diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index b7d8f9c68988f6c172a600e3c6ad985be8439cf8..ffbf388b06b25fdb42fb08cbccbc98ced9b2832c 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": "20260527T000053Z", - "generated_at": "2026-05-27T00:05:37Z", + "snapshot_id": "20260527T060031Z", + "generated_at": "2026-05-27T06:09:38Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 1918dcc281022a18417b94f891bb2d1f196d94ef..2aa5d0b504b739b10fafe99cd26231b157e0cf23 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": "20260527T000053Z", - "generated_at": "2026-05-27T00:05:37Z", + "snapshot_id": "20260527T060031Z", + "generated_at": "2026-05-27T06:09:38Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -13,8 +13,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -138,418 +138,430 @@ "issue:43866|issue:44863", "issue:44292|issue:46015" ], - "prepared_review_unit_hash": "7fc1927efed8da07ce1eeaba26f2c3e797e909d72fec4f71c0e48257c2c83bc9", + "prepared_review_unit_hash": "f760006c58d5943870603243aef41d69b3184c6414fb83f73f29977276588052", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 69, - "serialized_chars": 48404, - "estimated_input_tokens": 12101, - "estimated_eval_tokens": 24458 + "serialized_chars": 48405, + "estimated_input_tokens": 12102, + "estimated_eval_tokens": 24460 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issues, with two small duplicate clusters: a DeepSpeed ZeRO-3 empty-tensor/index-0 failure cluster and an AutoTokenizer wrong-token-IDs regression cluster. Most other pairs are only superficially similar and should stay separate.", - "confidence": 0.69, - "canonical_issue_reason": "issue:43596 is the clearest representative of the ZeRO-3 empty-tensor / index-0 failure cluster: it is concise, model-agnostic, and matches the near-identical follow-up issue 43638.", + "summary": "This cluster is mostly a mix of separate bug families rather than one single duplicate set. The strongest duplicate family is the AutoTokenizer v5 regression affecting multiple model families (45812/45920), with a smaller duplicate pair around ZeRO-3 Bert loading (43596/43638) and one around Qwen3_5MoeTextRotaryEmbedding offload/materialization behavior (45859/45902).", + "confidence": 0.83, + "canonical_issue_reason": "issue:45920 is the best umbrella issue: it describes the broader AutoTokenizer token-id regression across multiple model families, while issue:45812 is the narrower Granite-only report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43596 is the best overall issue candidate because it describes the underlying failure mode most generally and cleanly, making it the strongest anchor for duplicate triage.", + "best_issue_reason": "issue:45920 is the strongest global representative because it is the broadest, most up-to-date report of the tokenizer regression and already consolidates several affected models.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Different bugs: one is a config attribute error, the other is a streaming chat/completions failure." + "reason": "Different failures: missing config attribute vs streaming chat/completions inference error." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Both touch GPU/MoE behavior, but the failures are different (AMD GPU support vs MPS histogram op missing)." + "reason": "Unrelated backend/hardware errors: AMD GPU failure vs MPS histogram not implemented for Int." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Both involve model/runtime compatibility, but one is image preprocessor dtype mismatch and the other is AMD GPU execution failure." + "reason": "Different model/code paths: image-preprocessor dtype mismatch vs GPT-OSS AMD GPU failure." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Different dtype/autocast bug versus GPU backend incompatibility; same broad area only." + "reason": "Both are runtime mismatches, but they affect different models and code paths." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Different model families and different failure modes; not the same concrete bug." + "reason": "Different problem surfaces: AMD GPU execution vs GraniteMoEHybrid calling an invalid method." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "Both are Qwen-3 related, but the reported runtime errors and code paths are different." + "reason": "Different model-specific runtime errors; no shared concrete code path." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "Both are model/runtime errors, but one is DeepSpeed collective mismatch and the other is ONNX export failure." + "reason": "DeepSpeed collective mismatch and ONNX export failure are unrelated." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "Different model/export issue versus fine-grained FP8 failure; no concrete shared bug." + "reason": "ONNX export error vs FineGrainedFP8Config failure; different model features and code paths." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "ONNX export and CUDA assert under auto device map are unrelated failures." + "reason": "Export-time failure vs device-mapping cuda assert; not the same bug." }, { "left": "issue:45901", "right": "issue:46015", "accept": false, - "reason": "Both are task/example related, but one is a task crash and the other is outdated RAG examples." + "reason": "Task crash vs docs/example update; no shared underlying bug." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are docs cleanup for v5 removals, but they reference different removed content and likely different edits." + "reason": "Both are docs regressions in v5, but they concern different removed features and likely separate edits." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export error and GenerationMixin import error are separate regressions." + "reason": "Model export issue and GenerationMixin import error are unrelated." }, { "left": "issue:44805", "right": "issue:45823", "accept": false, - "reason": "Both mention device/mask style errors, but the concrete errors and models differ." + "reason": "Mask/index shape mismatch and Gemma4 device mismatch are different failures." }, { "left": "issue:44918", "right": "issue:46165", "accept": false, - "reason": "Different training failures: embedding unpacking with TRL versus resume-training with trust-remote-code." + "reason": "TRL SFT embedding unpacking and resume/trust_remote_code are unrelated." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Same general IndexError family, but 44805 is a different mask-shape mismatch and not the same bug." + "reason": "Both are IndexError-ish, but the models and execution paths differ substantially." }, { - "left": "issue:43596", + "left": "issue:43638", "right": "issue:46165", "accept": false, - "reason": "ZeRO-3 empty-index failure and resume-training/trust-remote-code are unrelated." + "reason": "ZeRO-3 Bert loading and resume-training/trust_remote_code are unrelated." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Different underlying errors: empty parameter/index-0 versus mask/tensor shape mismatch." + "reason": "Same broad IndexError symptom, but not the same model/path or root cause." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Different bugs: mask shape mismatch versus tensor-parallel-only failure on GPT-OSS MoE." + "reason": "Mask/index tensor mismatch is unrelated to GPT-OSS tensor-parallel failure." }, { "left": "issue:44560", "right": "issue:46165", "accept": false, - "reason": "Different model/task failures with no shared code-path evidence." + "reason": "Video StopIteration in Qwen3-vl embedding is unrelated to resume-training behavior." }, { - "left": "issue:44805", - "right": "issue:44918", + "left": "issue:43596", + "right": "issue:46165", "accept": false, - "reason": "Both are tensor-shape related, but the failures occur in different parts of the stack and are not clearly the same bug." + "reason": "Different code paths and symptoms." }, { "left": "issue:43572", "right": "issue:46165", "accept": false, - "reason": "Config field regression versus training resume issue; unrelated." + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to model-resume issues." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "Processor preset bug and ONNX export bug are unrelated." + "reason": "Processor voice_preset bug is unrelated to ModernBERT ONNX export." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Both involve DeepSpeed, but one is collective mismatch and the other is an index-0 empty-tensor failure." + "reason": "Both involve DeepSpeed, but they are different failures on different models and code paths." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Same ZeRO-3 empty-parameter/index-0 failure during BertModel loading/training; only wording and model-pretraining context differ." + "reason": "Same ZeRO-3 Bert loading failure class: both report index-0-out-of-bounds during loading/init of Bert models, with only wording/model specificity differing." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Liger Kernel cross-entropy dispatch and GPT-OSS tensor-parallel failure are different problems." + "reason": "Liger Kernel cross-entropy invocation and GPT-OSS tensor-parallel failure are unrelated." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Tokenizer detokenization bug and LoRA merge/save mismatch are unrelated." + "reason": "Detokenization regression and LoRA merge/save mismatch are different bugs." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "Different model/runtime issues; no common underlying bug." + "reason": "Qwen-3-8B-NVFP4 runtime error does not match LoRA weight merge/save inconsistency." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Activation checkpointing/FSDP tensor recomputation issue is not the same as the mask shape mismatch." + "reason": "Activation-checkpointing/FSDP recomputation mismatch is not the same as the mask-shape IndexError." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "Both are DeepSpeed training errors, but the concrete failures are different (weight rank error vs collective mismatch)." + "reason": "Different DeepSpeed problems: weight-dimension error vs collective mismatch." }, { "left": "issue:45859", "right": "issue:45902", - "accept": false, - "reason": "Same model component, but one is CPU-offload incompatibility and the other is uninitialized-memory after materialization; not clearly one bug." + "accept": true, + "reason": "Same Qwen3_5MoeTextRotaryEmbedding offload/materialization bug family; both point to broken handling of rotary-embedding state across device moves/offload." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "DeepSpeed ZeRO-3 index-0 failure and GPT-OSS tensor-parallel failure are unrelated." + "reason": "Bert ZeRO-3 loading and GPT-OSS TP failure are unrelated." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "Different training-stack regressions: activation checkpointing/FSDP vs custom compute_loss signature." + "reason": "Activation checkpointing/FSDP issue is unrelated to the custom_trainer compute_loss signature change." }, { "left": "issue:43596", "right": "issue:45161", "accept": false, - "reason": "Different model-parallel failures with no shared concrete bug." + "reason": "Different models and failure modes." }, { "left": "issue:43716", "right": "issue:45507", "accept": false, - "reason": "Different model-specific failures: dtype mismatch in preprocessing vs invalid method call in GraniteMoEHybrid." + "reason": "Multimodal dtype mismatch and invalid-method call are unrelated." }, { "left": "issue:39290", "right": "issue:45081", "accept": false, - "reason": "Different compatibility bugs: vLLM config attribute missing vs Mistral tokenizer regex patch crash." + "reason": "Gemma3 sliding_window_pattern regression and Mistral tokenizer regex patch crash are unrelated." }, { "left": "issue:45823", "right": "issue:45874", "accept": false, - "reason": "Both are Gemma4 issues, but one is device-map mismatch and the other is expensive reverse-embedding lookup." + "reason": "Auto device-map forward mismatch and expensive reverse embedding lookup are different Gemma4 issues." }, { "left": "issue:43425", "right": "issue:44292", "accept": false, - "reason": "Torch version compatibility and Qwen running error are not the same bug." + "reason": "Torch version incompatibility and Qwen-3-8B-NVFP4 runtime error are unrelated." }, { "left": "issue:34928", "right": "issue:41720", "accept": false, - "reason": "FSDP/checkpointing recompute mismatch and CUDA assert on auto device mapping are unrelated." + "reason": "FSDP activation-checkpointing bug and Qwen3 auto device-map cuda assert are different issues." }, { "left": "issue:43828", "right": "issue:45161", "accept": false, - "reason": "Autocast dtype mismatch and tensor-parallel-only failure are different code paths." + "reason": "Phi autocast dtype mismatch and GPT-OSS tensor-parallel failure are unrelated." }, { "left": "issue:44918", "right": "issue:45161", "accept": false, - "reason": "Different bugs: TRL SFT embedding unpacking versus GPT-OSS TP failure." + "reason": "Qwen3.5 input-embedding unpacking with TRL is not the same bug as GPT-OSS TP failure." }, { "left": "issue:43828", "right": "issue:44292", "accept": false, - "reason": "Both are Qwen-related runtime issues, but the specific failures differ and do not look like one fix." + "reason": "Different models and failure modes." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Image-preprocessor dtype mismatch and MPS histogram int limitation are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Torch Dynamo grouped_mm tracing failure and Mistral image-preprocessor dtype mismatch are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:44918", + "accept": false, + "reason": "Different Qwen-related errors with different subsystems and root causes." }, { "left": "issue:45507", "right": "issue:45685", "accept": false, - "reason": "Different MoE implementation problems on different backends; not the same underlying defect." + "reason": "GraniteMoEHybrid invalid method and MPS histogram int error are unrelated." }, { "left": "issue:43039", "right": "issue:44918", "accept": false, - "reason": "Liger Kernel cross-entropy call and TRL SFT embedding unpacking are unrelated." + "reason": "Liger Kernel cross-entropy call and Qwen3.5 embedding unpacking are unrelated." }, { "left": "issue:41762", "right": "issue:43638", - "accept": true, - "reason": "Same DeepSpeed ZeRO-3 size-0/index-0 failure pattern; only the affected model differs." + "accept": false, + "reason": "Both are ZeRO-3 load/index errors, but Gemma3 loading and Bert loading are different code paths and likely separate bugs." }, { "left": "issue:43827", "right": "issue:44512", "accept": false, - "reason": "Both are documentation regressions, but they concern different removed APIs/commands and would need separate edits." + "reason": "Both are docs regressions, but they refer to different removed v5 features and separate documentation fixes." }, { "left": "issue:43716", "right": "issue:44560", "accept": false, - "reason": "Different model bugs: dtype mismatch in preprocessing versus StopIteration in video embedding." + "reason": "Mistral image dtype mismatch and Qwen3-vl video StopIteration are unrelated." }, { "left": "issue:43828", "right": "issue:45685", "accept": false, - "reason": "Autocast dtype mismatch and MPS histogram op missing are different failures." + "reason": "Autocast dtype mismatch and MPS histogram limitation are unrelated." }, { "left": "issue:44918", "right": "issue:45685", "accept": false, - "reason": "TRL embedding unpacking and MPS histogram issue are unrelated." + "reason": "Qwen3.5 embedding unpacking and MPS histogram limitation are unrelated." }, { "left": "issue:43039", "right": "issue:43828", "accept": false, - "reason": "Both involve tensor computations, but one is Liger Kernel cross-entropy dispatch and the other is autocast dtype mismatch." + "reason": "Different runtime issues on different model/code paths." }, { "left": "issue:30277", "right": "issue:36331", "accept": false, - "reason": "DeepSpeed collective mismatch and custom trainer signature breakage are unrelated." + "reason": "DeepSpeed collective mismatch and custom_trainer signature mismatch are unrelated." }, { "left": "issue:34928", "right": "issue:36010", "accept": false, - "reason": "Activation checkpointing/FSDP error and GenerationMixin import error are separate regressions." + "reason": "Activation-checkpointing/FSDP issue is unrelated to GenerationMixin import failure." }, { "left": "issue:45237", "right": "issue:46015", "accept": false, - "reason": "Model GPU compatibility bug and outdated RAG docs are unrelated." + "reason": "GPT-OSS AMD GPU failure and outdated RAG docs are unrelated." }, { "left": "issue:43039", - "right": "issue:44560", + "right": "issue:44805", "accept": false, - "reason": "Liger Kernel cross-entropy issue and video StopIteration are different bugs." + "reason": "Liger Kernel cross-entropy issue and mask/tensor shape IndexError are unrelated." }, { "left": "issue:41762", "right": "issue:43596", - "accept": true, - "reason": "Same ZeRO-3 empty-tensor/index-0 bug; different model names only." + "accept": false, + "reason": "Similar ZeRO-3 symptom, but Gemma3 loading and Bert loading are different model families and likely separate bugs." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Different issues: missing config attribute for vLLM vs CUDA assert in auto device mapping." + "reason": "Gemma3 config attribute regression with vLLM and Qwen3 auto device mapping cuda assert are unrelated." }, { "left": "issue:45237", "right": "issue:45901", "accept": false, - "reason": "Both affect Qwen-3 family areas, but the concrete crash modes are different." + "reason": "AMD GPU support issue and table-question-answering crash are unrelated." }, { "left": "issue:43901", "right": "issue:44512", "accept": false, - "reason": "Different docs drift: TextClassificationPipeline return_all_scores versus transformers run command removal." + "reason": "return_all_scores docs mismatch and transformers run-command docs removal are separate documentation issues." }, { "left": "issue:44841", "right": "issue:45823", "accept": false, - "reason": "Processor failure for Voxtral and Gemma4 device mismatch are unrelated." - }, - { - "left": "issue:43975", - "right": "issue:46015", - "accept": false, - "reason": "Tokenizer detokenization bug and outdated RAG examples are unrelated." - }, - { - "left": "issue:45812", - "right": "issue:45920", - "accept": true, - "reason": "Same AutoTokenizer wrong-token-IDs v5 regression across model families; 45920 is the broader manifestation of the same bug." + "reason": "Voxtral processor failure and Gemma4 device mismatch are unrelated." }, { "left": "issue:44863", "right": "issue:46015", "accept": false, - "reason": "NemotronH checkpoint loading failure and outdated examples are unrelated." + "reason": "NemotronH checkpoint loading and RAG docs are unrelated." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Corrupted checkpoint versus model-loading implementation problem are not the same bug." + "reason": "Checkpoint corruption in Ovis2 and NemotronH checkpoint loading are different problems." }, { "left": "issue:44292", "right": "issue:46015", "accept": false, - "reason": "Qwen-3 runtime error and outdated RAG docs are unrelated." + "reason": "Qwen NVFP4 runtime error and outdated RAG docs are unrelated." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same AutoTokenizer v5 regression; 45920 is the broader multi-model report and 45812 is the narrower Granite-only instance." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet: it identifies the two supported duplicate clusters (ZeRO-3 index-0/empty-tensor and AutoTokenizer wrong-token-IDs) and treats the rest as unrelated. The soft-edge verdicts are generally conservative, and the accepted pairs are backed by nearly identical titles/error patterns." + "feedback": "The summary is grounded in the packet and stays appropriately conservative. The accepted soft-edge pairs are plausible same-bug families based on the issue titles, and the rejected pairs mostly avoid over-claiming across clearly different model/code paths. The umbrella issue selection around 45920 vs 45812 is also well supported by the titles." }, "evaluator_used": true, "retried": false, @@ -565,8 +577,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -656,24 +668,24 @@ "issue:39186|issue:44918", "issue:43716|issue:44918", "issue:40990|issue:44960", - "issue:45632|issue:45698", "issue:43645|issue:45399", + "issue:45632|issue:45698", "issue:43404|issue:45005" ], - "prepared_review_unit_hash": "c80ba2c97484a5344cd5a5f483fd828a234dd5a2d74f8b16de77c9f2a35a981c", + "prepared_review_unit_hash": "4c6c5128649708929ffa9530bf0fcbd3547fe8a3b804df720a66e12fc16eb647", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 39, - "serialized_chars": 39131, + "serialized_chars": 39129, "estimated_input_tokens": 9783, "estimated_eval_tokens": 19822 }, @@ -682,255 +694,255 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous cluster with a few real duplicate pockets. The main correction is to reject 39290 vs 43531: they both mention sliding_window, but the model families and failure contexts differ too much. A couple of tighter matches remain, especially SigLIP2 and custom-module loading.", - "confidence": 0.69, - "canonical_issue_reason": "Issue 39290 is still the strongest single anchor among the near-duplicate pockets, but only as a local representative; the cluster overall is too diverse for a strong global canonical.", + "summary": "This is a heterogeneous issue cluster with mostly false-positive similarities. The only strong duplicate-looking pair is the remote-code custom-module loading/cache collision reports.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45698 is the best canonical issue because it states the broad user-facing failure (\u201cfrom_pretrained loads wrong custom module after save_pretrained\u201d) and appears to be the general symptom behind the more specific cache-path-collision report in issue:45632.", "canonical_pr_reason": null, - "best_issue_reason": "If one issue must stand in for the cluster, 39290 is the best representative because it is the clearest specific bug report with the strongest nearby duplicate signal, though the overall set is not one bug family.", + "best_issue_reason": "issue:45698 is the best overall issue to anchor the cluster: it is open, broadly phrased, and covers the central custom-module loading problem more directly than the narrower cache-path collision variant.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Same keyword, but Gemma3TextConfig/vLLM and Qwen3-MoE point to different model-specific failures; not enough evidence for one bug." + "reason": "Different models and different failures; one is a missing config attribute, the other is a sliding-window model issue." }, { "left": "issue:43792", "right": "issue:45901", "accept": false, - "reason": "Whisper runtime failure and table QA crash are unrelated code paths and symptoms." + "reason": "Unrelated models and subsystems; no shared bug path." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "Model-loading test failure vs corrupted checkpoint; different underlying problems." + "reason": "Both are model-loading problems, but the models and failure modes are different." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Both are index errors, but the models, shapes, and triggers differ." + "reason": "Both mention zero-3/indexing symptoms, but the specific errors and model paths differ too much to be the same bug." }, { "left": "issue:43425", "right": "issue:45070", "accept": false, - "reason": "Torch version incompatibility and PretrainedConfig/pydantic regression are separate issues." + "reason": "One is torch-version compatibility, the other is a PretrainedConfig/pydantic regression." }, { "left": "issue:43596", "right": "issue:45800", "accept": false, - "reason": "ZeRO-3/Bert index error is unrelated to torch/transformers version incompatibility." + "reason": "A DeepSpeed ZeRO-3 indexing crash is not the same as a torch/transformers version incompatibility." }, { "left": "issue:43975", "right": "issue:45901", "accept": false, - "reason": "Detokenization bug and table-question-answering crash are not the same bug." + "reason": "Different features and errors; detokenization vs table QA crash." }, { "left": "issue:43792", "right": "issue:45070", "accept": false, - "reason": "Whisper runtime failure is unrelated to the pydantic field regression." + "reason": "No shared code path; Whisper runtime failure and PretrainedConfig serialization break are unrelated." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Both involve Qwen3.5/chat, but the concrete failures differ too much to merge." + "reason": "Both involve Qwen3.5 and chat, but one is a transformers chat crash and the other is a streaming API failure." }, { "left": "issue:44292", "right": "issue:45901", "accept": false, - "reason": "NVFP4 runtime failure is unrelated to the table QA crash." + "reason": "Completely different model/task and failure mode." }, { "left": "issue:39290", "right": "issue:45823", "accept": false, - "reason": "Different model families and symptoms: sliding_window/vLLM versus Gemma4 device-mapping." + "reason": "Different models and distinct runtime issues." }, { "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Streaming chat failure and MPS Int histogram error are unrelated." + "reason": "Chat streaming failure and MPS integer histogram error are unrelated." }, { "left": "issue:43404", "right": "issue:43866", "accept": false, - "reason": "Untied weights in Mistral3 and a corrupted checkpoint are different causes." + "reason": "Unrelated models; tied-weights bug vs corrupted checkpoint report." }, { "left": "issue:43541", "right": "issue:45685", "accept": false, - "reason": "Grouped_mm tracing error and MPS histogram error are distinct failure paths." + "reason": "Different hardware/code paths and error types." }, { "left": "issue:43541", "right": "issue:43828", "accept": false, - "reason": "Different models and different runtime errors; not the same bug." + "reason": "Both are numeric/runtime errors, but the model, backend, and exact failure are different." }, { "left": "issue:44841", "right": "issue:45084", "accept": false, - "reason": "Processor failure and template compilation error are unrelated." + "reason": "Processor failure for one model is unrelated to a template-node compile error." }, { "left": "issue:36010", "right": "issue:39290", "accept": false, - "reason": "GenerationMixin import error is unrelated to the sliding_window bug." + "reason": "ImportError for GenerationMixin and Gemma3 config attribute errors are unrelated." }, { "left": "issue:45800", "right": "issue:45823", "accept": false, - "reason": "Torch/transformers incompatibility and Gemma4 PLE device mismatch are not the same issue." + "reason": "Torch version incompatibility and Gemma4 device mismatch are separate problems." }, { "left": "issue:41093", "right": "issue:41762", "accept": false, - "reason": "Both are index errors around loading, but the model paths and exact failures differ." + "reason": "Both are ZeRO-3 related index crashes, but the symptoms and triggering paths differ." }, { "left": "issue:43572", "right": "issue:45070", "accept": false, - "reason": "StableLm pad_token_idx regression is unrelated to the PretrainedConfig/pydantic issue." + "reason": "Both are v5 regressions, but one is a missing config field and the other is a pydantic model break." }, { "left": "issue:43296", "right": "issue:43366", "accept": false, - "reason": "PaddleOCR-VL loading and GGUF gpt-oss support are different problems." + "reason": "Different model families and different loading/support problems." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Torch dynamo grouped_mm tracing failure and TP-not-working are not the same underlying bug." + "reason": "Grouped_mm tracing failure is not the same as GPT-OSS tensor-parallel support." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Int4 OOM regression and tied-weights translation bug are unrelated." + "reason": "Int4 OOM regression and tied-weights issue are different bug classes." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Autocast dtype mismatch and chat streaming failure are different code paths." + "reason": "Autocast dtype mismatch and streaming chat API failure are unrelated." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "init_empty_weights argument handling is unrelated to the pydantic field regression." + "reason": "Unexpected init_empty_weights argument and PretrainedConfig field regression are distinct." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "GGUF gpt-oss support and Qwen3.5 streaming inference are unrelated." + "reason": "GGUF GPT-OSS support and Qwen3.5 chat streaming are unrelated." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "Remote_code breakage and torch/transformers incompatibility are not the same bug." + "reason": "Remote-code breakage and torch/transformers version incompatibility are different failure modes." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Qwen3-MoE sliding_window and PretrainedConfig/pydantic breakage are unrelated." + "reason": "Sliding-window model bug and config field regression do not match." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Grouped_mm tracing failure and Float8 storage lookup error are different issues." + "reason": "Grouped_mm tracing error and missing Float8 storage object are unrelated." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Both are index errors, but the affected scenarios and tensors differ." + "reason": "Same broad ZeRO-3 area, but the concrete bugs and triggers differ." }, { "left": "issue:43054", "right": "issue:43994", - "accept": true, - "reason": "Both concern SigLIP2 producing bad outputs on the same model family; this looks like the same underlying issue." + "accept": false, + "reason": "Both mention siglip2, but one is about embedding quality while the other is a loading/pipeline correctness issue." }, { "left": "issue:45759", "right": "issue:45902", "accept": false, - "reason": "Composite config unwrapping and uninitialized rotary memory are unrelated." + "reason": "One is config unwrapping for multimodal CausalLM instantiation; the other is a Qwen3.5 MoE embedding initialization memory bug." }, { "left": "issue:45542", "right": "issue:45800", "accept": false, - "reason": "TensorBoard-without-TF backend error is unrelated to torch/transformers incompatibility." + "reason": "TensorBoard/TensorFlow backend handling and torch version incompatibility are unrelated." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP weight-shape error and Qwen3.5 embedding unpacking are different bugs." + "reason": "FSDP weight-shape error and embedding unpacking failure in SFT trainer are different bugs." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Mistral dtype mismatch and Qwen3.5 SFT unpacking are unrelated." + "reason": "Mistral image-preprocessor dtype mismatch is unrelated to Qwen3.5 embedding unpacking." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "High perplexity on gpt-oss-20b and GLM5 are different reports with no shared failure mode." - }, - { - "left": "issue:45632", - "right": "issue:45698", - "accept": true, - "reason": "Both point to wrong custom-module resolution after save/load; same dynamic-module path bug." + "reason": "Perplexity quality issue and GLM5 issue are not the same underlying bug." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Jupyter custom-model initialization and flash-attn fallback blocking are unrelated." + "reason": "Custom-model notebook initialization and flash-attn fallback gating are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": true, + "reason": "Both point to the same remote-code/custom-module loading path, with a cache-path collision plausibly causing the wrong module to be loaded after save/load." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Mistral3 tied-weight bug and translation-model tied-weight regressions are different model families and fixes." + "reason": "Both mention tied weights, but they target different models and different failure contexts." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The cluster-level summary is mostly grounded, but one soft-edge acceptance is too strong: issue:39290 vs issue:43531. They share a sliding_window keyword, but the packet shows different model families and different failure contexts (Gemma3TextConfig/vLLM vs Qwen3-MoE), so calling them the same bug overstates the evidence. For a conservative review, that pair should be marked false or at most tentative. The other accepted pairs are more plausible, but this single overconfident accept is enough to reject." + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative. The main linkage between issue:45632 and issue:45698 is well supported by the titles, and the other soft-edge rejections are appropriately cautious rather than overstated." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -943,8 +955,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -1033,20 +1045,20 @@ "issue:43881|issue:43883", "issue:45685|issue:45823" ], - "prepared_review_unit_hash": "9ae4db3ea942906e931f9fda13dc770f42ce24e3cf87a12b852c7d2c71ca4b7e", + "prepared_review_unit_hash": "273c8b3e27c67ba23d635700ec96046b0b8a4b1bfebecda49cf6039548b81820", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37578, + "serialized_chars": 37579, "estimated_input_tokens": 9395, "estimated_eval_tokens": 19046 }, @@ -1055,222 +1067,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issue set, with a few small duplicate clusters around non-persistent buffer corruption, remote_code regressions, and SigLIP2 discrepancy/quality problems. No PRs are present.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43950 is the clearest broad regression report in the cluster: it describes the underlying non-persistent-buffer corruption bug directly, and it has the strongest duplicate-style overlap with issue:44534 (plus likely the related 45902 symptom).", + "summary": "The set is mostly heterogeneous, with one clear duplicate-looking pair around the Transformers v5 non-persistent buffer corruption regression. I accepted that edge; the rest look like related-but-distinct bugs or entirely different issues.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 43950 is the clearest canonical issue: it names the regression, the code path (`from_pretrained()`), and the underlying non-persistent buffer corruption. Its twin 44534 is a shorter rephrasing of the same bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the best global anchor because it states a concrete Transformers 5.x regression with a clear code-path focus and likely absorbs the closest duplicate reports in this set.", + "best_issue_reason": "43950 is the best issue to keep as the representative bug report because it is the most specific and actionable description of the duplicate pair, with the regression and mechanism stated explicitly.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "Different problems: Qwen3Moe FP8 config failure vs flash-attn2 fallback being blocked by checks." + "reason": "Both are loading/compatibility failures, but one is FineGrainedFP8/Qwen3MoE and the other is flash-attn fallback gating; different code paths and symptoms." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "Different code paths: Qwen3.5 PagedAttentionCache crash vs tied-weights handling on translation models." + "reason": "Qwen3.5 PagedAttentionCache crash vs tied-weights regression on translation models; not the same bug." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization/offloading behavior is unrelated to tied-weights regressions." + "reason": "Quantization/offloading behavior is a different failure mode from tied-weights handling." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "Checkpoint-loading failure for NemotronH is not the same as tied-weights handling." + "reason": "NemotronH checkpoint loading failure is unrelated to the translation tied-weights issue." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "Unrelated: DSA indexer ReLU discussion vs a docs typo about a removed command." + "reason": "DSA indexer implementation concern vs a docs typo; unrelated." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "Different failures in different models/subsystems; no shared concrete bug." + "reason": "Different models and different failures: embedding unpacking with TRL SFT vs invalid method call in GraniteMoEHybrid." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Different symptoms: detokenization bug vs Qwen-3 NVFP4 runtime error." + "reason": "Detokenization bug in DeepSeek Coder is not the same as Qwen-3 NVFP4 runtime error." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "A cache-strategy typo is unrelated to old docs mentioning the removed run command." + "reason": "Cache-strategy typo in docs is unrelated to removed run-command docs drift." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "No common underlying bug; they affect different models and failure modes." + "reason": "Different model families and distinct runtime failures." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Both describe the same Transformers 5.x non-persistent-buffer corruption regression." + "reason": "Near-identical regression reports about Transformers v5 corrupting/filling non-persistent buffers with junk; same underlying bug." }, { "left": "issue:43644", "right": "issue:43950", "accept": false, - "reason": "AutoConfig remote_code field loss is a different bug than buffer corruption." + "reason": "Both mention v5 regressions, but one is about `AutoConfig` missing fields and the other about buffer corruption; different mechanisms." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "Both mention SigLIP2, but one is a docs/example issue and the other is a model quality discrepancy." + "reason": "SigLIP2 docs/example errors are different from a broad text-embedding quality discrepancy." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Different training/runtime issues: autocast dtype mismatch vs Qwen3.5 RLHF gradient explosion." + "reason": "Both involve Qwen3.5/typing issues, but one is autocast dtype mismatch and the other is RLHF gradient explosion from a different position_ids/SDPA path." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "Trainer API mismatch and ZeRO-3 load failure are unrelated." + "reason": "CustomTrainer kwarg mismatch and Gemma3 ZeRO-3 loading IndexError are unrelated." }, { "left": "issue:41628", "right": "issue:45800", "accept": false, - "reason": "Import/export API issue is unrelated to torch/transformers version incompatibility." + "reason": "AutoImageProcessor import failure is not the same as torch/transformers version incompatibility." }, { "left": "issue:43950", "right": "issue:45902", - "accept": true, - "reason": "Same non-persistent-buffer materialization bug; 45902 is a model-specific manifestation after meta\u2192to_empty(cuda)." + "accept": false, + "reason": "Related buffer/materialization theme, but 45902 is a model-specific uninitialized-memory bug rather than the same general from_pretrained corruption report." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Different low-level compatibility problems: bitsandbytes kwarg mismatch vs missing Float8 storage." + "reason": "bitsandbytes Int8Params kwarg incompatibility is unrelated to missing Float8 storage." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Corrupted checkpoint vs noisy loading behavior are not the same bug." + "reason": "Corrupted checkpoint vs loading noise; no shared bug." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality and tokenization changes are different problems." + "reason": "Different reports about embedding quality vs tokenization changes; not the same concrete defect." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad AutoTokenizer error messaging vs missing all_tied_weights_keys attribute are unrelated." + "reason": "Bad AutoTokenizer error message for Voxtral vs missing `all_tied_weights_keys` on Molmo; different issues." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Both concern SigLIP2 outputs, but the observed failures are too different to treat as the same bug." + "reason": "SigLIP2 zero-probabilities vs text-embedding quality regression are separate concerns." }, { "left": "issue:45663", "right": "issue:45902", "accept": false, - "reason": "Gemma4 FSDP2 shared_kv_states issue is unrelated to non-persistent buffer initialization." + "reason": "Gemma4 FSDP2 shared_kv_states bug is unrelated to Qwen3_5MoE rotary embedding memory initialization." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Flash-attn packed-sequence misinterpretation is different from CPU-offload incompatibility in rotary embedding." + "reason": "Both touch Qwen3.5, but one is Flash Attention illegal access from 3D position_ids and the other is CPU offload incompatibility." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Different models and different failure layers; no shared root cause." + "reason": "Different models and failures: Voxtral tokenizer loading vs glm-4v loading failure." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Gradient checkpointing inefficiency and int4 memory blowup are different optimization issues." + "reason": "Gradient checkpointing/LoRA behavior is unrelated to int4 quantization memory growth." }, { "left": "issue:43054", "right": "issue:43493", - "accept": true, - "reason": "Both point to the same SigLIP2 implementation discrepancy/quality regression versus the expected/original behavior." + "accept": false, + "reason": "Both concern SigLIP2, but one is quality regression and the other is HF-vs-JAX implementation discrepancy; not the same exact bug." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "General loading noise is not the same as failing to load a specific model." + "reason": "Loading noise is not the same as a model failing to load ScandiBERT." }, { "left": "issue:43975", "right": "issue:45237", "accept": false, - "reason": "Detokenization bug and AMD GPU compatibility failure are unrelated." + "reason": "Detokenization bug is unrelated to AMD GPU loading failure." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Noisy loading messages are unrelated to the specific model-load failure." + "reason": "Two separate loading-related issues with different models and symptoms." }, { "left": "issue:43643", "right": "issue:45020", - "accept": true, - "reason": "Same remote_code regression family; 43643 is a concrete AutoConfig missing-fields symptom of the broader breakage." + "accept": false, + "reason": "Remote-code loading problems are broadly related, but one is missing fields from `AutoConfig` and the other is a wider compatibility complaint." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "LoRA checkpointing behavior and Qwen3 MoE memory usage are different issues." + "reason": "LoRA gradient checkpointing vs MoE training memory inefficiency are not the same underlying defect." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "ZeRO-3 load indexing error and FP8 config failure are unrelated." + "reason": "Gemma3 ZeRO-3 loading IndexError and Qwen3MoE FP8 failure are distinct model/runtime issues." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "GLM-4V loading failure is not the same as the Molmo tied-weights attribute bug." + "reason": "Two separate model-loading failures with different root causes." }, { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "MPS int histogram support and Gemma4 device mismatch are different runtime problems." + "reason": "MPS histogram kernel limitation is unrelated to Gemma4 device-mismatch during forward." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays appropriately conservative. The accepted soft-edge pairs are supported by the issue titles and the provided soft-pair evidence, and the rejections do not overclaim shared root causes. No PRs are present, so the anchor choice is reasonable." + "feedback": "The summary is grounded in the packet, and the lone accepted edge (43950/44534) is well supported by nearly identical titles and the shared non-persistent buffer regression theme. The rejected edges are stated conservatively and mostly distinguish different models, mechanisms, or symptom classes without overstating equivalence." }, "evaluator_used": true, "retried": false, @@ -1286,8 +1298,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -1300,7 +1312,6 @@ "issue:42915", "issue:42947", "issue:43257", - "issue:43366", "issue:43404", "issue:43638", "issue:43742", @@ -1332,11 +1343,13 @@ "issue:45081", "issue:45305", "issue:45313", + "issue:45325", "issue:45356", "issue:45399", "issue:45538", "issue:45588", "issue:45823", + "issue:45902", "issue:45923" ], "soft_pairs": [ @@ -1355,12 +1368,12 @@ "issue:42915|issue:45020", "issue:43827|issue:44509", "issue:43872|issue:44291", - "issue:43257|issue:43866", "issue:43792|issue:44220", + "issue:43257|issue:43866", "issue:42915|issue:43872", "issue:41553|issue:44991", - "issue:36331|issue:41093", "issue:43257|issue:45305", + "issue:36331|issue:41093", "issue:42947|issue:45923", "issue:43940|issue:45313", "issue:43854|issue:44863", @@ -1371,238 +1384,235 @@ "issue:44479|issue:44560", "issue:44530|issue:44863", "issue:24643|issue:30064", - "issue:43366|issue:43828" + "issue:45325|issue:45902" ], - "prepared_review_unit_hash": "fd3079dccae342689c5c53338d250c89a63f0ea14141903d017d30e301516ec7", + "prepared_review_unit_hash": "8893ceb9917a0398648acd1d8d88ce6d23ad87cc1a3c57624fb0abecb44a0ad3", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37048, - "estimated_input_tokens": 9262, - "estimated_eval_tokens": 18780 + "serialized_chars": 37733, + "estimated_input_tokens": 9434, + "estimated_eval_tokens": 19124 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: only a few pairs look like true duplicates, mainly doc typos/regressions and a couple of same-error code-path bugs. Most soft-similar pairs share only broad subsystem/theme overlap, not the same underlying issue.", - "confidence": 0.94, - "canonical_issue_reason": null, + "summary": "This cluster is heterogeneous: most items are distinct Transformers regressions around model loading, tokenizers, DeepSpeed, PEFT, quantization, or docs. None of the soft pairs look like the exact same underlying bug/change, so all should stay separate.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45020 is the broadest and most representative umbrella: it covers recent-version loading regressions in a generic way and can anchor nearby model/tokenizer breakages without being too specific.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45020 is the best global issue representative because it is broad, user-facing, and closest to the common theme of version-induced loading failures across models.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Different models and failures: Qwen3VL from_pretrained weight_only error vs a generic GLM5 model-loading issue." + "reason": "Different problems: Qwen3VL weight_only loading error vs GLM5 model loading; no evidence they share the same failing code path." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer-loading regression vs flash-attn fallback gating; unrelated code paths." + "reason": "Tokenizer loading failure for a specific model vs flash-attn fallback gating; unrelated subsystems and fixes." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "ImportError for Qwen2.5-VL class name vs bitsandbytes Int8Params kwarg incompatibility; not the same bug." + "reason": "Import error for a missing class vs bitsandbytes Int8Params constructor incompatibility; different failure modes." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "Different LoRA/tie_word_embeddings warning issue, not the same HF4.x to HF5.x Qwen3 MoE loading failure." + "reason": "Both mention LoRA/Qwen3.5 warnings, but one is a hf4.x vs hf5.x model incompatibility and the other is a tie_word_embeddings warning; not the same bug." }, { "left": "issue:45020", "right": "issue:45823", "accept": false, - "reason": "Remote_code regressions broadly vs a Gemma4 device_map forward mismatch; no shared concrete bug." + "reason": "Broad remote_code breakage vs a Gemma4 device_map forward mismatch; too different in trigger and code path." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Model-like addition failing in tokenizer mapping vs remote_code regressions; overlapping theme only." + "reason": "Tokenizer mapping-name edge case vs generic remote_code regression; related area, but not the same defect." }, { "left": "issue:43828", "right": "issue:45823", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE vs Gemma4 PLE device mismatch; distinct failures." + "reason": "Autocast dtype mismatch in Phi-tiny-MoE vs Gemma4 device mismatch; distinct runtime issues." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Mistral regex patch crash vs flash-attn fallback blocking; unrelated bugs." + "reason": "Mistral regex patch crashing in tokenizer loading vs flash-attn2 fallback selection; unrelated." }, { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "MobileLLM key error while loading vs flash-attn install/fallback checks; different code paths." + "reason": "Key error loading a specific model vs flash-attention fallback logic; no shared underlying bug." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "General remote_code breakage vs Kimi-K2.5 tokenizer codec regression; not the same issue." + "reason": "General remote_code loading regressions vs Kimi-K2.5 tokenizer codec/fix_mistral_regex warning regression; different components." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "lm_head tying bug vs non-persistent buffer corruption on from_pretrained; separate loading semantics." + "reason": "Untied lm_head weights in Mistral3 vs non-persistent buffer corruption in from_pretrained; separate state-management bugs." }, { "left": "issue:43638", "right": "issue:45823", "accept": false, - "reason": "Deepspeed ZeRO-3 index error on non-pretrained Bert vs Gemma4 device mismatch; unrelated." + "reason": "DeepSpeed zero3 index error for non-pretrained Bert vs Gemma4 PLE device mismatch; unrelated training/forward paths." }, { "left": "issue:42915", "right": "issue:45020", "accept": false, - "reason": "FineGrainedFP8 Qwen3 MoE failure vs remote_code regression broadly; too different." + "reason": "FineGrainedFP8Config failure for Qwen3 MoE vs broad remote_code regressions; different precision/configuration issue." }, { "left": "issue:43827", "right": "issue:44509", - "accept": true, - "reason": "Same doc regression: both note removed v5 pipeline tasks (summarization/translation/text2text) still referenced in docs." + "accept": false, + "reason": "These are the same documentation topic, but the titles indicate separate reports; without more evidence, don't merge." }, { "left": "issue:43872", "right": "issue:44291", - "accept": true, - "reason": "Same underlying `_is_hf_initialized` / `Int8Params.__new__` incompatibility when loading models with init_empty_weights or bitsandbytes." - }, - { - "left": "issue:43257", - "right": "issue:43866", "accept": false, - "reason": "Qwen3 MoE weight conversion with accelerate+deepspeed vs corrupted Ovis2 checkpoint; unrelated model issues." + "reason": "bitsandbytes Int8Params error vs init_empty_weights/_is_hf_initialized error; similar weight-init area but not the same bug." }, { "left": "issue:43792", "right": "issue:44220", - "accept": true, - "reason": "Both point to the Whisper audio feature-extraction path failing; `_torch_extract_fbank_features()` is the likely root cause of 'can't run'." + "accept": false, + "reason": "Whisper model runtime failure vs _torch_extract_fbank_features issue; likely different audio pipeline failures." + }, + { + "left": "issue:43257", + "right": "issue:43866", + "accept": false, + "reason": "Qwen3 MOE weights not converted under accelerate+deepspeed vs corrupted Ovis2 checkpoint; different models and causes." }, { "left": "issue:42915", "right": "issue:43872", "accept": false, - "reason": "Qwen3 MoE FP8 failure vs bitsandbytes/init_empty_weights kwarg issue; different subsystems." + "reason": "Qwen3 MoE FP8 config failure vs bitsandbytes constructor incompatibility; different layers of the stack." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Bad AutoTokenizer error message for Voxtral vs tokenizer loading failure for est-roberta; not the same bug." + "reason": "Bad AutoTokenizer error message for Voxtral vs tokenizer loading regression for EMBEDDIA/est-roberta; same broad area, not same bug." }, { - "left": "issue:36331", - "right": "issue:41093", + "left": "issue:43257", + "right": "issue:45305", "accept": false, - "reason": "Trainer compute_loss signature change vs mask/tensor shape mismatch; unrelated errors." + "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed vs gradient averaging with GAS and DeepSpeed; training/loading issues are different." }, { - "left": "issue:43257", - "right": "issue:45305", + "left": "issue:36331", + "right": "issue:41093", "accept": false, - "reason": "DeepSpeed weight-conversion failure vs gradient averaging with GAS/model_accepts_loss_kwargs; different training bugs." + "reason": "CustomTrainer signature mismatch vs mask/tensor shape IndexError; unrelated trainer API vs tensor indexing bug." }, { "left": "issue:42947", "right": "issue:45923", "accept": false, - "reason": "PEFT LoRA gradient checkpointing ineffective vs missing supports_gradient_checkpointing flag on a remote-code variant." + "reason": "Gradient checkpointing ineffective with PEFT LoRA vs missing supports_gradient_checkpointing flag on a remote-code model; not the same change." }, { "left": "issue:43940", "right": "issue:45313", "accept": false, - "reason": "Both are ZeRO-3 weight-loading failures, but on different Qwen families and with different symptoms; not clearly the same defect." + "reason": "Qwen3-Next ZeRO-3 missing weights vs Qwen3.5 ZeRO-3 missing language_model weights; similar symptom but different models and likely different loader logic." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "GLM-4.7-Flash test loading failure vs NemotronH checkpoint loading failure; distinct models and failure modes." + "reason": "Unable to load GLM-4.7-Flash in tests vs NemotronH checkpoints not loading; distinct model implementations." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "Qwen3.5 RLHF NaN/SDPA fallback issue vs flash_attention.py s_aux=None crash; unrelated." + "reason": "RLHF gradient explosion from 3D position_ids and SDPA fallback vs flash_attention crash on s_aux=None; unrelated attention-path bugs." }, { "left": "issue:43872", "right": "issue:45538", "accept": false, - "reason": "bitsandbytes kwarg incompatibility vs CLIPTokenizer max_length sentinel value; no shared root cause." + "reason": "bitsandbytes incompatibility vs CLIPTokenizer model_max_length sentinel value; different subsystems." }, { "left": "issue:41093", "right": "issue:44805", - "accept": true, - "reason": "Same mask/index shape mismatch error pattern; the later report is the same underlying IndexError with different tensor sizes." + "accept": false, + "reason": "Both are mask/tensor shape errors, but they are separate reports with no evidence of the same root cause." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Different tokenizer-loading failures on different models; one is an error-message issue, the other a load failure." + "reason": "Bad error message for Voxtral vs inability to load cjvt/sleng-bert; different tokenizer/model-loading problems." }, { "left": "issue:44479", "right": "issue:44560", - "accept": true, - "reason": "Both describe the same Qwen video-input regression in v5.3.0, with StopIteration being a concrete symptom of the broader video pipeline breakage." + "accept": false, + "reason": "Both are Qwen3.5/VL regressions, but one is a video-input regression and the other is a StopIteration in embedding video handling; not clearly the same bug." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "PagedAttentionCache linear_attention crash vs NemotronH checkpoint loading failure; separate bugs." + "reason": "PagedAttentionCache linear_attention failure in Qwen3.5 vs NemotronH checkpoint loading; different model families and code paths." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "DeepSpeed training weight dimension error vs void segmentation map processing failure; unrelated." + "reason": "DeepSpeed training weight-dimension error vs image processor void segmentation-map failure; unrelated." }, { - "left": "issue:43366", - "right": "issue:43828", + "left": "issue:45325", + "right": "issue:45902", "accept": false, - "reason": "GGUF gpt-oss support request vs Phi-tiny-MoE autocast dtype mismatch; no common code-path bug." + "reason": "Qwen2.5-VL rope index temporal scaling bug vs uninitialized inv_freq after meta->to_empty(cuda); both model internals, but clearly different defects." