrained." + }, + { + "left": "issue:27957", + "right": "issue:34689", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 and Llama 3.2 vision model loading failures are unrelated." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel and torch.jit.script GeneratorExp export issues are different models and different root causes." + }, + { + "left": "issue:43746", + "right": "issue:46209", + "accept": false, + "reason": "PEFT local-checkpoint loading and FP8 inference on B200 are unrelated concerns." + }, + { + "left": "issue:37428", + "right": "issue:44987", + "accept": false, + "reason": "A missing flash-attention utility import and a model-loading failure are too generic to be the same bug." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype drift in eval and undefined video-token helper names are unrelated." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "float16 AMP loss crashes and a sink-less Flash Attention attribute error are different attention/precision bugs." + }, + { + "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." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Embedding dtype changes and gradient checkpointing in eval mode are separate issues." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Quantization offloading behavior and Gemma4 multi-GPU CUDA OOM are not the same failure." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Flash-attention generation failure and a stale num_return_sequences warning are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3 device-mapping cudaErrorAssert and Gemma4 multi-GPU OOM are distinct distributed-inference failures." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "A Python version compatibility issue and a DeepSeekV3MoE implementation divergence are different kinds of problems." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline top_k nesting behavior and Trainer GPU broadcast errors are unrelated." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer broadcast NoneType and deepspeed Adafactor OOM are different runtime failures." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer special-token behavior and Gemma3 sliding-window NaNs are unrelated bugs." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "A flash-attention attribute crash and a Gemma4 PLE device mismatch are different forward-pass issues." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Pegasus output drift and lm_head.weight serialization inconsistency are both v5 regressions, but not the same underlying bug." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export/training-guard failures and flash-attention sink-less crashes are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 30, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:27957", + "issue:29127", + "issue:29942", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43381", + "issue:43421", + "issue:43575", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43873", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44442", + "issue:44462", + "issue:44479", + "issue:44743", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45593", + "issue:45712", + "issue:45803", + "issue:46121", + "issue:46369" + ], + "soft_pairs": [ + "issue:45198|issue:45803", + "issue:45372|issue:45588", + "issue:42490|issue:43873", + "issue:33357|issue:39692", + "issue:45245|issue:45563", + "issue:29127|issue:34567", + "issue:45372|issue:45406", + "issue:43611|issue:46121", + "issue:43653|issue:44743", + "issue:43278|issue:43575", + "issue:44361|issue:45198", + "issue:44871|issue:44977", + "issue:42915|issue:42994", + "issue:44186|issue:45198", + "issue:44206|issue:44479", + "issue:43065|issue:45412", + "issue:43065|issue:43994", + "issue:43381|issue:43701", + "issue:45459|issue:45712", + "issue:29942|issue:30064", + "issue:45412|issue:46369", + "issue:43756|issue:45440", + "issue:44265|issue:45198", + "issue:45372|issue:45459", + "issue:43653|issue:44442", + "issue:43994|issue:45412", + "issue:43066|issue:44462", + "issue:27957|issue:29942", + "issue:29127|issue:33357", + "issue:45440|issue:45593", + "issue:43421|issue:45491" + ], + "prepared_review_unit_hash": "7562140989c232c3cb497573df58e41edb1be2b16dabd70a38bbea5ba36b9c5a", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36391, + "estimated_input_tokens": 9098, + "estimated_eval_tokens": 18452 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is highly heterogeneous; most soft pairs only share broad themes like tokenizers, quantization, or model loading, but not the same concrete bug. I would not merge any of the proposed soft edges. The closest thing to a hub is the Gemma 4 processor-loading issue, but even that only loosely relates to two other Gemma 4 reports.", + "confidence": 0.43, + "canonical_issue_reason": "issue:45372 is the most central concrete bug report in this set: it has a clear failure mode, related inbound references, and sits at the center of the only small sub-thread (Gemma 4 processor loading).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45372 is the best overall representative issue because it is the most actionable and specific bug report, while the rest are mostly isolated reports from unrelated subsystems.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Both are tokenizer/data-handling bugs, but one is Wav2Vec2 save/tokenization failure and the other is a bare-except bug in Fuyu conversion; different code paths and fixes." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Both mention Gemma 4-ish loading/runtime issues, but one is a missing ReasoningEffort import during processor loading and the other is a flash-attention crash on sink-less models; not the same bug." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Local-path import warning vs quantization/offloading behavior; same broad loading area but different underlying failures and fixes." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "Community CLIP bus error on macOS and SigLIP2 docs example errors are unrelated failures; one is runtime stability, the other is documentation/example correctness." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-count runtime error and a stale generate warning are unrelated issues with different triggers and code paths." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error messaging and TrainerState token-count tracking are unrelated subsystems and bugs." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both involve Gemma 4 processor loading, but one is an import dependency problem and the other is a missing _tokenizer attribute in serve; distinct fixes." + }, + { + "left": "issue:43611", + "right": "issue:46121", + "accept": false, + "reason": "Base-model-prefix loading regression and rope-validation TypeError are separate configuration/serialization bugs." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "Tokenizer special-token registration and Qwen3.5 recurrent-state reset are unrelated model/tokenization behaviors." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "Mixed precision dtype drift during eval and tensor-parallel OOM are different problems in different execution phases." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "MLukeTokenizer AttributeError and Wav2Vec2 save/tokenization failure are different tokenizer implementations with no shared concrete path." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Gemma-3 eos_token_id mismatch and Qwen3.5 flash-attention generation failure are unrelated model-specific bugs." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Both touch quantization, but one is FP8 config loading for Qwen3Moe and the other is quantized model saving; different stages and likely different fixes." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2 tokenization crashes and Wav2Vec2 tokenization/save failure are separate tokenizers and failure modes." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Feature extractor argument regression and Qwen video-input regression are unrelated code paths." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d placeholder in Sam3PixelDecoder and RT-DETR memory not released are not the same bug." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "A dummy layer implementation issue and SigLIP2 nonsensical outputs are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Eval-mode gradient checkpointing restriction and resume_from_checkpoint key mismatch are different training-loop bugs." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Hidden tokenizer errors from missing protobuf and leftover dummy classes are both cleanup/error-handling issues, but not the same underlying problem." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention 2 test failures and void segmentation-map processing are unrelated issues." + }, + { + "left": "issue:45412", + "right": "issue:46369", + "accept": false, + "reason": "Both are memory-related reports, but RT-DETR deletion GC behavior and SAM3-litetext overflow are different models and failure sources." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and DeepseekV3MoE divergence are separate architecture/implementation discrepancies." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export exportability and Wav2Vec2 tokenization/save failure do not share a concrete code path." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Gemma 4 processor loading failure and a protobuf-related tokenizer exception hiding bug are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and AutoTokenizer loading FastSpeech2ConformerTokenizer are different tokenizer-loading problems." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 inference corruption and RT-DETR memory leak are different model/runtime issues." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Wrong tokenizer decoder type in v5 and AutoTokenizer ignoring tokenizer.json are both tokenizer-loading related, but they are distinct bugs and fixes." + }, + { + "left": "issue:27957", + "right": "issue:29942", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 support and failing Flash Attention 2 tests are related only at a high level; not the same concrete issue." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 error clarity and CLIP bus error on macOS are entirely different bugs." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "DeepSeekV3MoE implementation divergence and D-FINE auxiliary-loss omission are unrelated model-specific issues." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime special-token post-processor updates and Gemma3 NaN embeddings from sliding-window attention are distinct tokenizer/model-runtime bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 31, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:27957", + "issue:29127", + "issue:33290", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43334", + "issue:43381", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43611", + "issue:43646", + "issue:43701", + "issue:43761", + "issue:43824", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44062", + "issue:44077", + "issue:44186", + "issue:44206", + "issue:44279", + "issue:44387", + "issue:44442", + "issue:44464", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44898", + "issue:44977", + "issue:45003", + "issue:45072", + "issue:45245", + "issue:45440", + "issue:45491", + "issue:45588", + "issue:45600", + "issue:45834", + "issue:45859", + "issue:45907", + "issue:46121", + "issue:46369", + "issue:46399" + ], + "soft_pairs": [ + "issue:43526|issue:45834", + "issue:29127|issue:35141", + "issue:43575|issue:43701", + "issue:43994|issue:46369", + "issue:42490|issue:43381", + "issue:43526|issue:44186", + "issue:44442|issue:45072", + "issue:43575|issue:44387", + "issue:43824|issue:44279", + "issue:42915|issue:43931", + "issue:43526|issue:44857", + "issue:43334|issue:43531", + "issue:46121|issue:46399", + "issue:43761|issue:45859", + "issue:27957|issue:29127", + "issue:44977|issue:45245", + "issue:27957|issue:33290", + "issue:43262|issue:44206", + "issue:33290|issue:34689", + "issue:45588|issue:45600", + "issue:44464|issue:45491", + "issue:43646|issue:44077", + "issue:33290|issue:35141", + "issue:33290|issue:34567", + "issue:43381|issue:43873", + "issue:44743|issue:44811", + "issue:44442|issue:44898", + "issue:45003|issue:45440", + "issue:42490|issue:44743", + "issue:44062|issue:45907", + "issue:34689|issue:43611" + ], + "prepared_review_unit_hash": "37983b0f1d58e5f8b212240f74b981bf672b05accdb41001eb474b918c10f4f7", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36862, + "estimated_input_tokens": 9216, + "estimated_eval_tokens": 18688 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated, model-specific bug reports with only superficial lexical overlap. None of the soft-similarity pairs look like the same underlying bug, so I would not merge any of them as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": "issue:43873 is the strongest representative issue: it is broad, user-facing, still open, and has inbound references; the rest are mostly narrow, model-specific regressions or errors.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best overall issue candidate because it has the widest apparent scope and the most ongoing signal among this mixed set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "Both are vision/OCR-related, but one is a label-reduction bug in BeitImageProcessorFast and the other is an OCR long-input index error in Kosmos2.5; different failure modes and code paths." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and token-embedding reinitialization are unrelated bugs in different components." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "One is an OOM when loading a large model with tensor parallelism, the other is a checkpoint key mismatch; no shared code-path bug." + }, + { + "left": "issue:43994", + "right": "issue:46369", + "accept": false, + "reason": "SigLIP2 wrong outputs via AutoModel/pipeline and SAM3-LiteText memory overflow are different model-specific issues." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "A local-path trailing-slash import warning is unrelated to gradient checkpointing being disallowed in eval mode." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and LayoutLMv2Tokenizer NER/padding crashes are separate preprocessing bugs." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "FastSpeech2ConformerTokenizer loading and dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention OOM, but one is a specific tensor-parallel load failure and the other is a quantization memory regression in 5.x; different problems." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Missing Qwen2.5-VL import symbol versus a vague dependency issue; not the same concrete bug." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "FineGrainedFP8Config failure in Qwen3-MoE and shape-mismatch loading in Qwen3-VL are different model/config issues." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "Label reduction in an image processor and AMP/CUDA crash in LwDetrImageLoss are unrelated." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Qwen3-VL missing pad_token_id and Qwen3-MoE sliding_window behavior are distinct model-family regressions." + }, + { + "left": "issue:46121", + "right": "issue:46399", + "accept": false, + "reason": "Rope-validation TypeError from JSON-loaded lists and a missing-decorator crash in Qwen3.5 kernels are different code paths." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression and Qwen3_5MoeTextRotaryEmbedding CPU-offload incompatibility are unrelated." + }, + { + "left": "issue:27957", + "right": "issue:29127", + "accept": false, + "reason": "Flash Attention 2 support for XLMRoberta and LayoutLMv3 error-message clarity are unrelated issues." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug and a categorical-limit runtime error are different failures." + }, + { + "left": "issue:27957", + "right": "issue:33290", + "accept": false, + "reason": "XLMRoberta Flash Attention 2 support and DeepSpeed Adafactor OOM do not share the same underlying bug." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Sampling-rate defaulting in audio chat templates and an unsupported `center` arg in LasrFeatureExtractor are separate audio bugs." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "AdaFactor/DeepSpeed OOM and Llama 3.2 Vision model-loading breakage are different failures despite both being resource/loading related." + }, + { + "left": "issue:45588", + "right": "issue:45600", + "accept": false, + "reason": "A flash-attention runtime AttributeError and stale auto_mappings references are unrelated; one is execution-time, the other is CI/config drift." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency and Gemma3 NaNs from sliding-window/all-padding windows are different bugs." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Custom model initialization breakage in Transformers 5.0.0 and patchtsmixer's optional post_init policy are not the same code-path problem." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "DeepSpeed OOM with Adafactor is unrelated to output-embedding reinitialization after resizing token embeddings." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "An optimizer memory issue and TrainerState token-count tracking bug are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing in eval mode and quantization offloading failures are different training/runtime behaviors." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Qwen3.5 recurrent-state reset and Whisper batch_decode skip_special_tokens handling are unrelated model/processor bugs." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading failure and Perceiver interpolation failure are different subsystems and failure modes." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils and DeepseekV3MoE divergence from remote implementation are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken keyword conflict and list_repo_files revision forwarding are separate API bugs." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both are loading regressions, but they concern different release versions and distinct loading mechanisms; 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": 21, + "review_unit_index": 32, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:25251", + "issue:28282", + "issue:42175", + "issue:42617", + "issue:42796", + "issue:42907", + "issue:43232", + "issue:43316", + "issue:43388", + "issue:43493", + "issue:43525", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43643", + "issue:43701", + "issue:43720", + "issue:43723", + "issue:43749", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44190", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44496", + "issue:44855", + "issue:44871", + "issue:44913", + "issue:44987", + "issue:45137", + "issue:45216", + "issue:45290", + "issue:45479", + "issue:45698", + "issue:45759", + "issue:45803", + "issue:46178", + "issue:46234", + "issue:46291", + "issue:46421", + "issue:46506" + ], + "soft_pairs": [ + "issue:43316|issue:43525", + "issue:42175|issue:46234", + "issue:43723|issue:45479", + "issue:43720|issue:44442", + "issue:43761|issue:44079", + "issue:44077|issue:44479", + "issue:43761|issue:44188", + "issue:43493|issue:43827", + "issue:43232|issue:43388", + "issue:44987|issue:46178", + "issue:42907|issue:45698", + "issue:44442|issue:45803", + "issue:43873|issue:44871", + "issue:43931|issue:45216", + "issue:43316|issue:45759", + "issue:43749|issue:43931", + "issue:44855|issue:45137", + "issue:43701|issue:46506", + "issue:43643|issue:44913", + "issue:44464|issue:46421", + "issue:44190|issue:44496", + "issue:46234|issue:46291", + "issue:42796|issue:45290", + "issue:43550|issue:43701", + "issue:25251|issue:28282", + "issue:42617|issue:43575", + "issue:43606|issue:43701" + ], + "prepared_review_unit_hash": "ce99dc6583b0835a8afa407e4de0eb78c4b2b894342f78a3e4626156cabad617", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35399, + "estimated_input_tokens": 8850, + "estimated_eval_tokens": 17956 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is highly heterogeneous; the soft links mostly connect superficially similar issues rather than true duplicates, so no PRs are present and no soft edge should be merged.", + "confidence": 0.69, + "canonical_issue_reason": "issue:43701 is the most central node in the soft-link graph and is a concrete, reproducible bug, but it only serves as a weak anchor because the surrounding issues span unrelated failure modes.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43701 is the best single representative for triage due to its higher connectivity and clear bug report, though it is not a true duplicate of the other items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config/API mismatches, but they concern different models and different missing fields; not the same underlying bug." + }, + { + "left": "issue:42175", + "right": "issue:46234", + "accept": false, + "reason": "One is about backend installation extras, the other about a missing import-utils symbol; different code paths and fixes." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer loading in v5 and degenerate classification loss are unrelated failures in different layers." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "Accelerate packed-weight loading vs. tokenizer loading; no shared code-path bug." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "A model forward missing hidden states is distinct from ModelOutput key assignment behavior." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "Optional post_init enforcement in one model has nothing to do with the Qwen video-input regression." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "Hidden-states output regression and torch.compile attention-kernel divergence are different problems." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation parity and stale docs referencing pipeline() are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs syncing and gather_for_metrics label dropping affect different subsystems and behaviors." + }, + { + "left": "issue:44987", + "right": "issue:46178", + "accept": false, + "reason": "Model loading failure and missing tensorflow_text backend mapping are separate dependency/import issues." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "One concerns dequantized model saving, the other custom-module resolution after save/load; similar theme but different bugs." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer loading and a broad exception in Fuyu tensor conversion are unrelated." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading behavior is not the same as eos_token_id configuration mismatch." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Weight-shape mismatch on load differs from incorrect checkpoint serialization; similar model family but distinct failures." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Config API discrepancy and multimodal text_config unwrapping are different configuration bugs." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and a specific model weight mismatch are not the same defect." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "A Python 3.13 TorchScript parsing issue is unrelated to a ZeRO3 deque underflow." + }, + { + "left": "issue:43701", + "right": "issue:46506", + "accept": false, + "reason": "Resume-checkpoint key mismatch and CPU/CUDA result divergence are different bug classes with no clear common fix." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Missing fields from AutoConfig with trust_remote_code and GPTNeoX rotary_pct reload behavior are separate config-loading issues." + }, + { + "left": "issue:44464", + "right": "issue:46421", + "accept": false, + "reason": "Both mention chunked generation/prefill, but one is inconsistent outputs and the other is spurious recompilation; not the same code-path bug." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in a training script and unrecognized model config are unrelated." + }, + { + "left": "issue:46234", + "right": "issue:46291", + "accept": false, + "reason": "A missing utility function and an import failure due to kernels version constraints are different import-time problems." + }, + { + "left": "issue:42796", + "right": "issue:45290", + "accept": false, + "reason": "Tokenizer BPE marker handling and chat-template tool-call crashes are distinct tokenizer behaviors." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile SDPA failure in Bamba is unrelated to checkpoint resume key mismatch." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k nesting behavior and an AutoModel PyTorch import error are completely unrelated." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "3d_parallel.py runtime failure and model-load OOM under tensor parallelism are different issues." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU-offload device mismatch and resume_from_checkpoint key mismatch 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": 21, + "review_unit_index": 33, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:27957", + "issue:32090", + "issue:41950", + "issue:42175", + "issue:43065", + "issue:43295", + "issue:43316", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43452", + "issue:43475", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43646", + "issue:43650", + "issue:43688", + "issue:43716", + "issue:43742", + "issue:43761", + "issue:43825", + "issue:44077", + "issue:44162", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44792", + "issue:44871", + "issue:44913", + "issue:45127", + "issue:45200", + "issue:45584", + "issue:45632", + "issue:45803", + "issue:46050", + "issue:46170", + "issue:46178", + "issue:46202", + "issue:46365", + "issue:46369", + "issue:46424" + ], + "soft_pairs": [ + "issue:43295|issue:43825", + "issue:43575|issue:43742", + "issue:27957|issue:32090", + "issue:16998|issue:22355", + "issue:43316|issue:43716", + "issue:43295|issue:44871", + "issue:42175|issue:43316", + "issue:44871|issue:45584", + "issue:44077|issue:44534", + "issue:45632|issue:46424", + "issue:45803|issue:46365", + "issue:44610|issue:46369", + "issue:44162|issue:44792", + "issue:43646|issue:44913", + "issue:44871|issue:46050", + "issue:43650|issue:46202", + "issue:45803|issue:46170", + "issue:41950|issue:43475", + "issue:43065|issue:44617", + "issue:43540|issue:44514", + "issue:43452|issue:44496", + "issue:43404|issue:43441", + "issue:43761|issue:44479", + "issue:43388|issue:43688", + "issue:45127|issue:45200", + "issue:43572|issue:46178" + ], + "prepared_review_unit_hash": "485c92a4353b26cb6d4309b1847f07053a8761787873ad251268f031ad779d31", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 34758, + "estimated_input_tokens": 8690, + "estimated_eval_tokens": 17636 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a set of soft-similarity false positives rather than true duplicates: the paired issues cover unrelated models, pipelines, config regressions, and import/runtime failures. No PRs are present, and no single issue is a clean canonical target.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the soft pairs are mostly unrelated bugs across different models and subsystems, so keeping one representative would be misleading.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue: there is no single issue that plausibly subsumes the rest of this mixed cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Different regressions: processor.tokenizer/image handling vs an error-message mismatch about translation tasks." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "Different loading failures for different models; one is TP/OOM, the other is a key error on MobileLLM." + }, + { + "left": "issue:27957", + "right": "issue:32090", + "accept": false, + "reason": "Unrelated problems: FlashAttention support for XLMRoberta vs Trainer broadcast failing on NoneType." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "A model max_length question is unrelated to a missing transformers.onnx module import error." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Different code paths and symptoms: Gemma3TextConfig API mismatch vs image preprocessor/model dtype mismatch." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Both are regressions, but they affect different APIs and model families; not the same underlying bug." + }, + { + "left": "issue:42175", + "right": "issue:43316", + "accept": false, + "reason": "Backend packaging issue for torch extras is unrelated to a Gemma config API discrepancy." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma eos_token_id inconsistency and Whisper empty-transcription failure are different bugs." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "One is about disallowing optional post_init on a specific model; the other is a buffer initialization regression." + }, + { + "left": "issue:45632", + "right": "issue:46424", + "accept": false, + "reason": "Both mention cache, but one is a local trust_remote_code path collision and the other is max_cache_length semantics." + }, + { + "left": "issue:45803", + "right": "issue:46365", + "accept": false, + "reason": "Bare except swallowing errors is unrelated to a Trainer import conflict with datasets." + }, + { + "left": "issue:44610", + "right": "issue:46369", + "accept": false, + "reason": "Different models and failures: OmDet-Turbo image size mismatch vs Sam3-LiteText memory overflow." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 breakage is unrelated to a Janus image generation test failure." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "A broad Transformers 5 initialization regression is not the same as GPTNeoX rotary_pct not persisting on reload." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "Gemma eos_token_id config mismatch and sinq save/load quantization breakage are different issues." + }, + { + "left": "issue:43650", + "right": "issue:46202", + "accept": false, + "reason": "The placeholder issue title is unrelated to a Gemma pipeline roundtrip failure." + }, + { + "left": "issue:45803", + "right": "issue:46170", + "accept": false, + "reason": "Exception handling in Fuyu conversion is unrelated to the Windows/ROCm torch.distributed import problem." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Both are video-adjacent, but one is a pipeline processor lookup bug and the other is a missing output attribute in Sam3Video." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Same family name, but one is a dummy Conv2d implementation detail and the other is a CUDA OOM symptom." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Different Qwen video/chat-template failures with different inputs and failure modes." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "Tokenizer/model loading with gguf_file is unrelated to an unrecognized model_type in config.json." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Both are Mistral-family, but one is tied weights and the other is FlashAttention compatibility; distinct bugs." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "CLIPVision hidden_states regression is unrelated to the Qwen video-input regression." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric gathering dropping labels is unrelated to auxiliary-loss normalization." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA/vocab-merge collapse and Gemma4 mm_token_type_ids defaulting are different model behaviors." + }, + { + "left": "issue:43572", + "right": "issue:46178", + "accept": false, + "reason": "StableLm pad_token_idx config and tensorflow_text backend registration are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 34, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:25251", + "issue:27957", + "issue:35141", + "issue:41720", + "issue:42490", + "issue:43065", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43550", + "issue:43644", + "issue:43673", + "issue:43761", + "issue:43827", + "issue:43931", + "issue:43976", + "issue:43994", + "issue:44060", + "issue:44077", + "issue:44190", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44933", + "issue:44964", + "issue:44977", + "issue:45245", + "issue:45362", + "issue:45381", + "issue:45406", + "issue:45542", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:46234", + "issue:46369", + "issue:46421" + ], + "soft_pairs": [ + "issue:43673|issue:44743", + "issue:45542|issue:46234", + "issue:43504|issue:45698", + "issue:43525|issue:43531", + "issue:44792|issue:44977", + "issue:44734|issue:45406", + "issue:44617|issue:45584", + "issue:44792|issue:45600", + "issue:44534|issue:44913", + "issue:44190|issue:44964", + "issue:43673|issue:46421", + "issue:43931|issue:44060", + "issue:45381|issue:46369", + "issue:35141|issue:42490", + "issue:43761|issue:44242", + "issue:45563|issue:45584", + "issue:44855|issue:45588", + "issue:43827|issue:43994", + "issue:44479|issue:44734", + "issue:43065|issue:44610", + "issue:43644|issue:44913", + "issue:41720|issue:45245", + "issue:44933|issue:45362", + "issue:43976|issue:44855", + "issue:44464|issue:45563", + "issue:43550|issue:44190", + "issue:44617|issue:46369", + "issue:25251|issue:27957", + "issue:44077|issue:44877" + ], + "prepared_review_unit_hash": "220fc7a238e7af3b81be61df4d16c7f351a728805d6fa1279dd8f30d9bd2c09a", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36200, + "estimated_input_tokens": 9050, + "estimated_eval_tokens": 18356 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly heterogeneous issue reports across different bugs and subsystems. All supplied soft pairs look like false positives; there is no pair here that clearly describes the same underlying bug/change. The only obvious duplicate theme in the item list is the v5 non-persistent-buffer regression, but that duplicate pair was not provided as a soft edge.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 43644 is the best canonical anchor among this set because it describes a concrete, broad regression and is the clearest representative of the only obvious duplicate theme in the list.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43644 is the strongest overall representative: specific enough to identify the bug, yet broad enough to stand in for the related v5 buffer-corruption reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both mention cache-related generation behavior, but one is a missing GenerationMixin cache in chunked prefill and the other is Qwen3.5 recurrent-state reset; different bugs and code paths." + }, + { + "left": "issue:45542", + "right": "issue:46234", + "accept": false, + "reason": "Completely different problems: a TensorBoard/TensorFlow backend import issue versus a missing import_utils helper for falcon-mamba availability." + }, + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Unrelated failures: BEiT semantic segmentation legacy-field loading versus from_pretrained selecting the wrong custom module after save_pretrained." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Different model bugs: missing pad_token_id on Llama4Config versus a sliding_window issue in Qwen3-MoE." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Both are model-generation failures, but on different models and different mechanisms: Janus test failure versus Qwen3.5 flash-attention behavior." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Serve KV-cache tensor indexing crash and Gemma4Processor missing _tokenizer are unrelated code paths and symptoms." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "CUDA OOM in Sam3Video versus Whisper empty-transcription failure; different models and failure modes." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "A Janus generate-images test failure and a CI break from removed Sam3LiteText configs are not the same bug." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "One is the v5 non-persistent-buffer corruption regression; the other is GPTNeoXConfig rotary_pct not persisting on reload." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading in a training script is unrelated to model-loading failure for microsoft/Phi-4-multimodal-instruct." + }, + { + "left": "issue:43673", + "right": "issue:46421", + "accept": false, + "reason": "Both involve chunked prefill/cache, but one is missing cache in v5.0.0 and the other is lazy cache init causing recompilation; distinct defects." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch and Qwen3-Next tied-weights warning are separate model/configuration issues." + }, + { + "left": "issue:45381", + "right": "issue:46369", + "accept": false, + "reason": "Qwen2.5-VL video position IDs regression and Sam3-LiteText memory overflow are different models and problems." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Embedding resize reinitialization and local-path trailing-slash import warning are unrelated." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "CLIPVision hidden_states regression and missing load-balancing loss when output_router_logits=False affect different components." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "A stale warning in paged generate versus Whisper failing on empty transcription after align_special_tokens; different bug surfaces." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 IndentationError in a scripted model file and a flash_attention AttributeError on s_aux=None are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() and SigLIP2 nonsensical outputs with AutoModel/pipeline are not the same issue." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Video-input regression across Qwen VL variants and a serve KV-cache indexing crash are different code paths." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder versus OmDet-Turbo processor output-size mismatch; not the same bug." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "The v5 non-persistent-buffer junk issue is unrelated to GPTNeoX rotary_pct not surviving reload." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3 auto device mapping cuda assert and the 2^24 category limit error are unrelated runtime failures." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "A missing import from image_utils and Qwen3.5-35B crashing in transformers chat are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "A Python-version support regression and a Python 3.13 parsing/IndentationError in a scripted function are different problems." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency and a stale num_return_sequences warning do not share the same underlying bug." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "Bamba-9B-v2 compile/SDPA failure and local dataset loading failure in an example script are unrelated." + }, + { + "left": "issue:44617", + "right": "issue:46369", + "accept": false, + "reason": "Both are memory-related, but one is Sam3Video OOM and the other is Sam3-LiteText memory overflow; different models and causes." + }, + { + "left": "issue:25251", + "right": "issue:27957", + "accept": false, + "reason": "Top-k nested-list output in a pipeline and XLMRoberta Flash Attention 2 support are unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "An optional post_init policy issue for patchtsmixer and a strict-config loading failure for granite_speech are separate configuration bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 35, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36246", + "issue:38617", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:42796", + "issue:43116", + "issue:43295", + "issue:43335", + "issue:43404", + "issue:43540", + "issue:43643", + "issue:43854", + "issue:44016", + "issue:44112", + "issue:44190", + "issue:44246", + "issue:44279", + "issue:44295", + "issue:44462", + "issue:44496", + "issue:44521", + "issue:44683", + "issue:44734", + "issue:44737", + "issue:44779", + "issue:44938", + "issue:44964", + "issue:45003", + "issue:45127", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45357", + "issue:45381", + "issue:45446", + "issue:45542", + "issue:45584", + "issue:45600", + "issue:45632", + "issue:45803", + "issue:45907", + "issue:46082", + "issue:46121", + "issue:46234", + "issue:46421", + "issue:46489", + "issue:46506" + ], + "soft_pairs": [ + "issue:45127|issue:45357", + "issue:43335|issue:44190", + "issue:44521|issue:45381", + "issue:44112|issue:45600", + "issue:43540|issue:44734", + "issue:44779|issue:46489", + "issue:43116|issue:44190", + "issue:44737|issue:45803", + "issue:44938|issue:45542", + "issue:43295|issue:45276", + "issue:44462|issue:45907", + "issue:45632|issue:46234", + "issue:45276|issue:45584", + "issue:42796|issue:44521", + "issue:36246|issue:41628", + "issue:36246|issue:38617", + "issue:46082|issue:46121", + "issue:45245|issue:45584", + "issue:44246|issue:45003", + "issue:42175|issue:44279", + "issue:43643|issue:44496", + "issue:44295|issue:46506", + "issue:44683|issue:45446", + "issue:44964|issue:45600", + "issue:44112|issue:44964", + "issue:43854|issue:44964", + "issue:43404|issue:45216", + "issue:41950|issue:44016", + "issue:45632|issue:46421" + ], + "prepared_review_unit_hash": "25f4afec95aea52d861ba402e2e70b40079134177296c7aaf4bc58ec5cbd3dfa", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36642, + "estimated_input_tokens": 9161, + "estimated_eval_tokens": 18578 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: mostly unrelated Transformers issues spanning model loading, tokenizers, multimodal processing, config validation, CI, and import/runtime bugs. I don\u2019t see any pair that clearly shares the same concrete underlying defect, so all soft edges should be rejected. If a single issue must be picked as the closest hub, the multimodal chat-template bug is the most central, but it is not a true duplicate anchor.", + "confidence": 0.84, + "canonical_issue_reason": "issue:44521 is the closest representative because it is a broad multimodal-input bug with multiple inbound references and sits near several other multimodal/processor-themed reports, but the cluster as a whole is not a single duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 has the strongest \u201chub\u201d characteristics in this set: open, well-discussed, and centered on multimodal processing behavior that overlaps thematically with several neighboring reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "Both involve Qwen3/Qwen2.5-family model behavior, but one is LoRA merge/tied embeddings collapse and the other is save_pretrained checkpoint key serialization; different code paths." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "Sparse-layer config creation and local dataset loading in an example script are unrelated bugs." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Both are multimodal Qwen issues, but one is chat-template assistant mask generation and the other is video position-id computation; not the same defect." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "These are separate CI failures in different areas: GraniteSpeech device-override testing vs auto-mappings referencing removed configs." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Qwen3OmniMoe video input validation and serve KV-cache continuation indexing are different runtime failures." + }, + { + "left": "issue:44779", + "right": "issue:46489", + "accept": false, + "reason": "Both are DeepSeek tokenizer regressions, but they target different model families/issues and the evidence does not show the same concrete tokenizer bug." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Example-script failures in classification and image-classification dataset loading are unrelated." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet CPU allocation behavior and Fuyu tensor conversion error handling are unrelated." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 import failure and missing TensorFlow backend dependency are different environment/setup problems." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Processor/tokenizer regression and Gemma4 embedding resize behavior are different model-code issues." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json and list_repo_files forwarding revision are unrelated repository-access behaviors." + }, + { + "left": "issue:45632", + "right": "issue:46234", + "accept": false, + "reason": "trust_remote_code cache collisions and a missing import_utils availability helper are unrelated." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 embedding resizing and Whisper empty transcription handling are unrelated." + }, + { + "left": "issue:42796", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer BPE marker handling and multimodal assistant-mask generation are different subsystems." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but one is Qwen2.5-VL image processor export and the other is top-level AutoImageProcessor export; not the same bug." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Qwen image-processor import and configuration_utils layer_type_validation import are unrelated missing-symbol issues." + }, + { + "left": "issue:46082", + "right": "issue:46121", + "accept": false, + "reason": "Llama head_dim validation and rope-validation ignore_keys type handling are different config-validation paths." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Category-cardinality limits and Whisper empty transcription errors are unrelated." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import slowness and unsafe sys.modules access are related only loosely; no evidence of the same underlying defect." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "A missing TensorFlow backend in extras installation and a generic dependency issue are too broad and not the same concrete problem." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "trust_remote_code field loss and unrecognized model_type/config.json loading failure are different config-loading bugs." + }, + { + "left": "issue:44295", + "right": "issue:46506", + "accept": false, + "reason": "Position_ids buffer reading and CPU/CUDA output divergence are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both mention flex_attention, but one is a compile failure on torch>=2.9 and the other is a version-check/import bug; different code paths." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Model loading for Phi-4 multimodal and stale auto-mappings references are unrelated." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "A GraniteSpeech CI test issue and a Phi-4 model-loading issue are unrelated." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Both are model-load failures, but they concern different models and distinct root causes." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Mistral3 weight tying and Qwen3.5 save_pretrained checkpoint correctness are different regression paths." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "A pipeline/image-processor bug and a notebook syntax error are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:46421", + "accept": false, + "reason": "trust_remote_code cache-path collision and lazy cache initialization recompilation are unrelated cache behaviors." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 36, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:27957", + "issue:28282", + "issue:32090", + "issue:35141", + "issue:42796", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43441", + "issue:43479", + "issue:43526", + "issue:43531", + "issue:43650", + "issue:43673", + "issue:43824", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44186", + "issue:44220", + "issue:44279", + "issue:44336", + "issue:44368", + "issue:44496", + "issue:44734", + "issue:44792", + "issue:44933", + "issue:44987", + "issue:45042", + "issue:45083", + "issue:45216", + "issue:45335", + "issue:45375", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45600", + "issue:45759", + "issue:46082", + "issue:46123", + "issue:46369", + "issue:46399", + "issue:46424", + "issue:46506", + "issue:46515" + ], + "soft_pairs": [ + "issue:44186|issue:44734", + "issue:27957|issue:35141", + "issue:43673|issue:46424", + "issue:43650|issue:44220", + "issue:44016|issue:44220", + "issue:43479|issue:45375", + "issue:32090|issue:35141", + "issue:45083|issue:46506", + "issue:43531|issue:45600", + "issue:28282|issue:35141", + "issue:43976|issue:44987", + "issue:43295|issue:45335", + "issue:43526|issue:43901", + "issue:44496|issue:45042", + "issue:43673|issue:44336", + "issue:42907|issue:43441", + "issue:43441|issue:45588", + "issue:44016|issue:44062", + "issue:46123|issue:46369", + "issue:43381|issue:45593", + "issue:44792|issue:46424", + "issue:43976|issue:44279", + "issue:43901|issue:44496", + "issue:43824|issue:44933", + "issue:44496|issue:45759", + "issue:43299|issue:45216", + "issue:44279|issue:46515", + "issue:45335|issue:45584", + "issue:44075|issue:44368", + "issue:42796|issue:43650", + "issue:46082|issue:46399" + ], + "prepared_review_unit_hash": "e8902f540aa1699b7c0fe74f8e76a92b5c990a5b368fbaee090e0dad9c84b33b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36965, + "estimated_input_tokens": 9242, + "estimated_eval_tokens": 18740 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a noisy false positive: the items span unrelated bugs across tokenizers, model configs, cache handling, training, docs, and import/export regressions. None of the soft-edge pairs look like the same underlying bug or a mergeable duplicate PR.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Different code paths and symptoms: LayoutLMv2 tokenizer NER/padding crash vs transformers serve KV-cache continuation indexing." + }, + { + "left": "issue:27957", + "right": "issue:35141", + "accept": false, + "reason": "XLMRoberta FlashAttention-2 support is unrelated to embedding reinitialization after resize_token_embeddings." + }, + { + "left": "issue:43673", + "right": "issue:46424", + "accept": false, + "reason": "Both involve cache concepts, but one is a missing GenerationMixin cache in chunked_prefill and the other is max_cache_length only being honored for quantized cache." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "No overlap in bug description or subsystem; one is an invalid title, the other is an audio feature extraction failure." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error is unrelated to _torch_extract_fbank_features()." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Different multimodal config bugs: defaulting None values vs a missing deepstack_visual_indexes field." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer _gpu_broadcast_one NoneType crash is unrelated to output embedding reinitialization after resizing token embeddings." + }, + { + "left": "issue:45083", + "right": "issue:46506", + "accept": false, + "reason": "Different model families and failure modes: helper length computation in qwen3_omni_moe vs CPU/CUDA divergence in PPDocLayoutV3." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to auto_mappings references to removed Sam3LiteText configs." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "Missing PyTorch dependency import error is not the same as embedding resize/post_init reinitialization." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Python version support regression and loading failure for a specific model are different issues." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Processor/tokenizer regression with images is unrelated to decoder.embed_tokens not updating after resize_token_embeddings." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "BEiT image processor reduce_labels bug and TextClassificationPipeline docs mismatch are different subsystems." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Unrecognized model config loading failure is unrelated to PIL image processors incorrectly requiring torchvision." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Cache-missing regression has no relation to ANSI codes emitted by loading_report." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Both mention Ministral/FlashAttention, but one is a save/dequantization problem and the other is a model loading failure under FlashAttention in v5 RC." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Different failure modes: model-specific FlashAttention regression vs a flash_attention.py AttributeError on s_aux=None for sink-less models." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error is unrelated to the AddedToken special keyword conflict." + }, + { + "left": "issue:46123", + "right": "issue:46369", + "accept": false, + "reason": "MaskGenerationPipeline partial-batch result loss and sam3-litetext memory overflow are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient-checkpointing eval-mode restriction is unrelated to D-FINE auxiliary-loss behavior when denoising is off." + }, + { + "left": "issue:44792", + "right": "issue:46424", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to cache max_cache_length handling." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python-version compatibility regression and a general dependency issue are not the same bug." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Pipeline documentation mismatch is unrelated to model config loading failure." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Both are import/export regressions, but they concern different missing symbols from different modules and code paths." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Unrecognized model config and AutoModelForCausalLM.from_config not unwrapping text_config are distinct loading bugs." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3VL MoE loading regression and Qwen3.5 save_pretrained checkpoint corruption are different code paths." + }, + { + "left": "issue:44279", + "right": "issue:46515", + "accept": false, + "reason": "Generic dependency issue is not the same as torch 2.4.1 incompatibility." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "resize_token_embeddings not updating decoder.embed_tokens is unrelated to Whisper empty-transcription generation failure." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Unused SGD optimizer args and a tie_word_embeddings warning are unrelated." + }, + { + "left": "issue:42796", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer BPE marker handling bug has nothing to do with the placeholder 'ADD THE DATA' issue." + }, + { + "left": "issue:46082", + "right": "issue:46399", + "accept": false, + "reason": "LlamaConfig head_dim validation and Qwen3.5 use_kernels registration crash are different model/config bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 37, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41950", + "issue:42994", + "issue:43295", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43479", + "issue:43525", + "issue:43577", + "issue:43650", + "issue:43653", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44242", + "issue:44279", + "issue:44291", + "issue:44351", + "issue:44462", + "issue:44466", + "issue:44521", + "issue:44617", + "issue:44683", + "issue:44779", + "issue:45020", + "issue:45081", + "issue:45103", + "issue:45200", + "issue:45216", + "issue:45276", + "issue:45278", + "issue:45325", + "issue:45357", + "issue:45372", + "issue:45405", + "issue:45542", + "issue:45593", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45907", + "issue:46082", + "issue:46209", + "issue:46512" + ], + "soft_pairs": [ + "issue:44279|issue:45405", + "issue:41950|issue:45020", + "issue:43577|issue:45216", + "issue:45103|issue:45372", + "issue:44351|issue:45278", + "issue:39692|issue:43901", + "issue:43295|issue:44779", + "issue:43299|issue:45357", + "issue:44242|issue:45200", + "issue:41950|issue:45823", + "issue:44683|issue:45542", + "issue:38175|issue:44617", + "issue:43976|issue:45405", + "issue:46082|issue:46512", + "issue:43650|issue:46209", + "issue:44779|issue:45812", + "issue:43299|issue:43854", + "issue:43479|issue:43525", + "issue:44521|issue:45325", + "issue:44466|issue:45276", + "issue:44462|issue:45081", + "issue:44016|issue:44291", + "issue:43653|issue:45593", + "issue:43441|issue:45834", + "issue:42994|issue:43317", + "issue:45081|issue:45907" + ], + "prepared_review_unit_hash": "00f174b81afb34d85958f0ec824c2f3920ed31abf532671b28a25ceb3c3b8675", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35038, + "estimated_input_tokens": 8760, + "estimated_eval_tokens": 17776 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a true duplicate set; it mixes many unrelated bugs across different models, tokenizer paths, docs, quantization, config, and loading/saving code. All soft edges look like false positives under strict duplicate triage.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue stands out because the cluster spans multiple independent problems rather than one underlying defect.", + "canonical_pr_reason": null, + "best_issue_reason": "None of the issues is a good global representative for the whole cluster; the closest thing to a broad regression still only covers one narrow code path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Both mention dependency/version problems, but one is a vague transformers dependency issue and the other is a specific PEFT version bump; different root causes." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video-classification processor lookup and remote_code breakage are unrelated code paths." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Both are regression-style model state issues, but one is about loaded dtype on Blip2 and the other about incorrect save_pretrained serialization for Qwen3.5." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "Auto-docstring annotation handling and Gemma 4 processor import failure are unrelated." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "A missing HybridCache symbol is not the same as broad import failures after a version upgrade." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs-related, but they concern different API examples and different mistakes." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Both involve tokenizer regressions, but they affect different behaviors and model families; not enough evidence of one shared bug." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3VL load failure and Qwen3.5 save_pretrained key serialization are different paths and symptoms." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "MoE load-balancing loss logic and Gemma 4 text-only mm_token_type_ids handling are different multimodal/model concerns." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Video pipeline processor selection and Gemma4 device-mismatch during forward are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Torch flex_attention compilation and tensorboard/TensorFlow backend setup have no common underlying bug." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "SigLIP2 zero probabilities and Sam3Video CUDA OOM are unrelated failures." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python version support regression and PEFT dependency version bump are different issues." + }, + { + "left": "issue:46082", + "right": "issue:46512", + "accept": false, + "reason": "LlamaConfig head_dim validation and stale config.dtype after from_config(dtype=...) are separate configuration bugs." + }, + { + "left": "issue:43650", + "right": "issue:46209", + "accept": false, + "reason": "The first is a non-descriptive title; the second is an FP8 inference problem on B200. No shared underlying change is visible." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they target different model families and likely different tokenizer code paths." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Qwen3VL loading and GLM-4.7-Flash unit-test loading are model-specific failures with different causes." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Phi4MultimodalConfig default subconfig initialization and Llama4Config missing pad_token_id are different config bugs." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Chat-template assistant masks and Qwen2.5-VL rope index scaling are separate multimodal bugs." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Incorrect lm_head serialization and Gemma4 embedding resize propagation are unrelated state vs resize problems." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json and Mistral regex patch crashing are different tokenizer-loading defects." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax and init_empty_weights argument mismatch are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss logic are entirely different subsystems." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "FlashAttention failure for Ministral-3 and OCR index error in Kosmos2.5 are unrelated." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized model saving and dequantized model loading/offload are different stages and likely different fixes." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regex patch crash and list_repo_files revision forwarding have no shared root cause." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 38, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:33290", + "issue:41628", + "issue:42175", + "issue:43441", + "issue:43525", + "issue:43643", + "issue:43673", + "issue:43723", + "issue:43761", + "issue:43784", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", + "issue:44188", + "issue:44242", + "issue:44246", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44361", + "issue:44373", + "issue:44496", + "issue:44589", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:45103", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45593", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45800", + "issue:45823", + "issue:46291", + "issue:46399" + ], + "soft_pairs": [ + "issue:44623|issue:44749", + "issue:33290|issue:45656", + "issue:44261|issue:44263", + "issue:44623|issue:45397", + "issue:45600|issue:45712", + "issue:44361|issue:45593", + "issue:45542|issue:45800", + "issue:44016|issue:45823", + "issue:42175|issue:43976", + "issue:43441|issue:44683", + "issue:43901|issue:43994", + "issue:44671|issue:44779", + "issue:43723|issue:44162", + "issue:43673|issue:43761", + "issue:46291|issue:46399", + "issue:43525|issue:43881", + "issue:43824|issue:43976", + "issue:43901|issue:44246", + "issue:44297|issue:45397", + "issue:43525|issue:43643", + "issue:44188|issue:44336", + "issue:44589|issue:45362", + "issue:43784|issue:45542", + "issue:44336|issue:44373", + "issue:41628|issue:43976", + "issue:44246|issue:44496", + "issue:45103|issue:46399", + "issue:45561|issue:45563", + "issue:44295|issue:45397", + "issue:44242|issue:45593" + ], + "prepared_review_unit_hash": "c2fd70d2a36a5eb5ef78bf065a139abb004f5e9d4406cb726e7b8158a61ad77c", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36530, + "estimated_input_tokens": 9133, + "estimated_eval_tokens": 18522 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a loose Transformers issue set with broad thematic overlap (loading/config/tokenizer/kernel regressions) but no true duplicate pairs. I would anchor the cluster on the generic loading regression in issue #43643, but all soft edges should be rejected.", + "confidence": 0.31, + "canonical_issue_reason": "Issue #43643 is the most broadly framed, concrete regression in the set: it describes a specific `AutoConfig.from_pretrained(..., trust_remote_code=True)` loading failure with missing fields, which is more reusable as an anchor than the model-specific or one-off reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #43643 is the best global representative because it is a clear API-level bug with a reproducible symptom and broad potential impact across remote-code models; it is closer to the cluster\u2019s loading/config theme than the other narrowly scoped reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Different bugs: missing files on `processor.save_pretrained` vs a slowdown/regression in filtering after upgrade." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Both mention DeepSpeed/optimizers, but one is OOM with Adafactor while the other is double optimizer stepping; not the same failure mode." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Unrelated components and symptoms: MLA layernorm precision/config issue vs `torch.split` return-value behavior in a MoE indexer." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "`processor.save_pretrained` missing files and Gemma-4 ZeRO-3 `from_pretrained` loading are different code paths and failure causes." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are repo hygiene/auto-object issues, but one is stale auto-mapping references and the other is leftover dummy classes; not the same bug." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "Different model/task bugs: MLuke tokenizer AttributeError vs D-FINE missing auxiliary losses when denoising is off." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "TensorFlow backend installation issue vs a torch/transformers version incompatibility; unrelated root causes." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax error vs a Gemma4 device-mapping forward mismatch; no common underlying defect." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "Backend packaging omission for `.[torch]` vs Python-version compatibility regression; not duplicates." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Both involve attention kernels, but one is Ministral-3 FlashAttention failure and the other is compiled flex_attention on torch>=2.9; different breakages." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs drift for `return_all_scores` vs incorrect model/pipeline outputs for SigLIP2; unrelated issues." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "CamemBERT masked-LM output regression and Deepseek tokenizer regression affect different model families and code paths." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "AutoTokenizer loading failure and ESM2 breakage are both model-loading problems, but they are distinct bugs with different symptoms." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache missing during chunked prefill vs CLIPVision hidden_states omission; different APIs and fixes." + }, + { + "left": "issue:46291", + "right": "issue:46399", + "accept": false, + "reason": "Both mention kernels, but one is import-time kernel version metadata failure and the other is a missing decorator on rotary-pos-emb registration; not the same change." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "`Llama4Config` missing `pad_token_id` and generic glm-4v-9b load failure are not concretely the same bug." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "ImportError for a missing model class vs Python 3.9/3.10 compatibility issue; unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Documentation mismatch and slow import performance are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save metadata mismatch vs Gemma-4 ZeRO-3 loading failure; different functionality." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Both concern missing config fields, but one is model-specific `pad_token_id` and the other is a broader `trust_remote_code` loading regression; not clearly the same bug." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "Attention-kernel divergence under `torch.compile` vs ANSI codes in loading report; no overlap beyond being library bugs." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Storage-object lookup error and Qwen3.5 chat crash are different runtime failures with no shared concrete code path." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Importing sentence-transformers breaks on missing `nn`, while the other is a packaging/backend dependency issue involving TensorFlow." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "Terminal ANSI output bug vs wrong docstring for `position_ids`; documentation/runtime mismatch but not the same issue." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "Missing `AutoImageProcessor` import and Python-version support regression are distinct issues." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Slow import performance and unrecognized-model config errors are unrelated." + }, + { + "left": "issue:45103", + "right": "issue:46399", + "accept": false, + "reason": "Auto-docstring crash with future annotations and Qwen3.5 kernel registration failure are different subsystems." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "A pytest-xdist race and a stale warning in paged generate are different test/runtime bugs." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "Buffer registration access error vs Gemma-4 ZeRO-3 loading failure; not the same underlying defect." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Load-balancing loss not added in MoE gating vs D-FINE auxiliary loss omission under denoising-off; similar-sounding loss issues but different models and logic." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 39, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36246", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:42886", + "issue:43257", + "issue:43262", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43673", + "issue:43901", + "issue:44038", + "issue:44162", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44351", + "issue:44485", + "issue:44496", + "issue:44561", + "issue:44734", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:44877", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45250", + "issue:45292", + "issue:45362", + "issue:45375", + "issue:45561", + "issue:45759", + "issue:45859", + "issue:46132", + "issue:46369", + "issue:46399", + "issue:46424", + "issue:46506" + ], + "soft_pairs": [ + "issue:42491|issue:43299", + "issue:43531|issue:45362", + "issue:43526|issue:45003", + "issue:41950|issue:44220", + "issue:44336|issue:44485", + "issue:44246|issue:44871", + "issue:44336|issue:46424", + "issue:43550|issue:46506", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:44734|issue:46424", + "issue:42886|issue:46132", + "issue:45003|issue:45561", + "issue:36246|issue:43596", + "issue:44779|issue:45292", + "issue:45859|issue:46399", + "issue:44295|issue:44749", + "issue:43475|issue:46369", + "issue:44242|issue:44857", + "issue:43901|issue:45003", + "issue:43673|issue:44561", + "issue:43352|issue:45250", + "issue:43257|issue:45137", + "issue:44496|issue:45003", + "issue:44162|issue:45200", + "issue:45375|issue:45759", + "issue:43262|issue:44246", + "issue:44877|issue:45375", + "issue:44038|issue:46506" + ], + "prepared_review_unit_hash": "85a7cc8694a95040016e97ea62965318d2e5b3ac42959b1ec389f7c629eda1db", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36274, + "estimated_input_tokens": 9069, + "estimated_eval_tokens": 18394 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a grab-bag of transformer regressions and import/config bugs, mostly centered on v5-era model-loading or pipeline issues. None of the soft pairs look like true duplicates; they share broad themes at most, not the same concrete bug.", + "confidence": 0.89, + "canonical_issue_reason": "issue:44561 is the broadest representative of the cluster\u2019s recurring v5 regression theme and captures a user-facing breakage pattern better than the narrower model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44561 is the best single issue anchor because it describes a clear, cross-cutting compatibility regression rather than a one-off model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen3-MoE loading regressions, but one is LoRA compatibility on hf4/hf5 and the other is Qwen3VL Moe loading on transformers 5.0; different failure modes." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Sliding-window behavior for Qwen3-MoE is unrelated to Qwen3.5-35B chat crashes; same family, different bug." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and unsafe sys.modules access are completely different code paths." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline image-processor lookup and audio fbank feature extraction are unrelated subsystems." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI-code output in loading_report has nothing to do with GLM-5 RoPE implementation." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import-performance complaints and Gemma eos_token_id config mismatch are unrelated bugs." + }, + { + "left": "issue:44336", + "right": "issue:46424", + "accept": false, + "reason": "Terminal-color handling in loading_report is unrelated to cache max length behavior." + }, + { + "left": "issue:43550", + "right": "issue:46506", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba and CPU-vs-CUDA differences in PPDocLayoutV3 are distinct issues." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different dependency surfaces." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Missing Qwen2_5_VLImageProcessor export and missing HybridCache export are separate import regressions." + }, + { + "left": "issue:44734", + "right": "issue:46424", + "accept": false, + "reason": "KV-cache continuation tensor indexing in serve is not the same as cache max_cache_length handling." + }, + { + "left": "issue:42886", + "right": "issue:46132", + "accept": false, + "reason": "Offline tokenizer cache loading and AttentionInterface.register behavior change are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Unsafe sys.modules access and pytest-xdist file-race behavior are different problems." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "Qwen2_5_VLImageProcessor import failure and DeepSpeed ZeRO3 BertModel index error do not share a concrete bug." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "DeepSeek tokenizer correctness and resize_token_embeddings not updating output embeddings are separate embedding/tokenizer bugs." + }, + { + "left": "issue:45859", + "right": "issue:46399", + "accept": false, + "reason": "Both involve Qwen3.5 rotary-related code, but one is CPU offload incompatibility and the other is missing decorator registration; not the same fix." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "Registering position_ids as a buffer and version-upgrade filtering slowdown are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:46369", + "accept": false, + "reason": "SAM 3 Video missing an attribute and SAM3 litetext memory overflow are different defects." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Missing load-balancing loss and AMP/CUDA crash in a different loss module are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Pipeline docs mismatch and sys.modules access bug are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Both are v5 regressions, but chunked_prefill cache handling and trust_remote_code breakage are different concrete code paths." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "A model-specific Flash Attention 2.0 unsupported error is not the same as the generic flash-attention topic issue." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed and empty-deque ZeRO3 failure are distinct DeepSpeed issues." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model/config loading and unsafe sys.modules access are unrelated." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 being broken and Gemma 4 mm_token_type_ids defaults are unrelated model-specific bugs." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both concern Qwen multimodal configs, but one drops a specific field under strict parsing while the other fails to unwrap text_config in from_config." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio chat-template sampling-rate defaults and import slowdown are unrelated." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Different models and different config-handling bugs; only a loose strict-config similarity." + }, + { + "left": "issue:44038", + "right": "issue:46506", + "accept": false, + "reason": "A Qwen3-VL-Moe v5 regression and CPU/CUDA result divergence in PPDocLayoutV3 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": 21, + "review_unit_index": 40, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:42886", + "issue:43065", + "issue:43408", + "issue:43475", + "issue:43540", + "issue:43653", + "issue:43906", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44188", + "issue:44297", + "issue:44361", + "issue:44368", + "issue:44466", + "issue:44479", + "issue:44483", + "issue:44561", + "issue:44623", + "issue:44749", + "issue:44877", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45292", + "issue:45335", + "issue:45397", + "issue:45412", + "issue:45459", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45656", + "issue:45685", + "issue:45701", + "issue:45987", + "issue:46082", + "issue:46123", + "issue:46170", + "issue:46202", + "issue:46399", + "issue:46421", + "issue:46431" + ], + "soft_pairs": [ + "issue:45292|issue:45584", + "issue:44561|issue:46399", + "issue:45292|issue:45701", + "issue:45412|issue:46123", + "issue:45137|issue:45600", + "issue:44466|issue:45335", + "issue:44987|issue:45459", + "issue:43475|issue:43540", + "issue:45397|issue:45656", + "issue:44995|issue:45200", + "issue:44060|issue:44077", + "issue:44479|issue:46170", + "issue:44368|issue:44977", + "issue:45685|issue:46431", + "issue:43653|issue:44162", + "issue:44995|issue:45588", + "issue:43065|issue:43408", + "issue:42886|issue:45229", + "issue:43906|issue:46202", + "issue:44368|issue:45563", + "issue:44877|issue:46170", + "issue:44483|issue:46202", + "issue:43931|issue:44368", + "issue:44188|issue:46202", + "issue:45987|issue:46082", + "issue:44188|issue:46421", + "issue:45584|issue:46170", + "issue:43065|issue:46123", + "issue:44297|issue:44623", + "issue:44361|issue:44749", + "issue:44749|issue:45656" + ], + "prepared_review_unit_hash": "f49a119a66ca0614963691cbba832aefa6a2428b9bdc01c777cd0ea811202673", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37028, + "estimated_input_tokens": 9257, + "estimated_eval_tokens": 18770 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is highly heterogeneous, with many unrelated bugs across tokenizers, vision models, generation, and distributed/runtime behavior. I would not merge any of the soft edges; the strongest topical mini-cluster is around save_pretrained/serialization, but even there the paired reports are not the same concrete bug.", + "confidence": 0.74, + "canonical_issue_reason": "issue:44297 is the most concrete, self-contained bug report in the set and sits in the clearest mini-cluster (tokenizer/serialization/save_pretrained behavior).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44297 is the best representative issue to anchor the cluster summary because it is specific, reproducible, and better scoped than the broader or more speculative reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Both involve embeddings/tokenization-adjacent behavior, but one is resize_token_embeddings on output embeddings and the other is Whisper generation after align_special_tokens; different failures and code paths." + }, + { + "left": "issue:44561", + "right": "issue:46399", + "accept": false, + "reason": "Unrelated problems: one is removal of is_torch_fx_available breaking trust_remote_code models, the other is a Qwen3.5 kernel registration/decorator error." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "resize_token_embeddings/output_embeddings is not the same bug as version-dependent tokenization changes; only a broad tokenization theme is shared." + }, + { + "left": "issue:45412", + "right": "issue:46123", + "accept": false, + "reason": "RT-DETR memory lifecycle leak and MaskGenerationPipeline partial-batch bug are unrelated code paths." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque failure and removed Sam3LiteText auto_mappings references are different CI/runtime issues." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both mention resizing/tied embeddings, but one is inconsistent serialization of lm_head.weight and the other is resize_token_embeddings not updating decoder.embed_tokens; not the same defect." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Loading a model from hub and hiding tokenizer errors when protobuf is missing are distinct tokenizer/loading behaviors." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Different model families and failures: missing fpn_position_embeddings in SAM3 video vs video input ValueError in Qwen3OmniMoe." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 ZeRO3 from_pretrained and deepspeed double optimizer.step are separate training/distributed issues." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Stale indexer cache on second forward pass is unrelated to mm_token_type_ids defaulting for text-only fine-tuning." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Incorrect tied-weights warning in Qwen3-Next and patchtsmixer post_init allowance are unrelated configuration issues." + }, + { + "left": "issue:44479", + "right": "issue:46170", + "accept": false, + "reason": "Video input regression in Qwen models and Windows ROCm import/type error are separate failures." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "A tied-embeddings warning during LoRA fine-tuning is not the same as flash-attention generation failure." + }, + { + "left": "issue:45685", + "right": "issue:46431", + "accept": false, + "reason": "MPS histogram support for Int in MOE is unrelated to missing chat template validation in a processor." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token decode bug and ESM2 breakage are different tokenizer/model issues." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Second-forward stale cache crash and flash_attention.py s_aux=None AttributeError are not the same code-path problem." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and sam3_video vs sam3_tracker type mismatch are separate SAM3 issues." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline cache loading for tokenizers and Gemma4 multi-GPU CUDA OOM are unrelated." + }, + { + "left": "issue:43906", + "right": "issue:46202", + "accept": false, + "reason": "An isolated reproduction of an older issue and a Gemma pipeline roundtrip failure on \u2581 are not clearly the same bug." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "Tie-word-embeddings warning and stale num_return_sequences warning are different generate-time warnings." + }, + { + "left": "issue:44877", + "right": "issue:46170", + "accept": false, + "reason": "Strict config rejection for granite_speech and a Windows ROCm import failure are unrelated." + }, + { + "left": "issue:44483", + "right": "issue:46202", + "accept": false, + "reason": "Chat/completions request rejection and Gemma \u2581 roundtrip failure are different user-facing bugs." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "StaticCache return type contract and LlamaConfig head_dim validation are unrelated API/config issues." + }, + { + "left": "issue:44188", + "right": "issue:46421", + "accept": false, + "reason": "torch.compile kernel divergence from branching and cache lazy init causing recompilation are different performance/correctness problems." + }, + { + "left": "issue:45584", + "right": "issue:46170", + "accept": false, + "reason": "Whisper empty transcription failure and Windows ROCm torch.distributed import issues are unrelated." + }, + { + "left": "issue:43065", + "right": "issue:46123", + "accept": false, + "reason": "SAM3 pixel decoder placeholder layer and mask generation batching behavior are not the same bug." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Both touch save_pretrained, but one is tokenizer_class mismatch and the other is missing processor files; similar area, different failure mode." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "MLukeTokenizer AttributeError and a performance slowdown after upgrade are unrelated." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer slowdown and double optimizer.step under deepspeed are different regressions." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 41, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:42994", + "issue:43232", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43704", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43873", + "issue:44038", + "issue:44162", + "issue:44188", + "issue:44222", + "issue:44351", + "issue:44623", + "issue:44749", + "issue:44869", + "issue:44871", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45406", + "issue:45478", + "issue:45656", + "issue:45753", + "issue:46132", + "issue:46202", + "issue:46326", + "issue:46399" + ], + "soft_pairs": [ + "issue:43646|issue:44162", + "issue:44623|issue:45656", + "issue:44749|issue:45310", + "issue:43232|issue:43673", + "issue:44871|issue:46132", + "issue:45230|issue:45397", + "issue:42994|issue:43482", + "issue:44869|issue:45356", + "issue:43475|issue:43643", + "issue:42371|issue:45092", + "issue:44038|issue:44623", + "issue:28282|issue:30990", + "issue:41628|issue:44933", + "issue:43475|issue:45381", + "issue:45375|issue:45478", + "issue:45310|issue:45375", + "issue:42994|issue:43525", + "issue:43784|issue:46399", + "issue:43723|issue:46326", + "issue:44964|issue:45137", + "issue:45406|issue:46202", + "issue:43232|issue:44188", + "issue:43873|issue:44945", + "issue:42673|issue:43704", + "issue:41628|issue:44351", + "issue:44945|issue:45753", + "issue:45081|issue:45103", + "issue:42994|issue:43824", + "issue:44222|issue:45310", + "issue:44038|issue:45310" + ], + "prepared_review_unit_hash": "de54b21be21979bfb4a2fa22a1b752e56aa7d7c4e13ec6ab95c522f64722347f", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36453, + "estimated_input_tokens": 9114, + "estimated_eval_tokens": 18484 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous transformer-related issues, with only a couple of clear duplicate pairs. The strongest exact duplicate is the Qwen3.5 MoE from_pretrained regression (45310/45478), and the Qwen3ForCausalLM VRAM leak reports (42673/43704) also look identical.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45310 is the clearest concrete regression report and has an exact duplicate in issue:45478; both describe the same Qwen3.5 MoE from_pretrained failure on transformers>=5.4.0.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is a strong cluster representative: it is specific, actionable, and clearly duplicated by issue:45478.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Different bugs: missing fields from trust_remote_code AutoConfig vs ESM2 breakage." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Different code paths: processor.save_pretrained file emission vs DeepSpeed optimizer step duplication." + }, + { + "left": "issue:44749", + "right": "issue:45310", + "accept": false, + "reason": "Different symptoms and models: general slowdown after upgrade vs Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Both are generation-related, but one is sync_gpus/update_kwargs behavior and the other is chunked_prefill cache handling; not the same bug." + }, + { + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "Different functionality: eos_token_id config mismatch vs AttentionInterface.register behavior change." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Generic bug report vs gemma-4 zero3 from_pretrained issue; no evidence they are the same bug." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure vs Qwen2.5-GGUF loading failure are different operations." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Whisper timestamp decode crash vs Kimi-K2.5 tokenizer codec regression; different code paths." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "SAM 3 video attribute error vs AutoConfig trust_remote_code missing fields; unrelated model families and failures." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 settings warning vs meta-initialization incompatibility for old InternVL2 checkpoints; not the same issue." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe transformers bug vs processor.save_pretrained missing files; different subsystems." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "AutoModel missing PyTorch ImportError vs Sentence Transformers loading hang; different failure modes." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Missing AutoImageProcessor export vs image_utils import failure; related area but not the same concrete bug." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Different models and symptoms: SAM 3 vision encoder attribute error vs qwen2.5-vl video position ids issue." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Different bugs: missing deepstack_visual_indexes field vs Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "One is a from_pretrained failure; the other is a config-field loss under @strict. Not the same bug." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized save failure vs Llama4Config missing pad_token_id attribute; different model/config issues." + }, + { + "left": "issue:43784", + "right": "issue:46399", + "accept": false, + "reason": "Sentence-transformers import NameError vs Qwen3.5 kernel registration ValueError; unrelated." + }, + { + "left": "issue:43723", + "right": "issue:46326", + "accept": false, + "reason": "Tokenizer loading error in v5 vs continue_final_message behavior for mistral_common tokenizers; not the same underlying bug." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Model loading regression vs DeepSpeed ZeRO3 deque IndexError; unrelated failure classes." + }, + { + "left": "issue:45406", + "right": "issue:46202", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer vs Gemma pipeline roundtrip on \u2581; different bugs." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Generation kwargs update issue vs attention kernel divergence under torch.compile; different code paths." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization offloading issue vs incorrect LLM output with pipeline parallelism; not the same problem." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact duplicate reports with the same title and same Qwen3ForCausalLM VRAM leak in multiple dataloader threads." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Different missing exports: AutoImageProcessor vs HybridCache." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline parallelism output mismatch vs Qwen3_5 infinite loop on a specific image; different symptoms." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Tokenizer regex crash vs auto_docstring crash with future annotations; unrelated bugs." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized saving failure vs missing Qwen2_5_VLForConditionalGeneration import; different issue types." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "FP8 save_pretrained moe vs Qwen3.5 MoE from_pretrained error; different operations." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Both mention Qwen/MoE, but the reported failures and model variants differ; not enough to treat as the same bug." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Same Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0; effectively identical duplicate reports." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative on the soft-edge pairs. The two accepted duplicates (42673/43704 and 45310/45478) are well supported by matching titles and descriptions, and the rejected pairs are framed cautiously without overclaiming. No evidence of overstated duplicate/bundle claims beyond what the packet supports." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 42, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42831", + "issue:42898", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43504", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43976", + "issue:44016", + "issue:44075", + "issue:44079", + "issue:44222", + "issue:44242", + "issue:44295", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44779", + "issue:44792", + "issue:44869", + "issue:44933", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45200", + "issue:45310", + "issue:45356", + "issue:45561", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45812", + "issue:46077", + "issue:46123", + "issue:46231" + ], + "soft_pairs": [ + "issue:43232|issue:44792", + "issue:36010|issue:41628", + "issue:42831|issue:44222", + "issue:42175|issue:44933", + "issue:42175|issue:45712", + "issue:44496|issue:45561", + "issue:43278|issue:45137", + "issue:43504|issue:45561", + "issue:45561|issue:45600", + "issue:43976|issue:44933", + "issue:46077|issue:46123", + "issue:30333|issue:33290", + "issue:30990|issue:35141", + "issue:43377|issue:44242", + "issue:42898|issue:44568", + "issue:44623|issue:45310", + "issue:42222|issue:45200", + "issue:43479|issue:44869", + "issue:43479|issue:44079", + "issue:43874|issue:45092", + "issue:44016|issue:46231", + "issue:42898|issue:45812", + "issue:43784|issue:43824", + "issue:43723|issue:44568", + "issue:42898|issue:44779", + "issue:44734|issue:45103", + "issue:44295|issue:45656", + "issue:44075|issue:46231", + "issue:42898|issue:45356" + ], + "prepared_review_unit_hash": "332e95d59ee7ec47214eecfc626fd653aa13e4d433122ed4df5c1017ee25bca5", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36508, + "estimated_input_tokens": 9127, + "estimated_eval_tokens": 18510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The candidate pairs are mostly superficial similarities across unrelated issues; I would reject all soft edges. The best issue anchor is a concrete tokenizer regression report rather than a broader or less specific packaging/import issue.", + "confidence": 0.8, + "canonical_issue_reason": "No single duplicate spans the set, but issue #44779 is the best anchor: it reports a concrete v4\u2192v5 tokenizer regression with a clear symptom and scope.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #44779 is the strongest standalone bug report here because it is specific, reproducible, and clearly describes the failing tokenizer path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation kwargs/caching bug vs a Janus image-generation test failure; different code paths and symptoms." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Two different missing imports (`GenerationMixin` vs `AutoImageProcessor`); same package, but not the same underlying defect." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "FineGrainedFP8 accuracy regression vs FP8 MoE save/load issue; different failure modes." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Packaging/backend dependency issue vs a nonexistent `image_utils` import; unrelated root causes." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Backend install metadata vs leftover dummy classes in `dummy_pt_objects.py`; not the same bug." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Model config loading error vs pytest-xdist file race; unrelated areas." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Dtype mismatch during evaluation vs DeepSpeed ZeRO3 deque underflow; different subsystems." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Legacy-field model loading failure vs xdist capture-file race; unrelated." + }, + { + "left": "issue:45561", + "right": "issue:45600", + "accept": false, + "reason": "Parallel-test file race vs stale auto_mappings references breaking CI; different causes." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python-version compatibility issue vs nonexistent import; not the same defect." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Attention-mask construction bug vs partial-batch pipeline result loss; separate code paths." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLFlow failure-reporting problem vs DeepSpeed Adafactor OOM; unrelated." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "SentenceTransformer loading stall vs token-embedding resize reinit bug; different functionality." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "MIMI padding-mask batching bug vs MoE load-balancing loss omission; not mergeable as one fix." + }, + { + "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; different bugs." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Missing files on `processor.save_pretrained` vs Qwen3.5 MoE `from_pretrained` error; unrelated." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Broken ViTPose models vs Gemma 4 multimodal token-type defaulting; different model families and issues." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Multimodal config default initialization vs Whisper word-timestamp decode crash; unrelated." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Config defaulting bug vs `ModelOutput` key assignment bug; distinct code paths." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Missing image-processor method vs remote-code/meta-init incompatibility; both multimodal, but different root causes." + }, + { + "left": "issue:44016", + "right": "issue:46231", + "accept": false, + "reason": "Notebook syntax error vs documentation parameter-name mismatch; not duplicates." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Generic tokenizer-space cleanup regression vs Granite token-ID regression; different tokenizer behaviors and fixes." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Import-time `nn` NameError vs missing Qwen2.5-VL class export; separate import failures." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "`AutoTokenizer.from_pretrained` load issue in v5 vs mDeBERTa special-token regression; same area but not the same bug." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but one is a global cleanup-behavior change and the other is DeepSeek-specific codec handling; not the same concrete defect." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Serve/KV-cache tensor indexing crash vs auto-docstring annotation handling; unrelated." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Buffer registration error vs optimizer stepping twice under DeepSpeed; different code paths." + }, + { + "left": "issue:44075", + "right": "issue:46231", + "accept": false, + "reason": "Optimizer argument wiring bug vs docs typo for training parameters; not a duplicate bug." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer cleanup behavior regression vs Kimi-K2.5 codec/warning regression; both tokenizers, but different underlying issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 43, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28282", + "issue:30333", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:41628", + "issue:42757", + "issue:42796", + "issue:42831", + "issue:42898", + "issue:43122", + "issue:43232", + "issue:43295", + "issue:43475", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:44060", + "issue:44206", + "issue:44222", + "issue:44263", + "issue:44297", + "issue:44373", + "issue:44448", + "issue:44484", + "issue:44485", + "issue:44623", + "issue:44933", + "issue:44977", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45245", + "issue:45310", + "issue:45381", + "issue:45405", + "issue:45656", + "issue:45712", + "issue:46082", + "issue:46132", + "issue:46209", + "issue:46305", + "issue:46424", + "issue:46431" + ], + "soft_pairs": [ + "issue:42898|issue:43295", + "issue:44060|issue:45125", + "issue:45405|issue:45712", + "issue:44222|issue:44297", + "issue:27957|issue:30333", + "issue:30333|issue:34689", + "issue:44933|issue:45712", + "issue:44297|issue:45656", + "issue:30333|issue:36331", + "issue:44222|issue:44623", + "issue:42831|issue:44623", + "issue:30333|issue:35141", + "issue:43479|issue:43644", + "issue:22355|issue:28282", + "issue:44448|issue:46132", + "issue:42796|issue:43723", + "issue:45245|issue:46132", + "issue:41628|issue:42757", + "issue:44297|issue:45310", + "issue:43475|issue:46305", + "issue:45071|issue:46082", + "issue:44263|issue:45092", + "issue:46209|issue:46431", + "issue:45381|issue:46305", + "issue:43122|issue:46209", + "issue:44373|issue:44485", + "issue:43723|issue:44977", + "issue:44222|issue:44484", + "issue:44206|issue:45103", + "issue:43232|issue:46424" + ], + "prepared_review_unit_hash": "b0a15ded78b8ed5e3b6e0bf11748342cc36ae62fa0ec4669c8f53d5c0eb3deea", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35894, + "estimated_input_tokens": 8974, + "estimated_eval_tokens": 18204 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are largely unrelated Transformers issues spanning tokenization, multimodal models, FP8, config validation, and import/runtime regressions. None of the soft pairs look like the same underlying bug or mergeable change, so all soft edges are rejected.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43644 is the broadest regression-style report in the set, describing a Transformers 5.0.0-wide buffer initialization problem rather than a narrow model-specific edge case.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43644 is a strong standalone bug report with a clear, version-scoped failure mode and broad user impact; it is the best representative issue here, though not a duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42898", + "right": "issue:43295", + "accept": false, + "reason": "Both are v5 regressions, but one is tokenization cleanup behavior and the other is processor/tokenizer image-handling breakage; different code paths." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Different Qwen model-specific bugs: tied-weights warning versus missing tensor-parallel plan." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "Unrelated: a dependency version bump issue versus leftover dummy class cleanup/import leakage." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both mention save_pretrained, but one is FP8 MoE checkpoint saving and the other is tokenizer_class metadata mismatch." + }, + { + "left": "issue:27957", + "right": "issue:30333", + "accept": false, + "reason": "No overlap in underlying problem: Flash Attention model support versus MLflow reporting after failed training." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "Different failures: MLflow run-state reporting versus Llama 3.2 Vision model loading regression." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Both touch imports/exports, but one is a missing image_utils import and the other is dummy class leakage in no-torch mode." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save metadata bug is unrelated to DeepSpeed double optimizer stepping." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "Training telemetry after failure is unrelated to the Trainer compute_loss signature change." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 MoE save_pretrained and processor.save_pretrained missing files are different serialization paths." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FP8 accuracy issue is unrelated to processor file saving." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "MLflow failure reporting is unrelated to embedding resize/post_init reinitialization." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Different problems: multimodal config defaults versus non-persistent buffer initialization." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Both are import errors, but one is missing transformers.onnx and the other is missing PyTorch for AutoModel." + }, + { + "left": "issue:44448", + "right": "issue:46132", + "accept": false, + "reason": "Different regressions: Pegasus output changes versus AttentionInterface.register behavior." + }, + { + "left": "issue:42796", + "right": "issue:43723", + "accept": false, + "reason": "Tokenizer BPE-marker handling is not the same as AutoTokenizer.from_pretrained failing in v5." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "No shared bug: category-count runtime error versus attention registration semantics." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Different missing imports from different packages; not the same failure." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer class metadata mismatch is unrelated to Qwen3.5 MoE from_pretrained errors." + }, + { + "left": "issue:43475", + "right": "issue:46305", + "accept": false, + "reason": "Different model components and symptoms: SAM 3 video attribute error versus Moonshine streaming mask off-by-one." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "Both are config-related, but one is type-checking regression and the other is head_dim validation logic." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "torch.split indexer behavior is unrelated to remote-code checkpoint meta-init incompatibility." + }, + { + "left": "issue:46209", + "right": "issue:46431", + "accept": false, + "reason": "FP8 inference on B200 and missing chat template are unrelated issues." + }, + { + "left": "issue:45381", + "right": "issue:46305", + "accept": false, + "reason": "Both are sequence-position/window bugs, but in different models and code paths." + }, + { + "left": "issue:43122", + "right": "issue:46209", + "accept": false, + "reason": "Tokenization output drift across versions is unrelated to FP8 inference failure on a specific GPU." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A docstring mistake is not the same as a RoPE implementation discussion/bug." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading failure and flash-attention generation bug are different problems." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "Checkpoint saving in FP8 MoE is unrelated to the max_shard_size default question." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Feature extractor argument validation and auto-docstring annotation handling are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:46424", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to cache max length handling." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 44, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:22355", + "issue:27957", + "issue:28218", + "issue:30990", + "issue:38175", + "issue:39401", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:43066", + "issue:43232", + "issue:43408", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43824", + "issue:44038", + "issue:44190", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44295", + "issue:44360", + "issue:44368", + "issue:44373", + "issue:44483", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44829", + "issue:44871", + "issue:44913", + "issue:45092", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45753", + "issue:45834", + "issue:46002", + "issue:46050", + "issue:46082", + "issue:46305", + "issue:46328", + "issue:46402", + "issue:46421" + ], + "soft_pairs": [ + "issue:44734|issue:46421", + "issue:44295|issue:44623", + "issue:43232|issue:45834", + "issue:44190|issue:46082", + "issue:43723|issue:44871", + "issue:44263|issue:44360", + "issue:44261|issue:45092", + "issue:44279|issue:46328", + "issue:45092|issue:45357", + "issue:42907|issue:46050", + "issue:30990|issue:38175", + "issue:43644|issue:46305", + "issue:42831|issue:45310", + "issue:44263|issue:44373", + "issue:44829|issue:46002", + "issue:44483|issue:45406", + "issue:42831|issue:44038", + "issue:41628|issue:43824", + "issue:43479|issue:44913", + "issue:43408|issue:46402", + "issue:22355|issue:27957", + "issue:42831|issue:45753", + "issue:42757|issue:45278", + "issue:22355|issue:28218", + "issue:39401|issue:44568", + "issue:27957|issue:30990", + "issue:42886|issue:43066", + "issue:44368|issue:46305" + ], + "prepared_review_unit_hash": "c803a54736ce0b8e2e15e582037e231037576805798f79a81987edd2dfca593f", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 35814, + "estimated_input_tokens": 8954, + "estimated_eval_tokens": 18164 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is highly heterogeneous: the soft-similarity pairs share at most broad subsystem overlap, but not the same concrete bug or change. I rejected all soft edges as duplicates, and the best representative issue is the broad import-regression report rather than any narrow symptom report.", + "confidence": 0.84, + "canonical_issue_reason": "If a single representative is needed, issue 45278 is the broadest umbrella for the import-error regression subset; however, the overall cluster is not a true duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45278 is the most general, user-facing regression report and best captures the broadest slice of the related import-breakage complaints in this mixed cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44734", + "right": "issue:46421", + "accept": false, + "reason": "Both mention cache-related serving paths, but one is a KV cache continuation crash and the other is lazy cache init causing recompilation; different bugs." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "One is about reading a bufferized position_ids tensor, the other about processor.save_pretrained missing files; unrelated failure modes." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation kwargs after sync_gpus and a Kosmos2.5 OCR index error are different code paths and symptoms." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "A local dataset loading script issue is not the same as a LlamaConfig head_dim validation error." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Tokenizer loading in v5 and Gemma-3 eos_token_id config mismatch are separate tokenizer/config bugs." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Both reference DSA indexer code, but one concerns torch.split outputs and the other a missing ReLU; not the same defect." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "Missing rms_norm_eps in MLA q_a_layernorm and remote-code meta-init incompatibility are different multimodal issues." + }, + { + "left": "issue:44279", + "right": "issue:46328", + "accept": false, + "reason": "A generic dependency issue is not the same as a padding performance regression." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "One is an old InternVL2 meta-init compatibility problem, the other is Qwen3.5 save_pretrained writing wrong visual encoder keys." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Different quantization/save-load failures: dequantized Ministrals/Devstrals versus sinq quantized models." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Sentence Transformers loading hangs and SigLIP2 zero probabilities are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:46305", + "accept": false, + "reason": "Non-persistent buffer junk and a streaming mask off-by-one are different bugs in different subsystems." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues and Qwen3.5 Moe from_pretrained failures are not the same problem." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "A runtime torch.split bug and a wrong docstring are not duplicates." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Training collapse with flash_attention_3 and NaNs on checkout/init point to different root causes." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat-completions request rejection and Gemma4Processor missing _tokenizer are separate serving failures." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "A numerical accuracy issue and a Qwen3-VL-Moe bug are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different symbols and model components; no evidence of the same missing export." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Default multimodal config initialization and GPTNeoX rotary_pct reload behavior are distinct config bugs." + }, + { + "left": "issue:43408", + "right": "issue:46402", + "accept": false, + "reason": "A sam3 tracker/model-type warning and missing Sam3Model text encoder weights are different SAM3 issues." + }, + { + "left": "issue:22355", + "right": "issue:27957", + "accept": false, + "reason": "Missing transformers.onnx and XLMRoberta Flash Attention 2 are unrelated." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FineGrainedFP8 accuracy and a Qwen3_5 image-specific infinite loop are different defects." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "The is_offline_mode import error may be one symptom in the broader upgrade report, but this pair does not clearly describe the same concrete bug." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "An import failure for transformers.onnx is unrelated to tokenizer spacing after added tokens." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Offset-mapping correctness and add_special_tokens BOS/EOS behavior are separate tokenizer bugs." + }, + { + "left": "issue:27957", + "right": "issue:30990", + "accept": false, + "reason": "Flash Attention 2 on XLMRoberta and Sentence Transformers loading hang are unrelated." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline cache loading and wrong tokenizer decoder type are both tokenizer issues, but not the same underlying bug." + }, + { + "left": "issue:44368", + "right": "issue:46305", + "accept": false, + "reason": "A tie_word_embeddings warning in Qwen3.5 LoRA and a streaming mask off-by-one are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 45, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42796", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43116", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43475", + "issue:43550", + "issue:43576", + "issue:43646", + "issue:43756", + "issue:43784", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44188", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44485", + "issue:44488", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:45200", + "issue:45278", + "issue:45290", + "issue:45593", + "issue:45600", + "issue:46002", + "issue:46132", + "issue:46326", + "issue:46369" + ], + "soft_pairs": [ + "issue:42796|issue:43122", + "issue:43232|issue:45593", + "issue:43408|issue:46369", + "issue:42886|issue:42994", + "issue:44188|issue:44373", + "issue:43576|issue:43784", + "issue:43550|issue:46132", + "issue:43646|issue:46002", + "issue:42371|issue:44263", + "issue:44261|issue:44373", + "issue:44295|issue:44297", + "issue:44117|issue:44704", + "issue:42907|issue:44945", + "issue:30990|issue:33357", + "issue:43756|issue:44060", + "issue:41628|issue:45278", + "issue:42907|issue:43937", + "issue:43976|issue:44945", + "issue:43784|issue:45600", + "issue:39401|issue:44488", + "issue:42907|issue:44964", + "issue:44704|issue:45290", + "issue:44360|issue:44485", + "issue:39401|issue:44521", + "issue:43122|issue:46326", + "issue:43122|issue:44521", + "issue:44038|issue:44297", + "issue:44625|issue:44829", + "issue:43116|issue:45200", + "issue:44038|issue:44295", + "issue:43408|issue:43475" + ], + "prepared_review_unit_hash": "5e7a24e52b5921dda04bd795d046baf80e207111352d7f44d7bd6db65aca3f2a", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36536, + "estimated_input_tokens": 9134, + "estimated_eval_tokens": 18524 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose semantic collection of unrelated Transformers bug reports. None of the soft pairs look like true duplicates or mergeable PRs; they span different models, APIs, and failure modes. The broadest umbrella issue is 45278 (general import regressions after upgrading to v5), but it is only a loose representative, not a real duplicate anchor.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 45278 is the broadest, most general regression report in the set and best serves as an umbrella for the v5 import-breakage subset, though the overall cluster is heterogeneous.", + "canonical_pr_reason": null, + "best_issue_reason": "45278 is the most suitable global representative because it describes a wide import-regression class rather than a single model-specific failure, so it is the closest thing to a canonical issue here.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42796", + "right": "issue:43122", + "accept": false, + "reason": "Different bugs: BPE marker handling for Ministral-3 vs tokenizer behavior changing across versions for another model." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Unrelated code paths: generation kwargs after sync_gpus vs D-FINE auxiliary losses when denoising is off." + }, + { + "left": "issue:43408", + "right": "issue:46369", + "accept": false, + "reason": "Different SAM3 problems: a model-type warning during tracker init vs a memory overflow in sam3-litetext." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Different failure modes: offline tokenizer cache loading vs quantized model saving." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "torch.compile kernel divergence vs a docstring typo; not the same underlying bug." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "One is a broken CLI env command; the other is a NameError on importing sentence-transformers." + }, + { + "left": "issue:43550", + "right": "issue:46132", + "accept": false, + "reason": "Bamba torch.compile/SDPA issue and AttentionInterface.register behavior change are separate bugs." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Both are v5-era regressions, but one is custom model init breakage and the other is NaNs after checkout/init." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 API warning and torch.split return values in a DSA indexer are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "Precision/config issue in MLA q_a_layernorm vs a wrong docstring; not the same bug." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Reading a registered position_ids buffer and tokenizer_class mismatch on save_pretrained are different issues." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES None in from_pretrained vs AutoProcessor not forwarding kwargs to cached_file." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Saving dequantized Ministrals/Devstrals and incorrect pipeline-parallel outputs are unrelated." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Both mention loading, but one is Sentence Transformers hanging and the other is a MacOS bus error with a CLIP model." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE layer count bug vs Qwen3-Next tied-weights warning; different model configs." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Related only at a broad v5-import level; missing AutoImageProcessor is not the same concrete bug as the wider import-error report." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Quantized/dequantized save failure vs invalid GenerationConfig validation; unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility breakage vs pipeline parallelism output corruption are distinct." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Import NameError in sentence-transformers vs stale auto_mappings references to removed Sam3LiteText configs." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Wrong offset_mapping for Qwen3 tokenizer vs inability to load cjvt/sleng-bert; different tokenizer/model problems." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Model saving failure vs model loading failure for Phi-4 multimodal instruct." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "AutoProcessor kwargs forwarding and apply_chat_template crashes with tool calls are separate code paths." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer ReLU issue vs GLM-5 RoPE implementation; unrelated." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer offset_mapping bug vs multimodal assistant_mask handling; not the same defect." + }, + { + "left": "issue:43122", + "right": "issue:46326", + "accept": false, + "reason": "Version-to-version tokenization differences vs continue_final_message behavior for mistral_common tokenizers." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Different symptoms and subsystems: tokenization drift vs apply_chat_template assistant mask generation." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe bug and tokenizer_class mismatch on save_pretrained are unrelated." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 vs flash_attention_3 causing degenerate training; different bugs." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Empty multi-label classification results vs mm_token_type_ids defaulting to zeros are not the same underlying issue." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Qwen3-VL-Moe generation bug vs position_ids buffer access error; different failures." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "SAM3 model-type warning and missing fpn_position_embeddings are related by family, 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": 21, + "review_unit_index": 46, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:29159", + "issue:36296", + "issue:38617", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:42907", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43278", + "issue:43441", + "issue:43452", + "issue:43493", + "issue:43611", + "issue:43824", + "issue:43825", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44075", + "issue:44188", + "issue:44222", + "issue:44351", + "issue:44355", + "issue:44485", + "issue:44625", + "issue:44704", + "issue:44779", + "issue:44936", + "issue:44938", + "issue:44945", + "issue:45072", + "issue:45216", + "issue:45278", + "issue:45357", + "issue:45446", + "issue:45478", + "issue:45753", + "issue:46002", + "issue:46231", + "issue:46328" + ], + "soft_pairs": [ + "issue:42907|issue:43452", + "issue:44038|issue:44222", + "issue:43824|issue:44351", + "issue:45216|issue:45478", + "issue:43122|issue:44779", + "issue:22355|issue:29159", + "issue:42757|issue:44351", + "issue:36296|issue:43116", + "issue:43278|issue:44945", + "issue:42175|issue:44938", + "issue:43493|issue:45753", + "issue:22355|issue:29127", + "issue:44188|issue:46328", + "issue:43937|issue:44485", + "issue:38617|issue:45278", + "issue:43066|issue:44779", + "issue:42673|issue:45072", + "issue:43992|issue:44704", + "issue:42371|issue:43874", + "issue:43611|issue:46002", + "issue:43116|issue:43493", + "issue:44355|issue:46231", + "issue:44038|issue:45478", + "issue:44625|issue:46002", + "issue:43976|issue:45446", + "issue:44075|issue:44936", + "issue:45357|issue:45478", + "issue:42222|issue:43441", + "issue:43116|issue:43825" + ], + "prepared_review_unit_hash": "3004ce0dbef9f429b44bf859bff8e2a5be75353cf12d3f552d15a2ec63d4ae72", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35872, + "estimated_input_tokens": 8968, + "estimated_eval_tokens": 18192 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is broadly heterogeneous. A few issues share high-level themes like import failures, tokenization regressions, or Qwen3.5 model regressions, but none are safe duplicate matches under strict triage. If forced to pick an anchor, issue 45278 is the broadest umbrella for import-regression reports.", + "confidence": 0.84, + "canonical_issue_reason": "issue 45278 is the best umbrella anchor because it describes a broad post-upgrade import-failure regression and can plausibly subsume related import-error reports better than the more narrow model-specific tickets.", + "canonical_pr_reason": null, + "best_issue_reason": "issue 45278 is the most generally representative issue in this set; it is broad enough to serve as a triage hub, even though the overall cluster is not a true duplicate group.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Different failure modes: saving dequantized models vs.from_pretrained/gguf loading." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Both mention Qwen3.5/MoE, but one is a general 5.0 regression and the other is FP8 save_pretrained; different code paths." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different modules." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Related to Qwen3.5 save/load, but one is save_pretrained checkpoint corruption and the other is a from_pretrained error." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both concern tokenizer behavior changes, but they refer to different models and different regressions." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "No module import failure for transformers.onnx is unrelated to slow vs fast tokenizer inconsistency." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Same broad import-error theme, but different missing symbols from different packages." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor-parallel training bug and empty multi-label classification results are different problems." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Different regressions: dtype drift in evaluation vs incorrect output under pipeline parallelism." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Backend packaging issue vs Python 3.14 load failure; not the same bug." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "Distinct model-specific regressions affecting different behaviors and likely different code paths." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "Transformers.onnx import failure is unrelated to LayoutLMv3 box-validation messaging." + }, + { + "left": "issue:44188", + "right": "issue:46328", + "accept": false, + "reason": "Attention-kernel divergence under torch.compile is unrelated to pad performance regression." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "Both are GLM-5-related, but one is GenerationConfig validation and the other is RoPE implementation." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "45278 is a broad import-error report, but 38617 is one specific missing import; insufficient evidence they are the same concrete bug." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Wrong tokenizer decoder type in v5 and Deepseek tokenization drift are different tokenizer regressions." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM and dtype mismatch CI issue are unrelated." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Different from_pretrained plumbing issues: missing weights vs kwargs not passed to cached_file." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 settings API change is unrelated to a missing image-patch method in GLM46V." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "Base-model-prefix loading breakage and NaNs at init/checkout are different classes of failure." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Classification script behavior and SigLIP2 implementation discrepancy are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled Python load errors and a documentation parameter-name mismatch are not the same issue." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5/MoE, but they describe different API paths and symptoms." + }, + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and NaNs in classification heads are different defects." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python version support failure and PyTorch version gate for AuxRequest are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Optimizer argument handling and trainer.evaluate() failing after train() are different training-stack bugs." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen3.5 regressions, but one is save_pretrained visual-key corruption and the other is a from_pretrained error." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are unrelated model-specific issues." + }, + { + "left": "issue:44355", + "right": "issue:46231", + "accept": false, + "reason": "Compiled-file loading errors are unrelated to a docs bug about set_training() parameters." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Multi-label classification empty output is unrelated to pipeline() translation-task messaging." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 47, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:33453", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42490", + "issue:42617", + "issue:42673", + "issue:42796", + "issue:43064", + "issue:43122", + "issue:43408", + "issue:43450", + "issue:43673", + "issue:43742", + "issue:43827", + "issue:43906", + "issue:43937", + "issue:43986", + "issue:44117", + "issue:44261", + "issue:44291", + "issue:44380", + "issue:44410", + "issue:44466", + "issue:44779", + "issue:44829", + "issue:44977", + "issue:45071", + "issue:45081", + "issue:45083", + "issue:45216", + "issue:45310", + "issue:45341", + "issue:45362", + "issue:45405", + "issue:45446", + "issue:45561", + "issue:45563", + "issue:45588", + "issue:46002", + "issue:46077", + "issue:46178", + "issue:46298", + "issue:46328", + "issue:46512" + ], + "soft_pairs": [ + "issue:45083|issue:45446", + "issue:42673|issue:42796", + "issue:43064|issue:43673", + "issue:43937|issue:45563", + "issue:33453|issue:39401", + "issue:44380|issue:46077", + "issue:43827|issue:45446", + "issue:44117|issue:45588", + "issue:43673|issue:46328", + "issue:42673|issue:43906", + "issue:44291|issue:45071", + "issue:39401|issue:44779", + "issue:44466|issue:46002", + "issue:43937|issue:44977", + "issue:43450|issue:43986", + "issue:45405|issue:45446", + "issue:44410|issue:45310", + "issue:45341|issue:46328", + "issue:38617|issue:41628", + "issue:45341|issue:45561", + "issue:44829|issue:45216", + "issue:44117|issue:45081", + "issue:45071|issue:46512", + "issue:42222|issue:46178", + "issue:43122|issue:46298", + "issue:43408|issue:43742", + "issue:42617|issue:45341", + "issue:42371|issue:44261", + "issue:45362|issue:46328", + "issue:42490|issue:42673" + ], + "prepared_review_unit_hash": "35960bc6f31cab6c44437f9ee52f21720f84cbb4e037bec87adfbe24d9ec3a66", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36085, + "estimated_input_tokens": 9022, + "estimated_eval_tokens": 18300 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most items are unrelated model-, tokenizer-, generation-, or testing/infrastructure-specific regressions. I did not find any soft edge pair that looks like the same underlying bug/change.", + "confidence": 0.61, + "canonical_issue_reason": "No clear canonical duplicate: the issues span many distinct code paths and product areas rather than one shared root cause.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43122 is the broadest, clearest regression report here (tokenization changed from 4.57.3 to 5.0) and is the best representative of the recurring tokenizer-regression theme.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Different subsystems and failures: qwen3_omni_moe feature-length helper vs flex_attention version gating." + }, + { + "left": "issue:42673", + "right": "issue:42796", + "accept": false, + "reason": "One is a VRAM leak in Qwen3ForCausalLM threading; the other is a tokenizer BPE-marker bug in Ministral." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "Different code paths: trainer/FSDP2 optimizer-state mismatch vs generation cache missing in chunked_prefill." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "Both mention generation, but one is invalid GenerationConfig loading and the other is a stale warning in paged generate()." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Both are tokenizer-related, but one is loading regression and the other is wrong offset_mapping for Qwen3." + }, + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "Attention-related, but different models and bugs: GPT2 backend scaling vs Gemma4 bidirectional mask construction." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Documentation cleanup for removed pipeline() APIs is unrelated to a PyTorch version check bug." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "Unrelated failures: tokenizer mapping None handling vs flash_attention crash on sink-less models." + }, + { + "left": "issue:43673", + "right": "issue:46328", + "accept": false, + "reason": "Generation cache bug and feature-extractor padding performance bug are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "43906 is only an isolated reproduction reference; no evidence it is the same VRAM leak as 42673." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Different breakages: init_empty_weights argument handling vs PretrainedConfig type-checking in v5.4.0." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they concern different models and symptoms (offset mapping vs general tokenization output)." + }, + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "Serialization of lm_head.weight is unrelated to NaNs in classification heads after checkout/init." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "Different generation issues: invalid GenerationConfig vs flash-attention malfunction for Qwen3.5." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Video-related but distinct: batched output shape bug vs missing torchvision crash in AutoProcessor." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "A dependency version bump problem is unrelated to an auxiliary-request import version check." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen model-family bugs: missing projections in qwen3next vs from_pretrained failure for Qwen3.5 MoE." + }, + { + "left": "issue:45341", + "right": "issue:46328", + "accept": false, + "reason": "Different issues: a testing_utils bug vs a feature-extractor padding performance regression." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different API surfaces." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "Shared file path is not enough; one is a generic testing_utils bug, the other is an xdist race on captured_info.txt." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "Different bugs: flash_attention_3 training degeneration vs incorrect save_pretrained checkpoint contents." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Both hit tokenizer loading, but one is a mapping lookup bug and the other is a Mistral regex patch crash." + }, + { + "left": "issue:45071", + "right": "issue:46512", + "accept": false, + "reason": "Type-checking breakage in v5.4.0 is unrelated to stale config.dtype after from_config(dtype=...)." + }, + { + "left": "issue:42222", + "right": "issue:46178", + "accept": false, + "reason": "Different areas entirely: vitpose model breakage vs backend registration error for tensorflow_text." + }, + { + "left": "issue:43122", + "right": "issue:46298", + "accept": false, + "reason": "Versioned tokenization regression is unrelated to a noisy logging message." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "Model-type warning for sam3_video/sam3_tracker is unrelated to a MobileLLM loading key error." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 API guidance and an MLA rms_norm_eps precision issue are not the same bug." + }, + { + "left": "issue:45362", + "right": "issue:46328", + "accept": false, + "reason": "Qwen3.5 chat crash and feature-extractor padding slowdown are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:42673", + "accept": false, + "reason": "Path-with-slash import warning is unrelated to the Qwen3ForCausalLM VRAM leak." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 48, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:27957", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42831", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43352", + "issue:43441", + "issue:43452", + "issue:43577", + "issue:43673", + "issue:43874", + "issue:43931", + "issue:43937", + "issue:43986", + "issue:44038", + "issue:44075", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44360", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44496", + "issue:44829", + "issue:44877", + "issue:44933", + "issue:45003", + "issue:45071", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45656", + "issue:46172", + "issue:46209", + "issue:46231" + ], + "soft_pairs": [ + "issue:44877|issue:46231", + "issue:44484|issue:45310", + "issue:39401|issue:46209", + "issue:38617|issue:44220", + "issue:43937|issue:44496", + "issue:44933|issue:45278", + "issue:42371|issue:42831", + "issue:44410|issue:45478", + "issue:42222|issue:45542", + "issue:44075|issue:45656", + "issue:43066|issue:43122", + "issue:42175|issue:42222", + "issue:43874|issue:44261", + "issue:44263|issue:44485", + "issue:43097|issue:44493", + "issue:43673|issue:45341", + "issue:43012|issue:43577", + "issue:41628|issue:45071", + "issue:45071|issue:45542", + "issue:27957|issue:43352", + "issue:44483|issue:45464", + "issue:44484|issue:45478", + "issue:43931|issue:44410", + "issue:44263|issue:46172", + "issue:44360|issue:46172", + "issue:43441|issue:44829", + "issue:43931|issue:44038", + "issue:43097|issue:43452", + "issue:43986|issue:45341", + "issue:45003|issue:45341", + "issue:43874|issue:44263" + ], + "prepared_review_unit_hash": "035ae003ff72d14a837e1a11e27540011ab63ee23ec4ea56ac04d60a9c1f9fa5", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36411, + "estimated_input_tokens": 9103, + "estimated_eval_tokens": 18462 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a loose semantic cluster of Transformers regressions, but the pairs are not strong enough to treat as duplicates. The closest subgroup is import/config breakage in v5/v5.5, with 45278 being the broadest representative issue.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45278 is the most representative single issue in this set: it captures a broad, user-visible regression pattern (many import errors after upgrading) that best anchors the loose import/config breakage subgroup.", + "canonical_pr_reason": null, + "best_issue_reason": "It is the best overall issue to keep as the cluster anchor because it is broader than the more specific model- or feature-level reports and most closely matches the recurring upgrade/regression theme.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44877", + "right": "issue:46231", + "accept": false, + "reason": "Strict config-loading regression vs a documentation parameter typo; unrelated symptoms and fixes." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "Max shard-size behavior question vs Qwen3.5 MoE from_pretrained failure; different code paths." + }, + { + "left": "issue:39401", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer offset_mapping bug vs FP8 inference failure on B200; unrelated models and subsystems." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "ImportError for a missing symbol vs audio feature-extraction issue; not the same bug." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Both involve model/config loading, but one is invalid GenerationConfig and the other is an unrecognized model/config.json issue." + }, + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "A specific missing import from image_utils is not clearly the same underlying bug as a broad \"many import errors\" report." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 API usage complaint vs FineGrainedFP8 accuracy regression; different topics." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Missing Qwen3Next projections vs Qwen3.5 MoE from_pretrained error; different model bugs." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "Broken vitpose models vs missing TensorFlow backend when only tensorboard is installed; unrelated." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "SGD args ignored vs optimizer.step called twice under deepspeed; different optimizer bugs." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Tokenizer decoder-type regression vs changed tokenization output; related area but not the same concrete failure." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "Package/backend install issue vs vitpose model breakage; no shared underlying bug." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Missing image patch helper vs missing rms_norm_eps in MLA; separate model implementation issues." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "DsaIndexer torch.split return-value issue vs GLM-5 RoPE implementation bug; different code paths." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removed tie_embeddings API vs unexpected position-id key warnings; different regressions." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Generation cache missing in chunked prefill vs a testing_utils bug; unrelated." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Warning during bfloat16 compilation vs Blip2 dtype not propagating to submodules; related dtype theme but not the same bug." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "Missing AutoImageProcessor import vs PreTrainedConfig type-checking regression; unrelated." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Config type-checking breakage vs undefined tf backend error; different failure modes." + }, + { + "left": "issue:27957", + "right": "issue:43352", + "accept": false, + "reason": "Both mention Flash Attention 2 support, but they concern different models and likely separate implementations." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Chat/completions request rejection vs streaming inference failure; same API family, but not a single concrete bug." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "Different issues: one is save_pretrained shard-size behavior, the other is Qwen3.5 MoE loading failure." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch vs Qwen3Next missing projections; different model families and defects." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "DSA indexer split semantics vs DeepSeek-V4 transpose/dtype issue; unrelated." + }, + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "Missing ReLU in DSA indexer vs DeepSeek-V4 tensor conversion/transposition bug; different changes." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "FlashAttention failure in Ministral-3 vs degenerate training with flash_attention_3; same broad feature, different concrete problems." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Different Qwen3-VL issues with distinct symptoms: weight shapes mismatch vs MoE model bug." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "API removal/change vs gguf_file loading breakage; not the same code path." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "AutoProcessor/torchvision dependency crash vs a testing utility bug; unrelated." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils vs a testing utility bug; unrelated." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Image processor helper missing vs GLM-MoE split-return issue; different components." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 49, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:38617", + "issue:42757", + "issue:42890", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43097", + "issue:43208", + "issue:43502", + "issue:43531", + "issue:43550", + "issue:43673", + "issue:43873", + "issue:43986", + "issue:43992", + "issue:44016", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44315", + "issue:44351", + "issue:44380", + "issue:44410", + "issue:44484", + "issue:44485", + "issue:44655", + "issue:44704", + "issue:44933", + "issue:44936", + "issue:44938", + "issue:44987", + "issue:45103", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45440", + "issue:45588", + "issue:45753", + "issue:46050", + "issue:46123", + "issue:46172", + "issue:46178", + "issue:46231", + "issue:46298", + "issue:46365", + "issue:46369", + "issue:46421" + ], + "soft_pairs": [ + "issue:45753|issue:46369", + "issue:44933|issue:46231", + "issue:43673|issue:46123", + "issue:44261|issue:44485", + "issue:43986|issue:44484", + "issue:44117|issue:44987", + "issue:43531|issue:45341", + "issue:45440|issue:46172", + "issue:43064|issue:44315", + "issue:43986|issue:45372", + "issue:38617|issue:42757", + "issue:43097|issue:45588", + "issue:43097|issue:46050", + "issue:44410|issue:45375", + "issue:44038|issue:44410", + "issue:42890|issue:46298", + "issue:38617|issue:44351", + "issue:43502|issue:43992", + "issue:44062|issue:44117", + "issue:43097|issue:43873", + "issue:44936|issue:46365", + "issue:44704|issue:45103", + "issue:43986|issue:44117", + "issue:44380|issue:46421", + "issue:44938|issue:46178", + "issue:43986|issue:44016", + "issue:42907|issue:44655", + "issue:43992|issue:44484", + "issue:43208|issue:43550", + "issue:43012|issue:43550", + "issue:42907|issue:45753" + ], + "prepared_review_unit_hash": "9e522707287745b720183f2638dd1bd528cbc1384aae766a416ee6053310724f", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36684, + "estimated_input_tokens": 9171, + "estimated_eval_tokens": 18598 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The soft pairs are superficially related by broad themes (imports, loading, cache, quantization, model families), but each points to a different concrete failure mode or code path. I would not merge any of these pairs as duplicates, and there is no clear canonical issue in this heterogeneous set.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45753", + "right": "issue:46369", + "accept": false, + "reason": "Both are model-runtime failures, but one is an infinite loop on a specific image and the other is a memory overflow on sam3-litetext; different symptoms and likely different code paths." + }, + { + "left": "issue:44933", + "right": "issue:46231", + "accept": false, + "reason": "One is a missing import from image_utils; the other is a documentation/parameter-name mismatch. Not the same bug." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Cache-related generation issue vs pipeline partial-batch result dropping. Different components and failure modes." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "MLA RMS norm epsilon precision problem vs GLM-5 RoPE implementation discussion. Same general area only at a very broad level." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "AutoProcessor video-model import crash without torchvision vs save_pretrained shard-size default question. Unrelated bugs." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling vs a model-loading failure on physical-intelligence/fast. Similar surface area, different root cause." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Qwen3-MoE sliding-window bug vs a testing utility bug. Not the same underlying issue." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "Different DeepSeek model-family correctness issues: remote-implementation divergence vs a tensor dtype/transpose bug." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Wrong optimizer states under FSDP2/PEFT/cpu_ram_efficient_loading vs Liger Kernel not applied with model_init. Different training paths." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Both affect processor/model loading, but one needs torchvision and the other fails on a missing mistral_common import. Distinct causes." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols from different modules; not the same missing export." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Tie-embeddings API removal vs flash_attention crash on s_aux=None. Separate regressions." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "API removal vs quantized save/load breakage. No shared code-path problem." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen configuration/model bugs: missing projections vs missing config field." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Both mention Qwen variants, but one is a broad transformers/Qwen3-VL-MoE bug and the other is a specific missing-projection issue." + }, + { + "left": "issue:42890", + "right": "issue:46298", + "accept": false, + "reason": "Integration-test flakiness from missing set_seed vs a noisy logging message. Not duplicates." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Both are import-name failures, but they involve different public symbols and different dependency layers." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "local_files_only still making API calls vs missing embed_tokens.weight in UMT5Encoder loading. Different bugs." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "AddedToken constructor argument conflict vs tokenizer-mapping None handling. Similar tokenizer area, but unrelated failures." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Tie_embeddings removal and quantization offloading are separate features with separate failure modes." + }, + { + "left": "issue:44936", + "right": "issue:46365", + "accept": false, + "reason": "Evaluation-after-training failure vs importing Trainer blocked by datasets. Different lifecycle stages and causes." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "AutoProcessor kwargs propagation bug vs auto_docstring crash under future annotations. Different subsystems." + }, + { + "left": "issue:43986", + "right": "issue:44016", + "accept": false, + "reason": "Processor loading crash vs notebook syntax error. Not the same issue." + }, + { + "left": "issue:44380", + "right": "issue:46421", + "accept": false, + "reason": "GPT2 attention scaling ignored under SDPA/FlashAttention vs chunked-prefill cache lazy initialization causing recompilation. Different mechanisms." + }, + { + "left": "issue:44938", + "right": "issue:46178", + "accept": false, + "reason": "Python 3.14 import/load failure vs BACKENDS_MAPPING validation error for tensorflow_text. Unrelated." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Saving dequantized models vs saving Pipeline objects. Different save_pretrained code paths." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing encoder weights on load vs shard-size default behavior on save. Not duplicates." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training blockers vs Bamba-9B-v2 torch.compile/SDPA failure. Different models and bugs." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "PyTorch bfloat16 precision warning vs compile-time SDPA failure. No shared root cause." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Quantized save/load failure vs image-specific infinite loop. Completely 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": 21, + "review_unit_index": 50, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42175", + "issue:42490", + "issue:42757", + "issue:42994", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43502", + "issue:43526", + "issue:43582", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43825", + "issue:43867", + "issue:43906", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44162", + "issue:44222", + "issue:44230", + "issue:44368", + "issue:44380", + "issue:44483", + "issue:44485", + "issue:44514", + "issue:44655", + "issue:44683", + "issue:45290", + "issue:45478", + "issue:45479", + "issue:45584", + "issue:45600", + "issue:45698", + "issue:45712", + "issue:46132", + "issue:46172", + "issue:46298", + "issue:46365" + ], + "soft_pairs": [ + "issue:43012|issue:44368", + "issue:43825|issue:44655", + "issue:43066|issue:43906", + "issue:45712|issue:46298", + "issue:43097|issue:46298", + "issue:43352|issue:44380", + "issue:43352|issue:43976", + "issue:44162|issue:46172", + "issue:45584|issue:46365", + "issue:44380|issue:46132", + "issue:44077|issue:46365", + "issue:44380|issue:45479", + "issue:44222|issue:45478", + "issue:36683|issue:43582", + "issue:43064|issue:44380", + "issue:42175|issue:46365", + "issue:42490|issue:43012", + "issue:43208|issue:43526", + "issue:43012|issue:44683", + "issue:42757|issue:43824", + "issue:38617|issue:43502", + "issue:45698|issue:46172", + "issue:44483|issue:44514", + "issue:45600|issue:46298", + "issue:44038|issue:44230", + "issue:43618|issue:44380", + "issue:43618|issue:43761", + "issue:43299|issue:43867", + "issue:42994|issue:44655", + "issue:44483|issue:45290", + "issue:36683|issue:43475", + "issue:43756|issue:44485" + ], + "prepared_review_unit_hash": "8fed415b2ae35c992507d2e6c89eb7dd1d529d56df608ab9d927a57e2712fef4", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37061, + "estimated_input_tokens": 9266, + "estimated_eval_tokens": 18788 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set spans many unrelated regressions across model loading, generation, packaging, attention backends, and serialization. None of the soft pairs looks like the same concrete bug/change closely enough to merge as duplicates.", + "confidence": 0.73, + "canonical_issue_reason": "issue:45698 is the strongest representative because it describes a concrete, reproducible save/load bug in a core path, is still open, and has the most discussion/activity among the candidates.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is the best standalone issue to keep as the cluster representative: it is specific, actionable, and broad enough to subsume nearby custom-module serialization failures.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Both mention warnings, but they are different warnings in different code paths: bfloat16 compilation vs tie_word_embeddings guidance during LoRA fine-tuning." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "One is about pipeline task support/error messaging, the other about saving Pipeline objects; related subsystem, but different underlying bugs." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "The first is a tokenizer decoder-type regression; the second is only an isolated reproduction reference and does not clearly identify the same bug." + }, + { + "left": "issue:45712", + "right": "issue:46298", + "accept": false, + "reason": "Dummy PT classes leaking into dir()/check_repo and a noisy logging message are unrelated maintenance issues." + }, + { + "left": "issue:43097", + "right": "issue:46298", + "accept": false, + "reason": "Removal of tie_embeddings_and_encoder_decoder and a noisy logging message are unrelated changes." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Both involve attention backends, but one is a model-level FlashAttention 2 support error and the other is GPT-2 scaling being ignored; not the same bug." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Nemotron FlashAttention support and a Python-version incompatibility are different failures with different causes." + }, + { + "left": "issue:44162", + "right": "issue:46172", + "accept": false, + "reason": "ESM2 breakage and a DeepSeek tensor transpose issue are distinct model-specific bugs." + }, + { + "left": "issue:45584", + "right": "issue:46365", + "accept": false, + "reason": "Whisper empty transcription handling and Trainer import failure due to datasets are unrelated." + }, + { + "left": "issue:44380", + "right": "issue:46132", + "accept": false, + "reason": "The first is about GPT2 attention scaling being ignored, while the second is about AttentionInterface.register changing function behavior; not the same defect." + }, + { + "left": "issue:44077", + "right": "issue:46365", + "accept": false, + "reason": "A patchtsmixer post_init validation issue is unrelated to a Trainer import conflict." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention backend scaling ignored and degenerate zero-loss sequence classification are different model behaviors." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "Both mention MoE/FP8/Qwen3.5, but one is about save_pretrained and the other about from_pretrained loading; not enough to be the same concrete bug." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "Gemma3Config missing vocab_size and Apple Silicon TypeError in caching_allocator_warmup are unrelated." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "Distributed optimizer-state corruption and GPT2 attention scaling are different code paths." + }, + { + "left": "issue:42175", + "right": "issue:46365", + "accept": false, + "reason": "One is backend selection after installing extras, the other is a Trainer import conflict with datasets; different packaging issues." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "Local-path import warning and bfloat16 compile precision warning do not share the same bug." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "xLSTM training blockers and BeitImageProcessorFast label reduction are unrelated model/component bugs." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "A PyTorch precision warning and a compiled flex_attention failure on torch>=2.9 are different regressions." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different symbols from different packages and likely different version mismatches." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "Missing layer_type_validation import and local_files_only still making API requests are unrelated failures." + }, + { + "left": "issue:45698", + "right": "issue:46172", + "accept": false, + "reason": "Wrong custom module after save_pretrained and a dtype/transpose bug in deepseek-v4 are different underlying issues." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "A /v1/chat/completions API acceptance problem and a batched apply_chat_template crash are different layers of the stack." + }, + { + "left": "issue:45600", + "right": "issue:46298", + "accept": false, + "reason": "Broken auto_mappings references and a noisy logging message are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Both involve Qwen3-VL-Moe, but one is a general loading bug and the other is FP8 inference support; not the same change." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIP attentions missing and GPT2 attention scaling ignored are different model regressions." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Both are CLIP regressions, but one affects attentions on CLIPOutput while the other drops hidden_states on CLIPVisionModel; not clearly the same bug." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Qwen3VL-Moe loading regression and a sorted-state_dict load error are not the same defect." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Quantized model saving failure and pipeline save_pretrained failure are different serialization paths." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Both are chat-related, but one is API request acceptance and the other is tool-call message handling in apply_chat_template." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Gemma3Config missing vocab_size and Sam3VisionEncoderOutput missing fpn_position_embeddings are different attribute errors in different models." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping and GLM-5 RoPE implementation discussion are related only at a very broad topic level, not the same concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 51, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:28282", + "issue:33453", + "issue:36010", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42757", + "issue:42890", + "issue:42898", + "issue:43010", + "issue:43064", + "issue:43066", + "issue:43352", + "issue:43525", + "issue:43575", + "issue:43611", + "issue:43618", + "issue:43867", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44075", + "issue:44077", + "issue:44246", + "issue:44355", + "issue:44492", + "issue:44625", + "issue:44655", + "issue:44855", + "issue:44857", + "issue:44938", + "issue:45071", + "issue:45216", + "issue:45276", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45478", + "issue:45561", + "issue:45593", + "issue:45850", + "issue:46050", + "issue:46077", + "issue:46082", + "issue:46202", + "issue:46285", + "issue:46365" + ], + "soft_pairs": [ + "issue:44625|issue:46365", + "issue:42898|issue:43066", + "issue:45276|issue:46077", + "issue:36010|issue:42757", + "issue:43352|issue:43937", + "issue:43611|issue:44655", + "issue:44655|issue:46050", + "issue:36683|issue:44077", + "issue:43867|issue:45216", + "issue:43976|issue:44938", + "issue:43352|issue:45593", + "issue:44625|issue:46077", + "issue:46077|issue:46082", + "issue:36683|issue:43525", + "issue:43618|issue:45216", + "issue:36683|issue:44625", + "issue:43992|issue:45310", + "issue:43064|issue:44492", + "issue:43010|issue:45278", + "issue:44075|issue:46365", + "issue:44355|issue:44855", + "issue:43867|issue:45375", + "issue:43867|issue:46202", + "issue:44938|issue:45071", + "issue:36683|issue:43064", + "issue:42890|issue:45561", + "issue:43010|issue:45850", + "issue:41669|issue:44246", + "issue:43867|issue:45357", + "issue:43992|issue:45478", + "issue:36296|issue:43575", + "issue:43618|issue:44077", + "issue:43010|issue:43976", + "issue:28282|issue:33453", + "issue:43618|issue:46285", + "issue:43618|issue:44857" + ], + "prepared_review_unit_hash": "4d202e58ff1d8ef6cd37920d3f039ad89e559f81a71ff69bfdf5ed8e5c54272b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38018, + "estimated_input_tokens": 9505, + "estimated_eval_tokens": 19266 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous mix of Transformers regression issues (imports, save/load, model configs, training, and tests). The only strong duplicate-like pair is the import-time slowdown reports.", + "confidence": 0.78, + "canonical_issue_reason": "issue:41669 is the clearest representative: it gives a concrete import-performance regression with a likely root cause (eager `import *` in models) and a measurable impact.