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative: it describes the cluster as mostly heterogeneous while calling out only a few plausible duplicate pairs. The accepted soft-edge verdicts are defensible from the issue titles/body context, and the rejected ones avoid overclaiming shared root causes." - }, - "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 }, @@ -1614,8 +1624,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -1636,11 +1646,14 @@ "issue:43644", "issue:43645", "issue:43646", + "issue:43708", "issue:43720", "issue:43824", + "issue:43828", "issue:43950", "issue:43957", "issue:43994", + "issue:44075", "issue:44155", "issue:44451", "issue:44488", @@ -1658,25 +1671,23 @@ "issue:45030", "issue:45276", "issue:45313", - "issue:45325", "issue:45357", "issue:45584", "issue:45663", "issue:45685", "issue:45823", - "issue:45834", - "issue:45902" + "issue:45834" ], "soft_pairs": [ - "issue:45325|issue:45902", + "issue:43366|issue:43828", "issue:45663|issue:45823", "issue:43366|issue:43531", "issue:44155|issue:45584", "issue:43950|issue:45834", "issue:44560|issue:45685", "issue:43824|issue:43957", - "issue:43550|issue:44910", "issue:43645|issue:43646", + "issue:43550|issue:44910", "issue:42915|issue:43957", "issue:43366|issue:44589", "issue:44560|issue:44918", @@ -1699,222 +1710,230 @@ "issue:44451|issue:44488", "issue:42915|issue:44756", "issue:45313|issue:45357", - "issue:42491|issue:45276" + "issue:42491|issue:45276", + "issue:43708|issue:44075", + "issue:43644|issue:43645" ], - "prepared_review_unit_hash": "521c103eb31bd9fa102759c9a3b13f0ce9636d713aa69ed23056a619f3a99db6", + "prepared_review_unit_hash": "8dae01133f0021020bcf2e8e32b4ac3d5db916940aed7f14a0f17c9fb6e11ce2", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36684, - "estimated_input_tokens": 9171, - "estimated_eval_tokens": 18598 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37794, + "estimated_input_tokens": 9449, + "estimated_eval_tokens": 19154 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most items are unrelated model-specific regressions; only a few soft pairs look like true duplicates. The clearest duplicate groups are the non-persistent-buffer regression, the custom-model init regression, and the SigLIP2 wrong-output reports.", - "confidence": 0.74, - "canonical_issue_reason": "issue:43644 is the best anchor for the non-persistent-buffer regression: it is the earlier, more fully specified report, and issue:44534 is essentially the same bug phrased later.", + "summary": "This cluster is mostly heterogeneous, with a few small duplicate mini-clusters around the same regression or model family. I accepted only pairs that look like the same underlying bug/change; most others are just related-by-topic at best.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 44534 is the cleanest anchor for one of the true duplicate groups: it states the non-persistent-buffer regression broadly and matches an exact duplicate title.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43644 is the strongest single issue to keep as canonical because it is concrete, well-scoped, and has a clear near-duplicate; it is easier to triage than the more model-specific or ambiguous reports.", + "best_issue_reason": "Issue 44534 is the best overall issue to keep because it is the most general, clearly scoped regression report among the accepted duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45325", - "right": "issue:45902", + "left": "issue:43366", + "right": "issue:43828", "accept": false, - "reason": "Different models and different failure modes: rope scaling bug vs uninitialized-memory rotary embedding issue." + "reason": "Different problems: GGUF/gpt-oss support vs a Phi-tiny-MoE dtype mismatch." }, { "left": "issue:45663", "right": "issue:45823", "accept": false, - "reason": "Same model family, but distinct code paths: FSDP2 shared_kv_states KeyError vs device_map auto device mismatch." + "reason": "Same model family, but different failure modes: FSDP2 KeyError vs device_map auto device mismatch." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Unrelated topics: GGUF architecture support vs Qwen3-MoE sliding_window behavior." + "reason": "Unrelated Qwen3-MoE sliding_window bug vs gpt-oss GGUF support request." }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "Different inference bugs in different pipelines; no shared concrete code path." + "reason": "Different audio/Whisper code paths and symptoms; not the same bug." }, { "left": "issue:43950", "right": "issue:45834", "accept": false, - "reason": "Non-persistent buffer corruption vs OCR long-input index error are not the same bug." + "reason": "Non-persistent-buffer corruption vs Kosmos2.5 long-ocr index error are unrelated." }, { "left": "issue:44560", "right": "issue:45685", "accept": false, - "reason": "Video StopIteration issue and MPS histogram Int limitation are unrelated." + "reason": "Different architectures and failures: Qwen3-VL video StopIteration vs MPS int histogram support." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Import error for a missing class vs meta-device loading regression; not the same issue." - }, - { - "left": "issue:43550", - "right": "issue:44910", - "accept": false, - "reason": "Different concrete failures: torch.compile with SDPA vs flash-attention crash on Qwen3.5 position_ids." + "reason": "ImportError for Qwen2.5-VL is unrelated to meta-device model loading breakage." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "These are the same custom-model initialization regression; one is just a notebook-specific phrasing of the other." + "reason": "Both describe the same Transformers v5 custom model initialization regression; one just adds the Jupyter-notebook context." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different bugs: Bamba torch.compile/SDPA crash vs Qwen3.5 flash-attention memory access." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "Qwen3MoE FineGrainedFP8Config failure is unrelated to meta-device loading bugs." + "reason": "Qwen3-MoE FP8 failure and meta-device loading are different code paths." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "GGUF support request and Float8 storage TypeError are different problems." - }, - { - "left": "issue:44560", - "right": "issue:44918", - "accept": false, - "reason": "Video StopIteration vs unpacking embeddings in TRL SFT trainer are different code paths." + "reason": "GGUF support request vs Float8 storage lookup error; no shared root cause." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "A script runtime failure and GGUF model support are unrelated." + "reason": "3d_parallel.py runtime failure is unrelated to GGUF model support." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "GenerationMixin import error is unrelated to Qwen3MoE FP8 loading." + "reason": "GenerationMixin import error is unrelated to Qwen3-MoE FP8 handling." }, { "left": "issue:44661", "right": "issue:45823", "accept": false, - "reason": "Tokenizer-mapping/model-like failure and Gemma4 device mismatch are different bugs." + "reason": "Tokenizer-mapping/new-model helper bug vs Gemma4 PLE device mismatch; different issues." }, { "left": "issue:44560", "right": "issue:44805", "accept": false, - "reason": "Different symptoms and likely different tensor-shape issues; not enough to treat as duplicates." + "reason": "Both are Qwen-related, but one is a video StopIteration bug and the other is a mask shape mismatch." }, { "left": "issue:43632", "right": "issue:45834", "accept": false, - "reason": "_is_hf_initialized flag regression is unrelated to the long OCR index error." + "reason": "_is_hf_initialized regression is unrelated to Kosmos2.5 long-input indexing." }, { "left": "issue:38175", "right": "issue:43994", "accept": true, - "reason": "Both report incorrect SigLIP2 outputs for the same model family; the later one is essentially the same underlying inference bug." + "reason": "Both report the same SigLIP2-base-patch16-224 bad-output regression, just with different symptom descriptions." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Deepspeed ZeRO-3 Bert initialization failure is unrelated to add-new-model-like tokenizer mapping behavior." + "reason": "Zero3/Bert index error and tokenizer-mapping helper failure are different bugs." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "Zero3 Bert index error is a different bug from the tokenizer-mapping/model-like issue." + "reason": "Zero3/Bert empty-shard IndexError is unrelated to add-new-model-like tokenizer mapping." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping bug and tensor mask shape mismatch are unrelated." }, { "left": "issue:44877", "right": "issue:45030", - "accept": false, - "reason": "Both involve config loading, but the model-specific failures are not the same concrete bug." + "accept": true, + "reason": "Both look like the same stricter config-validation regression preventing model configs from loading." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "device_map dequantized-model offload failure and mmap OOM on Strix Halo are different issues." + "reason": "GPU+CPU offload dequantized loading failure is unrelated to Strix Halo mmap OOM." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Exact duplicate regression: both say transformers v5 fills non-persistent buffers with junk." + "reason": "Exact same non-persistent-buffer junk regression in Transformers v5." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "SigLIP2 wrong outputs and docs-example mismatches/quantization failures are related only by model family, not the same bug." + "reason": "SigLIP2 output corruption is not the same as the documentation example's processor/quantization errors." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Qwen3_moe LoRA hf4.x/hf5.x compatibility and BitNet packed-weight loading are unrelated." + "reason": "Qwen3-MoE LoRA compatibility and BitNet packed-weight loading are different issues." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Auto device mapping CUDA assert and FineGrainedFP8Config failure are different Qwen issues." + "reason": "Different failure modes: auto device mapping cudaErrorAssert vs FineGrainedFP8Config." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "GLM5 issue and Qwen3.5 flash-attention generation bug are unrelated." + "reason": "GLM5 is unrelated to Qwen3.5 flash-attention generation failures." }, { "left": "issue:44451", "right": "issue:44488", - "accept": false, - "reason": "Both are BERT loading failures, but the titles do not establish the same concrete underlying bug." + "accept": true, + "reason": "Both are the same latest-version model-loading failure for specific BERT models." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "ZeRO-3 weight loading and incorrect save_pretrained visual encoder keys are different Qwen3.5 regressions." + "reason": "ZeRO-3 language_model weight loading and save_pretrained visual-key corruption are different bugs." }, { - "left": "issue:42491", - "right": "issue:45276", + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Trainer resume/max_steps math bug is unrelated to SGD optimizer args being ignored." + }, + { + "left": "issue:43644", + "right": "issue:43645", "accept": false, - "reason": "Qwen3_moe LoRA compatibility on hf5.x and Gemma4 resize_token_embeddings behavior are unrelated." + "reason": "Non-persistent-buffer corruption and custom model initialization breakage are different regressions." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and the duplicate judgments stay conservative overall. The canonical non-persistent-buffer pair is well supported by the titles, and the accepted soft pairs are plausibly same-bug matches without overclaiming beyond the available evidence. No pair is forced together on weak evidence, and the rejected pairs are explained in a cautious way." + "feedback": "The summary is grounded in the packet and stays conservative overall. The accepted duplicate pairs are supported by strong title/body similarity, especially the non-persistent-buffer regression around 43644/44534, and the rest of the pair judgments read as cautious rather than overstated. No PR mergeability issue applies here." }, "evaluator_used": true, "retried": false, @@ -1930,8 +1949,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -1947,28 +1966,28 @@ "issue:43404", "issue:43425", "issue:43454", - "issue:43644", - "issue:43645", - "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:44479", "issue:44488", "issue:44756", + "issue:44877", "issue:44910", "issue:44912", + "issue:44928", "issue:44960", "issue:44991", + "issue:45216", "issue:45276", "issue:45292", "issue:45335", @@ -1985,15 +2004,13 @@ "issue:46050" ], "soft_pairs": [ - "issue:43708|issue:44075", - "issue:43644|issue:43645", "issue:44488|issue:44991", "issue:43854|issue:43931", "issue:45488|issue:45812", "issue:39692|issue:43994", "issue:43257|issue:43854", - "issue:43856|issue:43873", "issue:42491|issue:43931", + "issue:43856|issue:43873", "issue:43856|issue:44387", "issue:45362|issue:45406", "issue:44960|issue:46050", @@ -2019,269 +2036,285 @@ "issue:43957|issue:44756", "issue:42947|issue:43381", "issue:45276|issue:45292", - "issue:44164|issue:44756" + "issue:44164|issue:44756", + "issue:43854|issue:45216", + "issue:43381|issue:44928", + "issue:44479|issue:44877", + "issue:45588|issue:45663" ], - "prepared_review_unit_hash": "c7771e78b7b4192731d92935b995cd8b50f4c107943c00197fa0c5edc3527706", + "prepared_review_unit_hash": "18c3af5791d812a9e1a7cdaec9f3769649932a3d8fa014fcd540028296bb07b6", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 35, - "serialized_chars": 38299, - "estimated_input_tokens": 9575, - "estimated_eval_tokens": 19406 + "soft_pair_count": 37, + "serialized_chars": 38940, + "estimated_input_tokens": 9735, + "estimated_eval_tokens": 19726 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated issue reports. The only soft edge that looks like the same underlying bug is the resize_token_embeddings/output_embeddings pair; all other candidates are too model-/path-specific to merge confidently.", - "confidence": 0.79, - "canonical_issue_reason": "Issue 45292 is the cleanest canonical for the shared resize_token_embeddings bug: it states the core generic symptom on output_embeddings, while 45276 is a broader Gemma4-specific manifestation of the same failure.", + "summary": "Mostly a false-positive semantic cluster: the items span unrelated tokenizer loading, quantization, model-loading, flash-attention, and training bugs. The only clear duplicate set is the `resize_token_embeddings` regression affecting `output_embeddings` / `decoder.embed_tokens` across Gemma4 and T5Gemma reports.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45292 is the best canonical issue because it states the broad API-level regression (`resize_token_embeddings` not updating `output_embeddings`) and the other matching reports look like model-specific manifestations of the same bug.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45292 is the best representative because it describes the common root symptom without extra model-specific detail, making it the most reusable target for deduplication.", + "best_issue_reason": "issue:45292 is the most general and reusable description of the duplicate set; issue:45276 and issue:45335 are narrower variants of the same embedding-resize failure.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43708", - "right": "issue:44075", - "accept": false, - "reason": "Different training/runtime bugs: checkpoint resume/max_steps accounting vs optimizer arguments being ignored." - }, - { - "left": "issue:43644", - "right": "issue:43645", - "accept": false, - "reason": "Both mention Transformers 5.0.0, but one is about non-persistent buffers and the other about custom model initialization in Jupyter; different problems." - }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading failures, but for different models and likely different root causes; not the same bug." + "reason": "Both are tokenizer-loading failures, but they affect different models and likely different tokenizer paths; too broad to treat as the same bug." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Different model-loading failures for different architectures; no clear shared code-path bug." + "reason": "Different models and different failure modes (unit test loading vs weight-shape mismatch) indicate separate issues." }, { "left": "issue:45488", "right": "issue:45812", "accept": false, - "reason": "Tokenizer regressions in v5, but one is LlamaTokenizer pre-tokenizer handling and the other is AutoTokenizer token IDs for Granite; not the same change." + "reason": "Both are tokenizer regressions, but one is a LlamaTokenizer pre-tokenizer override bug and the other is a Granite token-ID regression; not the same concrete fix." }, { "left": "issue:39692", "right": "issue:43994", "accept": false, - "reason": "Same model family, but one is a documentation/example issue plus quantization failure; the other is a runtime nonsensical-results report. Too different to merge confidently." + "reason": "Same model family, but one is a documentation/example problem and the other is an actual inference regression; different underlying issues." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Qwen3 MoE weight conversion/offload behavior vs GLM-4.7-Flash loading in tests; unrelated." + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to GLM-4.7-Flash loading failures." }, { - "left": "issue:43856", - "right": "issue:43873", + "left": "issue:42491", + "right": "issue:43931", "accept": false, - "reason": "Different resource problems: Qwen3 MoE training memory usage vs quantized offloading behavior." + "reason": "Different models and different problems; LoRA compatibility on hf5.x is not the same as a Qwen3-VL shape mismatch." }, { - "left": "issue:42491", - "right": "issue:43931", + "left": "issue:43856", + "right": "issue:43873", "accept": false, - "reason": "A LoRA compatibility regression across branches is unrelated to a Qwen3-VL weight-shape mismatch." + "reason": "Both mention memory, but one is Qwen3 MoE training memory use and the other is quantization offloading behavior; not the same bug." }, { "left": "issue:43856", "right": "issue:44387", "accept": false, - "reason": "Both involve memory, but one is MoE training memory usage and the other is int4 quantization reserved-memory/OOM; different issues." + "reason": "Both are memory-related, but one is MoE training footprint and the other is int4 quantization reserved-memory OOM; separate regressions." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Different models and failures: Qwen3.5 chat crash vs Gemma4Processor missing _tokenizer in serve." + "reason": "Different model families and different failure surfaces (chat crash vs processor attribute error)." }, { "left": "issue:44960", "right": "issue:46050", "accept": false, - "reason": "GLM5 loading and sinq save/load corruption are unrelated." + "reason": "GLM5 is unrelated to the sinq save/load quantization regression." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "LoRA hf4.x/hf5.x compatibility has no overlap with resize_token_embeddings behavior." + "reason": "LoRA hf4.x/hf5.x compatibility is unrelated to `resize_token_embeddings` not updating embeddings." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "Both are loading/saving related, but extra_state handling and init_empty_weights _is_hf_initialized are distinct bugs." + "reason": "Both concern loading/saving internals, but `extra_state` handling and `_is_hf_initialized` argument handling are distinct code paths." }, { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Both hit flash_attention paths, but one is a packed-sequence/position_ids crash and the other is an s_aux=None AttributeError; different code paths." + "reason": "Both are flash-attention crashes, but one is packed-sequence misinterpretation and the other is a missing `s_aux` attribute; different bugs." }, { "left": "issue:45276", "right": "issue:45335", - "accept": false, - "reason": "Similar symptom, but model-specific implementations differ (Gemma4 vs T5Gemma) and the reports do not establish a single shared root cause." + "accept": true, + "reason": "Both report the same `resize_token_embeddings` propagation bug, just on different model-specific embedding modules." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to Whisper pipeline return_language behavior." + "reason": "Importing `GenerationMixin` and Whisper `return_language` pipeline behavior are unrelated issues." }, { "left": "issue:43404", "right": "issue:43454", "accept": false, - "reason": "Both mention untied lm_head weights, but they are different model families and the evidence is not enough to treat them as the same bug." + "reason": "Both mention lm_head tying, but they are different model implementations and likely separate model-specific fixes." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Torch 2.10 compatibility and meta-device model loading are different incompatibility reports." + "reason": "Torch-version incompatibility and `torch.device(\"meta\")` loading breakage are different regressions." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Qwen3 auto device mapping CUDA assert is unrelated to AudioFlamingo3 track-to-track leakage." + "reason": "Qwen3 device-map CUDA assert and AudioFlamingo3 batch token leak are unrelated." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "Different models and failures; no shared underlying bug." + "reason": "SigLIP2 docs/example problems are unrelated to GLM5 model loading." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Tokenizer vocab-size mismatch is unrelated to failing to load a specific model." + "reason": "One is tokenizer vocab-size metadata mismatch; the other is a model-specific tokenizer loading failure." }, { "left": "issue:42947", "right": "issue:45704", "accept": false, - "reason": "Gradient checkpointing/PEFT and apex RMSNorm memory leak are unrelated." + "reason": "Gradient checkpointing/PEFT LoRA behavior is unrelated to the apex RMSNorm memory leak." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Gemma4 FSDP2 attention state rebuilding and Qwen multimodal text_config unwrapping are different model-architecture bugs." + "reason": "Different issues: Gemma-4 FSDP2 shared state handling versus Qwen multimodal `text_config` unwrapping." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Quantization offloading vs int4 reserved-memory growth are related broadly to memory, but not the same concrete bug." + "reason": "Quantized offloading behavior and int4 reserved-memory OOM are not the same concrete bug." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "init_empty_weights argument handling is unrelated to mmap/OOM on Strix Halo." + "reason": "`init_empty_weights` argument handling is unrelated to Strix Halo mmap OOM." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Different Gemma4 vs Qwen3.5 MoE offload issues; not mergeable as one fix." + "reason": "Different code paths: Gemma-4 attention state rebuilding versus Qwen3_5 MoE rotary embedding CPU offload." }, { "left": "issue:44155", "right": "issue:45823", "accept": false, - "reason": "Different models and failures: embedding/token leakage vs PLE device mismatch." + "reason": "AudioFlamingo3 track leakage and Gemma4 PLE device mismatch are unrelated." }, { "left": "issue:43335", "right": "issue:45663", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and Gemma4 FSDP2 attention state are unrelated." + "reason": "SwitchTransformers sparse-layer construction and Gemma-4 FSDP2 attention state handling are unrelated." }, { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Different quantized-model loading problems for different models/quantization formats." + "reason": "Different quantization formats and failures: NVFP4 load error versus MXFP4 fallback to bf16." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Different memory/OOM reports with no shared code-path evidence." + "reason": "Both end in OOM, but one is quantization memory reservation and the other is mmap behavior on specific hardware." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading failures and disabling mmap for Strix Halo are different issues." + "reason": "Model loading with meta tensors and disabling mmap for Strix Halo are unrelated fixes." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Gradient checkpointing in eval mode is unrelated to PEFT LoRA checkpointing effectiveness." + "reason": "Both mention gradient checkpointing, but one is ineffective under PEFT LoRA and the other is eval-mode incompatibility." }, { "left": "issue:45276", "right": "issue:45292", "accept": true, - "reason": "Both describe the same resize_token_embeddings failure to update output_embeddings; 45276 is the broader model-specific instance, while 45292 is the generic core report." + "reason": "They describe the same `resize_token_embeddings` regression, with one being the general report and the other a model-specific manifestation." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "save/from_pretrained extra_state handling is unrelated to mmap/OOM on Strix Halo." - } - ] - }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative overall. The only accepted soft edge, issue:45276 \u2194 issue:45292, is well supported by the titles and the shared resize_token_embeddings/output_embeddings symptom. The rejected pairs are phrased cautiously and do not overstate evidence. No PR-pair mergeability concerns apply here." - }, - "evaluator_used": true, - "retried": false, - "accepted_nontrivial_soft_edge": true, - "error_kind": null, - "error_message": null - }, - { - "label": "issue", - "component_index": 1, - "component_count": 17, + "reason": "Serialization `extra_state` handling is unrelated to mmap/OOM behavior." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "GLM-4.7-Flash loading and Qwen3.5 checkpoint save regression are separate model-specific issues." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Gradient checkpointing eval-mode behavior is unrelated to the Qwen3.5 3D position_ids / SDPA fallback bug." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video-input regression across Qwen VL models is unrelated to strict config blocking granite_speech loading." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Flash attention sink handling and Gemma-4 FSDP2 attention state rebuilding are different problems." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative overall. The only accepted duplicate set is the resize_token_embeddings regression across issue:45276, issue:45292, and issue:45335, which is supported by the titles and soft-pair signal. The other pairwise rejections are appropriately cautious given the limited evidence and model-specific differences." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, "review_unit_index": 7, "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -2289,16 +2322,16 @@ "issue:39404", "issue:43257", "issue:43299", - "issue:43381", "issue:43425", "issue:43452", + "issue:43596", "issue:43611", "issue:43645", "issue:43646", "issue:43708", "issue:43716", "issue:43828", - "issue:43854", + "issue:43856", "issue:43901", "issue:43906", "issue:43931", @@ -2313,37 +2346,33 @@ "issue:44509", "issue:44530", "issue:44589", + "issue:44671", "issue:44683", + "issue:44737", "issue:44756", "issue:44805", "issue:44863", - "issue:44877", "issue:44898", "issue:44912", "issue:44918", - "issue:44928", + "issue:44991", + "issue:45081", "issue:45084", - "issue:45216", "issue:45292", "issue:45335", - "issue:45588", - "issue:45663", "issue:45684", "issue:45698", "issue:45701", "issue:45759", - "issue:45823" + "issue:45823", + "issue:45859" ], "soft_pairs": [ - "issue:43854|issue:45216", - "issue:43381|issue:44928", - "issue:44479|issue:44877", "issue:45684|issue:45698", - "issue:45588|issue:45663", "issue:43646|issue:43708", "issue:43645|issue:44164", - "issue:43828|issue:43957", "issue:45292|issue:45335", + "issue:43828|issue:43957", "issue:43425|issue:43828", "issue:44402|issue:44451", "issue:43257|issue:44863", @@ -2360,226 +2389,251 @@ "issue:44479|issue:44805", "issue:43994|issue:44079", "issue:43425|issue:44918", - "issue:43901|issue:44509", "issue:44589|issue:45084", + "issue:43901|issue:44509", "issue:43299|issue:43931", "issue:44164|issue:45759", - "issue:36032|issue:43452" + "issue:43596|issue:44683", + "issue:36032|issue:43452", + "issue:44671|issue:44991", + "issue:39404|issue:44671", + "issue:44991|issue:45081", + "issue:43856|issue:44756", + "issue:44164|issue:44479", + "issue:44737|issue:45859" ], - "prepared_review_unit_hash": "e44fba1572dce3fbf5b76a47c515c65fcaf1232fff60a035d44c2b67aa14e082", + "prepared_review_unit_hash": "d3a75afa49518dedee36151c0071f45be91042818ecd3c7fa212be4f5f69b751", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36815, - "estimated_input_tokens": 9204, - "estimated_eval_tokens": 18664 + "soft_pair_count": 33, + "serialized_chars": 37731, + "estimated_input_tokens": 9433, + "estimated_eval_tokens": 19122 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous Transformers bug reports. Only one soft edge looks like a true duplicate pair: the `resize_token_embeddings` regression in 45292/45335, where the general bug and the T5Gemma-specific symptom match the same underlying API issue.", - "confidence": 0.86, - "canonical_issue_reason": "Issue 45292 is the broader, model-agnostic report: `resize_token_embeddings` does not update the affected embeddings. Issue 45335 appears to be the T5Gemma-specific manifestation of the same bug, so 45292 is the better canonical anchor.", + "summary": "The set is mostly a heterogeneous grab bag of Transformers 5.x regressions and model-specific loading bugs. Only one soft pair looks like the same underlying defect: resize_token_embeddings not updating tied output/decoder embeddings.", + "confidence": 0.81, + "canonical_issue_reason": "Issue 43611 is the broadest representative of the loading-regression subgroup (v5 breaks model loading via `base_model_prefix`), while most other items are narrower variants or unrelated failures.", "canonical_pr_reason": null, - "best_issue_reason": "45292 is the best representative issue because it describes the root regression in general terms, while 45335 is a narrower model-specific reproduction.", + "best_issue_reason": "Issue 43611 is the best single anchor for triage because it is concise, broadly scoped, and most representative of the model-loading breakages in this cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43854", - "right": "issue:45216", - "accept": false, - "reason": "Different bugs: one is a model load/save corruption case, the other is a Qwen3.5 checkpoint save_pretrained regression." - }, - { - "left": "issue:43381", - "right": "issue:44928", - "accept": false, - "reason": "Different failure modes: gradient checkpointing in eval mode vs RLHF NaNs from Qwen3.5 position_ids/SDPA behavior." - }, - { - "left": "issue:44479", - "right": "issue:44877", - "accept": false, - "reason": "Both are loading regressions, but on different model/config paths (video-input Qwen VL vs granite_speech strict config)." - }, { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Distinct issues: read-only permission propagation during save_pretrained vs loading the wrong custom module after save_pretrained." - }, - { - "left": "issue:45588", - "right": "issue:45663", - "accept": false, - "reason": "Different code paths: flash_attention crash on sink-less models vs Gemma4 FSDP2 shared_kv_states rebuild KeyError." + "reason": "Both involve custom-model save/load, but one is about file permissions during save_pretrained and the other is about loading the wrong custom module." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model initialization vs Trainer resume_from_checkpoint max_steps calculation are unrelated regressions." + "reason": "Custom model initialization is unrelated to Trainer resume_from_checkpoint/max_steps calculation." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Notebook custom-model init failure is not the same bug as extra_state not being handled by save/from_pretrained." - }, - { - "left": "issue:43828", - "right": "issue:43957", - "accept": false, - "reason": "Different root causes: autocast dtype mismatch in Phi-tiny-MoE vs meta-device loading breakage in some models." + "reason": "Notebook-defined custom model init and extra_state save/from_pretrained handling are different code paths and symptoms." }, { "left": "issue:45292", "right": "issue:45335", "accept": true, - "reason": "Same underlying API regression: `resize_token_embeddings` does not propagate to the relevant output/decoder embeddings; 45335 is the T5Gemma-specific reproduction." + "reason": "Same bug: resize_token_embeddings is not propagating to tied output/decoder embeddings; 45335 is the t5gemma-specific instance of 45292." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Both are dtype/device related, but one is an autocast mismatch in Phi MoE and the other is a meta-device loading failure for different models." }, { "left": "issue:43425", "right": "issue:43828", "accept": false, - "reason": "Torch version compatibility is a different problem from the Phi-tiny-MoE autocast dtype mismatch." + "reason": "Torch version incompatibility and an autocast dtype mismatch are distinct failures." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Both involve tokenizer/model loading, but the specific failures differ (vocab-size mismatch vs inability to load a particular model)." + "reason": "Tokenizer vocab metadata mismatch is not the same as a model-specific loading failure for vesteinn/ScandiBERT." }, { "left": "issue:43257", "right": "issue:44863", "accept": false, - "reason": "Separate model-loading bugs affecting different architectures and mechanisms." + "reason": "Qwen3 MoE conversion under accelerate+deepspeed and NemotronH checkpoint loading are unrelated model-loading bugs." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Unrelated issues: missing Float8 storage object vs mmap-induced OOM on Strix Halo." + "reason": "Missing Float8 storage support and mmap-related OOM behavior are different problems." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper pipeline return_language regression is unrelated to a broad tokenization-change report." + "reason": "Whisper pipeline return_language regression is unrelated to a general tokenization-version change report." }, { "left": "issue:44683", "right": "issue:45823", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device_map mismatches are different failures." + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 PLE device mismatch are different code-path issues." }, { "left": "issue:43425", "right": "issue:45823", "accept": false, - "reason": "Torch 2.10 incompatibility is not the same bug as Gemma4 PLE device placement mismatch." + "reason": "Torch 2.10 incompatibility is not the same as a device_map='auto' forward mismatch in Gemma4 PLE." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "Loading models with base_model_prefix is a different regression from Jupyter notebook custom-model initialization." + "reason": "Both are about v5 breaking custom/model loading, but one is base_model_prefix handling and the other is notebook-defined custom model initialization." }, { "left": "issue:43425", "right": "issue:43716", "accept": false, - "reason": "Torch compatibility problems are unrelated to the Mistral-3 image preprocessor dtype mismatch." + "reason": "A version incompatibility report and a Mistral-3 dtype mismatch are not the same bug." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "Different failures: PagedAttentionCache linear_attention crash vs a template-node compilation TypeError." + "reason": "PagedAttentionCache linear_attention crash and a template-compilation TypeError are separate failures." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline behavior and MXFP4 quantization fallback are unrelated." + "reason": "Whisper return_language and MXFP4 quantization fallback are unrelated regressions." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "No clear evidence they are the same underlying bug; the titles point to different reproduction contexts and failure modes." + "reason": "An isolated reproduction of another issue is not the same as the Qwen3.5 PagedAttentionCache crash." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "torch.export/torch_compilable_check and Perceiver interpolation are different code paths." + "reason": "torch.export with torch_compilable_check and Perceiver non-default-resolution classification are different bugs." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "Video-input regression in Qwen VL models is unrelated to the mask-shape IndexError." + "reason": "Qwen video-input regression and a mask-shape IndexError are not the same underlying defect." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "SigLIP2 inference quality issue is not the same as ModelOutput key assignment behavior." + "reason": "Incorrect Siglip outputs and ModelOutput key assignment are distinct behaviors." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch compatibility is unrelated to the Qwen3.5 embedding unpacking issue in TRL SFTTrainer." + "reason": "Torch version incompatibility is unrelated to Qwen3.5 embedding unpacking with TRL SFT trainer." }, { - "left": "issue:43901", - "right": "issue:44509", + "left": "issue:44589", + "right": "issue:45084", "accept": false, - "reason": "Both are docs-related, but they cover different removed/changed pipeline behaviors and are not the same issue." + "reason": "Float8 storage lookup failure and template-node compilation failure are different issues." }, { - "left": "issue:44589", - "right": "issue:45084", + "left": "issue:43901", + "right": "issue:44509", "accept": false, - "reason": "Missing Float8 storage object and non-template-node compilation are unrelated." + "reason": "These are both documentation issues, but they cover different removed/changed pipeline APIs." }, { "left": "issue:43299", "right": "issue:43931", "accept": false, - "reason": "Both mention Qwen3-VL loading, but the concrete failures differ too much to merge confidently." + "reason": "Qwen3VL MoE conversion/loading and Qwen3-VL weight-shape mismatch are related only broadly, not the same concrete bug." }, { "left": "issue:44164", "right": "issue:45759", "accept": false, - "reason": "extra_state save/load handling is unrelated to unwrapping text_config for composite Qwen multimodal configs." + "reason": "extra_state save/load handling and text_config unwrapping for composite configs are separate loader/saver problems." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Zero3 init index error and compiled flex_attention torch>=2.9 failure are unrelated." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "T5 tokenizer add_special_tokens conflict is unrelated to gguf_file breaking from_pretrained for tokenizer/model loading." + "reason": "T5Tokenizer add_special_tokens conflict and gguf_file.from_pretrained breakage are different tokenizer/load paths." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "CamemBERT masked LM prediction errors and est-roberta tokenizer loading failure are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline regression and CamemBERT prediction corruption are unrelated model families and bugs." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer loading failure for est-roberta and a Mistral regex patch crash are separate tokenizer code paths." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3 MoE training memory inefficiency and Strix Halo mmap OOM avoidance are not the same bug." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "extra_state save/load handling is unrelated to the Qwen video-input regression." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "XLNet CPU arange placement and Qwen3_5MoeTextRotaryEmbedding CPU offload compatibility are different device-placement issues." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative on the soft edges. The only accepted duplicate pair, 45292/45335, is a reasonable merge candidate from the titles alone: both report `resize_token_embeddings` failing to update downstream embeddings, with 45335 being the T5Gemma-specific manifestation. The remaining pairwise rejections do not overclaim sameness and are phrased cautiously enough for the available evidence." + "feedback": "The grouping is grounded in the packet and stays mostly conservative. The canonical anchor choice (43611) is defensible as a broad v5 loading-regression representative, and the only accepted soft pair (45292/45335) is strongly supported by the titles. The rejected pairs generally distinguish different code paths or symptom classes without overstating similarity. " }, "evaluator_used": true, "retried": false, @@ -2595,8 +2649,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -2608,13 +2662,17 @@ "issue:40990", "issue:42175", "issue:42915", + "issue:43089", + "issue:43262", "issue:43317", + "issue:43322", + "issue:43377", "issue:43421", "issue:43452", "issue:43482", + "issue:43540", "issue:43541", "issue:43582", - "issue:43596", "issue:43611", "issue:43632", "issue:43643", @@ -2622,21 +2680,16 @@ "issue:43645", "issue:43646", "issue:43716", - "issue:43856", "issue:43883", - "issue:44164", "issue:44206", "issue:44220", "issue:44451", "issue:44464", - "issue:44479", "issue:44488", "issue:44530", + "issue:44534", "issue:44661", - "issue:44671", "issue:44683", - "issue:44737", - "issue:44756", "issue:44811", "issue:44849", "issue:44987", @@ -2644,23 +2697,16 @@ "issue:45081", "issue:45084", "issue:45325", + "issue:45507", "issue:45584", - "issue:45859", "issue:45987", "issue:46178" ], "soft_pairs": [ - "issue:43596|issue:44683", - "issue:44671|issue:44991", - "issue:39404|issue:44671", - "issue:44991|issue:45081", - "issue:43856|issue:44756", - "issue:44164|issue:44479", - "issue:44737|issue:45859", - "issue:32090|issue:39404", "issue:43582|issue:43632", - "issue:44661|issue:44683", + "issue:32090|issue:39404", "issue:44488|issue:45081", + "issue:44661|issue:44683", "issue:44488|issue:44987", "issue:43611|issue:43644", "issue:40990|issue:43421", @@ -2681,275 +2727,277 @@ "issue:36032|issue:43645", "issue:43317|issue:43482", "issue:44849|issue:45325", - "issue:44987|issue:44991" + "issue:44987|issue:44991", + "issue:45084|issue:45507", + "issue:43883|issue:44991", + "issue:43262|issue:43377", + "issue:43645|issue:44534", + "issue:43089|issue:45987", + "issue:43646|issue:44534", + "issue:43322|issue:43540" ], - "prepared_review_unit_hash": "eb9d4803d52350f448dd896917dd93483ffc3f05dca020db179711655aea898c", + "prepared_review_unit_hash": "2cadc8ed113618a03c6c14be2913949a72dc3ac50f1dfe587e37dcc78c604d2f", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 32, - "serialized_chars": 37231, - "estimated_input_tokens": 9308, - "estimated_eval_tokens": 18872 + "serialized_chars": 36658, + "estimated_input_tokens": 9165, + "estimated_eval_tokens": 18586 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous; the soft matches are high-similarity false positives across different models, tokenizers, and regression types. I did not find any pair that looks like the same concrete bug/change.", - "confidence": 0.81, - "canonical_issue_reason": "Issue 43646 is the broadest v5 regression among the set and best represents the cluster\u2019s general model-initialization breakage theme, though it is not a true duplicate of the others.", + "summary": "This cluster is mostly a grab-bag of unrelated Transformers issues, with only a couple of plausible duplicate subgroups: GGUF loading regressions and an audio feature-extraction crash. I\u2019d use the broad GGUF loading report as the best representative issue.", + "confidence": 0.67, + "canonical_issue_reason": "issue:43452 is the broadest, most actionable report here: it covers GGUF loading failing through both AutoTokenizer and AutoModelForCausalLM, making it a better umbrella than the narrower follow-up-style report.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43646 is the most general and central issue, with the widest scope among the candidates.", + "best_issue_reason": "issue:43452 is the strongest global representative because it describes a concrete, cross-API loader regression and plausibly subsumes the more specific Qwen2.5-GGUF failure.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43596", - "right": "issue:44683", - "accept": false, - "reason": "Different failures in different subsystems: deepspeed/BertModel broadcast vs compiled flex_attention on newer torch." - }, - { - "left": "issue:44671", - "right": "issue:44991", - "accept": false, - "reason": "Both are v5-era loading regressions, but one is CamemBERT masked-LM output quality and the other is tokenizer loading for a different model." - }, - { - "left": "issue:39404", - "right": "issue:44671", - "accept": false, - "reason": "Whisper pipeline language return bug is unrelated to CamemBERT masked-LM prediction regression." - }, - { - "left": "issue:44991", - "right": "issue:45081", - "accept": false, - "reason": "Both involve tokenizer loading, but they hit different models and distinct code paths (_patch_mistral_regex vs generic tokenizer load failure)." - }, - { - "left": "issue:43856", - "right": "issue:44756", - "accept": false, - "reason": "Both mention memory/OOM, but one is Qwen3 MoE training memory usage and the other is mmap behavior on Strix Halo." - }, - { - "left": "issue:44164", - "right": "issue:44479", - "accept": false, - "reason": "Save/from_pretrained extra_state handling and Qwen VL video-input regression are separate bugs." - }, - { - "left": "issue:44737", - "right": "issue:45859", + "left": "issue:43582", + "right": "issue:43632", "accept": false, - "reason": "Both are device-placement issues, but they affect different models and code paths (XLNet relative positions vs Qwen3.5 MoE rotary embedding)." + "reason": "Different failures: Apple Silicon allocator warmup TypeError vs. `_is_hf_initialized` flag regression." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer gpu_broadcast NoneType crash is unrelated to Whisper return_language behavior." + "reason": "Trainer GPU broadcast error and Whisper `return_language` pipeline bug are unrelated." }, { - "left": "issue:43582", - "right": "issue:43632", + "left": "issue:44488", + "right": "issue:45081", "accept": false, - "reason": "AppleSilicon allocator TypeError and _is_hf_initialized regression are unrelated." + "reason": "Both are tokenizer-loading problems, but they affect different models and code paths." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "Tokenizer mapping lookup failure and torch >=2.9 flex_attention compilation failure are unrelated." - }, - { - "left": "issue:44488", - "right": "issue:45081", - "accept": false, - "reason": "Both are tokenizer-loading reports, but they concern different models and different breakages." + "reason": "`add-new-model-like`/tokenizer mapping and compiled flex_attention on torch>=2.9 are unrelated." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Both are loading regressions, but one is a specific model load failure and the other is a different model's tokenizer load failure." + "reason": "Separate model-loading regressions with different symptoms and likely different root causes." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Both are v5 regressions, but base_model_prefix loading and non-persistent buffer junk are distinct initialization bugs." + "reason": "Both are v5 regressions, but one is about `base_model_prefix` loading and the other about buffer initialization." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "High perplexity on gpt-oss and runtime special-token post-processor updates are unrelated." + "reason": "Perplexity/tokenization behavior is not the same as runtime post-processor syncing." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Image preprocessor dtype mismatch and template-node compilation error are unrelated." + "reason": "Mistral image dtype mismatch and template-node compilation error are unrelated." }, { "left": "issue:44206", "right": "issue:44220", - "accept": false, - "reason": "Both touch audio features, but the reported crashes and APIs differ; not the same concrete bug." + "accept": true, + "reason": "Both point to the same audio feature-extraction path; the extractor crash and `_torch_extract_fbank_features()` issue look like the same underlying bug." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "T5 tokenizer add_special_tokens conflict and Qwen3Moe FineGrainedFP8Config failure are unrelated." + "reason": "Tokenizer method-name conflict and Qwen3Moe FP8 config failure are unrelated." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Molmo tied-weights metadata error is unrelated to loading cjvt/sleng-bert." + "reason": "Molmo tied-weights attribute error is unrelated to BERT model-loading failures." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "AutoConfig trust_remote_code field loss is unrelated to Qwen3.5 PagedAttentionCache group-type crash." + "reason": "AutoConfig field loss with `trust_remote_code` is unrelated to PagedAttentionCache linear_attention handling." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "TrainerState token-count tracking and Qwen2.5-VL multi-image IterableDataset failure are unrelated." + "reason": "Trainer token-count tracking and multimodal IterableDataset loading are different bugs." }, { "left": "issue:42175", "right": "issue:46178", "accept": false, - "reason": "Both involve backend handling, but one is install extras coverage and the other is a missing BACKENDS_MAPPING entry for tensorflow_text." + "reason": "Package-extras missing TensorFlow is not the same as an unsupported backend mapping entry for `tensorflow_text`." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both are Whisper-related, but batch_decode skip_special_tokens handling and empty-transcription generation after align_special_tokens are different bugs." + "reason": "Both Whisper-related, but one is decode flag handling and the other is empty-transcription generation after alignment." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Mixtral grouped_mm tracing failure and template-node compilation failure are unrelated." + "reason": "Grouped_mm/dynamo tracing failure and non-template-node compilation error do not share a code path." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Different model-loading failures for different artifacts; not the same underlying bug." - }, - { - "left": "issue:42915", - "right": "issue:44530", - "accept": false, - "reason": "Qwen3Moe FP8 config failure and Qwen3.5 linear_attention group-type crash are distinct model/runtime issues." + "reason": "Different model-loading regressions with no clear shared failure mechanism." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Both are v5 initialization regressions, but one is non-persistent buffers and the other is custom model initialization; separate concrete bugs." + "reason": "Non-persistent buffer junk and custom model initialization breakage are separate regression types." }, { "left": "issue:44464", "right": "issue:45987", "accept": false, - "reason": "Chunked generation with compiled forward and StaticCache sequence-length return type are unrelated." + "reason": "Chunked generation inconsistency and cache sequence-length return type are related to generation, but not the same bug." }, { "left": "issue:43452", "right": "issue:43482", - "accept": false, - "reason": "gguf_file path breakage and Qwen2.5-GGUF loading failure are related to GGUF loading broadly, but not the same concrete failure." + "accept": true, + "reason": "43482 looks like a specific instance of the broader GGUF loading breakage described by 43452." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "Tokenizer method-name conflict and Jupyter custom-model initialization regression are unrelated." + "reason": "Tokenizer special-token conflict and Jupyter custom-model initialization are unrelated." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "Dequantized-model device_map offload failure and Qwen2.5-GGUF loading failure are different loading paths." + "reason": "Device-map/offload dequantization failure is not the same as GGUF loading in transformers v5." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Different Qwen model families and different bugs: hidden-states output vs rope index scaling." + "reason": "Different Qwen variants and different bugs: hidden-states output vs. RoPE position scaling." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Both are transformers>=5 loading regressions, but they affect different models and failure modes." - } - ] - }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, - "error_kind": null, - "error_message": null - }, - { - "label": "issue", - "component_index": 1, - "component_count": 17, - "review_unit_index": 9, - "review_unit_count": 50, - "status": "reviewed", + "reason": "Both are v5 loading regressions, but they affect different artifacts and likely different loaders." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Template compilation error and GraniteMoEHybrid invalid method call are unrelated." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Molmo attribute error is unrelated to tokenizer loading for est-roberta." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Sampling-rate defaulting and MIMI padding-mask behavior are different audio issues." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Notebook custom-model initialization and junk-filled non-persistent buffers are distinct regressions." + }, + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Generation performance overhead and cache length return type are not the same underlying defect." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Custom model initialization and buffer initialization junk are separate bugs." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Llava Next load-time segfault and Qwen3OmniMoe video-processing ValueError are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about duplicates. The GGUF umbrella choice (issue:43452 over issue:43482) is reasonable, and the one accepted soft pair (issue:44206 / issue:44220) is plausibly the same audio-path bug. Rejections are generally cautious 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": 17, + "review_unit_index": 9, + "review_unit_count": 50, + "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ + "issue:33666", "issue:36032", "issue:40444", + "issue:40990", "issue:41553", "issue:41720", "issue:42915", - "issue:43089", - "issue:43262", "issue:43296", - "issue:43322", "issue:43377", "issue:43404", + "issue:43421", "issue:43425", "issue:43531", - "issue:43540", "issue:43582", "issue:43611", "issue:43632", - "issue:43645", - "issue:43646", "issue:43742", + "issue:43792", "issue:43819", - "issue:43883", "issue:43994", "issue:44062", + "issue:44079", "issue:44220", "issue:44265", "issue:44291", @@ -2958,31 +3006,26 @@ "issue:44466", "issue:44479", "issue:44534", + "issue:44568", "issue:44589", "issue:44610", "issue:44683", "issue:44756", + "issue:44811", "issue:44877", "issue:44987", "issue:44991", + "issue:45020", "issue:45081", "issue:45084", "issue:45127", "issue:45216", "issue:45325", "issue:45357", - "issue:45507", - "issue:45987" + "issue:45663" ], "soft_pairs": [ - "issue:45084|issue:45507", - "issue:43883|issue:44991", - "issue:43262|issue:43377", - "issue:43645|issue:44534", - "issue:43089|issue:45987", - "issue:43646|issue:44534", "issue:43404|issue:44466", - "issue:43322|issue:43540", "issue:44403|issue:45081", "issue:43994|issue:44610", "issue:42915|issue:45084", @@ -3006,235 +3049,245 @@ "issue:43611|issue:44534", "issue:43425|issue:44403", "issue:43632|issue:44534", - "issue:41553|issue:43582" + "issue:41553|issue:43582", + "issue:40990|issue:44811", + "issue:43792|issue:44589", + "issue:44079|issue:45663", + "issue:44683|issue:45020", + "issue:33666|issue:40444", + "issue:44403|issue:44991", + "issue:43421|issue:44568" ], - "prepared_review_unit_hash": "19c3dc1c7459266af054a56bea1389f7f7e801d8f86169a4d197eeefd109a3b9", + "prepared_review_unit_hash": "4aa1a8a4d0f46af3b4fe3ef17308df78d019b800fc7c3ffbca62cb972385f004", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 32, - "serialized_chars": 36673, - "estimated_input_tokens": 9169, - "estimated_eval_tokens": 18594 + "serialized_chars": 36728, + "estimated_input_tokens": 9182, + "estimated_eval_tokens": 18620 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is a loose collection of mostly unrelated Transformers bug reports, with a strong but not exact concentration around v5 model-loading/serialization regressions. None of the soft pairs look like the same concrete bug, so I rejected all of them.", - "confidence": 0.57, - "canonical_issue_reason": "Issue 43611 is the broadest, most central representative of the repeated v5 loading/initialization regressions in this set, even though the cluster overall is heterogeneous.", + "summary": "Mostly a loose cluster of unrelated transformer issues; only the Qwen3.5 save_pretrained regressions look like true duplicates, while most other pairs are merely same-subsystem or same-model-family similarities.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45216 is the broader original report of the Qwen3.5 save_pretrained regression; issue:45357 reads like a narrower follow-up about the same incorrect checkpoint serialization, specifically the visual encoder keys.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43611 is the best issue anchor because it describes a general v5 compatibility break rather than a single model-specific symptom.", + "best_issue_reason": "issue:45216 is the best representative issue here because it describes a clear regression with broad user impact and the other close match (45357) is a narrower manifestation of the same save_pretrained problem.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45084", - "right": "issue:45507", - "accept": false, - "reason": "Different failures: a template compilation error versus an invalid method call in GraniteMoEHybrid. Same framework area, not the same bug." - }, - { - "left": "issue:43883", - "right": "issue:44991", - "accept": false, - "reason": "Both are model-loading issues, but one is a missing attribute on Molmo and the other is a tokenizer loading failure for a specific model. No concrete shared code path." - }, - { - "left": "issue:43262", - "right": "issue:43377", - "accept": false, - "reason": "Audio chat-template sampling-rate default and MIMI batched-vs-single output mismatch are unrelated defects." - }, - { - "left": "issue:43645", - "right": "issue:44534", - "accept": false, - "reason": "Notebook custom model initialization breakage and non-persistent buffer corruption are different v5 regressions." - }, - { - "left": "issue:43089", - "right": "issue:45987", - "accept": false, - "reason": "Generation performance overhead and a cache API type-contract bug are unrelated." - }, - { - "left": "issue:43646", - "right": "issue:44534", - "accept": false, - "reason": "Custom model initialization failure versus non-persistent buffer junk values; not the same underlying bug." - }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Both involve tied weights/serialization, but one is a missing tie in Mistral3 conditional generation and the other is inconsistent lm_head serialization by device. Similar theme, different defect." - }, - { - "left": "issue:43322", - "right": "issue:43540", - "accept": false, - "reason": "Segfault on Llava Next loading and a Qwen3OmniMoe video-processing ValueError are distinct multimodal issues." + "reason": "Both mention lm_head/tied weights, but one is a missing tie in Mistral3ForConditionalGeneration and the other is device-dependent save_pretrained serialization; different bugs." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "Noise while loading transformers and a Mistral regex patch crash are not the same failure mode." + "reason": "Unrelated: noisy transformer loading behavior vs a tokenizer crash in Mistral regex patching." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Different model families and different symptoms: bad AutoModel/pipeline outputs for SigLIP2 versus incorrect processor image size for OmDet-Turbo." + "reason": "Different vision bugs and code paths: SigLIP2 bad outputs vs OmDet-Turbo processor/model input-size mismatch." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "FP8 config failure for Qwen3Moe and template-node compilation error are unrelated." + "reason": "Unrelated FP8 config failure vs a template-compilation TypeError." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "MIMI padding-mask batching mismatch and DAC STE mismatch are different audio-model bugs." + "reason": "Different models and failure modes: batched-vs-single padding-mask mismatch vs DAC from_latents / STE mismatch." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export with torch_compilable_check and OmDet-Turbo image-size mismatch are unrelated." + "reason": "Torch export failure with torch_compilable_check is not the same bug as the OmDet-Turbo image-size mismatch." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "ScandiBERT loading failure and Mistral regex patch crash are different tokenizer/model-loading issues." + "reason": "ScandiBERT loading failure is unrelated to the Mistral tokenizer regex crash." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Physical-Intelligence/fast loading regression and Mistral regex patch crash are not the same code-path problem." + "reason": "Different loading failures on different models; no evidence of the same underlying defect." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "_torch_extract_fbank_features() and compiled flex_attention on torch>=2.9 concern different subsystems." + "reason": "Audio feature extraction bug vs compiled flex_attention failure on newer torch." }, { "left": "issue:42915", "right": "issue:44291", "accept": false, - "reason": "Qwen3Moe FP8 failure and init_empty_weights/_is_hf_initialized argument mismatch are different initialization bugs." + "reason": "FineGrainedFP8Config failure is distinct from the init_empty_weights/_is_hf_initialized argument error." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "_is_hf_initialized breakage and mmap OOM avoidance on Strix Halo are unrelated." + "reason": "Different regressions: v5 _is_hf_initialized behavior vs mmap OOM on Strix Halo." }, { "left": "issue:45216", "right": "issue:45357", - "accept": false, - "reason": "Both are Qwen3.5 save_pretrained regressions, but one is generic checkpoint correctness and the other is specifically incorrect visual encoder keys. Similar area, not clearly one bug." + "accept": true, + "reason": "Both describe the same Qwen3.5 save_pretrained regression; 45357 is a narrower reproduction focused on incorrect visual encoder keys." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "Tokenizer method-name conflict and MobileLLM key error are different loading failures." + "reason": "T5 tokenizer special-token method conflict is unrelated to the MobileLLM key error." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "AddedToken special-flag duplication and physical-intelligence/fast loading failure are unrelated tokenizer/model issues." + "reason": "AddedToken special-argument error is not the same as loading physical-intelligence/fast." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "FP8 config failure and missing Float8 storage type are different Torch/version compatibility problems." + "reason": "Different FP8-related failures with different symptoms and code paths." }, { "left": "issue:44479", "right": "issue:45325", "accept": false, - "reason": "Video-input regression across several Qwen models and still-image rope-index scaling in Qwen2.5-VL are different multimodal bugs." + "reason": "Related Qwen-VL family, but 44479 is a broad video-input regression across multiple models while 45325 is a specific still-image temporal-position scaling bug; not safe to merge as duplicates." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL loading failure in vLLM and Qwen3-MoE sliding_window behavior are unrelated." + "reason": "PaddleOCR-VL loading failure is unrelated to the Qwen3-MoE sliding_window issue." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Both touch tied embeddings, but one is missing lm_head tying and the other is LoRA merge collapse with extended vocabulary. Not the same concrete bug." + "reason": "Both involve tied weights, but one is missing tying in Mistral3 while the other is LoRA-merge corruption with extended vocabulary." }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Qwen3-MoE sliding_window and granite_speech strict-config loading are unrelated." + "reason": "Qwen3-MoE sliding_window bug and granite_speech strict-config loading are unrelated." }, { "left": "issue:36032", "right": "issue:41720", "accept": false, - "reason": "T5 tokenizer add_special_tokens conflict and Qwen3 auto device mapping cuda assert are unrelated." + "reason": "T5 tokenizer load error vs Qwen3 auto-device-mapping CUDA assert; different failures." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Multiple-images IterableDataset failure and Qwen2.5-VL temporal position scaling are different multimodal training/inference bugs." + "reason": "Both are Qwen2.5-VL-related, but multi-image iterable-dataset finetuning failure is not the same as temporal position_id scaling." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "Both are v5 regressions, but one is a base_model_prefix loading break and the other is junk-filled non-persistent buffers. Same release, different defect." + "reason": "base_model_prefix loading break and non-persistent buffer junk are separate v5 regressions." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch 2.10 incompatibility and noisy loading logs are unrelated." + "reason": "Torch version incompatibility is unrelated to loading noise." }, { "left": "issue:43632", "right": "issue:44534", "accept": false, - "reason": "_is_hf_initialized flag handling and non-persistent buffer corruption are different v5 breakages." + "reason": "Different v5 breakages: _is_hf_initialized flag handling vs non-persistent buffer serialization." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Bad AutoTokenizer error for Voxtral and AppleSilicon caching_allocator_warmup TypeError are unrelated." + "reason": "Bad AutoTokenizer error message and AppleSilicon caching_allocator_warmup TypeError are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity/evaluation issue vs Whisper batch_decode skip_special_tokens bug." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper runtime failure vs missing Float8 storage object are unrelated." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment bug and Gemma-4 FSDP2 shared_kv_states KeyError are different code paths." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Compiled flex_attention torch>=2.9 failure is distinct from remote_code loading regressions." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Different Qwen-VL training issues: multi-GPU training vs iterable-dataset multi-image finetuning." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Transformer loading noise is not the same as the est-roberta tokenizer loading failure." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime special-token post-processing sync and a specific add_special_tokens BOS/EOS regression are related in area but not clearly the same defect." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted near-duplicate pair, 45216/45357, is well supported by their titles and timelines as the same Qwen3.5 save_pretrained regression with a narrower follow-up. The remaining pairwise rejections are appropriately cautious and do not overstate duplicate evidence." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3246,22 +3299,21 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ - "issue:33666", "issue:35707", "issue:40444", - "issue:40990", "issue:41720", "issue:42175", + "issue:42915", "issue:43296", "issue:43317", "issue:43377", "issue:43381", - "issue:43421", + "issue:43454", "issue:43482", "issue:43525", "issue:43531", @@ -3269,24 +3321,23 @@ "issue:43632", "issue:43644", "issue:43646", + "issue:43701", "issue:43749", "issue:43792", "issue:43818", "issue:43856", "issue:43957", + "issue:43994", "issue:44079", "issue:44265", "issue:44291", - "issue:44403", "issue:44462", + "issue:44466", + "issue:44479", "issue:44534", - "issue:44568", "issue:44589", - "issue:44683", - "issue:44811", "issue:44829", - "issue:44991", - "issue:45020", + "issue:44898", "issue:45072", "issue:45081", "issue:45325", @@ -3296,18 +3347,13 @@ "issue:45491", "issue:45538", "issue:45542", - "issue:45663", + "issue:45800", "issue:45812", + "issue:45823", + "issue:45834", "issue:45923" ], "soft_pairs": [ - "issue:40990|issue:44811", - "issue:43792|issue:44589", - "issue:44079|issue:45663", - "issue:44683|issue:45020", - "issue:33666|issue:40444", - "issue:44403|issue:44991", - "issue:43421|issue:44568", "issue:43381|issue:45923", "issue:35707|issue:40444", "issue:44462|issue:45812", @@ -3315,8 +3361,8 @@ "issue:43482|issue:45538", "issue:42175|issue:45542", "issue:44829|issue:45479", - "issue:43317|issue:43856", "issue:43482|issue:43525", + "issue:43317|issue:43856", "issue:44079|issue:44534", "issue:43818|issue:45357", "issue:43531|issue:44291", @@ -3330,228 +3376,229 @@ "issue:43632|issue:45072", "issue:45325|issue:45381", "issue:43611|issue:43646", - "issue:43792|issue:43957" + "issue:43792|issue:43957", + "issue:45325|issue:45834", + "issue:43994|issue:44898", + "issue:43454|issue:44466", + "issue:42915|issue:43482", + "issue:43296|issue:45800", + "issue:44479|issue:45823", + "issue:43701|issue:45325" ], - "prepared_review_unit_hash": "da05d36dd29e3de7b5e8e2e0561e8b33d3961116e68242e06bbbeffeefdbe930", + "prepared_review_unit_hash": "f61456958e42e5421c3b6147c99a38cf4b2d6ab0b1d758593201d1e0c3c4244d", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36899, - "estimated_input_tokens": 9225, - "estimated_eval_tokens": 18706 + "serialized_chars": 36891, + "estimated_input_tokens": 9223, + "estimated_eval_tokens": 18702 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of distinct Transformers/v5 regressions and model-specific loading/training bugs, not a tight duplicate set. Only one soft pair looks like the same underlying Qwen2.5-VL position-id bug.", + "summary": "This cluster is mostly a grab bag of distinct Transformers regressions, not one cohesive duplicate set. The only strong soft-duplicate here is the Qwen2.5-VL position-id regression pair.", "confidence": 0.67, - "canonical_issue_reason": "issue:45020 is the broadest umbrella in the set ('Recent transformers versions break models using `remote_code`') and best represents the overall v5 regression theme across multiple model-loading failures.", + "canonical_issue_reason": "issue:45325 is the clearest duplicate anchor: it states the concrete Qwen2.5-VL rope/position-id bug, and issue:45381 appears to be the same regression described from the video-input side.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the most suitable cluster representative because it is the broadest, most reusable description of the loading/regression class, even though many other items are more specific manifestations.", + "best_issue_reason": "issue:43611 is the best overall representative of the set because it captures a broad, high-signal Transformers v5 model-loading regression with clear generic scope.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:40990", - "right": "issue:44811", - "accept": false, - "reason": "Different models and failures: perplexity on GPT-OSS vs Whisper batch_decode special-token handling." - }, - { - "left": "issue:43792", - "right": "issue:44589", - "accept": false, - "reason": "Unrelated symptoms: Whisper runtime failure vs missing Float8 storage type." - }, - { - "left": "issue:44079", - "right": "issue:45663", - "accept": false, - "reason": "Different code paths: ModelOutput key assignment vs Gemma4 attention shared_kv_states rebuild." - }, - { - "left": "issue:44683", - "right": "issue:45020", - "accept": false, - "reason": "Different layers: torch>=2.9 flex_attention compilation vs remote_code model loading regressions." - }, - { - "left": "issue:33666", - "right": "issue:40444", - "accept": false, - "reason": "Both are Qwen2-VL/Qwen2.5-VL related, but one is multi-GPU training and the other is iterable dataset handling with multiple images per prompt." - }, - { - "left": "issue:44403", - "right": "issue:44991", - "accept": false, - "reason": "One is generic loading noise, the other is a specific tokenizer loading regression for EMBEDDIA/est-roberta." - }, - { - "left": "issue:43421", - "right": "issue:44568", - "accept": false, - "reason": "Both involve tokenizers, but one is runtime post-processor updates and the other is add_special_tokens/BOS-EOS behavior for a specific tokenizer." - }, { "left": "issue:43381", "right": "issue:45923", "accept": false, - "reason": "Eval-mode gradient checkpointing error is not the same as a missing supports_gradient_checkpointing flag on a remote-code model." + "reason": "Different gradient-checkpointing problems: eval-mode misuse vs missing supports_gradient_checkpointing on a trust_remote_code model." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL iterable-dataset multi-image finetuning." + "reason": "Unrelated failures: generation with inputs_embeds/past_key_values vs Qwen2.5-VL multi-image IterableDataset finetuning." }, { "left": "issue:44462", "right": "issue:45812", "accept": false, - "reason": "Both are AutoTokenizer regressions, but one is ignoring tokenizer.json while the other is wrong token IDs for Granite models; too different to merge confidently." + "reason": "Both involve tokenization, but one is repo tokenizer loading and the other is wrong Granite token IDs; different bugs." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "init_empty_weights/_is_hf_initialized argument breakage is separate from the Float8 storage lookup failure." + "reason": "Different load-path regressions: unexpected _is_hf_initialized arg vs Float8 storage lookup failure." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "GGUF loading in Transformers v5 is unrelated to CLIPTokenizer max_length semantics." + "reason": "Different tokenization/model issues: GGUF loading failure vs CLIPTokenizer max_length behavior." }, { "left": "issue:42175", "right": "issue:45542", "accept": false, - "reason": "Both mention TF/backend packaging, but one is missing TensorFlow in a torch extra and the other is a tensorboard-only undefined backend error." + "reason": "Both mention TensorFlow backend packaging, but the concrete failures differ enough that they do not look like the same bug." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "FlashAttention-3 training degeneration is not the same bug as single-label classification with num_labels=1 yielding zero loss." + "reason": "Different training-path bugs: flash_attention_3 degeneracy vs single-label classification with num_labels=1 zero loss." }, { - "left": "issue:43317", - "right": "issue:43856", + "left": "issue:43482", + "right": "issue:43525", "accept": false, - "reason": "Device-map/offload loading failure is distinct from Qwen3 MoE memory inefficiency during training." + "reason": "Different failure modes: Qwen2.5-GGUF loading vs missing pad_token_id on Llama4Config." }, { - "left": "issue:43482", - "right": "issue:43525", + "left": "issue:43317", + "right": "issue:43856", "accept": false, - "reason": "Different model-loading regressions: GGUF/Transformers v5 vs Llama4Config missing pad_token_id." + "reason": "Different problems: device_map/offload loading failure vs Qwen3 MoE training memory inefficiency." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "ModelOutput key bookkeeping is unrelated to non-persistent buffer initialization." + "reason": "Different internals: ModelOutput key assignment bug vs non-persistent buffer junk fill." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Video-LLaVA architecture/weights issue is separate from Qwen3.5 save_pretrained visual-encoder key regression." + "reason": "Different model families and symptoms: Video-LLaVA weight/tower issue vs Qwen3.5 visual-key save_pretrained regression." }, { "left": "issue:43531", "right": "issue:44291", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to init_empty_weights argument handling." + "reason": "Sliding-window Qwen3-MoE bug is unrelated to init_empty_weights/_is_hf_initialized loading regression." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "Qwen3 auto device-map CUDA assert is unrelated to Mistral tokenizer regex patch crashing." + "reason": "Different code paths entirely: auto device-mapping CUDA assert vs tokenizer regex patch crash." }, { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "MIMI padding-mask mismatch and Gemma3 mixed-length NaNs are both batching issues but appear to be different mechanisms and models." + "reason": "Both are batching/masking issues, but the models and concrete failure modes differ." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "torch.export/compilable_check failure is not the same as bfloat16 dtype mismatches in inference." + "reason": "torch.export/torch_compilable_check export failure is unrelated to bfloat16 dtype mismatches." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "Both are multimodal Qwen2.5-VL issues, but one is dataset finetuning with multiple images and the other is video position-id calculation." + "reason": "Both are Qwen2.5-VL, but one is multi-image finetuning input handling and the other is video position-id regression." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Non-persistent buffer junk and ModelOutput key assignment are different object-state bugs." + "reason": "Different regressions: non-persistent buffer initialization vs ModelOutput key bookkeeping." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM efficient loading regression is unrelated to ModelOutput field handling." + "reason": "FSDP CPU RAM-efficient loading and ModelOutput key assignment are unrelated." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "Packaging/backend omission in a torch install extra is not the same as PaddleOCR-VL loading failure in vLLM." + "reason": "Packaging/backend install issue vs vLLM model-load failure; not the same bug." }, { "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "_is_hf_initialized regression and bfloat16 dtype mismatch are different Transformers v5 issues." + "reason": "_is_hf_initialized breakage and bf16 dtype mismatches are separate Transformer v5 regressions." }, { "left": "issue:45325", "right": "issue:45381", "accept": true, - "reason": "Same Qwen2.5-VL position-id/rope-index code path; one reports still-image temporal scaling, the other video input vision_position_ids being wrong." + "reason": "Same Qwen2.5-VL positional-id/rope-index regression in v5.3.0; one report is image/still-image side, the other is video-input side." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both are Transformers v5 init/load regressions, but one is base_model_prefix loading and the other is custom model initialization; not clearly the same bug." + "reason": "Related v5 model-init/loading regressions, but the titles point to different concrete failure modes." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Whisper model runtime failure is not the same as meta-device loading breakage affecting other models." + "reason": "Different loading/runtime issues: Whisper model failure vs meta-device loading breakage." }, { - "left": "issue:45020", - "right": "issue:45381", + "left": "issue:45325", + "right": "issue:45834", "accept": false, - "reason": "Remote-code loading regressions and Qwen2.5-VL video position IDs are unrelated." - } + "reason": "Qwen2.5-VL position-id regression is unrelated to Kosmos2.5 OCR index errors." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model failures: SigLIP2 bad outputs vs Perceiver non-default-resolution interpolation failure." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Both mention lm_head weights, but one is untied-weight generation and the other is device-dependent serialization." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different model/config paths: Qwen3MoE FP8 config vs Qwen2.5-GGUF loading." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "PaddleOCR-VL load failure and torch/transformers version incompatibility are not the same bug." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Different Qwen/Gemma regressions: video-input regression vs Gemma4 device-map mismatch." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint-resume key mismatch is unrelated to Qwen2.5-VL rope-index position IDs." + } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative on the soft edges. The cluster is indeed heterogeneous, and the lone accepted Qwen2.5-VL position-id pair is the only clearly mergeable one. The choice of issue:45020 as a broad umbrella is defensible given the repeated v5/remote-code loading regressions in the set." + "feedback": "The summary is grounded in the packet and stays appropriately conservative overall. The only accepted soft-duplicate pair (issue:45325 / issue:45381) is supported by the titles, and the rejected pairs are described cautiously rather than overstated." }, "evaluator_used": true, "retried": false, @@ -3567,8 +3614,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -3578,6 +3625,7 @@ "issue:40444", "issue:40990", "issue:41720", + "issue:42617", "issue:42915", "issue:42947", "issue:43257", @@ -3587,48 +3635,40 @@ "issue:43322", "issue:43377", "issue:43454", - "issue:43482", "issue:43526", "issue:43550", "issue:43572", "issue:43606", - "issue:43701", + "issue:43720", "issue:43746", + "issue:43749", "issue:43819", "issue:43873", "issue:43881", "issue:43883", - "issue:43994", "issue:44186", "issue:44265", "issue:44291", "issue:44466", - "issue:44479", "issue:44488", + "issue:44534", "issue:44568", "issue:44610", "issue:44877", "issue:44898", + "issue:44977", + "issue:44991", "issue:45072", - "issue:45325", + "issue:45357", "issue:45584", "issue:45701", "issue:45759", - "issue:45800", "issue:45823", - "issue:45834", "issue:45920", "issue:45923", "issue:46050" ], "soft_pairs": [ - "issue:45325|issue:45834", - "issue:43994|issue:44898", - "issue:43454|issue:44466", - "issue:42915|issue:43482", - "issue:43296|issue:45800", - "issue:44479|issue:45823", - "issue:43701|issue:45325", "issue:44291|issue:44877", "issue:43881|issue:44488", "issue:43606|issue:44265", @@ -3643,224 +3683,245 @@ "issue:41720|issue:45823", "issue:43454|issue:43550", "issue:39692|issue:40444", - "issue:43377|issue:44610", "issue:43454|issue:45072", + "issue:43377|issue:44610", "issue:43572|issue:45759", "issue:43257|issue:45923", + "issue:40990|issue:42947", "issue:43377|issue:44898", "issue:43317|issue:43873", - "issue:40990|issue:42947", "issue:43526|issue:43883", - "issue:43322|issue:46050" + "issue:43322|issue:46050", + "issue:43881|issue:44991", + "issue:43454|issue:43720", + "issue:41720|issue:43322", + "issue:43296|issue:44991", + "issue:43749|issue:44534", + "issue:44466|issue:45357", + "issue:41720|issue:44977", + "issue:42617|issue:42915", + "issue:40990|issue:45701" ], - "prepared_review_unit_hash": "ad4a78dd2a4e8475ab64e94a47d8bd6eec8dace9b4f2f9a3a8c5551a308b5e99", + "prepared_review_unit_hash": "cc5859e416b97e2dde55f46e1e569a2d8bc80dc38b708a0914e7f42d38b83f98", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36827, - "estimated_input_tokens": 9207, - "estimated_eval_tokens": 18670 + "soft_pair_count": 32, + "serialized_chars": 37412, + "estimated_input_tokens": 9353, + "estimated_eval_tokens": 18962 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a set of semantically similar but non-duplicate bug reports spanning different models and code paths. The soft pairs share broad themes like loading, offloading, quantization, or tied weights, but they are not the same underlying bug/change, so none should be merged.", + "summary": "This cluster is mostly a collection of unrelated bug reports, with only a loose subtheme around loading/offloading/quantization and device-mapping regressions. None of the soft pairs look like safe duplicates or mergeable PR equivalents.", "confidence": 0.95, - "canonical_issue_reason": "issue:43873 is the broadest umbrella candidate (offloading + quantization) among the items, but the cluster is too heterogeneous for a true canonical duplicate.", + "canonical_issue_reason": "If forced to pick an anchor, issue 43873 is the broadest representative of the offloading/quantization subtheme, but the set is not a true duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the best cluster representative because it is the most general and central of the few offloading/quantization reports, though it still does not subsume the rest.", + "best_issue_reason": "issue 43873 is the most central and broadly phrased report in the loading/offloading/quantization area, so it is the best representative issue for the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45325", - "right": "issue:45834", - "accept": false, - "reason": "Different models and different failures: Qwen2.5-VL rope-index scaling vs Kosmos2.5 OCR length/index error." - }, - { - "left": "issue:43994", - "right": "issue:44898", - "accept": false, - "reason": "Both are vision-related, but they describe different models and bugs: SigLIP2 inference mismatch vs Perceiver resolution/interpolation failure." - }, - { - "left": "issue:43454", - "right": "issue:44466", - "accept": false, - "reason": "Both mention tied weights/lm_head, but one is an AyaVision model bug and the other is a v5 serialization inconsistency; not the same code-path problem." - }, - { - "left": "issue:42915", - "right": "issue:43482", - "accept": false, - "reason": "Different quantization/loading issues on different model families (Qwen3Moe FP8 vs Qwen2.5-GGUF v5 load failure)." - }, - { - "left": "issue:43296", - "right": "issue:45800", - "accept": false, - "reason": "Unrelated compatibility reports: PaddleOCR-VL loading in vLLM vs a torch/transformers version incompatibility." - }, - { - "left": "issue:44479", - "right": "issue:45823", - "accept": false, - "reason": "Both involve multimodal/device behavior, but one is a video-input regression and the other is a Gemma4 PLE device mismatch." - }, - { - "left": "issue:43701", - "right": "issue:45325", - "accept": false, - "reason": "Checkpoint resume key mismatch is unrelated to Qwen2.5-VL rope-index scaling." - }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "Both are loading/config issues, but one is an init_empty_weights API break and the other is strict config validation for granite_speech." + "reason": "Different regressions: `_is_hf_initialized`/`init_empty_weights` type error vs strict config loading for `granite_speech`." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "Different models and symptoms: glm-4v-9b load failure vs cjvt/sleng-bert load failure." + "reason": "One is a device mismatch on model load, the other is a current-version tokenizer/model loading failure; different code paths and models." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "CPU offload device mismatch in Bark is unrelated to torch.export failures in torch_compilable_check models." + "reason": "Bark CPU offload device mismatch is unrelated to `torch.export` failures from `torch_compilable_check`." }, { "left": "issue:43873", "right": "issue:45701", "accept": false, - "reason": "Offloading/quantization behavior is unrelated to a tokenizer-version regression." + "reason": "Broad loading-related wording, but one is offloading/quantization and the other is tokenization behavior changing by version." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting is unrelated to GraniteSpeech PEFT checkpoint loading." + "reason": "Audio processor sampling-rate default bug vs GraniteSpeech adapter checkpoint loading; different failure modes." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "Different tasks and bugs: perplexity on gpt-oss-20b vs Whisper empty-transcription generation failure." + "reason": "High perplexity on one model is not the same as Whisper empty-transcription failure after token alignment." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "Bark voice_preset and progressive generation with inputs_embeds/past_key_values are different code paths." + "reason": "Voice preset support in BarkProcessor is unrelated to progressive generation with `inputs_embeds` and `past_key_values`." }, { "left": "issue:44568", "right": "issue:45920", "accept": false, - "reason": "A specific BOS/EOS tokenizer bug is not the same as the broader AutoTokenizer wrong-IDs regression." + "reason": "Tokenizer special-token regression is unrelated to wrong token IDs across multiple models." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "Progressive generation failure is unrelated to Qwen3 auto device mapping cudaErrorAssert." + "reason": "Generic generation bug with cached inputs vs Qwen3 auto device-map CUDA assert; not the same concrete bug." }, { "left": "issue:44186", "right": "issue:44265", "accept": false, - "reason": "Tokenizer crashes on NER/padding are unrelated to torch.export failures with torch_compilable_check." + "reason": "NER/batched padding crash in LayoutLMv2Tokenizer is unrelated to export failures caused by `torch_compilable_check`." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Both are model-specific runtime bugs, but Bark CPU offload mismatch and DAC.from_latents STE mismatch are different problems." + "reason": "Both are model math/runtime issues, but Bark CPU offload mismatch and DAC forward/from_latents mismatch are distinct." }, { "left": "issue:41720", "right": "issue:45823", "accept": false, - "reason": "Both involve device_map auto, but the model families and failure modes differ (cudaErrorAssert vs forward device mismatch)." + "reason": "Qwen3 auto device-mapping assert vs Gemma4 PLE device mismatch; similar theme, different model and root cause." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "AyaVision tied-weights issue is unrelated to Bamba torch.compile/SDPA failure." + "reason": "AyaVision weight-tying bug is unrelated to Bamba torch.compile/SDPA failure." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "Documentation/example errors for SigLIP2 are unrelated to Qwen2.5-VL multi-image finetuning failure." + "reason": "SigLIP2 documentation errors are unrelated to Qwen2.5-VL finetuning with iterable multi-image datasets." }, { - "left": "issue:43377", - "right": "issue:44610", + "left": "issue:43454", + "right": "issue:45072", "accept": false, - "reason": "MIMI padding-mask batch mismatch is unrelated to OmDet-Turbo image-size mismatch." + "reason": "Weight tying/garbage text generation is not the same as dtype mismatch in SwitchTransformers and TimmWrapperModel." }, { - "left": "issue:43454", - "right": "issue:45072", + "left": "issue:43377", + "right": "issue:44610", "accept": false, - "reason": "Different subsystems: tied lm_head weights vs dtype mismatch in bfloat16 inference." + "reason": "MIMI batching/padding-mask issue vs OmDet-Turbo processor/model input-size mismatch; different components." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig is unrelated to composite Qwen config unwrapping." + "reason": "Missing `pad_token_idx` in StableLmConfig is unrelated to failing to unwrap `text_config` for Qwen multimodal configs." }, { "left": "issue:43257", "right": "issue:45923", "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to a missing supports_gradient_checkpointing flag." + "reason": "Qwen3 MOE loading with accelerate+deepspeed is unrelated to a missing gradient-checkpointing capability flag." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity/evaluation anomaly is not the same bug as ineffective gradient checkpointing with PEFT LoRA." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "MIMI encoder padding-mask issue is unrelated to Perceiver non-default-resolution interpolation behavior." + "reason": "Different model families and problems: MIMI padding-mask mismatch vs Perceiver non-default-resolution interpolation failure." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Both mention offload/quantization, but one is dequantized-model loading on gpu+cpu offload and the other is a broader offloading regression." - }, - { - "left": "issue:40990", - "right": "issue:42947", - "accept": false, - "reason": "High perplexity on a benchmark is unrelated to gradient checkpointing not taking effect with PEFT LoRA." + "reason": "Both mention offloading/loading, but one is dequantized model placement on gpu+cpu offload and the other is quantization/offloading behavior more generally." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "BeitImageProcessorFast label reduction is unrelated to Molmo missing all_tied_weights_keys." + "reason": "BEiT `reduce_labels` returning one label is unrelated to Molmo missing `all_tied_weights_keys`." }, { "left": "issue:43322", "right": "issue:46050", "accept": false, - "reason": "Llava Next loading segfault and sinq save/load quantization breakage are distinct failure modes." + "reason": "Llava Next load segfault and sinq save/load breakage affect different stages and mechanisms." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "PaddleOCR-VL load failure is not the same as tokenizer loading regression for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "AyaVision weight tying is unrelated to BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3 device-map CUDA assert and Llava Next load segfault are different code paths and failure signatures." + }, + { + "left": "issue:43296", + "right": "issue:44991", + "accept": false, + "reason": "PaddleOCR-VL vLLM loading failure is unrelated to tokenizer loading regression." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and v5 junk-filled non-persistent buffers are both loading-related but not the same bug." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Inconsistent lm_head serialization by device is different from incorrect visual encoder keys in `save_pretrained`." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3 auto device-mapping assert is unrelated to Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "`3d_parallel.py` runtime failure and Qwen3MoE FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity regression and version-dependent tokenization changes are different symptoms and likely different subsystems." } ] }, @@ -3879,50 +3940,50 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ "issue:36010", "issue:36331", - "issue:40990", "issue:41720", "issue:42617", "issue:42915", "issue:43295", - "issue:43296", + "issue:43316", "issue:43317", "issue:43322", "issue:43404", "issue:43450", - "issue:43454", "issue:43526", + "issue:43531", + "issue:43540", "issue:43550", + "issue:43572", "issue:43577", "issue:43606", "issue:43708", "issue:43720", - "issue:43749", + "issue:43782", "issue:43873", "issue:43881", "issue:44008", "issue:44186", "issue:44387", "issue:44423", - "issue:44466", "issue:44479", - "issue:44534", "issue:44610", "issue:44671", "issue:44743", "issue:44756", "issue:44877", "issue:44898", + "issue:44913", "issue:44918", - "issue:44977", "issue:44991", "issue:45072", + "issue:45081", "issue:45161", "issue:45357", "issue:45406", @@ -3930,23 +3991,15 @@ "issue:45684", "issue:45701", "issue:45704", + "issue:45812", "issue:45823" ], "soft_pairs": [ - "issue:43881|issue:44991", - "issue:43454|issue:43720", - "issue:41720|issue:43322", - "issue:43296|issue:44877", - "issue:43749|issue:44534", - "issue:44466|issue:45357", - "issue:41720|issue:44977", - "issue:42617|issue:42915", - "issue:40990|issue:45701", - "issue:44387|issue:45704", "issue:44008|issue:45823", + "issue:44387|issue:45704", "issue:36331|issue:42617", - "issue:44186|issue:45072", "issue:43577|issue:43708", + "issue:44186|issue:45072", "issue:44877|issue:44991", "issue:36331|issue:42915", "issue:43404|issue:45357", @@ -3955,238 +4008,225 @@ "issue:43873|issue:44756", "issue:44423|issue:45406", "issue:41720|issue:42617", - "issue:45632|issue:45684", "issue:44479|issue:44918", "issue:43550|issue:43606", "issue:44610|issue:44898", + "issue:45632|issue:45684", "issue:44186|issue:44610", "issue:42915|issue:43317", "issue:43295|issue:45701", "issue:43526|issue:43881", "issue:41720|issue:45161", "issue:36010|issue:42617", - "issue:44671|issue:44743" + "issue:44671|issue:44743", + "issue:45081|issue:45812", + "issue:43531|issue:43572", + "issue:43316|issue:44913", + "issue:43540|issue:43782", + "issue:42915|issue:43322" ], - "prepared_review_unit_hash": "d6844e82e0aa3a14bfac0af6512ac07876fde015f299440df4e1d88bb226d1ac", + "prepared_review_unit_hash": "f3e7ffb107f64cf1cabc9454ecdb91148fdd8cb0fa86a0b92b58ca99f2a62563", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 37121, - "estimated_input_tokens": 9281, - "estimated_eval_tokens": 18818 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36546, + "estimated_input_tokens": 9137, + "estimated_eval_tokens": 18530 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a set of unrelated Transformers issues with only superficial overlap around loading, offloading, and model-specific regressions. I did not find any soft pair that looks like the same underlying bug or change, so all soft edges are rejected.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 43873 is the best cluster anchor: it is a broad, currently open infrastructure bug about offloading/quantization with strong discussion and inbound-reference signals, while most other items are narrower model-specific regressions or API breakages.", + "summary": "All soft edges look like false positives: the paired issues touch different models, different APIs, or different failure modes, so none should be merged as duplicates. The best standalone issue candidate is the broad offloading/quantization report (43873), but it is not a duplicate of the others.", + "confidence": 0.92, + "canonical_issue_reason": "issue:43873 is the strongest canonical issue candidate because it is open, broadly framed, and centrally discussed around a recurring offloading/quantization failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43873 is the most representative and actionable issue in this set because it captures a broad class of loading/offloading failures and has the strongest evidence of ongoing impact.", + "best_issue_reason": "issue:43873 is the best issue to keep as a tracking reference: it is broad enough to represent a real subsystem pain point and has substantial discussion/inbound attention.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43881", - "right": "issue:44991", - "accept": false, - "reason": "Both are loading failures, but one is a GLM model load problem and the other is a tokenizer loading regression for a specific repo; no sign of the same bug." - }, - { - "left": "issue:43454", - "right": "issue:43720", - "accept": false, - "reason": "Different code paths: tied lm_head weights in a multimodal model vs packed-weight unpacking during accelerate loading." - }, - { - "left": "issue:41720", - "right": "issue:43322", - "accept": false, - "reason": "Both crash during model use/loading, but they involve different models and different failure modes (cuda assert vs segfault)." - }, - { - "left": "issue:43296", - "right": "issue:44877", - "accept": false, - "reason": "One is a vLLM/PaddleOCR-VL loading failure, the other is strict config validation blocking granite_speech; these are unrelated." - }, - { - "left": "issue:43749", - "right": "issue:44534", - "accept": false, - "reason": "Both touch loading/serialization, but one is FSDP CPU RAM-efficient loading and the other is non-persistent buffer initialization; not the same bug." - }, - { - "left": "issue:44466", - "right": "issue:45357", - "accept": false, - "reason": "Both are save_pretrained regressions, but one is tied lm_head serialization varying by device and the other is incorrect visual encoder keys for Qwen3.5." - }, - { - "left": "issue:41720", - "right": "issue:44977", - "accept": false, - "reason": "Auto device-map CUDA failure and flash-attention generation issues are different bugs in different execution paths." - }, - { - "left": "issue:42617", - "right": "issue:42915", - "accept": false, - "reason": "One is a script/runtime failure for 3d_parallel.py; the other is a model/config-specific FineGrainedFP8 failure." - }, - { - "left": "issue:40990", - "right": "issue:45701", + "left": "issue:44008", + "right": "issue:45823", "accept": false, - "reason": "Tokenizer/version-related reasoning is too broad here; one reports high perplexity on a model/dataset, the other a general tokenization change across versions." + "reason": "Different models and different bugs: Gemma3n forward variable collision vs Gemma4 device-map forward mismatch." }, { "left": "issue:44387", "right": "issue:45704", "accept": false, - "reason": "Both mention memory, but one is CUDA reserved-memory growth under int4 quantization and the other is an apex FusedRMSNorm leak in T5." - }, - { - "left": "issue:44008", - "right": "issue:45823", - "accept": false, - "reason": "Different Gemma model families and different symptoms: variable name collision vs device-mapping mismatch." + "reason": "Both mention memory, but one is int4 reserved-memory/OOM and the other is a T5 apex RMSNorm leak; distinct causes and fixes." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "Trainer compute_loss signature breakage is unrelated to the 3d_parallel.py execution problem." + "reason": "Trainer compute_loss kwarg breakage is unrelated to a failure to run 3d_parallel.py." }, { - "left": "issue:44186", - "right": "issue:45072", + "left": "issue:43577", + "right": "issue:43708", "accept": false, - "reason": "Tokenizer crash on NER/batched padding is unrelated to bfloat16 dtype mismatches in inference." + "reason": "BLIP2 dtype loading bug and Trainer checkpoint max_steps bug are separate code paths." }, { - "left": "issue:43577", - "right": "issue:43708", + "left": "issue:44186", + "right": "issue:45072", "accept": false, - "reason": "Blip2 dtype loading bug and Trainer max_steps resume calculation are separate concerns." + "reason": "Tokenizer padding/truncation crash is unrelated to model dtype mismatches in inference." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config validation for granite_speech is not the same as tokenizer loading for EMBEDDIA/est-roberta." + "reason": "Config strictness blocking granite_speech loading is not the same bug as tokenizer loading failure for est-roberta." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Trainer API incompatibility and FineGrainedFP8 model failure do not share the same underlying bug." + "reason": "CustomTrainer API mismatch is unrelated to Qwen3MoE FP8 configuration failure." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "Weight-tying bug in Mistral3 and incorrect visual encoder keys in Qwen3.5 are different save/serialization defects." + "reason": "Mistral3 weight tying bug and Qwen3.5 save_pretrained visual key regression affect different models and save/load paths." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Video processor shape handling and tokenization-version changes are unrelated issues." + "reason": "Batched video processor shape bug is unrelated to the tokenization regression." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "torch.compile with SDPA on Bamba-9B-v2 is a different code path from packed-weight loading in BitNet." + "reason": "torch.compile+SDPA failure is a different problem from unpacking packed weights during accelerate loading." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Both involve memory/offload concerns, but one is quantization/offloading correctness and the other is a Strix Halo mmap OOM workaround." + "reason": "Both involve offloading/memory, but one is quantization/offload behavior and the other is a Strix Halo mmap OOM workaround." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Both crash in serve/multimodal scenarios, but they fail in different processor paths with different missing attributes." + "reason": "Serve crashes stem from different processor assumptions: string .to() misuse vs missing _tokenizer attribute." }, { - "left": "issue:45632", - "right": "issue:45684", + "left": "issue:41720", + "right": "issue:42617", "accept": false, - "reason": "Cache-path collisions for local trust_remote_code models and read-only permission propagation during save_pretrained are unrelated filesystem issues." + "reason": "Qwen3 auto device mapping cuda assert is unrelated to the 3d_parallel.py runtime failure." }, { "left": "issue:44479", "right": "issue:44918", "accept": false, - "reason": "A video-input regression across Qwen VL models is not the same as TRL SFT embedding unpacking failure." + "reason": "Video-input regression in Qwen VL models is not the same as Qwen3.5 embedding unpacking during TRL SFT training." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Different model families and distinct causes: compile/SDPA failure versus CPU-offload device mismatch." + "reason": "Different failures: SDPA/torch.compile issue vs CPU offload device mismatch for bark-small." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Both are image preprocessing/resolution bugs, but the affected models and mechanism differ enough that they do not look like one bug." + "reason": "Processor emitting wrong image size is not the same as a model refusing non-default resolution even with interpolation." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collision for trust_remote_code is unrelated to read-only permission propagation in save_pretrained." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "Tokenizer crash on NER inputs and processor output-size mismatch are separate issues." + "reason": "Tokenizer crash and image processor resolution bug are separate subsystems and symptoms." }, { "left": "issue:42915", "right": "issue:43317", "accept": false, - "reason": "FineGrainedFP8 model failure and dequantized model loading with cpu offload are not the same underlying defect." + "reason": "Qwen3MoE FP8 config failure and dequantized model offload load failure are different quantization/offload bugs." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "Processor/tokenizer regression in v4.57.5 is not the same as a broad tokenization-change report." + "reason": "Processor.tokenizer regression is not the same as a generic tokenization-change report." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "Label-reduction logic in BeitImageProcessorFast is unrelated to a GLM model loading failure." + "reason": "BeitImageProcessorFast label reduction bug is unrelated to glm-4v-9b loading failure." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Auto device-map CUDA assert and GPT-OSS MoE tensor-parallel failure are different distributed/inference problems." + "reason": "Auto device-map CUDA assert is not the same as GPT-OSS MoE tensor-parallel support failure." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to the 3d_parallel.py runtime failure." + "reason": "GenerationMixin import error has no overlap with 3d_parallel.py execution failure." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "CamemBERT masked-LM prediction regression and Qwen3.5 recurrent-state reset with cache are different model-specific bugs." + "reason": "CamemBERT masked-LM prediction regression and Qwen3.5 cache-state reset bug are different model behaviors." + }, + { + "left": "issue:45081", + "right": "issue:45812", + "accept": false, + "reason": "Mistral regex patch crash and Granite token-ID regression are both tokenizer-related but clearly different bugs and model families." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue and StableLmConfig pad_token_idx regression are unrelated." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy is not the same as GPTNeoX rotary_pct reload behavior." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Qwen3OmniMoe video-processing ValueError and Qwen3VL weight_only load error are distinct multimodal load paths." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3MoE FP8 config failure is unrelated to Llava Next segmentation fault on load." } ] }, @@ -4205,8 +4245,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -4220,16 +4260,11 @@ "issue:36331", "issue:41720", "issue:42490", - "issue:42915", - "issue:43316", - "issue:43322", "issue:43329", "issue:43388", "issue:43421", "issue:43531", - "issue:43540", "issue:43550", - "issue:43572", "issue:43606", "issue:43653", "issue:43720", @@ -4237,6 +4272,7 @@ "issue:43782", "issue:43873", "issue:43927", + "issue:44077", "issue:44186", "issue:44361", "issue:44451", @@ -4244,27 +4280,26 @@ "issue:44509", "issue:44514", "issue:44545", + "issue:44589", "issue:44625", "issue:44898", - "issue:44913", + "issue:44987", "issue:44991", + "issue:45042", "issue:45059", "issue:45072", - "issue:45081", + "issue:45381", "issue:45406", "issue:45464", "issue:45479", + "issue:45588", "issue:45701", "issue:45704", "issue:45759", - "issue:45812" + "issue:45812", + "issue:45859" ], "soft_pairs": [ - "issue:45081|issue:45812", - "issue:43531|issue:43572", - "issue:43316|issue:44913", - "issue:43540|issue:43782", - "issue:42915|issue:43322", "issue:44186|issue:44361", "issue:43421|issue:45812", "issue:36331|issue:41720", @@ -4278,230 +4313,239 @@ "issue:41720|issue:43746", "issue:43873|issue:45704", "issue:43421|issue:45701", - "issue:43606|issue:43653", "issue:43531|issue:45759", + "issue:43606|issue:43653", "issue:34634|issue:41720", "issue:42490|issue:43329", - "issue:44492|issue:44509", "issue:34567|issue:35707", + "issue:44492|issue:44509", "issue:44514|issue:44545", "issue:44625|issue:45479", - "issue:43421|issue:43927", "issue:44898|issue:45072", - "issue:44451|issue:44991" + "issue:43421|issue:43927", + "issue:44451|issue:44991", + "issue:45759|issue:45859", + "issue:44077|issue:44625", + "issue:43531|issue:44589", + "issue:43782|issue:45381", + "issue:44987|issue:45042", + "issue:45588|issue:45759" ], - "prepared_review_unit_hash": "08eb5397ce4e9e7ffffae1b8ca8d68e1c74cb9d05ec3221f4611af875c1c0bd2", + "prepared_review_unit_hash": "5649b820861e38e3ff7500da17e03dd063f9afa8e67e398e67e04be538c9f7fa", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36630, - "estimated_input_tokens": 9158, - "estimated_eval_tokens": 18572 + "soft_pair_count": 30, + "serialized_chars": 36953, + "estimated_input_tokens": 9239, + "estimated_eval_tokens": 18734 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Only one soft pair is a true duplicate: the two Qwen2_5_VLProcessor.apply_chat_template crashes are the same bug. All other pairs are related at a broad subsystem level but describe different failures, models, or code paths, so they should stay separate.", - "confidence": 0.97, - "canonical_issue_reason": "issue:44514 is the best canonical issue: it is the earlier, more complete report for the exact same batched-input/padding=False crash, while issue:44545 is a near-verbatim duplicate.", + "summary": "The set is mostly a loose tokenizer/model-loading/multimodal-bug bucket rather than a duplicate cluster. None of the soft pairs look like the same underlying defect, so all candidate edges should stay split. If one issue must represent the group, the Granite AutoTokenizer regression (45812) is the strongest concrete tokenizer bug, but it is not a duplicate of the others.", + "confidence": 0.72, + "canonical_issue_reason": "Issue 45812 is the clearest, most concrete tokenizer regression in the set: it describes a silent v4\u2192v5 token-ID break for a broad model family and is more actionable than the vaguer tokenizer reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44514 is the strongest representative overall because it has the same concrete failure mode as issue:44545 and appears to be the primary report that the later issue duplicates.", + "best_issue_reason": "45812 is the best single issue to anchor this cluster because it is specific, reproducible, and has a clear regression boundary; several others are adjacent tokenizer problems, but not the same bug.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45081", - "right": "issue:45812", - "accept": false, - "reason": "Both involve tokenizer regressions, but they are different bugs: Mistral regex patching vs Granite token ID corruption." - }, - { - "left": "issue:43531", - "right": "issue:43572", - "accept": false, - "reason": "Different models and different config issues; one is sliding_window handling, the other is missing pad_token_idx after v5." - }, - { - "left": "issue:43316", - "right": "issue:44913", - "accept": false, - "reason": "Both are config-serialization discrepancies, but they affect different fields and model families." - }, - { - "left": "issue:43540", - "right": "issue:43782", - "accept": false, - "reason": "Different Qwen multimodal failures with different inputs and error modes; not the same underlying bug." - }, - { - "left": "issue:42915", - "right": "issue:43322", - "accept": false, - "reason": "Different loading/runtime failures in different architectures; one is FP8 config, the other is Llava Next segfault." - }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "Tokenizer crashes on different models and tasks; not one shared defect." + "reason": "Both are tokenizer failures, but they affect different tokenizer classes and different failure modes (LayoutLMv2 batching/NER vs MLuke task AttributeError)." }, { "left": "issue:43421", "right": "issue:45812", "accept": false, - "reason": "Both concern tokenizer behavior after special-token changes, but one is runtime post-processor refresh and the other is a Granite token-ID regression." + "reason": "Runtime special-token/post-processor refresh is a different bug from Granite models getting wrong token IDs after the v4\u2192v5 tokenizer change." }, { "left": "issue:36331", "right": "issue:41720", "accept": false, - "reason": "Unrelated failures: Trainer custom loss signature vs Qwen3 device-mapping cuda assert." + "reason": "Trainer API signature mismatch and Qwen3 device-mapping cudaErrorAssert are unrelated code paths and symptoms." }, { "left": "issue:29159", "right": "issue:36032", "accept": false, - "reason": "Different tokenizer problems: slow/fast inconsistency versus T5 tokenizer load-time method conflict." + "reason": "Generic slow-vs-fast tokenizer inconsistency is not the same as the T5Tokenizer load-time add_special_tokens name conflict." }, { "left": "issue:43388", "right": "issue:45059", "accept": false, - "reason": "Different subsystems and symptoms; metric-gathering label trimming is not related to SAM3 PCS text/bbox behavior." + "reason": "Different subsystems: metric gathering for tuple labels vs SAM3 text/bbox prompt behavior." }, { "left": "issue:45406", "right": "issue:45464", "accept": false, - "reason": "Different runtime paths and models: Gemma4Processor missing _tokenizer vs Qwen3.5 streaming chat/completions failure." + "reason": "Processor missing _tokenizer in transformers serve is distinct from Qwen3.5 streaming chat/completions failure." }, { "left": "issue:30064", "right": "issue:34634", "accept": false, - "reason": "Different processor bugs affecting different modalities; segmentation-map handling is unrelated to Bark voice_preset." + "reason": "Void segmentation-map processing and Bark voice_preset handling are separate processor bugs." }, { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "Completely different issues: MacOS bus error on CLIP vs generation with inputs_embeds/past_key_values." + "reason": "MacOS CLIP bus error and inputs_embeds/past_key_values generation issues do not share the same underlying defect." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "Both are CI/offload-related, but one is a CPU-offload device mismatch and the other is packed-weight unpacking in accelerate loading." + "reason": "Both mention loading/offload, but one is bark-small CPU offload device mismatch and the other is BitNet packed-weight unpacking during accelerate loading." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "Different bugs in different areas: torch.compile/SDPA vs tokenizer special-token registration." + "reason": "SDPA/torch.compile failure in Bamba-9B-v2 is unrelated to BigBirdTokenizer mask-token registration and empty decode output." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "Different loading problems: auto device mapping crash vs PEFT adapter local-checkpoint loading." + "reason": "Qwen3 auto device-map assertion and GraniteSpeech PEFT local-checkpoint loading are different model-loading problems." }, { "left": "issue:43873", "right": "issue:45704", "accept": false, - "reason": "Different features and failure modes: quantization offloading vs T5/apex RMSNorm memory leak." + "reason": "Quantization/offloading behavior and T5 apex.FusedRMSNorm memory leak are different bugs with no shared code path." }, { "left": "issue:43421", "right": "issue:45701", "accept": false, - "reason": "One is a specific TokenizersBackend post-processor update bug; the other is a broader tokenization-version-change report." + "reason": "A tokenizer backend not refreshing post-processors at runtime is not the same as a broad report that versions change tokenization." }, { - "left": "issue:43606", - "right": "issue:43653", + "left": "issue:43531", + "right": "issue:45759", "accept": false, - "reason": "Different bug classes: offload device mismatch vs tokenizer special-token decode behavior." + "reason": "Sliding-window handling in Qwen3-MoE and from_config not unwrapping text_config for Qwen multimodal configs are separate model/config issues." }, { - "left": "issue:43531", - "right": "issue:45759", + "left": "issue:43606", + "right": "issue:43653", "accept": false, - "reason": "Different model/config issues; sliding_window in Qwen3-MoE is unrelated to AutoModelForCausalLM unwrapping text_config in Qwen 3.5/3.6." + "reason": "These are different CI failures in different models: device mismatch during CPU offload vs mask-token registration in a tokenizer." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "Different symptom and subsystem: BarkProcessor voice_preset vs Qwen3 auto device mapping crash." + "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping are unrelated bugs in different components." }, { "left": "issue:42490", "right": "issue:43329", "accept": false, - "reason": "Both involve loading-related behavior, but one is path parsing/import warning and the other is a multimodal token-counting bug." + "reason": "A trailing-slash local path import warning and undefined video-branch helpers in multimodal token counting are not the same issue." }, { - "left": "issue:44492", - "right": "issue:44509", + "left": "issue:34567", + "right": "issue:35707", "accept": false, - "reason": "Unrelated docs/typo issues: cache strategy typo vs pipeline docs mentioning removed tasks." + "reason": "TrainerState token-count tracking and progressive generation with embeds/past KV are unrelated." }, { - "left": "issue:34567", - "right": "issue:35707", + "left": "issue:44492", + "right": "issue:44509", "accept": false, - "reason": "TrainerState token-count tracking is unrelated to progressive generation with inputs_embeds and past_key_values." + "reason": "A typo in cache-strategy docs is not the same as stale pipeline-task documentation for removed v5 tasks." }, { "left": "issue:44514", "right": "issue:44545", - "accept": true, - "reason": "Same exact crash report: Qwen2_5_VLProcessor.apply_chat_template fails on batched input when padding=False; 44545 is a duplicate of 44514." + "accept": false, + "reason": "These are effectively the same title, but the pair was not presented with evidence of one underlying report; treat as separate records unless explicit duplication is confirmed." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Different classification bugs: config propagation in Qwen3.5 vs degenerate zero loss across sequence-classification models." + "reason": "Propagating num_labels in Qwen3.5 configs is different from the single-label-classification zero-loss bug across sequence-classification models." }, { - "left": "issue:43421", - "right": "issue:43927", + "left": "issue:44898", + "right": "issue:45072", "accept": false, - "reason": "Both are tokenizer/config state issues, but one is post-processor refresh and the other is DiaConfig persistence of custom token IDs." + "reason": "Perceiver non-default resolution interpolation and bfloat16 dtype mismatches in unrelated models are not the same defect." }, { - "left": "issue:44898", - "right": "issue:45072", + "left": "issue:43421", + "right": "issue:43927", "accept": false, - "reason": "Different model/runtime failures: Perceiver non-default resolution vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + "reason": "Runtime special-token/post-processor updates differ from DiaConfig losing custom token IDs on save/load." }, { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "Different tokenizer load failures for different models; no evidence of the same underlying bug." + "reason": "Loading a specific model failing and tokenizer loading for EMBEDDIA/est-roberta are different model-specific loading bugs." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Qwen multimodal config unwrapping and Qwen3.5-MoE rotary embedding CPU-offload compatibility are separate problems." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "A patchtsmixer post_init policy issue is unrelated to Qwen3.5 num_labels propagation." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Sliding-window handling and missing Float8 storage registration are unrelated." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "weight_only loading for Qwen3VL and video vision_position_ids in Qwen2.5-VL are different multimodal bugs." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Loading a specific model and PIL backend image processors requiring torchvision are different regressions." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink-less s_aux handling and Qwen multimodal config unwrapping do not share the same code path." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative on the soft-edge pairs. The accepted duplicate pair, issue:44514/issue:44545, is strongly supported by matching titles and closely aligned failure descriptions, with 44514 appearing to be the earlier and fuller report. I don\u2019t see any overstatement in the other pair judgments based on the provided evidence." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -4513,8 +4557,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -4524,55 +4568,50 @@ "issue:40444", "issue:41720", "issue:42886", + "issue:43295", "issue:43381", "issue:43421", "issue:43531", + "issue:43540", + "issue:43550", "issue:43596", + "issue:43618", "issue:43645", "issue:43653", + "issue:43697", "issue:43720", "issue:43746", - "issue:43782", "issue:43819", "issue:43844", "issue:43873", "issue:43994", + "issue:44008", "issue:44062", - "issue:44077", "issue:44186", "issue:44220", "issue:44368", "issue:44451", - "issue:44589", "issue:44610", - "issue:44625", "issue:44661", "issue:44671", "issue:44843", "issue:44918", "issue:44936", - "issue:44987", - "issue:45042", + "issue:45072", "issue:45081", + "issue:45183", "issue:45356", "issue:45357", - "issue:45381", - "issue:45588", + "issue:45479", "issue:45663", "issue:45704", - "issue:45759", + "issue:45834", "issue:45859", "issue:45907", "issue:45923", "issue:46050" ], "soft_pairs": [ - "issue:45759|issue:45859", - "issue:44077|issue:44625", - "issue:43531|issue:44589", - "issue:43782|issue:45381", - "issue:44987|issue:45042", - "issue:45588|issue:45759", "issue:44062|issue:44451", "issue:44368|issue:45357", "issue:39692|issue:44610", @@ -4581,238 +4620,244 @@ "issue:45704|issue:45923", "issue:43421|issue:43653", "issue:45663|issue:45923", - "issue:43381|issue:43844", "issue:42886|issue:44843", + "issue:43381|issue:43844", "issue:45704|issue:45859", "issue:43531|issue:43645", "issue:41720|issue:45704", "issue:43421|issue:44186", "issue:43994|issue:44671", "issue:43746|issue:43873", - "issue:39692|issue:43873", "issue:45081|issue:45356", + "issue:39692|issue:43873", "issue:43596|issue:44220", "issue:40444|issue:43746", "issue:45356|issue:45907", "issue:30064|issue:36010", "issue:43720|issue:43819", "issue:44220|issue:44661", - "issue:43653|issue:43720" + "issue:43653|issue:43720", + "issue:43540|issue:44008", + "issue:43618|issue:43697", + "issue:43550|issue:45072", + "issue:45183|issue:45479", + "issue:43295|issue:45834", + "issue:43550|issue:43819" ], - "prepared_review_unit_hash": "0067a6f5ed62946114e3b99c12f3a10c08c63da3b58aa98e5b2e2adaba0ef8f0", + "prepared_review_unit_hash": "74e715da9d84727c01f78c894715af87e7b1123bacae28f47d30a568566bc176", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 36772, - "estimated_input_tokens": 9193, - "estimated_eval_tokens": 18642 - }, - "overflow_policy": "truncate_then_skip", - "trimmed": false, - "aggressively_trimmed": false, - "split": true, - "analyst_result": { - "summary": "No duplicates or mergeable equivalents stand out here. The soft pairs mostly share broad areas like tokenizers, multimodal models, offloading, or training, but each describes a different concrete failure mode or root cause, so all edges should be rejected.", - "confidence": 0.97, - "canonical_issue_reason": null, - "canonical_pr_reason": null, - "best_issue_reason": null, - "best_pr_reason": null, - "soft_edge_verdicts": [ - { - "left": "issue:45759", - "right": "issue:45859", - "accept": false, - "reason": "Different bugs: one is config unwrapping for composite Qwen multimodal models, the other is CPU offload incompatibility in rotary embedding forward." - }, - { - "left": "issue:44077", - "right": "issue:44625", - "accept": false, - "reason": "Unrelated failures: optional post_init policy in patchtsmixer vs num_labels propagation in Qwen3.5 config." - }, - { - "left": "issue:43531", - "right": "issue:44589", - "accept": false, - "reason": "Different subsystems and symptoms: Qwen3-MoE sliding window handling vs missing Float8 storage type on load." - }, - { - "left": "issue:43782", - "right": "issue:45381", - "accept": false, - "reason": "Both are Qwen/VL-related, but one is a weight_only loading error and the other is a video vision_position_ids bug; not the same defect." - }, - { - "left": "issue:44987", - "right": "issue:45042", - "accept": false, - "reason": "Separate problems: loading a specific model fails vs PIL image processors incorrectly requiring torchvision." - }, - { - "left": "issue:45588", - "right": "issue:45759", - "accept": false, - "reason": "Different code paths: flash_attention sink-less s_aux handling vs from_config not unwrapping text_config." - }, + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37462, + "estimated_input_tokens": 9366, + "estimated_eval_tokens": 18988 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a grab bag of unrelated issues across tokenizers, vision-language models, training/inference, and offline loading. None of the soft-edge pairs look like the same underlying bug or change, so I reject all links.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the set is not a duplicate cluster, but a collection of unrelated issues spanning different models and code paths.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44843 is the strongest standalone report: it has a clear failure mode, a specific code path (_patch_mistral_regex), and an actionable offline-mode regression.", + "best_pr_reason": null, + "soft_edge_verdicts": [ { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "No shared bug: AddedToken keyword duplication error vs a model loading failure for ScandiBERT." + "reason": "Both involve tokenizers, but one is an AddedToken constructor error and the other is a model loading failure for ScandiBERT; different bugs." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "Distinct Qwen issues: warning about tie_word_embeddings vs incorrect visual encoder keys during save_pretrained." + "reason": "Different Qwen3.5-related problems: one is a training warning about tie_word_embeddings, the other is a save_pretrained visual-encoder key regression." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "Different model/docs problems: SigLIP2 example errors vs OmDet-Turbo processor/model input-size mismatch." + "reason": "Both mention processor/model mismatches, but they affect different models and different failure modes." }, { "left": "issue:43746", "right": "issue:46050", "accept": false, - "reason": "Not the same issue: local PEFT checkpoint loading for GraniteSpeech vs quantized sinq save/load corruption." + "reason": "Different persistence/loading bugs: PEFT local checkpoint loading vs quantized model save_pretrained/from_pretrained behavior." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Different failures: SFT trainer input embedding unpacking vs trainer.evaluate() breaking after train()." + "reason": "Both mention trainers, but one is about unpacking Qwen3.5 input embeddings and the other about evaluate() after train()." }, { "left": "issue:45704", "right": "issue:45923", "accept": false, - "reason": "Unrelated: apex FusedRMSNorm memory leak in T5 vs missing supports_gradient_checkpointing flag." + "reason": "Completely different issues: a T5 RMSNorm memory leak versus a missing gradient-checkpointing flag in a trust_remote_code model." }, { "left": "issue:43421", "right": "issue:43653", "accept": false, - "reason": "Different tokenizer bugs: runtime special-token post-processor refresh vs BigBird mask token registration/decode output." + "reason": "Both touch tokenizer special tokens, but one is runtime post-processor syncing and the other is BigBirdTokenizer mask-token registration." }, { "left": "issue:45663", "right": "issue:45923", "accept": false, - "reason": "Distinct issues: FSDP2 shared_kv_states rebuilding per layer vs trust_remote_code gradient-checkpointing flag missing." + "reason": "Different model/training bugs: Gemma-4 FSDP2 shared_kv_states handling vs gradient-checkpointing support flag." }, { - "left": "issue:43381", - "right": "issue:43844", + "left": "issue:42886", + "right": "issue:44843", "accept": false, - "reason": "Both involve training behavior, but one is eval-mode gradient checkpointing support and the other is gradient explosion under ZeRO-3." + "reason": "Both involve HF_HUB_OFFLINE, but 42886 is cache loading and 44843 is an unconditional model_info() call in mistral regex patching." }, { - "left": "issue:42886", - "right": "issue:44843", + "left": "issue:43381", + "right": "issue:43844", "accept": false, - "reason": "Related to offline tokenizer loading, but not the same bug: one is generic cache/offline failure, the other is a specific _patch_mistral_regex model_info call." + "reason": "Unrelated training behavior issues: gradient checkpointing in eval mode vs gradient growth under ZeRO-3." }, { "left": "issue:45704", "right": "issue:45859", "accept": false, - "reason": "Different models and symptoms: memory leak from fused norm use vs CPU offload incompatibility in Qwen rotary embedding." + "reason": "Different models and code paths: T5 norm memory leak vs Qwen3.5-MoE rotary embedding CPU offload incompatibility." }, { "left": "issue:43531", "right": "issue:43645", "accept": false, - "reason": "Distinct regressions: sliding_window handling in Qwen3-MoE vs Jupyter custom model definition/initialization in Transformers 5." + "reason": "No shared bug: Qwen3-MoE sliding_window behavior versus Jupyter custom model initialization regression." }, { "left": "issue:41720", "right": "issue:45704", "accept": false, - "reason": "No overlap beyond being model-runtime bugs: cudaErrorAssert on Qwen3 auto device mapping vs T5 fused RMSNorm memory leak." + "reason": "Different subsystems and failure modes: Qwen3 auto device mapping CUDA assert versus T5 norm memory leak." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Different tokenizer paths: special-token post-processor update issue vs LayoutLMv2Tokenizer crashes on NER/padding." + "reason": "Both are tokenizer-related, but one is special-token post-processor updates and the other is LayoutLMv2 NER/padding crashes." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "Separate model correctness bugs: SigLIP2 AutoModel/pipeline output regression vs CamemBERT masked LM predictions in v5." + "reason": "Different model regressions: SigLIP2 pipeline/AutoModel output issues versus CamemBERT masked-LM prediction changes." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Different areas: PEFT local checkpoint loading vs quantization/offloading behavior." + "reason": "Both mention loading/offloading, but one is PEFT checkpoint loading and the other is quantization/offloading behavior." }, { - "left": "issue:39692", - "right": "issue:43873", + "left": "issue:45081", + "right": "issue:45356", "accept": false, - "reason": "One is a documentation/example mismatch, the other is a quantization offloading issue; not the same bug." + "reason": "Both mention fix_mistral_regex, but one is a crash in _patch_mistral_regex and the other is a Kimi-K2.5 codec/warning regression." }, { - "left": "issue:45081", - "right": "issue:45356", + "left": "issue:39692", + "right": "issue:43873", "accept": false, - "reason": "Both touch mistral regex logic, but one is a crash on fix_mistral_regex and the other is a Kimi tokenizer regression with codec handling/warning." + "reason": "Quantization is only a broad theme here; SigLIP2 docs errors and offloading-with-quantization are different issues." }, { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "Different runtime errors in different components: ZeRO-3/BertModel indexing vs audio fbank feature extraction." + "reason": "Different code paths entirely: DeepSpeed zero3/BertModel versus audio feature extraction." }, { "left": "issue:40444", "right": "issue:43746", "accept": false, - "reason": "Unrelated multimodal training failure vs local checkpoint loading with PEFT adapters." + "reason": "Different domains and bugs: Qwen2.5-VL multi-image IterableDataset training versus GraniteSpeech checkpoint loading." }, { "left": "issue:45356", "right": "issue:45907", "accept": false, - "reason": "Completely different: tokenizer regression vs hub API list_repo_files not forwarding revision." + "reason": "No relation: tokenizer codec/fix_mistral_regex regression versus list_repo_files missing revision forwarding." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Different domains entirely: image processor segmentation-map handling vs GenerationMixin import error." + "reason": "Completely unrelated: image processor segmentation-map handling versus an ImportError for GenerationMixin." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "Separate bugs: packed BitNet weights not unpacked on accelerate load vs DAC.from_latents not matching forward pass without STE." + "reason": "Different quantization-related model bugs with distinct mechanisms: packed-weight loading versus missing STE in DAC.from_latents." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Different code paths: torch fbank feature extraction vs add-new-model-like failing inside TOKENIZER_MAPPING_NAMES." + "reason": "Audio feature extraction and tokenizer-mapping model registration are unrelated." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "No shared underlying defect: BigBird special-token decode bug vs BitNet packed-weight loading." + "reason": "Tokenizer special-token decoding bug versus BitNet packed-weight loading bug; no shared underlying defect." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Both are multimodal, but one is Qwen3OmniMoe video input handling and the other is Gemma 3n variable-name collision with audio masks." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "Both are output regressions in v5, but they affect different models and different tensors/paths." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Different inference/compiler issues in different models: SDPA/torch.compile failure versus bfloat16 dtype mismatches." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "One is validation/error-message logic for input_ids vs inputs_embeds; the other is a zero-loss classification bug." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Both involve processor/tokenizer-style workflows, but the first is a regression around processor.tokenizer/images and the second is a long-OCR index error." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Completely different bugs: Bamba torch.compile/SDPA versus DAC latent reconstruction with missing STE." } ] }, @@ -4831,29 +4876,30 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ "issue:30064", + "issue:34634", + "issue:36010", "issue:36331", "issue:39692", "issue:41720", - "issue:43295", "issue:43322", "issue:43335", "issue:43388", "issue:43421", "issue:43531", "issue:43540", - "issue:43550", "issue:43577", - "issue:43618", + "issue:43606", "issue:43643", + "issue:43644", "issue:43653", - "issue:43697", "issue:43716", + "issue:43749", "issue:43756", "issue:43782", "issue:43818", @@ -4864,6 +4910,7 @@ "issue:44186", "issue:44442", "issue:44448", + "issue:44479", "issue:44488", "issue:44671", "issue:44811", @@ -4876,9 +4923,7 @@ "issue:45072", "issue:45081", "issue:45084", - "issue:45183", "issue:45381", - "issue:45479", "issue:45584", "issue:45685", "issue:45834", @@ -4886,18 +4931,12 @@ "issue:46032" ], "soft_pairs": [ - "issue:43540|issue:44008", - "issue:43618|issue:43697", - "issue:43550|issue:45072", - "issue:45183|issue:45479", - "issue:43295|issue:45834", - "issue:43550|issue:43819", "issue:43388|issue:46032", "issue:44062|issue:44488", "issue:43716|issue:44008", "issue:43335|issue:43756", - "issue:43322|issue:45685", "issue:44186|issue:44898", + "issue:43322|issue:45685", "issue:43653|issue:43819", "issue:43531|issue:43643", "issue:43950|issue:45987", @@ -4908,209 +4947,201 @@ "issue:44448|issue:44671", "issue:43421|issue:44811", "issue:41720|issue:44991", - "issue:39692|issue:45834", "issue:43577|issue:43818", + "issue:39692|issue:45834", "issue:43782|issue:44977", "issue:44821|issue:45084", "issue:45584|issue:45834", - "issue:44186|issue:44442" + "issue:44186|issue:44442", + "issue:43644|issue:43749", + "issue:43606|issue:45072", + "issue:44479|issue:45834", + "issue:34634|issue:36010" ], - "prepared_review_unit_hash": "17ca86f12c6443af6d5748b803825e884b962b795653e5cfb7be2b99641f824e", + "prepared_review_unit_hash": "c7bdcca96b9abf9391bef14ad15a75609003f3565b7355facf4b62ee253e5aad", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 36510, - "estimated_input_tokens": 9128, - "estimated_eval_tokens": 18512 + "soft_pair_count": 26, + "serialized_chars": 35744, + "estimated_input_tokens": 8936, + "estimated_eval_tokens": 18128 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose grab-bag of unrelated Transformers issues across tokenizers, multimodal models, generation, and cache logic. None of the soft pairs looks like the same concrete bug; the closest is the two Mistral regex issues, but they describe different failure modes.", - "confidence": 0.93, - "canonical_issue_reason": "No strong canonical issue: the items are heterogeneous and do not form a single duplicate set. If one must be chosen as an anchor, issue 44843 is the closest because it is the most specific and directly tied to the Mistral tokenizer patch path.", + "summary": "This cluster is mostly heterogeneous issue reports with no true duplicate group. The only borderline pair is the two Mistral regex/tokenizer-loading reports, but they describe different failure modes and should stay separate.", + "confidence": 0.94, + "canonical_issue_reason": "issue:44843 is the closest thing to a cluster anchor: it is the broadest and most central tokenizer-loading regression here, and it has the only plausible near-match with another item in the set.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44843 is the best representative only in a weak sense: it is specific, reproducible, and the only item in a near-related soft pair, but it still does not unify the cluster.", + "best_issue_reason": "issue:44843 is the best representative issue overall because it targets a concrete, recurring loading-path regression and is broader than the more specific follow-up crash in issue:45081.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43540", - "right": "issue:44008", - "accept": false, - "reason": "Different models and failure modes: Qwen3OmniMoe video input validation vs Gemma 3n attribute collision in forward()." - }, - { - "left": "issue:43618", - "right": "issue:43697", - "accept": false, - "reason": "Both are regressions, but on different models/code paths: CLIP attentions assignment vs RTDetrV2 output drift." - }, - { - "left": "issue:43550", - "right": "issue:45072", - "accept": false, - "reason": "Unrelated bugs: torch.compile+SDPA on Bamba vs dtype mismatch in SwitchTransformers/TimmWrapperModel." - }, - { - "left": "issue:45183", - "right": "issue:45479", - "accept": false, - "reason": "Different core problems: input validation XOR logic vs degenerate single-label classification loss." - }, - { - "left": "issue:43295", - "right": "issue:45834", - "accept": false, - "reason": "One is a processor/tokenizer regression; the other is a long-ocr index error in Kosmos2.5." - }, - { - "left": "issue:43550", - "right": "issue:43819", - "accept": false, - "reason": "Different models and bugs: SDPA/compile failure vs DAC forward/from_latents mismatch." - }, { "left": "issue:43388", "right": "issue:46032", "accept": false, - "reason": "Different training/runtime paths: gather_for_metrics batching vs Mamba2 cache correctness." + "reason": "Different subsystems: evaluation gathering vs Mamba2 cache behavior." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "Both involve tokenizer loading, but one is an AddedToken keyword conflict and the other is a model-specific load failure." + "reason": "One is an AddedToken API error; the other is a tokenizer loading failure for a specific model." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Different subsystems: dtype mismatch in image preprocessing vs Gemma 3n audio mask name collision." + "reason": "Different code paths and failures: image preprocessor dtype mismatch vs Gemma3n variable collision." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "Distinct model/config bugs: sparse-layer construction vs RoPE layer dropping." + "reason": "Unrelated model/config bug vs RoPE-layer counting issue." }, { - "left": "issue:43322", - "right": "issue:45685", + "left": "issue:44186", + "right": "issue:44898", "accept": false, - "reason": "No shared code path: Llava Next loading segfault vs MPS histogram not implemented for Int." + "reason": "Tokenizer crash for NER/padding is unrelated to Perceiver image-size handling." }, { - "left": "issue:44186", - "right": "issue:44898", + "left": "issue:43322", + "right": "issue:45685", "accept": false, - "reason": "Different model families and symptoms: LayoutLMv2 tokenizer crashes vs Perceiver interpolation failure." + "reason": "Segfault in Llava loading is unrelated to an MPS histogram/int dtype error." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "Tokenizer special-token registration bug vs DAC STE mismatch; unrelated." + "reason": "Tokenizer special-token registration bug is unrelated to DAC latent/STE mismatch." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Different areas: Qwen3-MoE sliding window vs AutoConfig trust_remote_code field loss." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to AutoConfig trust_remote_code field loss." }, { "left": "issue:43950", "right": "issue:45987", "accept": false, - "reason": "Both are cache/state-related regressions, but one corrupts non-persistent buffers and the other violates get_seq_length return type." + "reason": "Non-persistent buffer corruption is unrelated to StaticCache return-type issues." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "Both are config-loading issues, but they concern different mechanisms and model families." + "reason": "Both are config-loading issues, but they affect different mechanisms and models." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Both mention _patch_mistral_regex, but the failures differ: offline model_info access vs missing backend_tokenizer attribute." + "reason": "Same broad file/function area, but different bugs: offline model_info call vs missing backend_tokenizer attribute." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Completely unrelated: image segmentation-map processing vs CustomTrainer loss signature mismatch." + "reason": "Different training/image-processing failures with no shared code path." }, { "left": "issue:43540", "right": "issue:45381", "accept": false, - "reason": "Both involve Qwen variants, but the bugs are different: video input handling vs flash-attention generation behavior." + "reason": "Both mention Qwen video, but the concrete failures and likely fixes are different models/code paths." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Separate model regressions with different outputs and different tasks: summarization vs masked LM." + "reason": "Different model regressions; same version does not make them duplicates." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Related to tokenization/special tokens in a broad sense, but the concrete bugs are different: runtime post-processor refresh vs Whisper batch_decode skip_special_tokens handling." + "reason": "Runtime special-token/post-processor update bug is not the same as batch_decode skip_special_tokens behavior." }, { "left": "issue:41720", "right": "issue:44991", "accept": false, - "reason": "Unrelated failures: Qwen3 device-mapping cuda assert vs tokenizer load failure for est-roberta." + "reason": "Unrelated issues: device-mapping crash vs tokenizer loading failure for a different model." }, { - "left": "issue:39692", - "right": "issue:45834", + "left": "issue:43577", + "right": "issue:43818", "accept": false, - "reason": "Different multimodal/documentation issues: SigLIP2 example problems vs Kosmos2.5 OCR index error." + "reason": "Blip2 dtype initialization and Video-LLaVA weight/attention structure are unrelated." }, { - "left": "issue:43577", - "right": "issue:43818", + "left": "issue:39692", + "right": "issue:45834", "accept": false, - "reason": "Different model loading bugs: BLIP2 dtype propagation vs Video-LLaVA video_tower/temporal-attention issue." + "reason": "Doc-example/model mismatch is unrelated to long OCR index errors in Kosmos2.5." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Different Qwen model families and failures: weight_only load error vs flash-attention generation problem." + "reason": "Different Qwen generation regressions with different triggers and code paths." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "No shared underlying bug: AutoImageProcessor-from-URL loading vs template-node compilation error." + "reason": "AutoImageProcessor URL loading is unrelated to a template compilation TypeError." }, { "left": "issue:45584", "right": "issue:45834", "accept": false, - "reason": "Different input domains and failure modes: Whisper empty transcription after token alignment vs Kosmos2.5 long OCR index error." + "reason": "Whisper empty-transcription handling is unrelated to Kosmos2.5 long-input indexing." }, { "left": "issue:44186", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer bugs affecting different models and APIs; not the same underlying issue." + "reason": "Tokenizer crash in LayoutLMv2 is unrelated to FastSpeech2Conformer tokenizer loading." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Both involve loading/state, but not the same bug or fix." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Device offload mismatch and bf16 dtype mismatch are different issues." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Video regression in Qwen models is unrelated to Kosmos2.5 OCR indexing." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "Bark voice_preset behavior and GenerationMixin import failure are unrelated." } ] }, @@ -5129,23 +5160,21 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ "issue:28218", "issue:29159", "issue:34634", - "issue:36010", "issue:36331", "issue:42175", "issue:43262", + "issue:43329", "issue:43526", "issue:43577", "issue:43582", - "issue:43606", - "issue:43644", "issue:43650", "issue:43653", "issue:43697", @@ -5161,7 +5190,6 @@ "issue:44060", "issue:44291", "issue:44368", - "issue:44479", "issue:44485", "issue:44509", "issue:44554", @@ -5175,35 +5203,33 @@ "issue:45020", "issue:45072", "issue:45081", + "issue:45084", "issue:45325", "issue:45406", + "issue:45538", + "issue:45600", "issue:45701", - "issue:45834", "issue:45987", "issue:46192", "issue:46209" ], "soft_pairs": [ - "issue:43644|issue:43749", - "issue:43606|issue:45072", - "issue:44479|issue:45834", - "issue:44561|issue:45020", - "issue:34634|issue:36010", "issue:43577|issue:43582", + "issue:44561|issue:45020", "issue:43825|issue:43994", "issue:44805|issue:44936", "issue:45987|issue:46192", "issue:44671|issue:45701", - "issue:44485|issue:44509", "issue:44610|issue:45072", "issue:28218|issue:29159", + "issue:44485|issue:44509", "issue:43720|issue:45072", - "issue:43825|issue:43827", "issue:43782|issue:45325", "issue:43906|issue:46209", + "issue:43825|issue:43827", "issue:42175|issue:45081", - "issue:43749|issue:44291", "issue:43582|issue:44857", + "issue:43749|issue:44291", "issue:45081|issue:45406", "issue:43262|issue:43819", "issue:34634|issue:36331", @@ -5212,216 +5238,188 @@ "issue:44060|issue:44368", "issue:43653|issue:44912", "issue:43742|issue:45020", - "issue:43526|issue:44554" + "issue:43526|issue:44554", + "issue:45084|issue:45538", + "issue:43329|issue:45600" ], - "prepared_review_unit_hash": "15a17dbba50b445dff34ac4f6c746368c730b9a88e82671d3935e780e1100d6a", + "prepared_review_unit_hash": "84197617e6f879eb57bc66e0ab4a700f941c00a39ae78d1677c075a3cedccd60", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36600, - "estimated_input_tokens": 9150, - "estimated_eval_tokens": 18556 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35476, + "estimated_input_tokens": 8869, + "estimated_eval_tokens": 17994 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issue mix; only the tokenizer added-token spacing reports look like the same underlying bug. The rest are similar at a subsystem level but not concrete duplicates.", + "summary": "This cluster is mostly a set of unrelated Transformers bug reports. The only strong duplicate-looking pair is the remote-code regression pair (44561 / 45020); the rest are too model-, feature-, or subsystem-specific to merge as duplicates.", "confidence": 0.78, - "canonical_issue_reason": "issue:28218 is the clearest concrete report of the tokenizer added-token spacing bug, and it best matches the one plausible duplicate in the set.", + "canonical_issue_reason": "Issue 45020 is the best umbrella issue: it is broad, current, and matches the clearest duplicate candidate about remote-code breakage in recent Transformers versions.", "canonical_pr_reason": null, - "best_issue_reason": "issue:28218 is the strongest representative issue: specific, self-contained, and directly tied to the duplicate tokenizer behavior.", + "best_issue_reason": "45020 has the broadest scope and strongest overlap with a likely duplicate, while the other issues are more specific or clearly unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43644", - "right": "issue:43749", - "accept": false, - "reason": "Both concern model/loading behavior, but one is about non-persistent buffer initialization and the other about FSDP CPU RAM efficient loading; not the same bug." - }, - { - "left": "issue:43606", - "right": "issue:45072", - "accept": false, - "reason": "Different models and failures: CPU offload device mismatch vs bfloat16 dtype mismatch. Same broad inference area, not the same issue." - }, - { - "left": "issue:44479", - "right": "issue:45834", + "left": "issue:43577", + "right": "issue:43582", "accept": false, - "reason": "Both are multimodal input bugs, but they affect different models and different failure modes." + "reason": "Different failure modes: one is a model dtype-loading mismatch, the other is an Apple Silicon allocator warmup TypeError." }, { "left": "issue:44561", "right": "issue:45020", - "accept": false, - "reason": "issue:44561 is a specific trust_remote_code regression from a removed helper; issue:45020 is a broad umbrella report. Related, but not the same concrete bug." - }, - { - "left": "issue:34634", - "right": "issue:36010", - "accept": false, - "reason": "One is a BarkProcessor preset bug, the other an import error for GenerationMixin; no duplicate evidence." - }, - { - "left": "issue:43577", - "right": "issue:43582", - "accept": false, - "reason": "Different domains: BLIP2 dtype loading vs Apple Silicon allocator warmup TypeError." + "accept": true, + "reason": "Both describe the same underlying v5/regression class: recent Transformers changes breaking trust_remote_code/remote_code models." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "One is a v5 pipeline error-message mismatch; the other is a bad AutoModel/pipeline result for a specific model. Not the same bug." + "reason": "One is an error-message/docs issue around translation pipelines; the other is a model output quality problem with AutoModel/pipeline." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "Generic index-shape error vs trainer evaluate-after-train failure; too little concrete overlap to treat as duplicate." + "reason": "An IndexError in masking/indexing is unrelated to trainer.evaluate() failing after train()." }, { "left": "issue:45987", "right": "issue:46192", "accept": false, - "reason": "One is a StaticCache return-type contract bug, the other silently ignores return_tensors when text_kwargs is passed." + "reason": "Different processor/cache contract bugs: seq-length return type vs return_tensors being ignored when text_kwargs is passed." }, { "left": "issue:44671", "right": "issue:45701", "accept": false, - "reason": "CamemBERT masked-LM regression and a general tokenization-version complaint are not the same code-path problem." - }, - { - "left": "issue:44485", - "right": "issue:44509", - "accept": false, - "reason": "One discusses GLM-5 RoPE implementation, the other stale docs about removed pipeline tasks." + "reason": "A CamemBERT masked-LM regression is not the same underlying bug as a broad tokenization-change report." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "Different processors/models and different defects: input-size mismatch vs dtype mismatch." + "reason": "These are unrelated model/processor bugs: input-size mismatch vs dtype mismatches in inference." }, { "left": "issue:28218", "right": "issue:29159", - "accept": true, - "reason": "Both describe the same tokenizer added-token spacing inconsistency, with slow/fast tokenizer behavior and an extra space after added tokens." + "accept": false, + "reason": "Both are tokenizer-related, but the titles do not clearly point to the same concrete bug; one is about added-token spacing, the other a broad slow/fast inconsistency report." }, { - "left": "issue:43720", - "right": "issue:45072", + "left": "issue:44485", + "right": "issue:44509", "accept": false, - "reason": "BitNet packed-weight loading failure is unrelated to dtype mismatch issues in other models." + "reason": "GLM-5 RoPE discussion is unrelated to docs still referencing removed pipeline tasks." }, { - "left": "issue:43825", - "right": "issue:43827", + "left": "issue:43720", + "right": "issue:45072", "accept": false, - "reason": "Both mention v5 pipeline removals, but one is an error-message issue and the other is stale docs; different artifacts and fixes." + "reason": "Packed-weight loading in BitNet is a different code path from dtype mismatches in SwitchTransformers/TimmWrapperModel." }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "Different Qwen-VL bugs: weight_only loading error vs still-image temporal position scaling." + "reason": "Qwen3VL weight_only load failures and Qwen2.5-VL RoPE index scaling are distinct issues." }, { "left": "issue:43906", "right": "issue:46209", "accept": false, - "reason": "A reproduction of an older issue is not the same bug as an FP8/B200 inference failure." + "reason": "An isolated reproduction of an older issue is unrelated to FP8 inference failures on B200." }, { - "left": "issue:42175", - "right": "issue:45081", + "left": "issue:43825", + "right": "issue:43827", "accept": false, - "reason": "Packaging/backend selection issue vs tokenizer regex patch crash; unrelated." + "reason": "Both mention v5 pipeline removals, but one is runtime messaging and the other is documentation drift; not the same bug." }, { - "left": "issue:43749", - "right": "issue:44291", + "left": "issue:42175", + "right": "issue:45081", "accept": false, - "reason": "Both are loading-related, but one is FSDP CPU RAM efficient loading and the other is init_empty_weights argument handling; distinct failures." + "reason": "Backend dependency installation and Mistral regex patching are unrelated." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "Apple Silicon warmup TypeError and CUDA AMP float16 crash are different platform-specific bugs." + "reason": "Apple Silicon allocator warmup and CUDA AMP loss crashes are different failures." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading and init_empty_weights argument errors are separate loading regressions." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Tokenizer regex patch crash and Gemma4Processor missing _tokenizer are separate component bugs." + "reason": "Tokenizer regex patching crash and Gemma4Processor missing _tokenizer are different processor/tokenizer bugs." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and DAC latent reconstruction mismatch are unrelated." + "reason": "Audio processor sampling-rate defaults and DAC from_latents/STE mismatch are unrelated." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "Processor voice_preset failure and CustomTrainer compute_loss signature mismatch do not describe the same defect." + "reason": "BarkProcessor voice_preset and CustomTrainer compute_loss signature changes do not share an underlying code path." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "Different vision bugs: label reduction returning one label vs RTDetrV2 output drift in v5." + "reason": "BEiT reduce_labels behavior and RTDetrV2 output drift are different model/processor regressions." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "No meaningful overlap; one is an empty placeholder title, the other a reproduction issue." + "reason": "A vague request for data and an isolated reproduction of another issue are not duplicates." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "Both involve Qwen tie-word-embedding warnings, but one is a wrong tied-weights mapping bug and the other is a user-facing warning during LoRA fine-tuning." + "reason": "Both mention tied-weights warnings, but the model-specific warning behavior and the alleged cross-layer tying bug are not clearly the same issue." }, { "left": "issue:43653", "right": "issue:44912", "accept": false, - "reason": "Tokenizer special-token registration and MXFP4 quantization loading are unrelated." + "reason": "A tokenizer special-token registration bug is unrelated to MXFP4 quantization loading fallback." }, { "left": "issue:43742", "right": "issue:45020", "accept": false, - "reason": "A specific MobileLLM load key error is not the same underlying problem as a broad remote_code regression." - }, - { - "left": "issue:43526", - "right": "issue:44554", - "accept": false, - "reason": "BeitImageProcessorFast label reduction and MPS attention correctness are different issues." + "reason": "A key error while loading a specific model is not the same as a broad remote_code regression report." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded and appropriately conservative. The summary stays within the packet evidence, and the only accepted soft-edge pair is plausibly the same tokenizer spacing bug based on the issue titles. The other pairings are correctly rejected as subsystem-level similarities rather than clear duplicates." + "feedback": "The summary is well grounded in the packet and stays conservative. The only duplicate-like pair it elevates is 44561/45020, which is supported by the titles, and the rest are appropriately treated as unrelated or too specific. No evidence appears overstated." }, "evaluator_used": true, "retried": false, @@ -5437,11 +5435,12 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ + "issue:29159", "issue:34567", "issue:36010", "issue:36331", @@ -5469,6 +5468,7 @@ "issue:43819", "issue:43854", "issue:43927", + "issue:43931", "issue:44387", "issue:44479", "issue:44568", @@ -5482,32 +5482,27 @@ "issue:44977", "issue:44991", "issue:45042", - "issue:45084", "issue:45356", "issue:45459", - "issue:45538", - "issue:45600", "issue:45759", "issue:45803", "issue:45859" ], "soft_pairs": [ - "issue:45084|issue:45538", - "issue:43329|issue:45600", "issue:44589|issue:45042", "issue:43638|issue:44936", "issue:43746|issue:44811", "issue:43452|issue:44991", "issue:44387|issue:44617", "issue:43550|issue:44912", - "issue:36010|issue:43645", "issue:43819|issue:43927", + "issue:36010|issue:43645", "issue:43257|issue:43329", "issue:43854|issue:44977", "issue:39692|issue:44912", - "issue:38175|issue:44610", "issue:43720|issue:44610", "issue:45459|issue:45803", + "issue:38175|issue:44610", "issue:36010|issue:36331", "issue:43454|issue:43577", "issue:43531|issue:45042", @@ -5520,220 +5515,219 @@ "issue:43329|issue:43450", "issue:43749|issue:44756", "issue:34567|issue:40990", - "issue:44568|issue:45356" + "issue:44568|issue:45356", + "issue:43257|issue:43931", + "issue:29159|issue:34567" ], - "prepared_review_unit_hash": "fb0fadd3e3b2d7fa119080baef4663f42585049a6f58372a1750cae28223fc3a", + "prepared_review_unit_hash": "f5382abc69d6b255bb3663a1bdacca79b85f15e93238d5ea9f60c74e5a37e8f0", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36622, - "estimated_input_tokens": 9156, - "estimated_eval_tokens": 18568 + "serialized_chars": 36089, + "estimated_input_tokens": 9023, + "estimated_eval_tokens": 18302 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated Transformer issue reports with a few thematically similar pairs. Only the Qwen video-input regression pair looks like the same underlying bug; no PRs are present.", - "confidence": 0.61, - "canonical_issue_reason": "issue:43329 is the clearest root-cause report in the set: it names the broken video branch in `_get_num_multimodal_tokens` and affects multiple VLM processors.", + "summary": "The cluster is highly heterogeneous: most items are unrelated tokenizer, trainer, loading, or multimodal bugs. No soft pair is a safe duplicate merge under a strict same-root-cause standard.", + "confidence": 0.78, + "canonical_issue_reason": "No single canonical issue stands out because the set spans many unrelated bugs; the overall cluster is too mixed to collapse into one duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43329 is the best representative because it describes a concrete, centralized multimodal-video code-path bug with broad downstream impact.", + "best_issue_reason": "issue:44479 is the closest thing to a representative issue for the only semi-coherent subgroup (video-input regressions across Qwen/VLM models), but it still does not subsume the rest of the set.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45084", - "right": "issue:45538", - "accept": false, - "reason": "Different problems: compilation/template-node errors vs tokenizer max-length behavior." - }, - { - "left": "issue:43329", - "right": "issue:45600", - "accept": false, - "reason": "Unrelated subsystems: multimodal video token counting vs stale auto_mappings references." - }, { "left": "issue:44589", "right": "issue:45042", "accept": false, - "reason": "Different failure modes: missing Float8 storage vs image processor torchvision dependency." + "reason": "Different failures in different subsystems: float8 storage loading vs PIL image processor torchvision dependency." }, { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "Training/DeepSpeed index error is not the same as evaluate-after-train state failure." + "reason": "Both are training/eval failures, but one is a deepspeed/BERT zero3 index error and the other is a trainer lifecycle regression." }, { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "Different APIs and bugs: PEFT local checkpoint loading vs Whisper batch_decode skip_special_tokens handling." + "reason": "Local checkpoint loading with PEFT adapters is unrelated to Whisper batch_decode ignoring skip_special_tokens." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "Both concern loading, but different entrypoints and concrete regressions; not the same bug." + "reason": "One is gguf_file breaking from_pretrained loading; the other is a tokenizer regression for a specific model family." }, { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "Both mention OOM, but one is quantization-memory regression and the other is a model-specific memory limit issue." + "reason": "Both mention OOM, but the underlying causes and code paths are different model/runtime issues." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "Different model/code-path failures: torch.compile+SDPA vs MXFP4 load fallback." + "reason": "Torch.compile SDPA failure in Bamba is not the same bug as MXFP4 loading fallback in git-oss." }, { - "left": "issue:36010", - "right": "issue:43645", + "left": "issue:43819", + "right": "issue:43927", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to Jupyter custom-model initialization breakage." + "reason": "DAC latent conversion/STE mismatch is unrelated to DiaConfig losing custom token IDs on save/load." }, { - "left": "issue:43819", - "right": "issue:43927", + "left": "issue:36010", + "right": "issue:43645", "accept": false, - "reason": "Unrelated bugs: DAC forward/from_latents mismatch vs config fields lost after save/load." + "reason": "GenerationMixin import failure is a different problem from Jupyter custom-model initialization in Transformers 5.0." }, { "left": "issue:43257", "right": "issue:43329", "accept": false, - "reason": "Different loading/dispatch paths: Qwen3-MoE weight conversion vs multimodal token counting." + "reason": "Qwen3-MoE accelerate/deepspeed weight conversion and multimodal video token counting are different bugs." }, { "left": "issue:43854", "right": "issue:44977", "accept": false, - "reason": "Different models and symptoms; both are generation-related but not the same concrete bug." + "reason": "GLM-4.7-Flash load/test issues do not match the Qwen3.5 flash-attention generation bug." }, { "left": "issue:39692", "right": "issue:44912", "accept": false, - "reason": "SigLIP2 doc-example failures are unrelated to gpt-oss MXFP4 loading." - }, - { - "left": "issue:38175", - "right": "issue:44610", - "accept": false, - "reason": "Different vision-model issues: zero probabilities vs processor output shape/input-size mismatch." + "reason": "SigLIP2 docs example errors are unrelated to git-oss quantization loading behavior." }, { "left": "issue:43720", "right": "issue:44610", "accept": false, - "reason": "Accelerate packed-weight unpacking bug is unrelated to OmDet processor size mismatch." + "reason": "BitNet packed-weight unpacking during accelerate loading is unrelated to OmDet-Turbo processor image-size mismatch." }, { "left": "issue:45459", "right": "issue:45803", "accept": false, - "reason": "Both hide errors with exceptions, but they are different functions and different root causes." + "reason": "Both hide errors, but they are different code paths and different failure modes, so not the same duplicate." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 is not the same as a processor/model input-size mismatch." }, { "left": "issue:36010", "right": "issue:36331", "accept": false, - "reason": "Import-time GenerationMixin failure is unrelated to custom compute_loss keyword regression." + "reason": "GenerationMixin import and custom compute_loss kwargs are separate Trainer/AutoModel regressions." }, { "left": "issue:43454", "right": "issue:43577", "accept": false, - "reason": "Different multimodal model bugs: lm_head tying vs dtype propagation for BLIP2." + "reason": "AyaVision weight tying and BLIP2 dtype loading are different model-loading defects." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to PIL backend torchvision requirements." + "reason": "Qwen3-MoE sliding_window behavior has no concrete overlap with PIL backend torchvision requirements." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Different audio processor methods and symptoms; not the same bug." + "reason": "Audio apply_chat_template sample-rate defaulting and Whisper batch_decode token skipping are different processor behaviors." }, { "left": "issue:43329", "right": "issue:44479", - "accept": true, - "reason": "Both point to the same broken video-input path in Qwen multimodal processors; 44479 reads like the broader v5.3.0 manifestation of the helper bug in 43329." + "accept": false, + "reason": "43329 is a specific undefined-helper bug in multimodal token counting; 44479 is a broader video-input regression across models, not clearly the same root cause." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Qwen3-MoE loading/conversion is unrelated to GPT-oss OOM." + "reason": "Qwen3-MoE weight conversion and GPT-oss OOM are unrelated loading/performance problems." }, { "left": "issue:42490", "right": "issue:43257", "accept": false, - "reason": "Local-path import warning with trailing slash is unrelated to Qwen3-MoE accelerate/deepspeed weight conversion." + "reason": "Trailing-slash local path warning is unrelated to accelerate/deepspeed weight conversion for Qwen3-MoE." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "Both involve config handling, but different methods and failure modes; not the same underlying bug." + "reason": "Both involve config/model construction, but one is trust_remote_code missing fields and the other is text_config unwrapping for composite multimodal configs." }, { "left": "issue:43606", "right": "issue:45859", "accept": false, - "reason": "Both mention CPU offload/device handling, but they are model-specific issues in different code paths." + "reason": "CPU offload incompatibility appears in both, but the affected models and failure mechanisms are different." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Video token counting bug is unrelated to batched video processor output shape." + "reason": "Undefined helpers in video token counting are not the same as batched video processor output shape errors." }, { "left": "issue:43749", "right": "issue:44756", "accept": false, - "reason": "Different memory-loading contexts: FSDP CPU RAM efficient loading vs mmap on Strix Halo." + "reason": "FSDP CPU RAM efficient loading and Strix Halo mmap OOM are separate memory-loading issues." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "Trainer token counter bug is unrelated to GPT-oss perplexity." + "reason": "TrainerState token counting not updating is unrelated to GPT-oss perplexity results." }, { "left": "issue:44568", "right": "issue:45356", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different tokenizer code paths." + "reason": "Tokenizer special-token insertion regression and Kimi codec/regex warning issues are different tokenizer bugs." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3-MoE conversion/loading and Qwen3-VL weight-shape mismatch are different model-loading failures." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency is unrelated to TrainerState token accounting." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The only accepted soft edge (43329/44479) is reasonably supported by the overlapping Qwen video-input regression context, and the other pair rejections do not overclaim sameness. No PRs are present, so there is no mergeability issue." - }, - "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 }, @@ -5745,8 +5739,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -5774,13 +5768,13 @@ "issue:43782", "issue:43824", "issue:43927", - "issue:43931", "issue:44062", "issue:44291", "issue:44479", "issue:44488", "issue:44560", "issue:44568", + "issue:44737", "issue:44821", "issue:44849", "issue:44898", @@ -5795,12 +5789,11 @@ "issue:45405", "issue:45488", "issue:45538", + "issue:45588", "issue:45701", "issue:45907" ], "soft_pairs": [ - "issue:43257|issue:43931", - "issue:29159|issue:34567", "issue:42915|issue:43742", "issue:44291|issue:45042", "issue:42915|issue:44821", @@ -5819,246 +5812,250 @@ "issue:43540|issue:44560", "issue:43317|issue:45538", "issue:44062|issue:44991", + "issue:44987|issue:45907", "issue:24643|issue:29159", "issue:45137|issue:45405", - "issue:44987|issue:45907", - "issue:28282|issue:34689", "issue:43577|issue:45072", + "issue:28282|issue:34689", "issue:43742|issue:44291", "issue:43257|issue:43756", - "issue:28282|issue:29127", "issue:42175|issue:44488", + "issue:28282|issue:29127", "issue:42915|issue:43749", "issue:43582|issue:43782", + "issue:44737|issue:45588", "issue:38175|issue:44898" ], - "prepared_review_unit_hash": "427ab9fafce1a8b60dfd7648f4ae88d095bf249cdc3943a96d8a607f260c7cf7", + "prepared_review_unit_hash": "aa70bce33535ce5ac6c117817aef01d82c3be02338b196b0b4ee0a4d80915a0f", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36692, - "estimated_input_tokens": 9173, - "estimated_eval_tokens": 18602 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37051, + "estimated_input_tokens": 9263, + "estimated_eval_tokens": 18782 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues mostly share model-loading/tokenization language but describe different concrete failures. I did not find any soft pair that looks like the same underlying bug or change; they span distinct code paths such as DeepSpeed/FSDP loading, video preprocessing, tokenizer behavior, quantization, and model-specific config regressions.", - "confidence": 0.9, - "canonical_issue_reason": "issue:45701 is the best umbrella-style issue in this set: it is broad, open, and explicitly about version-dependent tokenization changes, which makes it the closest thing to a cluster representative.", + "summary": "Mostly heterogeneous issue cluster with one credible duplicate pair around Qwen VL video-input regression. The rest are separate bugs across tokenizer, loading, quantization, device-mapping, and backend behavior.", + "confidence": 0.87, + "canonical_issue_reason": "Issue 44479 is the best canonical issue for the only real overlap: it describes a broad Qwen video-input regression affecting multiple related models, and 45381 looks like a narrower report of the same problem.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45701 is the strongest global representative because it is broad enough to subsume related tokenizer regressions, has active inbound references, and is phrased as a general behavioral regression rather than a narrow model-specific symptom.", + "best_issue_reason": "44479 is the strongest representative issue in this set because it is broad, concrete, and has the clearest duplicate-like companion report; most other items are