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41669 is the best global issue candidate because it is specific, actionable, and directly matches the shared import-slowdown theme better than the broader symptom-only report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44625", + "right": "issue:46365", + "accept": false, + "reason": "Different failures: Qwen3.5 `num_labels` propagation vs `Trainer` import being affected by `datasets`." + }, + { + "left": "issue:42898", + "right": "issue:43066", + "accept": false, + "reason": "Both concern tokenization/decoder behavior, but they describe different regressions and different code paths." + }, + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Same model family (Gemma4) but unrelated bugs: embedding resize propagation vs attention-mask construction." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different dependencies; no same underlying bug is evident." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Unrelated model-specific errors: Flash Attention support gating vs invalid `GenerationConfig`." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "One is loading via `base_model_prefix`, the other is saving `Pipeline` objects; different serialization paths." + }, + { + "left": "issue:44655", + "right": "issue:46050", + "accept": false, + "reason": "Both mention `save_pretrained`, but one is pipeline saving and the other is quantized model roundtrip; not the same bug." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "Different models and failures: missing `vocab_size` in Gemma3Config vs `patchtsmixer` post-init policy." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "Both affect model loading/saving, but one is a generic sorted-state_dict load failure and the other is Qwen3.5 checkpoint regression." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version compatibility reports, but they target different versions and likely different breakages." + }, + { + "left": "issue:43352", + "right": "issue:45593", + "accept": false, + "reason": "Flash Attention support gating and missing auxiliary losses are unrelated model-behavior issues." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "Different Qwen3.5 vs Gemma4 bugs; no shared concrete code-path problem." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Gemma4 attention-mask bug vs LlamaConfig validation bug are separate configuration/runtime issues." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different missing-config-attribute failures in different models (`vocab_size` vs `pad_token_id`)." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIP output field removal and Qwen3.5 save_pretrained regression are unrelated." + }, + { + "left": "issue:36683", + "right": "issue:44625", + "accept": false, + "reason": "Different missing-attribute / propagation bugs in different model configs." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Both involve Qwen3.5-related loading, but one is missing weights and the other is a from_pretrained error; too little evidence they are the same bug." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Distributed training optimizer-state bug vs a typo in cache strategies; unrelated." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Decorator semantics for cache updates vs broad import-error regressions; not the same underlying issue." + }, + { + "left": "issue:44075", + "right": "issue:46365", + "accept": false, + "reason": "Optimizer argument handling and an import failure tied to `datasets` are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Both are import-time failures, but one is about compiled Python files and the other is a TorchScript parse/indentation issue." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "Generic load-model-state_dict issue vs a missing Qwen3_5MoeVisionConfig field; not a duplicate." + }, + { + "left": "issue:43867", + "right": "issue:46202", + "accept": false, + "reason": "Both are model roundtrip/load failures, but they concern different models and different failure modes." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python 3.14 import failure and PretrainedConfig type-checking regression are different compatibility problems." + }, + { + "left": "issue:36683", + "right": "issue:43064", + "accept": false, + "reason": "Config attribute bug vs distributed training optimizer-state corruption; unrelated." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Missing `set_seed()` in tests and xdist file-race in captured_info are different test-infrastructure issues." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "Specific cache-update decorator change vs a broad report that v5 breaks too many models; not the same bug." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "Both point to the same import-time slowdown/regression in `transformers` import, with 41669 giving the concrete likely cause (`import *` in models)." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "Generic load failure vs Qwen3.5 visual-encoder key regression; distinct serialization bugs." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Both reference Qwen3.5-ish loading problems, but one is missing embed weights and the other is a `from_pretrained` error; not enough to treat as the same bug." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug vs TP OOM on Qwen2-57B; different symptoms and likely different causes." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIP output field regression and `patchtsmixer` post_init policy are unrelated." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache update no-grad semantics and Python-version support are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "PyTorch missing ImportError for AutoModel vs tokenizer-loading regression; different layers and failures." + }, + { + "left": "issue:43618", + "right": "issue:46285", + "accept": false, + "reason": "CLIP attentions regression vs CLIP text-model structural inconsistency; related area, but not the same concrete bug." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "CLIP attentions missing and an AMP/CUDA loss crash are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall and conservative. The canonical/best choice around 41669 vs 44246 is well supported by the packet: both are import-time slowdown regressions, with 41669 providing the concrete cause. The other soft-edge rejections mostly stay within the evidence and do not overclaim duplication." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 52, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:27957", + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:42548", + "issue:42757", + "issue:42831", + "issue:42890", + "issue:43010", + "issue:43065", + "issue:43232", + "issue:43299", + "issue:43502", + "issue:43576", + "issue:43723", + "issue:43824", + "issue:43827", + "issue:43874", + "issue:44062", + "issue:44112", + "issue:44222", + "issue:44230", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44387", + "issue:44485", + "issue:44556", + "issue:44655", + "issue:44749", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:44964", + "issue:45071", + "issue:45125", + "issue:45375", + "issue:45440", + "issue:45468", + "issue:45478", + "issue:45600", + "issue:46132" + ], + "soft_pairs": [ + "issue:44360|issue:45440", + "issue:42890|issue:44964", + "issue:43502|issue:44062", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:44387|issue:44556", + "issue:44222|issue:45125", + "issue:43723|issue:44556", + "issue:43010|issue:43065", + "issue:43576|issue:46132", + "issue:43502|issue:44485", + "issue:45125|issue:45478", + "issue:44355|issue:44749", + "issue:42548|issue:44351", + "issue:30333|issue:33453", + "issue:45125|issue:45375", + "issue:43010|issue:43232", + "issue:43502|issue:44222", + "issue:30990|issue:31795", + "issue:43502|issue:44961", + "issue:42831|issue:43874", + "issue:43299|issue:44556", + "issue:43827|issue:44961", + "issue:30333|issue:36296", + "issue:44355|issue:45071", + "issue:44355|issue:44655", + "issue:33453|issue:34689", + "issue:27957|issue:33453", + "issue:42757|issue:43576", + "issue:43576|issue:45071", + "issue:44315|issue:44360", + "issue:43576|issue:43824", + "issue:31795|issue:38617", + "issue:44961|issue:45468", + "issue:44485|issue:44961", + "issue:43576|issue:44938", + "issue:44908|issue:44961", + "issue:44222|issue:44230" + ], + "prepared_review_unit_hash": "9a0e9bfae9427cd7e1fe3a1f31da0897fa70cc79d711f2731da8b8497c45cb88", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 38, + "serialized_chars": 37943, + "estimated_input_tokens": 9486, + "estimated_eval_tokens": 19228 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous Transformers issues with a few loose v5/loading-regression similarities, but no pair is a safe duplicate match. I would not merge any of the soft edges.", + "confidence": 0.83, + "canonical_issue_reason": "issue:44556 is the broadest representative of the loading/regression theme seen across the set, especially version-upgrade breakages; however, the overall cluster is still too mixed to treat as one true duplicate group.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44556 is the best umbrella issue for the repeated v5 reload/loading-breakage reports, but it is not a strong canonical duplicate for the whole cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "Different model families and bugs: DSA indexer ReLU vs DeepSeekV3 implementation divergence." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "CI test flakiness from missing set_seed is unrelated to Phi-4 multimodal loading failure." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "local_files_only network violation and AddedToken keyword collision are different code paths." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Different test failures in different models/subsystems." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Seeded test instability vs stale auto_mappings reference; not the same bug." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Both are version-regression reports, but one is int4 OOM memory behavior and the other is checkpoint reload failure." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained bug and missing _tp_plan/tensor-parallel support are related only broadly, not the same defect." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Tokenizer loading regression and general checkpoint reload regression are not the same concrete failure." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Cache no_grad decoration and Sam3PixelDecoder dummy Conv2d are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Broken env CLI and AttentionInterface.register behavior change are different features." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Network requests with local_files_only and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but one is tensor-parallel metadata and the other is a from_pretrained loading error; not enough to merge." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled Python file errors and slowed filtering after upgrade are different issues." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Different missing exports: PreTrainedModel vs HybridCache." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLFlow reporting and tokenizer loading are unrelated." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen3.5 config/model bugs: missing _tp_plan vs dropped config field." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Cache/layer no_grad change and generation kwargs sync bug are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Local-files-only network leakage and FP8 save_pretrained are different failures." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Loading hang and documentation confusion do not match." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "Network access bug and unrelated issue title." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FP8 accuracy issue and GLM46V patch-count AttributeError are different bugs." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Both involve loading regressions, but different model/version scenarios and likely different root causes." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs still referencing pipeline() is unrelated." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLFlow reporting and tensor-parallel training bug are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled-file load errors and PretrainedConfig type-checking regression are different." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled-file errors and save_pretrained pipeline serialization are unrelated." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Different loading regressions affecting different components/models." + }, + { + "left": "issue:27957", + "right": "issue:33453", + "accept": false, + "reason": "Flash Attention 2 support and tokenizer loading regression are unrelated." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Import error from huggingface_hub and env CLI breakage are different issues." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Broken env command and PretrainedConfig typing regression are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel model_init behavior and DSA indexer ReLU are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "env CLI bug and Qwen2.5-VL import error are different failure modes." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation clarification and missing symbol import are unrelated." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "Unrelated issues with no shared code path." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation and unrelated placeholder title." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug and unrelated title." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "Both mention FP8/MoE, but one is save_pretrained and the other is inference support; too broad to merge." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 53, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:36296", + "issue:41628", + "issue:41669", + "issue:42371", + "issue:42548", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43525", + "issue:43526", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43901", + "issue:43976", + "issue:44038", + "issue:44230", + "issue:44246", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44393", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44704", + "issue:44908", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45341", + "issue:45362", + "issue:45478", + "issue:45542", + "issue:45753", + "issue:45850", + "issue:46095", + "issue:46143", + "issue:46144", + "issue:46209", + "issue:46218", + "issue:46328" + ], + "soft_pairs": [ + "issue:46143|issue:46144", + "issue:36296|issue:43901", + "issue:41628|issue:42548", + "issue:36296|issue:43526", + "issue:44315|issue:46209", + "issue:43673|issue:44492", + "issue:45362|issue:45850", + "issue:44623|issue:45230", + "issue:44273|issue:45542", + "issue:44038|issue:45125", + "issue:44393|issue:45071", + "issue:44273|issue:45003", + "issue:44393|issue:45753", + "issue:43576|issue:44556", + "issue:44230|issue:46095", + "issue:43525|issue:45850", + "issue:41669|issue:43673", + "issue:44556|issue:45092", + "issue:43572|issue:45850", + "issue:44492|issue:45850", + "issue:45278|issue:45850", + "issue:42371|issue:44492", + "issue:41669|issue:46218", + "issue:42890|issue:44908", + "issue:43408|issue:44315", + "issue:43352|issue:44393", + "issue:43576|issue:44908", + "issue:44246|issue:44273", + "issue:44295|issue:45230", + "issue:44297|issue:45230", + "issue:43976|issue:44393", + "issue:44908|issue:45071", + "issue:44704|issue:44908", + "issue:41669|issue:44351", + "issue:44492|issue:46328", + "issue:41669|issue:45341", + "issue:43976|issue:45850", + "issue:44393|issue:45478", + "issue:43976|issue:46218" + ], + "prepared_review_unit_hash": "0b13650cd6f70556cb3a167d36056aba6e43bca1437fbc790d07f276578801c2", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 39, + "serialized_chars": 38154, + "estimated_input_tokens": 9539, + "estimated_eval_tokens": 19334 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose mix of Transformers v5 regression reports, import failures, and model-specific compatibility bugs. None of the soft pairs look like the same concrete defect, so all soft edges are rejected.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45278 is the broadest and most representative report in the set: it captures multiple import regressions after upgrading to v5.5.0, which matches the dominant theme across the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best global issue representative because it is concrete enough to summarize the upgrade/regression class while still broad enough to cover several nearby reports in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "Related RoFormer area, but one is a kwargs passthrough bug and the other is an attention-interface implementation issue; not the same fix." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training bug and TextClassificationPipeline docs mismatch are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Both are import errors, but they fail on different symbols and likely different missing exports/code paths." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor parallel training and BeitImageProcessorFast label reduction are unrelated subsystems." + }, + { + "left": "issue:44315", + "right": "issue:46209", + "accept": false, + "reason": "Liger-kernel/model_init behavior and FP8 inference on B200 are different runtime problems." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Chunked-prefill cache regression and a cache-strategy typo are not the same underlying bug." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "The Qwen3.5 chat crash may be one example of broader breakage, but the reports are not specific enough to treat as the same concrete bug." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Processor save_pretrained missing files is a specific serialization bug; the other issue is too generic to match it." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading failure and TensorBoard/TF backend initialization error are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Both mention Qwen3.5/Moe, but one is a general model bug and the other is specifically missing tensor-parallel plan support." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "A Qwen3-VL bounding-box generation issue and a PretrainedConfig type-checking regression are unrelated." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Lazy loading problems and unsafe sys.modules access are not the same defect." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Both involve Qwen3/VL outputs, but one is bounding-box hallucination/error and the other is an infinite loop on a specific image." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Both are v5 compatibility complaints, but one is about the env command and the other about loading old checkpoints; different code paths." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 inference support and a DeepSeekV4 weight-conversion script request are different issues." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config is a specific config regression, not the same as a generic v5 breakage report." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star cleanup/performance and GenerationMixin cache missing in chunked_prefill are different underlying changes." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both concern v5 reload compatibility, but they affect different model families and likely different failure points." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression is a specific config bug, not the same as a broad model-breakage umbrella." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "A typo in cache-strategy text is unrelated to general v5 breakage." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "Both are broad v5 complaints, but one is specifically about import errors after upgrade and the other is a generic 'many models break' report; not the same defect." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 API settings and a cache-strategy typo are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:46218", + "accept": false, + "reason": "Both touch imports, but one is import-time slowdown from star imports and the other is a module import failure for a specific model." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Test seeding flakiness and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Model-type warning for sam3 tracker and Liger kernel application are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash-Attn support for Nemotron and Qwen3-VL bounding-box errors are different model-specific bugs." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "Env command breakage and inverse_sqrt scheduler kwargs omission are unrelated." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Slow import and lazy-loading malfunction overlap superficially, but the reports do not identify the same concrete failure mode." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "Position_ids buffer access error and a generic bug report are not the same issue." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch is a specific serialization bug, not the same as the generic issue." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python-version compatibility and Qwen3-VL output errors are unrelated." + }, + { + "left": "issue:45071", + "right": "issue:45092", + "accept": false, + "reason": "PretrainedConfig type-checking regression and old InternVL2 meta-init incompatibility are different bugs." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor kwargs forwarding and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:44908", + "accept": false, + "reason": "Unsafe sys.modules access and scheduler kwargs handling do not share the same code path." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star cleanup/performance and HybridCache import failure are different issues." + }, + { + "left": "issue:44492", + "right": "issue:46328", + "accept": false, + "reason": "Cache-strategy typo and SequenceFeatureExtractor.pad inefficiency are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Model import cleanup and a testing_utils bug are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45850", + "accept": false, + "reason": "Python version support breakage and generic model breakage are not the same concrete defect." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL bounding-box hallucination and Qwen3.5-Moe from_pretrained failure are different symptoms and likely different fixes." + }, + { + "left": "issue:43976", + "right": "issue:46218", + "accept": false, + "reason": "Python-version compatibility and a specific module import failure 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": 21, + "review_unit_index": 54, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:43066", + "issue:43298", + "issue:43352", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43606", + "issue:43618", + "issue:43704", + "issue:43827", + "issue:44077", + "issue:44230", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44655", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44938", + "issue:44998", + "issue:45230", + "issue:45290", + "issue:45310", + "issue:45468", + "issue:45850", + "issue:46018", + "issue:46095", + "issue:46172", + "issue:46218" + ], + "soft_pairs": [ + "issue:44077|issue:44861", + "issue:43827|issue:44908", + "issue:31795|issue:33453", + "issue:44315|issue:44829", + "issue:43576|issue:44273", + "issue:44360|issue:45468", + "issue:44655|issue:44861", + "issue:45230|issue:45310", + "issue:44273|issue:44908", + "issue:43576|issue:46218", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:41669|issue:43704", + "issue:46018|issue:46172", + "issue:31795|issue:34689", + "issue:31515|issue:31795", + "issue:43704|issue:44485", + "issue:44485|issue:45468", + "issue:42371|issue:43704", + "issue:45850|issue:46218", + "issue:43502|issue:43519", + "issue:43519|issue:44623", + "issue:43298|issue:46095", + "issue:43519|issue:43704", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:44261|issue:45468", + "issue:44351|issue:46218", + "issue:43352|issue:46218", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44492|issue:44998", + "issue:43618|issue:44998", + "issue:44998|issue:45850", + "issue:44938|issue:44998", + "issue:44861|issue:45290", + "issue:43525|issue:44861", + "issue:31515|issue:38617", + "issue:43572|issue:44861", + "issue:44230|issue:44623", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:43298|issue:44230" + ], + "prepared_review_unit_hash": "fd05a2d2cf11d669c06bdadac264010b3c06d4f8f400e9c7cdc93b16d8d6e2ec", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 45, + "serialized_chars": 40347, + "estimated_input_tokens": 10087, + "estimated_eval_tokens": 20430 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is noisy and mostly heterogeneous. A few items touch v5 loading/tokenizer regressions, but the soft pairs are only superficially similar and none should be merged as duplicates.", + "confidence": 0.34, + "canonical_issue_reason": "issue:33453 is the clearest, concrete representative of the small loading/regression subset; it is more specific and actionable than the broader/vaguer v5 complaints.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:33453 best fits as the cluster anchor because it describes a real, generic regression class without being overly broad or model-specific.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Optional post_init policy and an AttributeError in tied-weight key handling are different code paths and failure modes." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still mentioning pipeline() is unrelated to inverse_sqrt scheduler kwargs being ignored." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Forward-doc confusion about input_ids/past_key_values is not the same bug as tokenizer loading regression." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel not applying under model_init and flash_attention_3 causing degenerate training are different training issues." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "A broken env command and lazy loading failures are unrelated symptoms." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer ReLU discussion and Gemma-4 audio positional encoding are unrelated model internals." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Pipeline save_pretrained serialization and tied-weight key AttributeError are different subsystems." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "A generic bug report is not enough to treat this as the same as a specific Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading behavior and scheduler kwarg handling are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:46218", + "accept": false, + "reason": "Broken env command and missing Qwen3_5 import are not the same failure." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are different processor/model bugs." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Timestamp math in a processor is unrelated to scheduler kwargs being dropped." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star performance cleanup and a VRAM leak in multithreaded inference are unrelated." + }, + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeek shared-expert gating and a comb.to(...).transpose bug do not match the same code path." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Forward API documentation confusion is not the same as a model-loading breakage for Llama 3.2 Vision." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Slow checkpoint loading and forward docs confusion are unrelated." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leakage and RoPE implementation issues are different problems." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "GLM-5 RoPE and Gemma-4 audio positional encoding are model-specific and unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 API migration guidance has nothing to do with a VRAM leak." + }, + { + "left": "issue:45850", + "right": "issue:46218", + "accept": false, + "reason": "A vague claim that v5 breaks many models is too broad to merge with a specific Qwen3_5 import failure." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "local_files_only network requests and timestamp calculation are different bugs." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp math is unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Continuous batching not supporting audio models and deepseek-v4 weight conversion are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Timestamp calculation and VRAM leakage are unrelated." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return handling and Gemma-4 audio positional encoding are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch and Gemma-4 audio positional encoding are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "Missing rms_norm_eps in MLA q_a_layernorm is not the same as Gemma-4 audio positional encoding." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "Both are import errors, but they are for different symbols and likely different root causes." + }, + { + "left": "issue:43352", + "right": "issue:46218", + "accept": false, + "reason": "Flash Attention 2 support limitation and Qwen3_5 import failure are different issues." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Both mention tokenizer/v5, but one is a behavior change and the other is a wrong decoder type; not enough to call them duplicates." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs handling is unrelated to the placeholder 'Unemployment' issue." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "A typo in cache strategies is unrelated to a placeholder bug report." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions assignment and a placeholder issue are unrelated." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "A placeholder title does not establish the same underlying bug as a broad v5-breakage complaint." + }, + { + "left": "issue:44938", + "right": "issue:44998", + "accept": false, + "reason": "Python 3.14 load failure and a placeholder issue are unrelated." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weight key crashes and chat-template tool-call crashes are different code paths." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config and tied-weight key AttributeError are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow from_pretrained checkpoint loading and an ImportError for layer_type_validation are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig and tied-weight key handling are unrelated." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 support for Qwen3-VL/Qwen3.5 MoE and missing processor save files are different issues." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLFlow reporting on failed training jobs and slow checkpoint loading are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Slow checkpoint loading and a tensor-parallel training bug are not the same failure." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 inference support and a from_pretrained error for Qwen3.5 MoE are different symptoms." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Checkpoint load slowness is not the same as tokenizer-loading regression." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Audio continuous batching support and FP8 Qwen3-VL/Qwen3.5 support are different feature requests/bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 21, + "review_unit_index": 55, + "review_unit_count": 55, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-464", + "nodes": [ + "issue:37934", + "issue:42936", + "issue:43240", + "issue:43668", + "issue:43717", + "issue:43813", + "issue:44031", + "issue:46018", + "issue:46144" + ], + "soft_pairs": [ + "issue:43717|issue:46144", + "issue:42936|issue:46018", + "issue:43668|issue:44031", + "issue:42936|issue:43240", + "issue:43240|issue:46144", + "issue:42936|issue:44031", + "issue:42936|issue:46144", + "issue:42936|issue:43813", + "issue:43240|issue:43668", + "issue:43668|issue:43813", + "issue:37934|issue:43668", + "issue:42936|issue:43668" + ], + "prepared_review_unit_hash": "ee3df0b014496603d4a02edf8497a317d8ceaecb8d65830888ae6640943b8e8b", + "component_budget": { + "node_count": 464, + "item_count": 464, + "soft_pair_count": 1705, + "serialized_chars": 884247, + "estimated_input_tokens": 221062, + "estimated_eval_tokens": 442380 + }, + "budget": { + "node_count": 9, + "item_count": 9, + "soft_pair_count": 12, + "serialized_chars": 8610, + "estimated_input_tokens": 2153, + "estimated_eval_tokens": 4562 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the issues cover unrelated model, tokenizer, config, and typo problems. None of the soft pairs look like the same underlying bug, so all should be rejected.", + "confidence": 0.91, + "canonical_issue_reason": "No true canonical duplicate issue emerges from this set; if one issue must anchor the cluster, issue 42936 is the most concrete functional bug report, but it is still distinct from the others.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 42936 is the strongest representative only in the sense that it describes a specific runtime/model bug with clear behavior, but the cluster overall is not a duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43717", + "right": "issue:46144", + "accept": false, + "reason": "Different model areas and failures: Mamba init-weights behavior vs RoFormer attention interface usage." + }, + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2Former ignore_value handling and DeepSeek-V4 expert gating are unrelated bugs." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "One is a config type-hint issue; the other is a tokenizer regex warning regression." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2Former ignore_value usage and loss-function kwargs propagation are separate code paths." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss kwargs forwarding and RoFormer attention interface usage do not describe the same defect." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Different subsystems and symptoms: segmentation model ignore_value vs tokenizer warning behavior." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "No overlap in concrete bug or code path." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "A Mask2Former behavior bug and a quantization-typo issue are unrelated." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Loss kwargs forwarding and a config type annotation mistake are distinct issues." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "These are different kinds of fixes: type-hint correction vs typo in a string/path reference." + }, + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Llama4 norm semantics and ModernBERT config typing are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Mask2Former ignore_value behavior and ModernBERT norm_eps typing 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": 2, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10519-3", + "nodes": [ + "issue:10519", + "issue:11005", + "issue:11307" + ], + "soft_pairs": [ + "issue:10519|issue:11307", + "issue:10519|issue:11005", + "issue:11005|issue:11307" + ], + "prepared_review_unit_hash": "a83a8558adb7482c5210b7a13c9baeefd6d1a27062875064cc9edae94c81ee04", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2876, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2876, + "estimated_input_tokens": 719, + "estimated_eval_tokens": 1694 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The three issues are not duplicates: one is about tokenizer truncation direction, one requests ReduceLROnPlateau-like scheduler behavior, and one asks for Wav2Vec2 word boundary offsets. The title overlap is superficial only.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue for the cluster: the topics are unrelated and do not describe the same underlying bug or feature request.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 10519 is the clearest and most self-contained report, but it is not a cluster representative because the other issues are unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:10519", + "right": "issue:11307", + "accept": false, + "reason": "Different domains and requests: truncation behavior vs. audio alignment offsets." + }, + { + "left": "issue:10519", + "right": "issue:11005", + "accept": false, + "reason": "Unrelated feature requests: tokenizer truncation direction vs. learning-rate scheduler behavior." + }, + { + "left": "issue:11005", + "right": "issue:11307", + "accept": false, + "reason": "No shared underlying bug or change; one is about schedulers, the other about Wav2Vec2 token timing offsets." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 3, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-13244-3", + "nodes": [ + "issue:13244", + "issue:15980", + "issue:16618" + ], + "soft_pairs": [ + "issue:13244|issue:16618", + "issue:15980|issue:16618" + ], + "prepared_review_unit_hash": "3973c443d761b27691eb5eb2cb0042fdd0640e2180cecbfd84698b65e918506f", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2479, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2479, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These issues do not appear to describe the same underlying bug: one is about TAPAS tokenization differences, one about a misleading private-model download/auth error, and one about failing to load a specific public model.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No clear canonical issue: the three reports concern different code paths and failure modes, so this cluster does not look like a true duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Different problem domains: tokenization parity vs model loading failure for bert-base-uncased." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "One is an auth/download error for private models; the other is a generic inability to load a public model." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-26350-2", + "nodes": [ + "issue:26350", + "issue:33260" + ], + "soft_pairs": [ + "issue:26350|issue:33260" + ], + "prepared_review_unit_hash": "aa222752ae273a9c94064ddade8900882b303de330a3dc02d872ee6a5fc701a5", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated feature-request issues: one for Flash Attention 2 support on more architectures, the other for GGUF support on more architectures. They share a broad theme of architecture support but not the same underlying change or bug, so they should not be clustered as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No single canonical issue fits because the requests target different features and code paths; if forced, issue 26350 is simply the older and more discussed thread, not a duplicate representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different feature requests (Flash Attention 2 vs GGUF) with different implementation targets; same broad topic of architecture support is not enough to treat them as the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 5, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-27088-8", + "nodes": [ + "issue:27088", + "issue:28936", + "issue:38435", + "issue:43834", + "issue:44365", + "issue:44370", + "issue:46222", + "issue:46226" + ], + "soft_pairs": [ + "issue:27088|issue:44370", + "issue:27088|issue:44365", + "issue:27088|issue:43834", + "issue:27088|issue:38435", + "issue:27088|issue:28936", + "issue:28936|issue:43834", + "issue:28936|issue:44365", + "issue:28936|issue:38435", + "issue:28936|issue:44370", + "issue:38435|issue:44370", + "issue:38435|issue:44365", + "issue:38435|issue:43834", + "issue:38435|issue:46222", + "issue:38435|issue:46226", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:43834|issue:46222", + "issue:43834|issue:46226", + "issue:44365|issue:44370", + "issue:44365|issue:46222", + "issue:44365|issue:46226", + "issue:44370|issue:46222", + "issue:44370|issue:46226", + "issue:46222|issue:46226" + ], + "prepared_review_unit_hash": "e513404bcbfc646d88d74b7586f9e02416545d838e79b9d66935df80043586a1", + "component_budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13529, + "estimated_input_tokens": 3383, + "estimated_eval_tokens": 7022 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 24, + "serialized_chars": 13529, + "estimated_input_tokens": 3383, + "estimated_eval_tokens": 7022 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster is mostly repetitive i18n translation-template noise. Only the 46222/46226 pair has enough evidence to accept as a true duplicate; the other pairs are just soft textual matches without concrete shared targets, filenames, or diff evidence.", + "confidence": 0.72, + "canonical_issue_reason": "issue:46222 is the safest canonical anchor because it is part of the only exact duplicate pair in the packet and is the earlier of the two identical issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:46222 is the strongest issue representative here: it has an exact duplicate in 46226, while the rest of the cluster is only template-level similarity without stronger evidence.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Both are only textually similar translation-template issues; there is no concrete evidence they are the same work item." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "High template similarity alone is not enough to collapse these into one issue." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "The titles are too generic to prove duplication, and there is no supporting target or diff evidence." + }, + { + "left": "issue:27088", + "right": "issue:38435", + "accept": false, + "reason": "The locale-tagged titles indicate related theme, but there is still no evidence they are the same issue." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "These have different locale tags in the title and no evidence of the same underlying request." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "The similarity is only at the template level; no concrete shared issue target is shown." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "No evidence beyond repetitive wording suggests these are duplicates." + }, + { + "left": "issue:28936", + "right": "issue:38435", + "accept": false, + "reason": "Different locale-tagged translation requests; nothing in the packet ties them to the same work item." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Soft similarity alone is insufficient to merge these as duplicates." + }, + { + "left": "issue:38435", + "right": "issue:44370", + "accept": false, + "reason": "Near-identical template text is not enough without shared targets, files, or other concrete linkage." + }, + { + "left": "issue:38435", + "right": "issue:44365", + "accept": false, + "reason": "The packet only shows repeated template wording, not the same underlying issue." + }, + { + "left": "issue:38435", + "right": "issue:43834", + "accept": false, + "reason": "These are too similar textually, but there is no concrete evidence they are the same issue." + }, + { + "left": "issue:38435", + "right": "issue:46222", + "accept": false, + "reason": "The titles/body lengths are similar, but there is no evidence that they refer to the same request." + }, + { + "left": "issue:38435", + "right": "issue:46226", + "accept": false, + "reason": "Template similarity alone does not justify merging these as duplicates." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": false, + "reason": "Both are generic placeholder-style issues, but the packet does not prove they are the same underlying item." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": false, + "reason": "Very similar wording, but no concrete evidence of a shared issue target or duplicate report." + }, + { + "left": "issue:43834", + "right": "issue:46222", + "accept": false, + "reason": "The titles are generic and similar, but that is not enough to confirm duplication." + }, + { + "left": "issue:43834", + "right": "issue:46226", + "accept": false, + "reason": "No evidence beyond soft textual similarity supports treating these as the same issue." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": false, + "reason": "Near-identical template text is suggestive, but still not enough to accept as a duplicate." + }, + { + "left": "issue:44365", + "right": "issue:46222", + "accept": false, + "reason": "The packet gives only similarity scores, not concrete shared issue evidence." + }, + { + "left": "issue:44365", + "right": "issue:46226", + "accept": false, + "reason": "Soft similarity alone is insufficient to collapse these into one issue." + }, + { + "left": "issue:44370", + "right": "issue:46222", + "accept": false, + "reason": "There is no concrete link beyond repetitive template language." + }, + { + "left": "issue:44370", + "right": "issue:46226", + "accept": false, + "reason": "The available data does not substantiate that these are the same underlying request." + }, + { + "left": "issue:46222", + "right": "issue:46226", + "accept": true, + "reason": "Same title and identical body length with the strongest similarity score in the packet make this the only grounded duplicate pair." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The 46222/46226 duplicate call is grounded, but the rest of the analysis is not conservative enough. The packet only gives titles, lengths, dates, and soft-similarity scores; it does not substantiate claims that the placeholder-title issues are separate Turkish/Spanish/Romanian locale-specific work items. Several of the false-pair rationales overstate evidence by inferring locale separation from data that isn\u2019t present. The canonical choice of 43834 is also not strongly supported over the other nearly identical placeholder issues." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-36817-16", + "nodes": [ + "issue:36817", + "issue:36879", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322", + "issue:46133" + ], + "soft_pairs": [ + "issue:36817|issue:41211", + "issue:36817|issue:43671", + "issue:36817|issue:36879", + "issue:36879|issue:41211", + "issue:36879|issue:43909", + "issue:36879|issue:44121", + "issue:36879|issue:43837", + "issue:36879|issue:43671", + "issue:37015|issue:43837", + "issue:37015|issue:44121", + "issue:37015|issue:43909", + "issue:36879|issue:37015", + "issue:37015|issue:42503", + "issue:41115|issue:43837", + "issue:41115|issue:44121", + "issue:41115|issue:42503", + "issue:41115|issue:45027", + "issue:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42738|issue:45027", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43671|issue:43909", + "issue:43837|issue:43909", + "issue:43837|issue:45322", + "issue:43837|issue:44121", + "issue:43909|issue:44205", + "issue:44121|issue:45322", + "issue:43837|issue:45027", + "issue:37015|issue:45027", + "issue:43837|issue:45177", + "issue:45322|issue:46133" + ], + "prepared_review_unit_hash": "e5a3cb9e75d9cb920d5d5f3cea96c458fb40e835d5eef274e93ac8517472ad65", + "component_budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21655, + "estimated_input_tokens": 5414, + "estimated_eval_tokens": 11084 + }, + "budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21655, + "estimated_input_tokens": 5414, + "estimated_eval_tokens": 11084 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are all separate model-request issues across different model families and modalities (text, vision, audio, OCR, detection). The similarities are mostly generic 'add model/support' phrasing, not the same underlying change, so none of the soft edges should be merged.", + "confidence": 0.97, + "canonical_issue_reason": "No true canonical duplicate exists here; the issues are distinct requests. If a hub-like representative is needed, issue 43671 is the closest because it is an active speech-model request with the strongest thematic overlap to the other audio-related proposals.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43671 is the strongest cluster anchor by discussion/activity and recurring overlap with related audio-model requests, but it still should not subsume the other issues as duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Different model requests: EuroBert config vs DEIMv2. Shared 'add model' wording only." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert config and Qwen3-TTS are unrelated models and code paths." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert vs RF-DETR are separate architectures and tasks; not the same change." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are different detection/vision models, not one duplicate request." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR is a detection model; LFM2.5 Audio 1.5B is an audio model request." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "RF-DETR and a weight-sparse transformer are unrelated model families." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR and Qwen3-ASR are different modalities and implementation paths." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "RF-DETR vision detection and Qwen3-TTS speech support are not the same underlying change." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are distinct model additions, not duplicates." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT and circuit-sparsity transformer support are different model/code-path requests." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT is a text model; LFM2.5 Audio 1.5B is an audio model request." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are unrelated models with no shared concrete fix." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT models are different model requests, not one duplicate." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support and Qwen3-ASR support target different models and tasks." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 and the weight-sparse transformer request are separate changes." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 and ModernVBERT are different model additions." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 is a model architecture request; Voxtral-4B-TTS is a speech model support request." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated vision vs speech requests." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT and Qwen3-ASR are different model families and modalities." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are distinct model support requests." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT is a concrete model request; BERT-like models with RoPE is a broad architectural theme." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "Generic BERT-with-RoPE discussion and Voxtral TTS support are not the same change." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Same Qwen3 family, but TTS vs ASR are different modalities and code paths." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-TTS and Voxtral TTS are different model integrations, not one duplicate fix." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated speech vs OCR model requests." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-TTS and the sparse transformer request target different model architectures." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio 1.5B are separate audio-model support requests." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio 1.5B are different audio-related models, not the same change." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "ASR support and EUPE vision-language encoder support are unrelated." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR and sparse transformer support are different concrete requests." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "Audio model support and SAM3-LiteText are different modalities and model paths." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Sparse transformer request and EUPE encoder request are unrelated architectures." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral TTS are different speech models with separate integrations." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral TTS are distinct models in different domains." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "ASR support and OCR2 support are different model types and fixes." + }, + { + "left": "issue:45322", + "right": "issue:46133", + "accept": false, + "reason": "EUPE and TIPSv2 are both vision-language related, but they are different models and not one mergeable duplicate change." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer support and EUPE encoder support are different underlying changes." + } + ] + }, + "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": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-38740-2", + "nodes": [ + "issue:38740", + "issue:46443" + ], + "soft_pairs": [ + "issue:38740|issue:46443" + ], + "prepared_review_unit_hash": "89a25fef005fa7647b044253b990e08e4d9aa793199d3a724bd951cf1a01d306", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1505, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1505, + "estimated_input_tokens": 377, + "estimated_eval_tokens": 1010 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are not duplicates: one is a docs request about adding Pruna as an optimization framework, while the other is an open Vision RoPE refactor. They do not describe the same bug, feature, or code path.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical representative because they concern unrelated topics and should not be clustered together.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38740", + "right": "issue:46443", + "accept": false, + "reason": "Reject: the titles and contexts are unrelated (documentation/framework listing vs. Vision RoPE refactor), so they 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": 8, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41084-2", + "nodes": [ + "issue:41084", + "issue:43037" + ], + "soft_pairs": [ + "issue:41084|issue:43037" + ], + "prepared_review_unit_hash": "b95fddc452f4489288b4a793ed551a4aba374dea21a84097767a4a8a005d82fe", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1479, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1479, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are not duplicates: one is a vague request about block decoding, while the other is a specific model-support request for DeepSeek v3.2. They do not appear to describe the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": "issue #43037 is the more concrete, actionable report because it names a specific feature/support target (DeepSeek v3.2), whereas #41084 is too vague to serve as a clear canonical issue.", + "canonical_pr_reason": null, + "best_issue_reason": "#43037 is the better representative issue for the cluster simply because it is specific and scoped; #41084 is too ambiguous to anchor triage.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "The titles and available metadata point to unrelated concerns: generic block decoding vs DeepSeek v3.2 support. No shared targets, files, or explicit linkage suggest the same underlying problem." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 9, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43598-2", + "nodes": [ + "issue:43598", + "issue:43600" + ], + "soft_pairs": [ + "issue:43598|issue:43600" + ], + "prepared_review_unit_hash": "d48756b022ca5936c9106a0c03a3cf181e98a64603e3c6094c84d00d3db3cd3c", + "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 only at a broad Trainer/customization level, but they describe different problems: one is about `remove_unused_column` behavior, the other about FLOPS estimation for custom trainers. They should not be merged as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 43600 is the more concrete, user-facing bug report: it targets FLOPS estimation behavior for custom trainers with a specific failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43598 is the broader customization concern around `remove_unused_column` and may be the better umbrella topic for Trainer configurability.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "They mention custom trainers, but the underlying issues are different: unused-column removal customizability vs FLOPS counting logic. Not the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 10, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43684-2", + "nodes": [ + "issue:43684", + "issue:44405" + ], + "soft_pairs": [ + "issue:43684|issue:44405" + ], + "prepared_review_unit_hash": "25bfc1bb83cc8cb1810e6fe7a863ff60be65759e705ff74aa6fcae11c244e7d9", + "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 model-support requests, but they target different integrations: one asks for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, while the other asks for AutoModelForSequenceClassification support for Qwen3.5. They should not be merged as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "Issue #43684 is the broader model-registration request and is more likely to represent the core automation gap, but it is not the same underlying request as #44405.", + "canonical_pr_reason": null, + "best_issue_reason": "#43684 is the better canonical issue because it is the broader and more general API registration request; #44405 is a separate task for sequence classification support on a different model family/config.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different target APIs and model families: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration versus Qwen3.5 sequence classification support. Related area, but not the same bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-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": "8dadb17f96972fb5e481efd11aa998c0b19e74adfd9ed77af096ca7ea126a8b2", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Only the three \"huggingface\" issues have plausible cluster affinity, and that affinity is title-level only. With no body text, filenames, targets, or references, it is too weak to assert they are true duplicates; 43845 is just the earliest plausible representative among them. \"Fett\" looks unrelated.", + "confidence": 0.58, + "canonical_issue_reason": "issue:43845 is the earliest of the three title-similar \"huggingface\" issues, so it is the best provisional representative if this cluster is later confirmed.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43845 is the best global issue candidate because it is the earliest and cleanest title match in the small title-only \"huggingface\" group.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "The titles are different and there is no concrete overlap beyond weak soft similarity, so this is not enough to treat them as the same issue." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Different titles with no shared body details, targets, or filenames; the evidence is too thin to call them duplicates." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "\"Fett\" does not have enough visible overlap with the title-only \"huggingface\" reports to justify a same-issue merge." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": false, + "reason": "Same short title, but no body text or other concrete matching evidence is available; title similarity alone is not enough to confirm a duplicate." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "The titles match, but the packet provides no additional evidence of the same underlying bug, so this should stay a soft cluster only." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "Identical-looking titles are suggestive, but without body details or references this is still insufficient to confidently merge as duplicates." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The clustering is plausible, but the writeup is a bit too strong for the packet. The only shared evidence for the three \u201chuggingface\u201d issues is title-level soft similarity; all pair records have deterministic_accept=false and no body text, filenames, targets, or references. So calling them duplicates/same topic and accepting all three pairings is not fully conservative. The claim that 43845 is the earliest representative is grounded, but the duplicate conclusion should be softer." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-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": "eeb65e7f966c79e59dbe8f869b0ae66b57a4a0cd0119764e5a7cf23ab4afaaed", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All five items are low-signal and appear unrelated; no soft duplicate pairs are convincing. Issue 44418 is the only one that looks like a plausible substantive report, but it is still too vague to anchor a duplicate cluster with confidence.", + "confidence": 0.84, + "canonical_issue_reason": "Issue 44418 is the least spam-like and the only title that reads like a real user request, so it is the best available issue anchor among otherwise noisy items.", + "canonical_pr_reason": null, + "best_issue_reason": "44418 is the best issue candidate because it appears to be an actual documentation enhancement request, whereas the others look like spam or unrelated text.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are titled '' but there is no concrete bug, feature, or target described to indicate the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Generic spam titles with no shared technical context; similarity is purely superficial." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam and the other is a documentation enhancement suggestion; they do not describe the same problem." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "The titles are unrelated and contain no shared actionable subject matter." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "A documentation request and a spam post are not the same underlying issue." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "No evidence of a common bug or change request beyond both being low-detail closed issues." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Both are spam-labeled titles, but there is no indication they refer to the same report or problem." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "The titles are unrelated; one is placeholder spam and the other is a specific quoted phrase." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "No shared concrete issue is visible; the similarity signal is not enough to treat them as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45145-2", + "nodes": [ + "issue:45145", + "issue:45151" + ], + "soft_pairs": [ + "issue:45145|issue:45151" + ], + "prepared_review_unit_hash": "1f59f6538556bf6eaae22229e8f3d890cf4a30d84edcfab3e97d2bf02abd6746", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues have identical titles and similar lengths, but there is no body-level evidence that they are true duplicates. They look like a possible duplicate / likely same topic, not a confirmed duplicate edge.", + "confidence": 0.69, + "canonical_issue_reason": "#45145 is the earlier report and the best representative for this topic, but the evidence only supports a likely duplicate, not a confirmed one.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best issue candidate because it is the earliest instance and has the same title as #45151, making it the most suitable cluster representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": false, + "reason": "Identical titles and close body lengths suggest the same topic, but there is no content or metadata evidence proving the same underlying issue; this is only a soft similarity." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The packet supports only a soft similarity judgment: identical titles and close body lengths, but no body/content evidence showing the issues are truly duplicates. The analyst overstates the evidence by calling them the same underlying issue and saying the earlier issue \u2018fully represents\u2019 the cluster. A more conservative verdict would be \u2018possible duplicate / likely same topic\u2019 rather than an accepted duplicate edge." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45295-2", + "nodes": [ + "issue:45295", + "issue:45373" + ], + "soft_pairs": [], + "prepared_review_unit_hash": "f5fd95fada2f530ab0999e343c17c16266cde0c7182bb22509972ca44827cd49", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues describe the same missing Gemma 4 sequence-classification support; the later, more specific open issue is the better cluster representative.", + "confidence": 0.94, + "canonical_issue_reason": "issue #45373 is the clearest canonical tracker: it is open, explicitly names the missing class `Gemma4ForSequenceClassification`, and states the module-level gap relative to Gemma 2/3.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #45373 best captures the concrete request and likely remaining work, while #45295 is a shorter earlier duplicate with less detail and is already closed.", + "best_pr_reason": null, + "soft_edge_verdicts": [] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45672-2", + "nodes": [ + "issue:45672", + "issue:45674" + ], + "soft_pairs": [ + "issue:45672|issue:45674" + ], + "prepared_review_unit_hash": "c2805542e07c879f9fe3946da965e2ba2d8808513c9a9eeae8328f21a48c7067", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1673, + "estimated_input_tokens": 419, + "estimated_eval_tokens": 1094 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1673, + "estimated_input_tokens": 419, + "estimated_eval_tokens": 1094 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate 4-bit inference issues: one is a Gemma4 audio-module uint8/torch.finfo TypeError, the other is a BitsAndBytesConfig lm_head exclusion regression causing an AssertionError. They share the broad theme of quantized inference failures but not the same code path or fix.", + "confidence": 0.94, + "canonical_issue_reason": "issue #45674 is the more general, configuration-level regression and is likely the better representative of the cluster than the model-specific Gemma4 audio-module bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #45674 is the strongest standalone issue because it describes a clear, reproducible 4-bit inference regression tied to a specific configuration change and has broader applicability.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Different underlying bugs: one is a torch.finfo uint8 failure in Gemma4 audio modules, the other is a BitsAndBytesConfig skip-modules/ lm_head exclusion assertion. Same general area, but not the same code-path 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": 16, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45762-2", + "nodes": [ + "issue:45762", + "issue:45763" + ], + "soft_pairs": [ + "issue:45762|issue:45763" + ], + "prepared_review_unit_hash": "2485bd27610dfaf21844ae8ba07e8427c4fe689a03c6c3cf0ced418fc320c0f1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are just closed placeholder-like entries with different personal-name titles and different dates; there is no evidence they describe the same underlying bug or request, so they should not be clustered as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical representative because there is no duplicate relationship to resolve; the titles point to different entities and there is no body evidence tying them together.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "Different names and dates in the titles, with no body text, links, or shared evidence indicating the same issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 17, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45764-2", + "nodes": [ + "issue:45764", + "issue:45765" + ], + "soft_pairs": [ + "issue:45764|issue:45765" + ], + "prepared_review_unit_hash": "9ba07b8f4e5f12357788f874257cbf99b4a297bd7a65c4d0f64e643820a828ac", + "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": "The two issues are unrelated non-technical placeholders (personal names plus dates) with no shared problem signal beyond superficial similarity.", + "confidence": 0.98, + "canonical_issue_reason": "Neither issue is a meaningful canonical technical report; both look like unrelated personal-entry stubs rather than duplicates of the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No issue in this cluster is a good representative for duplicate triage, since neither contains actionable problem details.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "Different names and dates; no evidence they describe the same underlying bug or request." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 18, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45810-2", + "nodes": [ + "issue:45810", + "issue:46295" + ], + "soft_pairs": [ + "issue:45810|issue:46295" + ], + "prepared_review_unit_hash": "cff6777a29e1aad5a880bc76cfa3ff7a7da57ebe0a78fec0abf4f92495d9137a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1540, + "estimated_input_tokens": 385, + "estimated_eval_tokens": 1026 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1540, + "estimated_input_tokens": 385, + "estimated_eval_tokens": 1026 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are unrelated feature requests; this is not a duplicate cluster.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No single canonical issue fits the cluster: 45810 asks for a Qwen3_5 token-classification/value-model class, while 46295 requests FP8 training support.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45810", + "right": "issue:46295", + "accept": false, + "reason": "Different feature requests in different areas (model class support vs FP8 training support); they do not describe the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 19, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-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": "70f6389fcf624b5fe6e08c4a76c5cb39ee0346bd354e9feeb4547458d073d85c", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5728, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5728, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The items split into two clear duplicate pairs: 45870/45871 are the same example request, and 45872/45873 are the same DataForge scraping-service request. 45869 is a separate dataset promo/request and does not match the others.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue for the full cluster: there are two separate duplicate topics (45870/45871 and 45872/45873), while 45869 is unrelated.", + "canonical_pr_reason": null, + "best_issue_reason": "45871 and 45873 are the best representatives of their respective duplicate pairs; 45869 should not be treated as canonical for either topic.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Different topics: a dataset bundle promo/request vs an example about financial time series prediction." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Different topics: a dataset bundle promo/request vs an example about financial time series prediction." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Same title and apparently same example request; these look like exact duplicates." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Different requests: one is a prediction example, the other is a financial time series plus scraping service topic." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Same title and apparently same service/request; these look like exact duplicates." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The analysis is grounded in the packet: 45870/45871 and 45872/45873 have identical titles and body lengths, while 45869 has a clearly different title/topic. The conclusion that the cluster splits into two duplicate pairs plus one unrelated issue is conservative and supported by the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 20, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46439-2", + "nodes": [ + "issue:46439", + "issue:46441" + ], + "soft_pairs": [ + "issue:46439|issue:46441" + ], + "prepared_review_unit_hash": "91b021c40e97535b87037ad4af00bbcfba461216da485858be3efdac0de954e4", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1636, + "estimated_input_tokens": 409, + "estimated_eval_tokens": 1074 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are in the same static-cache / linear-attention area, but they describe different failure modes: one is shape corruption during `early_initialization`, the other is a Qwen3.5-specific `KeyError: 'linear_attention'`. Related, but not the same bug.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 46439 is the better representative of the broader static-cache/linear-attention corruption problem because it states a concrete shape/runtime failure and a likely root-cause area.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46439", + "right": "issue:46441", + "accept": false, + "reason": "Different concrete failures and likely different code paths: wrong-shape `RuntimeError` in early initialization vs Qwen3.5 `KeyError` for missing `linear_attention` handling. Similar subsystem, but not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 21, + "component_count": 21, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46551-2", + "nodes": [ + "issue:46551", + "issue:46552" + ], + "soft_pairs": [ + "issue:46551|issue:46552" + ], + "prepared_review_unit_hash": "849ab6d615658470f0ead7b6cc29bb1579e3c13aadbdfee6244274ee6d1951c5", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1707, + "estimated_input_tokens": 427, + "estimated_eval_tokens": 1110 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1707, + "estimated_input_tokens": 427, + "estimated_eval_tokens": 1110 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both artifacts are the same issue report: identical title and essentially identical content about the Wav2Vec2CTCTokenizer v5 word-delimiter bug. They should be clustered together as duplicates/reposts of the same underlying bug.", + "confidence": 0.99, + "canonical_issue_reason": "Issue #46551 is a closed instance with a bit more discussion activity, making it the more natural anchor for the duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "#46551 is the best issue representative because it is the closed/original-looking report and already carries the discussion context for the bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46551", + "right": "issue:46552", + "accept": true, + "reason": "Identical title and matching bug description indicate the same underlying issue, not just a related problem." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster judgment is well grounded: the two issues have identical titles, identical body length, and a soft-similarity score/jaccard of 1.0, so treating them as duplicates is conservative and supported by the packet. The representative/canonical choice is also reasonable since #46551 is closed and has slightly more discussion activity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-39895-14", + "nodes": [ + "pull_request:39895", + "pull_request:41855", + "pull_request:42668", + "pull_request:43251", + "pull_request:43254", + "pull_request:43424", + "pull_request:43823", + "pull_request:43995", + "pull_request:44827", + "pull_request:45170", + "pull_request:45213", + "pull_request:45655", + "pull_request:46048", + "pull_request:46267" + ], + "soft_pairs": [ + "pull_request:39895|pull_request:43424", + "pull_request:41855|pull_request:43424", + "pull_request:41855|pull_request:43823", + "pull_request:42668|pull_request:43823", + "pull_request:43251|pull_request:43823", + "pull_request:43251|pull_request:44827", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:43995", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:46267", + "pull_request:44827|pull_request:46048", + "pull_request:44827|pull_request:45655", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "ffdb47284d0535eda03297df83dd065aa26e82d555f6bdee4b839ab1b508607f", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 38086, + "estimated_input_tokens": 9522, + "estimated_eval_tokens": 19300 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 19, + "serialized_chars": 38086, + "estimated_input_tokens": 9522, + "estimated_eval_tokens": 19300 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous PRs from unrelated model/features. The only true duplicate-like pair is the two fixed_cross_entropy changes (43251 and 43254), which address the same code path and can plausibly be merged.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43251 is the clearest representative of the shared cross-entropy kwargs fix: it is issue-linked, concrete, and overlaps directly with 43254 in the same function and behavior change.", + "best_issue_reason": null, + "best_pr_reason": "PR 43251 is the best PR candidate because it most directly captures the underlying bug fix and has the explicit tracking issue, while 43254 is a near-duplicate variant.", + "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." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Different subsystems: Mistral tokenizer methods vs. executorch export testing." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Unrelated changes: tokenizer API additions vs. adding MobileLLM model support." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Different processor robustness fixes vs. new MobileLLM model implementation." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Cross-entropy loss kwargs fix is unrelated to adding MobileLLM." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Different areas: loss utility change vs. Mistral4 test/model updates." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Executorch dynamic-shape export test is unrelated to Mistral4 tests/model changes." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Executorch exportability test vs. MobileLLM model addition are unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "New MobileLLM model support is unrelated to config resolution ordering fix." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "MobileLLM model addition and Falcon output-interface refactor are different code paths." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactor is unrelated to the model-creation skill/docs PR." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test/model work is unrelated to the layrnorm->layernorm rename fix." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46267", + "accept": false, + "reason": "Mistral4 testing/model changes do not match the chat parsing overhaul." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Different features: Mistral4 fixes vs. adding a TDT loss kernel integration." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 test/model updates are unrelated to config resolution ordering." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 fixes are unrelated to the model-creation skill and related scaffolding." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "Adding MobileLLM is unrelated to the model-creation skill PR." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactor and model-creation skill changes are unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Same underlying fix in the same function: both expose/pass supported cross_entropy kwargs (weight, label_smoothing) through fixed_cross_entropy, and could be merged into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays appropriately conservative. The only accepted soft-edge pair, 43251 vs 43254, is well supported by the diffs: both modify the same fixed_cross_entropy helper in the same file, target the same issue, and implement the same supported-kwargs pass-through with only minor signature/implementation differences. The other pairwise rejections are also conservative and match the packet content." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 36, + "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": "fef3ecf1a4e7cd878c7bc490d66dcb45480b9211ede3d94810480ee1b3f09b73", + "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 pull requests, but they implement different scopes: one adds the core model, the other adds a realtime variant with acoustic-tokenizer/generation changes. Shared files reflect the same package, not the same fix.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #40546 is the cleaner representative of the core VibeVoice addition; #42781 is a separate realtime extension and draft, so it should not be treated as the same change.", + "best_issue_reason": null, + "best_pr_reason": "PR #40546 is the best single representative for this cluster because it targets the base VibeVoice implementation and is more complete/stable than the draft realtime follow-up.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Reject: they share VibeVoice scaffolding, but #40546 is the base implementation while #42781 adds a distinct realtime path with tokenizer/generation changes; not the same concrete bug or mergeable change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 3, + "component_count": 36, + "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": "ae3796b8d76991d426520b4b6ebcc31fb8bed1be482bf5597984c42a70e6b2d7", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4674, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4674, + "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 PRs are unrelated feature additions in different model families, so they should not be clustered as duplicates.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical duplicate PR: #40755 adds TimesFM covariate forecasting, while #41224 adds DINOv3 image classification support; they target different code paths and products.", + "best_issue_reason": null, + "best_pr_reason": "#41224 is the better representative only in the sense that it is a focused, review-mature feature PR; #40755 is a separate, larger TimesFM change and not a duplicate.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different models and implementations: TimesFM covariate forecasting vs DINOv3 image classification. They do not fix the same bug or add the same change and would not be merged into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 4, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-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": 36, + "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": "51efc0afe41677b7f85d339899d636edbc9d54f80df4de758442bed478fc0fa5", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10436, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10436, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three items are separate model-addition PRs (DeepSeek OCR, Isaac, Kimi2-6) that share framework wiring patterns but target different model families and code paths, so they are not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45630 is the strongest representative if one must be chosen: it is the most mature and broad model-addition PR, with the widest set of model/auto/processor updates and the highest review activity.", + "best_issue_reason": null, + "best_pr_reason": "PR 45630 is the best cluster representative on completeness and review depth, but it is still unrelated to the other PRs at the model level.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different model additions: DeepSeek OCR vs Isaac. They share generic auto-registration/docs scaffolding, but not the same underlying model or code path." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Different model additions: Isaac vs Kimi2-6. Similar broad framework files are touched, but the concrete model implementations, processors, and tests are unrelated." + } + ] + }, + "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": 36, + "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": "da90dfc1d067b105136eb92141170731f884ce0197f215b83316707553d0cc36", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5934, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5934, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are workflow maintenance upgrades, but they change different action families for different reasons: #42975 focuses on Node 24 compatibility for several GitHub Actions, while #42976 upgrades Docker build actions and a few other actions to newer releases. They overlap in files touched, but not in the underlying change.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "#42975 is the more specific and narrowly scoped PR, centered on a clear Node 24 compatibility upgrade across workflows; #42976 is broader and less directly tied to that concrete compatibility goal.", + "best_issue_reason": null, + "best_pr_reason": "#42975 is the better representative PR because its intent is explicit and cohesive, whereas #42976 is a looser general version bump across workflow tooling.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "Reject: both touch GitHub Actions workflows, but they update different actions for different purposes and are not the same underlying fix/change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42978-5", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45490", + "pull_request:46562" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43451|pull_request:45490", + "pull_request:45490|pull_request:46562" + ], + "prepared_review_unit_hash": "c6cecf8af4738a73a63ace929149104524046d4f90acac52501eda46df2263d0", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 7, + "serialized_chars": 17511, + "estimated_input_tokens": 4378, + "estimated_eval_tokens": 9012 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 7, + "serialized_chars": 17511, + "estimated_input_tokens": 4378, + "estimated_eval_tokens": 9012 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are all model-addition PRs for different architectures and model families; the shared files are mostly boilerplate registry/docs updates, not evidence of the same underlying change. No soft edges should be merged.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43451 (Add Molmo2) is the most substantial representative here: it has the broadest integration surface and the highest review activity, so it best stands in for the cluster if one PR must be chosen.", + "best_issue_reason": null, + "best_pr_reason": "#43451 is the best representative PR for this cluster because it is the most complete and actively reviewed model-addition change, whereas the others are separate model introductions.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different models: ViT NEPA vs Molmo. They share generic docs/auto-registry updates, but not the same code-path fix or mergeable change set." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "Different model families: ViT NEPA vs Molmo2. Shared auto/docs boilerplate is incidental; these are separate model implementations." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "ViT NEPA and CTSM are unrelated architectures (vision vs time-series). Similar framework wiring does not indicate the same underlying change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are distinct model additions with different files and APIs; they are not the same concrete bug/change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo vs CTSM are unrelated model additions. Overlap is limited to shared registry/documentation plumbing." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM target different model families and code paths; the shared auto-registration files are too generic to treat them as duplicates." + }, + { + "left": "pull_request:45490", + "right": "pull_request:46562", + "accept": false, + "reason": "CTSM is a time-series model, while Step3p7 is a multimodal/vision model addition. They are separate implementations, not one 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": 8, + "component_count": 36, + "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-43291-5", + "nodes": [ + "pull_request:43291", + "pull_request:43665", + "pull_request:44815", + "pull_request:46380", + "pull_request:46539" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43291|pull_request:46380", + "pull_request:43665|pull_request:46539", + "pull_request:43665|pull_request:44815", + "pull_request:43665|pull_request:46380", + "pull_request:44815|pull_request:46539", + "pull_request:44815|pull_request:46380", + "pull_request:46380|pull_request:46539" + ], + "prepared_review_unit_hash": "832bab5aa24d3d6a9b7b1784349fbf4e963626cb2ade4b07b95508b83d7f23d5", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 14125, + "estimated_input_tokens": 3532, + "estimated_eval_tokens": 7320 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 14125, + "estimated_input_tokens": 3532, + "estimated_eval_tokens": 7320 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are not true duplicates: they span unrelated areas (Whisper decoding, CLIP/ViT tests, FP8 dequant/loading, generation watermark validation, PhiMoE integration tests). The strongest standalone production fix is the FP8/loading change, but the cluster as a whole should not be merged together.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44815 is the most concrete, self-contained production fix in the cluster: it adjusts core loading and FP8 kernel behavior for a specific dequant/loading bug, unlike the test-only or broad/mixed-change PRs.", + "best_issue_reason": null, + "best_pr_reason": "PR 44815 is the best overall PR candidate because it addresses a clear code-path problem with a focused implementation and is less ambiguous than the test-heavy or multi-area patches.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Different targets and bugs: Whisper tokenizer decoding vs. unrelated CLIP/ViT test edits. Shared wording like \"fix\" and test changes is not enough." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper decode/tokenization logic is unrelated to FP8 dequant/core loading. No shared code path or mergeable single fix." + }, + { + "left": "pull_request:43291", + "right": "pull_request:46380", + "accept": false, + "reason": "Both mention Whisper only incidentally, but the actual fixes are different: tokenizer decode behavior vs. generation watermark/config handling and assorted test updates." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46539", + "accept": false, + "reason": "Both are test-oriented, but they affect different models and failure modes. These do not look like the same underlying bug." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "CLIP/ViT test assertions are not the same underlying issue as FP8 loading/dequant logic." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46380", + "accept": false, + "reason": "The overlap is only generic test edits. The substantive code paths differ completely, so they are not duplicates." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46539", + "accept": false, + "reason": "FP8/core loading fixes and PhiMoE integration-test setup are unrelated code paths and could not plausibly be one PR." + }, + { + "left": "pull_request:44815", + "right": "pull_request:46380", + "accept": false, + "reason": "Different concrete bugs: FP8 dequant/loading vs. generation watermark validation plus other model/test changes. Too broad to treat as one fix." + }, + { + "left": "pull_request:46380", + "right": "pull_request:46539", + "accept": false, + "reason": "Generation config/watermark handling and PhiMoE integration-test behavior are unrelated; they only share a generic \"fix\" title." + } + ] + }, + "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": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43488-2", + "nodes": [ + "pull_request:43488", + "pull_request:43532" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:43532" + ], + "prepared_review_unit_hash": "6c7765ff877e740da4b78916d5aabcdf565ccb3b6dcf73e665139beaa48d28be", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4563, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4563, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are not duplicates: one is a small packaging/version/checking tweak in setup.py and ViT tests, while the other rewires workflow/testing utilities and also touches multiple files. They share only a test filename, not the same underlying change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "No clear canonical PR for duplication: these PRs target different code paths and different kinds of repository-bot/demo changes.", + "best_issue_reason": null, + "best_pr_reason": "PR #43532 is the better representative item in this cluster because it is the broader, more substantive change; #43488 is a narrower setup/version tweak with a different purpose.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Reject: they do not fix the same concrete bug/change. #43488 edits setup/version/dependency metadata and a ViT test, while #43532 changes CI/workflow and notification/testing utilities with a different scope." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 10, + "component_count": 36, + "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": "cf7bcfa0cd2825b8797cf2e1759ebe655ea23ced3f7244339ad6c58d29d52861", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs add different new pipelines: one for promptable concept segmentation and one for promptable visual segmentation. They share documentation and registry scaffolding, but they target different tasks, models, and test suites, so they are not duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43613 is the better representative because it is the broader visual-segmentation pipeline change and has more model/test coverage, making it the more substantial member of this cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR #43613 is the strongest standalone candidate: it implements a concrete pipeline addition across multiple segmentation-related models and tests, so it is the best exemplar of this class of changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Different underlying features and code paths: concept segmentation vs visual segmentation. They overlap in docs/registry plumbing, but the pipeline implementations, model coverage, and tests are distinct, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 11, + "component_count": 36, + "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": "1ced6893ee561db094e8665a772b10c51a3bfabcd4038dc8eaf7693239852798", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate Trainer extensibility PRs in the same file: one changes gradient-accumulation loss scaling, the other refactors signature-column filtering. They do not fix the same underlying bug or implement the same change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43651 is the more focused behavior change around training loss scaling, but it is not a duplicate of PR 43663.", + "best_issue_reason": null, + "best_pr_reason": "No single PR should be treated as the duplicate representative here; 43651 is only slightly more self-contained, while 43663 addresses a different Trainer override point.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Different code paths and goals: 43651 adds _loss_is_scaled_for_ga for gradient-accumulation loss handling, while 43663 adds _get_signature_columns for column filtering. Shared file alone is not enough to merge them as one change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 12, + "component_count": 36, + "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": "719d2fdd11e1eb45ae7058e256b93e72573102321ce7adedc0ee60753c66f042", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3663, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs touch the same CLI file and reference a TypeAdapter NameError, but they appear to fix different failure modes: one is a broader annotation/runtime typing cleanup, the other is specifically about pydantic being absent. They are not safe duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR stands out as the same fix for both items; 43842 is the tighter, more explicit match to the TypeAdapter/pydantic NameError, while 43656 looks broader and likely addresses a different annotation-evaluation problem.", + "best_issue_reason": null, + "best_pr_reason": "43842 is the better representative of the cluster because its title and diff directly target the TypeAdapter NameError when pydantic is not installed, with a narrowly scoped change in serve.py.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Reject: they share a file and broad CLI context, but the fixes are not clearly the same bug. 43656 appears to address runtime annotation evaluation via __future__ annotations, while 43842 explicitly fixes missing TypeAdapter import handling when pydantic is absent; they also target different issues." + } + ] + }, + "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": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-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": 36, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44066|pull_request:44071", + "pull_request:44018|pull_request:44068", + "pull_request:44066|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44019|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44086|pull_request:44722", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44116", + "pull_request:44068|pull_request:44722", + "pull_request:44018|pull_request:44116", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44085", + "pull_request:44018|pull_request:44086", + "pull_request:44068|pull_request:44086", + "pull_request:44071|pull_request:44085", + "pull_request:44059|pull_request:44086", + "pull_request:44085|pull_request:44086", + "pull_request:44098|pull_request:44101", + "pull_request:44059|pull_request:44722", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:43997|pull_request:44073", + "pull_request:44071|pull_request:44086", + "pull_request:44017|pull_request:44071", + "pull_request:44073|pull_request:44101", + "pull_request:44073|pull_request:44161", + "pull_request:44059|pull_request:44071", + "pull_request:44056|pull_request:44101", + "pull_request:44018|pull_request:44059", + "pull_request:44059|pull_request:44068", + "pull_request:44101|pull_request:44161", + "pull_request:43997|pull_request:44017", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44018", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:44018|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44010|pull_request:44073", + "pull_request:44018|pull_request:44129", + "pull_request:43997|pull_request:44074", + "pull_request:43997|pull_request:44129", + "pull_request:44010|pull_request:44018", + "pull_request:44071|pull_request:44129", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154" + ], + "prepared_review_unit_hash": "6b8b64c85a0fc08dc9b10057510e0b4cfdfbe88d0911d5075a74264144ecb2c2", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84903, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 + }, + "budget": { + "node_count": 27, + "item_count": 27, + "soft_pair_count": 72, + "serialized_chars": 75098, + "estimated_input_tokens": 18775, + "estimated_eval_tokens": 37806 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a broad set of model-specific PRs that all refactor output tracing to the standardized decorator-based interface. The only strong duplicate candidates are same-model pairs within GPT-J and GPT-Neo; cross-model pairs share the theme and tracking issue but not the same concrete code path.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "44066 is the best canonical PR: it is explicitly tied to #43979 and contains a focused, concrete GPT-J/CodeGen output-tracing refactor that sits at the center of the genuine duplicate chain.", + "best_issue_reason": null, + "best_pr_reason": "44722 is the strongest single PR candidate in the duplicate set: it targets the same GPT-J/CodeGen output-tracing change with the clearest same-file overlap to the other accepted GPT-J PRs.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different model families/files; the shared issue target only indicates the same refactor theme, not the same underlying code path." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Different models and implementations; both are output-tracing refactors, but not the same bug or change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Both are GPT-J output-tracing refactors in the same file, so they look like the same underlying change/iteration." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "ResNet and EfficientNet are different code paths; same issue target is not enough to make them duplicates." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet and GPT-J are unrelated model implementations, so these are not the same change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen and EfficientNet are different model families; no shared code path." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa v2 are unrelated implementations, despite the same refactor pattern." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen and MGP-STR do not share the same concrete fix, only the broad output-tracing theme." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen and MPT are different model code paths; not a duplicate pair." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Both touch GPT-Neo output tracing in the same file and look like duplicate iterations of the same refactor." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J and GPT-Neo are separate model implementations, so these are not the same concrete change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are unrelated code paths." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are different models; same refactor theme is too broad." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are different model implementations." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo and GPT-J are different model code paths, so this is not the same underlying change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are unrelated implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44086", + "accept": false, + "reason": "MPT and MGP-STR are different models and fix different code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44085", + "accept": false, + "reason": "MPT and GPT-J are not the same underlying bug/change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 and MGP-STR are different model implementations." + }, + { + "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 are different model families; no same-code-path match." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT2 and GPT-J are different model code paths, so this is not a duplicate change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different models; same refactor theme is too broad." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "Segformer and LongT5 are unrelated implementations." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44722", + "accept": false, + "reason": "VisualBert and GPT-J are different code paths." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44161", + "accept": false, + "reason": "VisualBert and LongT5 are different models and fixes." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT2 and MPT are distinct model implementations." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44073", + "accept": false, + "reason": "TextNet and VisualBert are different code paths." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44101", + "accept": false, + "reason": "MPNet and Flaubert are unrelated model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and Segformer are different models; the shared refactor theme is not enough." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are separate model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT and ImageGPT are different models and changes." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert are unrelated implementations." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 do not share the same concrete fix." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are different model code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are unrelated implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44071", + "accept": false, + "reason": "RegNet and MPT are different models." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are unrelated model code paths." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert and VisualBert are different implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are unrelated model code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44059", + "accept": false, + "reason": "GPT-Neo and GPT2 are different model implementations." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBert and GPT-Neo are not the same underlying change." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are different model code paths." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "CVT/FNet and LongT5 are unrelated implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "CVT/FNet and Segformer are different model families." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert and DeBERTa v2 are different code paths." + }, + { + "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 in the same files; 44722 looks like a direct follow-up or revision of 44066." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Both are GPT-J output-tracing refactors and share the same underlying file-level change." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-J and SpeechT5 are different models, so this is not the same fix." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": false, + "reason": "CVT/FNet and ResNet are different model implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": false, + "reason": "CVT/FNet and SpeechT5 are unrelated code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": false, + "reason": "DeBERTa v2 and SpeechT5 are different models and changes." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44154", + "accept": false, + "reason": "DeBERTa v2 and VITS are unrelated implementations." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays grounded in the packet and the soft-edge calls are generally conservative. The accepted duplicate pairs all have a plausible same-file/same-model basis, especially the GPT-J and GPT-Neo iterations. The cross-model pairs are rejected appropriately. One naming oddity remains in 44085, but the verdict does not overstate that evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 36, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44024", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44024|pull_request:44076", + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "daf4fec55da030acfa937262405a8b9b72405b90817768ac33748de47fe86dfd", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84903, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 6, + "serialized_chars": 17126, + "estimated_input_tokens": 4282, + "estimated_eval_tokens": 8820 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These PRs are all model-specific refactors to standardized output tracing/capture, but they touch different architectures and code paths, so they are thematically similar rather than true duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No exact duplicate PR stands out in this set; if a representative is needed, PR #44129 is the clearest because it has an explicit issue target and a well-scoped standardized output capture refactor.", + "best_issue_reason": null, + "best_pr_reason": "PR #44129 is the best representative of the cluster because it is the most explicitly tied to a tracking issue and clearly centers on the same standardized output-capture migration pattern.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "Both are output-tracing refactors, but FocalNet and ImageGPT are different model implementations with separate forward paths and capture hooks; they are not the same underlying change." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "VisualBert and TextNet are unrelated model families; the PRs share a refactor theme but not a common concrete code-path bug or a mergeable single fix." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "MPNet and VisualBert each introduce model-specific output-capture plumbing in different files and layers; this is the same pattern, not the same change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "These touch different model backends (MPNet vs TextNet) and different forward/output mechanics, so they should not be merged as duplicates." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated subsystems; both refactor output handling, but they are separate model-local changes with no shared bug fix." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert attention tracing and SpeechT5 output capture are model-specific refactors in different architectures; they do not fix the same concrete problem." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 15, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43998-13", + "nodes": [ + "pull_request:43998", + "pull_request:43999", + "pull_request:44000", + "pull_request:44001", + "pull_request:44002", + "pull_request:44003", + "pull_request:44004", + "pull_request:44025", + "pull_request:44026", + "pull_request:44027", + "pull_request:44028", + "pull_request:44029", + "pull_request:44030" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44028", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030" + ], + "prepared_review_unit_hash": "347d6bb2975d2e8b06e16f0f25354f6ac7c2544206b0f89190de89df8f51add7", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40172, + "estimated_input_tokens": 10043, + "estimated_eval_tokens": 20342 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40172, + "estimated_input_tokens": 10043, + "estimated_eval_tokens": 20342 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a batch of separate model-specific output-tracing refactors, not duplicate bug fixes. The PRs share a common implementation pattern (`can_return_tuple` / output-capture cleanup) but target different model families and code paths.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44003 is the most representative of the shared refactor theme: it spans two related model files (`falcon_mamba` and `mamba`) and adds the new output-capturing plumbing, making it a good cluster anchor.", + "best_issue_reason": null, + "best_pr_reason": "PR 44003 is the strongest global representative because it is broader than the single-file refactors and most clearly shows the common output-tracing change pattern.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families (`timm_backbone` vs `upernet`); same refactor motif only, not the same concrete code path." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "`timm_backbone` and `vision_text_dual_encoder` are unrelated model paths; both refactor output handling but do different changes." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Separate models (`timm_backbone` vs `univnet`); no shared underlying bug or mergeable fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "`timm_backbone` and `mamba/falcon_mamba` are distinct code paths; similarity is only the generic output-tracing cleanup." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Different subsystems (`timm_backbone` vs `rwkv`); same style of refactor, not the same change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "`mobilenet_v1` and `univnet` are unrelated model implementations; not a duplicate fix." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families (`mobilenet_v1` vs `upernet`); same output-tracing pattern only." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "`mobilenet_v1` and `depth_anything` touch different forward paths and model architectures." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Separate encoder models with different call paths; similar refactor theme is too broad to treat as duplicate." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "`mobilenet_v1` vs `speech_encoder_decoder` are different model stacks; not the same bug/change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "`vision_text_dual_encoder` and `univnet` are unrelated; they only share the refactor style." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models and code paths (`vision_text_dual_encoder` vs `upernet`); not a single underlying fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "`vision_text_dual_encoder` and `mamba/falcon_mamba` are separate architectures; same theme, different change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "`vision_text_dual_encoder` vs `depth_anything`; both adjust output handling but are not the same concrete PR." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "`univnet` and `upernet` are unrelated model code paths; no duplicate defect or shared fix." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "`univnet` vs `mamba/falcon_mamba`; only the generic output-tracing pattern overlaps." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families (`univnet` vs `depth_anything`); not mergeable as one underlying change." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "`upernet` and `rwkv` are different architectures and forward paths; same refactor motif only." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "`upernet` vs `mamba/falcon_mamba`; distinct code paths and different output plumbing." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "`mamba/falcon_mamba` and `superpoint` are unrelated model implementations; similar tracing cleanup is not a duplicate." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "`mamba` and `codegen` affect different forward implementations and return conventions; not the same fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "`timm_backbone` and `codegen` are different model families; the overlap is only high-level refactoring." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "`univnet` and `codegen` touch unrelated model code paths; not a duplicate change." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "`upernet` vs `codegen`; same generic output-refactor theme, but different concrete models." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "`vision_text_dual_encoder` and `codegen` are separate implementations; no shared underlying bug." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "`depth_anything` and `vision_encoder_decoder` are different model families; not the same change." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "`depth_anything` vs `speech_encoder_decoder`; distinct code paths, only similar refactor style." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "`depth_anything` and `superpoint` are unrelated models; not a duplicate fix." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "`depth_anything` and `rwkv` are separate architectures; same output-tracing theme only." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "`depth_anything` vs `dpr`; different model code paths and return types, not the same bug/change." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "`vision_encoder_decoder` and `rwkv` are unrelated implementations; overlap is only the refactor pattern." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "`vision_encoder_decoder` vs `speech_encoder_decoder`; similar wrapper pattern, but different concrete code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "`vision_encoder_decoder` and `superpoint` are different model families; not a duplicate." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "`vision_encoder_decoder` and `dpr` affect different model wrappers and outputs; not the same fix." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "`speech_encoder_decoder` and `superpoint` are unrelated model paths; same style of cleanup only." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "`speech_encoder_decoder` vs `rwkv`; distinct architectures and code paths." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "`speech_encoder_decoder` and `dpr` are separate model wrappers; not a duplicate change." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "`superpoint` and `rwkv` are unrelated models; only the tracing/refactor motif matches." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "`superpoint` vs `dpr`; different code paths and return handling, not the same underlying change." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "`rwkv` and `dpr` are distinct models; same broad output-tracing cleanup, but not a shared 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": 16, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44662-2", + "nodes": [ + "pull_request:44662", + "pull_request:45401" + ], + "soft_pairs": [ + "pull_request:44662|pull_request:45401" + ], + "prepared_review_unit_hash": "64d4a613513c07698063e39b7479adf643d6664cf2194682d5c4d311e91287e0", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6550, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6550, + "estimated_input_tokens": 1638, + "estimated_eval_tokens": 3532 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated model-addition PRs that only overlap on shared Transformers scaffolding (auto registration, docs toc, package init). They do not appear to fix the same code-path problem or be mergeable into one PR.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "If a PR must represent the cluster, #45401 is the cleaner candidate because it is narrower, has an explicit issue target, and is a self-contained Voxtral TTS implementation.", + "best_issue_reason": null, + "best_pr_reason": "#45401 is the better standalone PR in this pair: it has explicit tracking, a focused scope, and clearer feature intent. #44662 is a separate PenguinVL implementation and not the same change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Reject: PenguinVL and Voxtral TTS are different model implementations with different codepaths, files, and purposes; the overlap is only generic framework registration/docs, not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 36, + "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": "0dc1e2e980dff3d8317e915f1eddb60e316c5daeeadc5da8b7a0eb4f1b8b6696", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4407, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4407, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are about model/module fusion, but they target different layers and concrete changes: PR #44979 introduces a general Module Fusion API and tests, while PR #45363 wires n-to-1 kernel fusion through KernelConfig across existing integration points. They are related feature work, not the same underlying change.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45363 is the broader, more end-to-end feature implementation: it modifies core model loading/integration code, adds a user-facing example, and connects fusion through KernelConfig. That makes it the better representative for the cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR #45363 is the strongest standalone representative because it is the more complete integration of the fusion feature and touches the actual runtime path, whereas PR #44979 is mainly the API foundation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "They share the general fusion theme, but PR #44979 defines a generic Module Fusion API while PR #45363 implements n-to-1 kernel fusion via KernelConfig in different files and code paths. They could not plausibly be merged as one duplicate PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 36, + "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": "9d90558a1d93abd68ad85f06781dba68e61481665578f9edd6d0bc22452d3ea1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests address the same Mistral regex AttributeError by changing the pre-tokenizer access from backend_tokenizer.pre_tokenizer to tokenizer.pre_tokenizer; 45317 is the fuller version with tests.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "45317 is the better canonical PR because it contains the same fix as 45086 plus test coverage and a clearer description/title.", + "best_issue_reason": null, + "best_pr_reason": "45317 is the strongest representative of the cluster: same underlying code change as 45086, but more complete and validated by tests.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same concrete bug fix in the same file for the same issue target; 45317 is essentially an expanded version of 45086 with tests." + } + ] + }, + "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": 36, + "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-45115-4", + "nodes": [ + "pull_request:45115", + "pull_request:45569", + "pull_request:46332", + "pull_request:46565" + ], + "soft_pairs": [ + "pull_request:45115|pull_request:45569", + "pull_request:45569|pull_request:46565", + "pull_request:46332|pull_request:46565" + ], + "prepared_review_unit_hash": "2aa0f234b08791945081aeba8fac3acaadda999a2776cbb01c59cf5671f631b8", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 3, + "serialized_chars": 13472, + "estimated_input_tokens": 3368, + "estimated_eval_tokens": 6992 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 3, + "serialized_chars": 13472, + "estimated_input_tokens": 3368, + "estimated_eval_tokens": 6992 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are related by the Nemotron family and some shared auto/doc plumbing, but they target different concrete changes: NemotronH refactoring, NemotronH dense/sparse split, Nemotron ASR streaming, and Nemotron 3.5 ASR streaming. I do not see duplicate-level overlap.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46332 is the clearest standalone feature implementation in the cluster: it adds Nemotron ASR streaming end-to-end (model, losses, masking, auto mappings, docs). The others are either a different model family/version or a separate architectural refactor, so they are not better canonical duplicates.", + "best_issue_reason": null, + "best_pr_reason": "PR #46332 is the strongest representative PR for this cluster because it has the most complete, coherent change set and the richest implementation surface. PR #46565 is a related follow-up for Nemotron 3.5 ASR, while #45115 and #45569 concern NemotronH architecture/model refactoring rather than ASR streaming.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Both touch NemotronH code, but #45115 is a structural inheritance/weight-mapping refactor and #45569 is a broader model split into NemotronH dense/sparse with docs and auto updates. They are related, not the same concrete change." + }, + { + "left": "pull_request:45569", + "right": "pull_request:46565", + "accept": false, + "reason": "#45569 is about NemotronH dense/sparse model architecture, while #46565 adds Nemotron 3.5 ASR streaming. Shared auto/docs files do not make them the same underlying change." + }, + { + "left": "pull_request:46332", + "right": "pull_request:46565", + "accept": false, + "reason": "Both are ASR-related, but they target different model versions and code paths: Nemotron ASR versus Nemotron 3.5 ASR. This is a family overlap, not a 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": 36, + "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": "bf8cb3894f8d056a3256f4c6ae658bf796efc2501728ab8aae4bd1faa6b63a76", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5466, + "estimated_input_tokens": 1367, + "estimated_eval_tokens": 2990 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5466, + "estimated_input_tokens": 1367, + "estimated_eval_tokens": 2990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related but distinct PRs: one introduces per-layer heterogeneous configuration, the other builds on that to add heterogeneous modeling/cache support. They are not duplicates and should not be merged as the same change.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45333 is the more foundational, self-contained change: it adds the per-layer configuration machinery that the modeling PR builds on.", + "best_issue_reason": null, + "best_pr_reason": "PR #45333 is the best representative of the cluster because it scopes a single coherent feature and provides the config layer needed by the broader modeling work.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Different layers of the feature: #45333 adds heterogeneous per-layer config, while #45332 adds model/cache/masking behavior that consumes that config. Related, but not the same underlying change and not something to merge as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 21, + "component_count": 36, + "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": "a34123231e7373f0c71addcb9c1e0fcb3da0eb84bd40ff0e2b878f99fcdc9581", + "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": "Both PRs are typing-related, but they are not the same change: #45415 is a broad type-checking sweep across many `src/transformers/*` files, while #45425 adds focused typing helpers/protocols in `_typing.py` and related modules. The overlap in shared files looks incidental, not like one concrete fix.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "#45415 is the broader umbrella typing PR, spanning many modules plus type-checking config, so it is the clearest representative of the cluster\u2019s theme.", + "best_issue_reason": null, + "best_pr_reason": "#45415 is the best representative PR because it is the larger, more central typing pass rather than a narrow helper-type cleanup.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Both are typing-related, but they address different scopes and code paths: a wide repo type-checking sweep versus targeted typing protocol/alias additions. Shared files alone are not enough to treat them as one mergeable PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 22, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45513-2", + "nodes": [ + "pull_request:45513", + "pull_request:46084" + ], + "soft_pairs": [ + "pull_request:45513|pull_request:46084" + ], + "prepared_review_unit_hash": "84e144de4093796624d772b2e4941e54e33f87e0ec7a4768f6e4248f66a79a4e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4858, + "estimated_input_tokens": 1215, + "estimated_eval_tokens": 2686 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4858, + "estimated_input_tokens": 1215, + "estimated_eval_tokens": 2686 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are not duplicates: one fixes multi-token cached forward in Qwen3.5/Qwen3-next/Olmo hybrid linear attention paths, while the other fixes Mamba2 cache handling when seq_len > 1. The underlying code paths and bugs are different.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "No issue artifacts are present. Among the two PRs, #46084 is the cleaner canonical representative because it has a single model focus, an explicit linked issue, and a narrowly scoped cache fix.", + "best_issue_reason": null, + "best_pr_reason": "#46084 is the best representative PR: it targets one concrete bug in Mamba2 use_cache behavior for seq_len > 1 and is tightly scoped to that fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45513", + "right": "pull_request:46084", + "accept": false, + "reason": "Reject: they address different model families and different failure modes. #45513 changes cached multi-token forward logic for Qwen3.5/Qwen3-next/Olmo hybrid linear attention, while #46084 fixes Mamba2 cache state handling when seq_len > 1. They do not look mergeable into one PR for the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 23, + "component_count": 36, + "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": "35b5004541f5c489e237098d8312aa7c42d9761e810f5c8763ba3f44c22d1265", + "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": "Both PRs are about expert parallelism, but they fix different bugs on different code paths: 45548 targets EP + DeepSpeed ZeRO-3 loading/config serialization, while 45662 targets EP + FSDP2 where experts can be overwritten by rank-0 broadcast. Similar area, not the same change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "45662 is the stronger representative of the cluster because it addresses the more concrete EP correctness bug across loading/sharding paths; 45548 is a narrower loading/serialization fix and not a duplicate.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Different underlying bugs and fixes: one is ZeRO-3/accelerate loading and config handling, the other is FSDP2 expert overwriting via broadcast. Shared EP context is too broad to treat as the same PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 36, + "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": "5432cedf76c6798b21c4de4449d0cd579720fbe78f0d56a27103ebeef210721a", + "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-related captured_info logging collision in the testing debug path. PR 45639 fixes the log writer to use worker-specific output, while PR 45645 applies the same core fix and also updates CI log collection to read multiple captured_info files, so they look like alternative/overlapping implementations of one bugfix.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45645 is the better canonical PR because it contains the core xdist-safe output fix plus the CI/workflow adjustment needed to consume the split log files, making it the more complete end-to-end change for this bug.", + "best_issue_reason": null, + "best_pr_reason": "PR 45645 is the strongest representative because it covers both the producer-side collision fix and the downstream log aggregation behavior; PR 45639 is a narrower subset of the same change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Same underlying xdist captured_info collision bug; both modify the same testing debug log path, and 45645 extends the same fix with workflow handling for the split files. They could 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": 25, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45670-3", + "nodes": [ + "pull_request:45670", + "pull_request:46320", + "pull_request:46401" + ], + "soft_pairs": [ + "pull_request:45670|pull_request:46320", + "pull_request:45670|pull_request:46401", + "pull_request:46320|pull_request:46401" + ], + "prepared_review_unit_hash": "58eb74f1dd5f00f7b784df1817c9f8568ab94a3d6e538b607397bf28f4d1e2a9", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7580, + "estimated_input_tokens": 1895, + "estimated_eval_tokens": 4046 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7580, + "estimated_input_tokens": 1895, + "estimated_eval_tokens": 4046 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster mixes one unrelated AutoModel mapping nit with two separate PRs that both add attn_implementation dispatch support, but for different model families. The RecurrentGemma and BLIP-2/InstructBLIP/InstructBLIPVideo PRs are thematically similar yet not the same underlying change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46401 is the best representative of the meaningful subgroup: it makes a concrete attn_implementation dispatch change across several closely related multimodal models and has the broader, more complete implementation surface. PR #46320 is similar in theme but targets a different model family, and #45670 is unrelated.", + "best_issue_reason": null, + "best_pr_reason": "PR #46401 is the strongest standalone PR in the set because it addresses a concrete code-path change, spans multiple affected models in one coherent change, and has substantial implementation/test coverage. It is more representative than the small AutoModel registration nit in #45670.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45670", + "right": "pull_request:46320", + "accept": false, + "reason": "These are different changes: one adds a GLMASR AutoModel registration entry, the other adds attn_implementation dispatch for RecurrentGemma. Same repo area only at a very high level, not the same bug or patch." + }, + { + "left": "pull_request:45670", + "right": "pull_request:46401", + "accept": false, + "reason": "#45670 is a model-mapping nudge for GLMASR, while #46401 implements attn_implementation dispatch for BLIP-2/InstructBLIP variants. They do not fix the same concrete code-path problem." + }, + { + "left": "pull_request:46320", + "right": "pull_request:46401", + "accept": false, + "reason": "Both are attn_implementation-related, but they target different model families and different model classes. This is shared feature work, not a duplicate or mergeable single fix for the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 26, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45720-2", + "nodes": [ + "pull_request:45720", + "pull_request:46259" + ], + "soft_pairs": [ + "pull_request:45720|pull_request:46259" + ], + "prepared_review_unit_hash": "ca26dbb2efe28df53d4c0eafe8fb21c6804340ecad70603550010173f78267f9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7248, + "estimated_input_tokens": 1812, + "estimated_eval_tokens": 3880 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7248, + "estimated_input_tokens": 1812, + "estimated_eval_tokens": 3880 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The pair is not a duplicate: PR 45720 implements TRF003 by adjusting forward return-tuple behavior across model forward methods, while PR 46259 implements TRF018 by refactoring model weight initialization. The overlap in filenames is incidental to broad model coverage, not the same underlying change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR for a duplicate cluster here; the two PRs target different lint rules and different code paths, so neither should stand in for the other.", + "best_issue_reason": null, + "best_pr_reason": "If one PR must be highlighted, 45720 is the more self-contained change: it addresses a single forward-path behavior across models. 46259 is a separate, broader initialization-rule refactor.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45720", + "right": "pull_request:46259", + "accept": false, + "reason": "Reject: TRF003 (forward return handling) and TRF018 (_init_weights refactor) are distinct changes affecting different behaviors, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 27, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45737-2", + "nodes": [ + "pull_request:45737", + "pull_request:45738" + ], + "soft_pairs": [ + "pull_request:45737|pull_request:45738" + ], + "prepared_review_unit_hash": "cc2ade73a49ec6a8883a7a24c433988645588bcd049e8aad2bce56b1b24daa1c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same MusicgenMelody cache initialization bug in `forward()`: replacing `EncoderDecoderCache(DynamicCache, DynamicCache)` with a single `DynamicCache`. PR #45737 is the more complete duplicate because it bundles the fix with a regression test, while #45738 is the fix-only variant.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45737 is the best canonical PR because it contains the same code fix as #45738 plus a regression test, making it the most complete representation of the bug fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #45737 is the strongest overall candidate: it fixes the concrete cache bug and adds a targeted regression test, whereas #45738 only changes the implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45737", + "right": "pull_request:45738", + "accept": true, + "reason": "Same underlying code-path bug in `musicgen_melody` cache setup; the diffs match on the core fix and the test-only addition in #45737 could plausibly live in the same PR as #45738." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: both PRs make the same `musicgen_melody` cache initialization change, and #45737 only adds a regression test on top of that fix. The soft-edge accept is conservative enough because the diffs are mergeable into one PR for the same bugfix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 28, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45932-2", + "nodes": [ + "pull_request:45932", + "pull_request:46100" + ], + "soft_pairs": [ + "pull_request:45932|pull_request:46100" + ], + "prepared_review_unit_hash": "acc6dee16e01f25991922f6264e2167a2f9ef3e63e3d44d45c1b2b65c3ff98c9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5083, + "estimated_input_tokens": 1271, + "estimated_eval_tokens": 2798 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5083, + "estimated_input_tokens": 1271, + "estimated_eval_tokens": 2798 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both items are unrelated open PRs that only look similar because they involve tests/infrastructure. PR 45932 is a CI/telemetry and model-code experiment, while PR 46100 is a ViT test migration; they should not be deduplicated.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 46100 is the cleaner, self-contained change focused on test migration. PR 45932 is an experimental 'DO NOT MERGE' PR that mixes CI secrets and unrelated model-file edits, so it is a poorer representative.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "Reject: these are different changes. 45932 touches CI/workflow secrets and multiple model implementations; 46100 adds/migrates ViT tests. They do not fix the same concrete code-path problem and would not plausibly be merged into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 29, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46073-3", + "nodes": [ + "pull_request:46073", + "pull_request:46154", + "pull_request:46196" + ], + "soft_pairs": [ + "pull_request:46073|pull_request:46154", + "pull_request:46073|pull_request:46196", + "pull_request:46154|pull_request:46196" + ], + "prepared_review_unit_hash": "a77679dd8f7deeecaf1791f6636b377248ae20b11389faa8574673ac810f9d44", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 8401, + "estimated_input_tokens": 2101, + "estimated_eval_tokens": 4458 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 8401, + "estimated_input_tokens": 2101, + "estimated_eval_tokens": 4458 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three PRs are about adding SDPA/attn_implementation dispatch, but they target different model families and code paths (CTRL, T5, wav2vec2-derived audio models). They are thematically similar, not the same underlying change, so they should not be clustered as duplicates.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single PR is a true canonical duplicate for the others; if one representative is needed, [T5] Support attn_implementation=sdpa dispatch is the cleanest and most self-contained example of the pattern.", + "best_issue_reason": null, + "best_pr_reason": "PR 46154 is the best representative because it is narrowly scoped to one model family and clearly implements the SDPA dispatch change without the broader multi-model scope of 46196.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46073", + "right": "pull_request:46154", + "accept": false, + "reason": "Different models and implementation targets (CTRL vs T5). Same feature theme, but not the same bug/change and not plausibly one merged PR." + }, + { + "left": "pull_request:46073", + "right": "pull_request:46196", + "accept": false, + "reason": "CTRL support vs wav2vec2/seamless audio models are separate code paths and separate model families; too broad to treat as duplicates." + }, + { + "left": "pull_request:46154", + "right": "pull_request:46196", + "accept": false, + "reason": "Both add SDPA dispatch, but T5 and wav2vec2/audio model stacks are unrelated enough that they are 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": 30, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46126-2", + "nodes": [ + "pull_request:46126", + "pull_request:46269" + ], + "soft_pairs": [ + "pull_request:46126|pull_request:46269" + ], + "prepared_review_unit_hash": "b05910510a2cfe418a40350d92a3bc77f07e16217919502bf7b9fb3451d0899b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5572, + "estimated_input_tokens": 1393, + "estimated_eval_tokens": 3042 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5572, + "estimated_input_tokens": 1393, + "estimated_eval_tokens": 3042 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related distributed-transformers PRs, but they do not look like the same underlying change: one is a focused EP/Trainer integration PR, the other is a broad distributed branch base touching many unrelated areas.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": "No issue in the cluster. As a PR representative, #46269 is the broader umbrella change and better captures the wider distributed refactor set in this cluster.", + "best_issue_reason": null, + "best_pr_reason": "#46269 is the best single representative because it is the larger distributed base PR and subsumes more of the shared infrastructure touched here, whereas #46126 is a narrower EP/Trainer feature integration.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46126", + "right": "pull_request:46269", + "accept": false, + "reason": "Shared distributed files are not enough: #46126 is specifically EP + Trainer integration, while #46269 is a broad distributed branch-base PR spanning many modules. They do not appear mergeable as one concrete code-path 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": 31, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46138-9", + "nodes": [ + "pull_request:46138", + "pull_request:46432", + "pull_request:46501", + "pull_request:46502", + "pull_request:46503", + "pull_request:46504", + "pull_request:46554", + "pull_request:46558", + "pull_request:46563" + ], + "soft_pairs": [ + "pull_request:46138|pull_request:46554", + "pull_request:46138|pull_request:46558", + "pull_request:46138|pull_request:46563", + "pull_request:46138|pull_request:46504", + "pull_request:46138|pull_request:46432", + "pull_request:46432|pull_request:46503", + "pull_request:46432|pull_request:46501", + "pull_request:46432|pull_request:46502", + "pull_request:46432|pull_request:46554", + "pull_request:46432|pull_request:46558", + "pull_request:46501|pull_request:46503", + "pull_request:46501|pull_request:46502", + "pull_request:46501|pull_request:46504", + "pull_request:46501|pull_request:46554", + "pull_request:46502|pull_request:46503", + "pull_request:46502|pull_request:46504", + "pull_request:46502|pull_request:46554", + "pull_request:46503|pull_request:46504", + "pull_request:46503|pull_request:46554", + "pull_request:46504|pull_request:46554", + "pull_request:46504|pull_request:46558", + "pull_request:46554|pull_request:46558", + "pull_request:46554|pull_request:46563", + "pull_request:46558|pull_request:46563", + "pull_request:46504|pull_request:46563", + "pull_request:46432|pull_request:46563" + ], + "prepared_review_unit_hash": "26bb94552e9796fbede1a2203edb3a5494b9bf6e8b8410a0c2188a2fba9c3d99", + "component_budget": { + "node_count": 9, + "item_count": 9, + "soft_pair_count": 26, + "serialized_chars": 23361, + "estimated_input_tokens": 5841, + "estimated_eval_tokens": 11938 + }, + "budget": { + "node_count": 9, + "item_count": 9, + "soft_pair_count": 26, + "serialized_chars": 23361, + "estimated_input_tokens": 5841, + "estimated_eval_tokens": 11938 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are all workflow-maintenance PRs, but they mostly touch different files or make different changes within the same workflow. The only near-cluster is the pr-ci-security-gate.yml trio, and those edits address different concerns, so there are no true duplicates here.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46554 is the most central representative of the only related subgroup: it makes the substantive security-gate behavior change in .github/workflows/pr-ci-security-gate.yml, while the other same-file PRs change different mechanics (permissions/approval flow and workflow reference notes).", + "best_issue_reason": null, + "best_pr_reason": "PR 46554 is the best global representative because it is the clearest, most substantive workflow security change and has the strongest overlap with the other same-file PRs, even though they are not duplicates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46138", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow files and different automation paths; only a generic similarity in CI-related maintenance." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46558", + "accept": false, + "reason": "Unrelated workflow file change; no shared concrete bug or code path." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46563", + "accept": false, + "reason": "Unrelated workflow file change; same broad area but different workflow and purpose." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflow file and different behavior; not the same underlying change." + }, + { + "left": "pull_request:46138", + "right": "pull_request:46432", + "accept": false, + "reason": "Different workflow files and distinct fixes; not a duplicate pair." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46503", + "accept": false, + "reason": "Different workflow files with separate changes; no same concrete bug/change." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46501", + "accept": false, + "reason": "Different workflow files and unrelated edits; only superficial similarity." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46502", + "accept": false, + "reason": "Different workflow files; these are separate workflow tweaks, not duplicates." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46554", + "accept": false, + "reason": "One updates ssh-runner, the other pr-ci-security-gate; different workflow behavior and no shared bug." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46558", + "accept": false, + "reason": "Different workflow files and distinct workflow logic; not the same change." + }, + { + "left": "pull_request:46432", + "right": "pull_request:46563", + "accept": false, + "reason": "Different workflow files; no evidence they fix the same issue." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46503", + "accept": false, + "reason": "Both are generic workflow updates, but they change different files and behaviors." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46502", + "accept": false, + "reason": "Different workflows with different fixes; not the same underlying problem." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflow files and unrelated CI behavior changes." + }, + { + "left": "pull_request:46501", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow files; no concrete overlap beyond general CI maintenance." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46503", + "accept": false, + "reason": "Different workflow files and separate behavior changes." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflow files; these are not the same fix." + }, + { + "left": "pull_request:46502", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow files and different concerns; not a duplicate." + }, + { + "left": "pull_request:46503", + "right": "pull_request:46504", + "accept": false, + "reason": "Different workflow files and unrelated CI logic changes." + }, + { + "left": "pull_request:46503", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow files; no shared bug or change target." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46554", + "accept": false, + "reason": "Different workflow files and different operational behavior." + }, + { + "left": "pull_request:46504", + "right": "pull_request:46558", + "accept": false, + "reason": "Different workflow files; only broad CI context overlaps." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46558", + "accept": false, + "reason": "Same workflow file, but the changes are about different concerns: security checks versus permissions/approval gating." + }, + { + "left": "pull_request:46554", + "right": "pull_request:46563", + "accept": false, + "reason": "Same workflow file, but one adds active security checks while the other swaps in a reusable workflow and comments; different underlying change." + }, + { + "left": "pull_request:46558", + "right": "pull_request:46563", + "accept": false, + "reason": "Same workflow file, but these edit different parts of the gate and do not fix the same concrete issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 32, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46152-2", + "nodes": [ + "pull_request:46152", + "pull_request:46207" + ], + "soft_pairs": [ + "pull_request:46152|pull_request:46207" + ], + "prepared_review_unit_hash": "c8cea507182c64838bb0ce43a734c0ad32433aa8b33ebe6204e345eca569eb06", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4275, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4275, + "estimated_input_tokens": 1069, + "estimated_eval_tokens": 2394 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are related at a theme level (both replace direct file-open source inspection), but they change different code paths for different reasons and should not be treated as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 46207 is the cleaner representative of the cluster\u2019s theme: it has a focused source-inspection refactor in one code path and includes tests, while PR 46152 is a separate package-resource traversal fix in import_utils.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46152", + "right": "pull_request:46207", + "accept": false, + "reason": "Reject: they address different underlying behaviors in different modules (_resolve_traversable vs _can_set_attn_implementation). The shared idea of avoiding open() is too broad to imply the same bug or mergeable change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 33, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46173-3", + "nodes": [ + "pull_request:46173", + "pull_request:46436", + "pull_request:46474" + ], + "soft_pairs": [ + "pull_request:46436|pull_request:46474", + "pull_request:46173|pull_request:46474" + ], + "prepared_review_unit_hash": "e2a420a32c3e8c964156e024e94d866058597753744bb3270ae39557a8fe7507", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5921, + "estimated_input_tokens": 1481, + "estimated_eval_tokens": 3218 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5921, + "estimated_input_tokens": 1481, + "estimated_eval_tokens": 3218 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs are the same StaticLayer type-hint change in cache_utils.py and should be clustered together; the generation cache max-len fix is a separate change in a different code path.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46474 is the cleanest representative of the StaticLayer.get_seq_length type-hint change: it has the same patch and target as #46173, but a simpler title and identical scope. PR #46436 is unrelated in implementation and bug surface, so it should not be the canonical PR for this cluster.", + "best_issue_reason": null, + "best_pr_reason": "#46474 is the best single PR for the duplicated StaticLayer annotation fix because it directly matches the code change in cache_utils.py and the same tracking issue. #46173 is effectively the same change; #46436 addresses a different static-cache length handling bug in generation/utils.py.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46436", + "right": "pull_request:46474", + "accept": false, + "reason": "Different fixes: #46436 changes static cache max length selection in generation/utils.py, while #46474 only updates the StaticLayer.get_seq_length type annotation in cache_utils.py. Same broad subsystem, but not the same underlying bug or mergeable into one PR." + }, + { + "left": "pull_request:46173", + "right": "pull_request:46474", + "accept": true, + "reason": "Same underlying change: both PRs update StaticLayer.get_seq_length return typing in src/transformers/cache_utils.py and target the same issue #45987. The diffs are effectively identical, so they are duplicates." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative: the packet supports that #46173 and #46474 are duplicate diffs on the same cache_utils.py type-hint change and same issue target, while #46436 is a separate generation/utils.py change with no shared files or target. The soft-edge decisions are appropriately cautious." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 34, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46293-2", + "nodes": [ + "pull_request:46293", + "pull_request:46478" + ], + "soft_pairs": [ + "pull_request:46293|pull_request:46478" + ], + "prepared_review_unit_hash": "7bd39ae48d659fdee338d62b65f52e0c5fa39368b02b874770cb97e7816d0b13", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4443, + "estimated_input_tokens": 1111, + "estimated_eval_tokens": 2478 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4443, + "estimated_input_tokens": 1111, + "estimated_eval_tokens": 2478 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs target the same hub-kernels area and share issue #46291, but they implement different fixes: one updates kernel repo/version metadata for kernels>=0.15.0, while the other pins LayerRepository revisions to main and adds broader config/test changes. They are related, but not the same duplicate change.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46478 is the better canonical PR because it is broader and more complete: it updates multiple hub-kernel references, touches kernel config, and includes test coverage, making it the stronger representative of the fix set.", + "best_issue_reason": null, + "best_pr_reason": "PR #46478 is the best overall PR in the cluster because it addresses the problem more comprehensively and is more suitable as the merge target for the shared hub-kernel compatibility work.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46293", + "right": "pull_request:46478", + "accept": false, + "reason": "Both mention the same issue and file, but the concrete changes differ: #46293 adjusts repo/version compatibility metadata, while #46478 pins revisions to main and extends config/tests. They are related fixes, not the same PR-worthy code change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 35, + "component_count": 36, + "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-46310-2", + "nodes": [ + "pull_request:46310", + "pull_request:46546" + ], + "soft_pairs": [ + "pull_request:46310|pull_request:46546" + ], + "prepared_review_unit_hash": "2e95f980edea99a22e3a77a3d0f8b8dfd14626ebc379ee5a88d84dcb3fb9d659", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4851, + "estimated_input_tokens": 1213, + "estimated_eval_tokens": 2682 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4851, + "estimated_input_tokens": 1213, + "estimated_eval_tokens": 2682 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs fix the same off-by-one bug in the Moonshine streaming sliding-window mask: the right window should include the boundary element (`<= right_window_size` instead of `<`). PR 46546 is the stronger canonical PR because it includes the Moonshine fix plus test coverage, while 46310 is the narrower earlier patch.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46546 is the better canonical PR because it contains the same Moonshine streaming boundary fix as 46310, plus test coverage and a broader related adjustment, making it the more complete change set.", + "best_issue_reason": null, + "best_pr_reason": "PR 46546 is the best overall PR candidate: it addresses the same concrete bug, is more complete, and adds validation. PR 46310 is essentially a narrower version of the same fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46310", + "right": "pull_request:46546", + "accept": true, + "reason": "Same underlying Moonshine streaming right-window boundary bug; both change the same mask logic from `< right_window_size` to `<= right_window_size`, so they are duplicate fixes for the same code path." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 36, + "component_count": 36, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46376-2", + "nodes": [ + "pull_request:46376", + "pull_request:46417" + ], + "soft_pairs": [ + "pull_request:46376|pull_request:46417" + ], + "prepared_review_unit_hash": "70e123cd3ad46e00f5ffcdb6e523b2bade9966e5bc87f351330ddaa9d14315ae", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5938, + "estimated_input_tokens": 1485, + "estimated_eval_tokens": 3226 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5938, + "estimated_input_tokens": 1485, + "estimated_eval_tokens": 3226 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two unrelated pull requests: one adds multimodal support to continuous batching infrastructure, the other adds the HunYuan VL model package, docs, and tests. They do not fix the same code-path problem and should not be deduplicated.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46417 is the better canonical representative because it is a concrete, self-contained model addition with docs and tests, whereas PR 46376 is an infrastructure change for continuous batching.", + "best_issue_reason": null, + "best_pr_reason": "PR 46417 is the strongest standalone PR in the cluster: it has the clearest deliverable, broadest implementation scope, and the most complete supporting test/documentation set.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46376", + "right": "pull_request:46417", + "accept": false, + "reason": "Different changes: continuous batching multimodal support vs. adding a specific HunYuan VL model. They touch different subsystems and are not the same underlying fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/manifest.json b/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..65aa4862abedf8bf4c867bc604c46fc96eb0b021 --- /dev/null +++ b/snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260612t000041z", + "artifacts": { + "hybrid": "snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-06-12T00:09:12Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260612T000041Z", + "variant": "hybrid" +} diff --git a/snapshots/20260612T000041Z/manifest.json b/snapshots/20260612T000041Z/manifest.json index 89297d123a53578397a73ee696a008e8d8ea3863..d288a214ece87fc8670ad009d5bb266816292ead 100644 --- a/snapshots/20260612T000041Z/manifest.json +++ b/snapshots/20260612T000041Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 308 }, "extracted_at": "2026-06-12T00:00:41Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260612t000041z", + "runs": { + "hybrid-model-20260612t000041z": { + "analysis_id": "hybrid-model-20260612t000041z", + "artifacts": { + "hybrid": "snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260612T000041Z/analysis-runs/hybrid-model-20260612t000041z/manifest.json", + "model": null, + "published_at": "2026-06-12T00:09:12Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260612T000041Z", "watermark": {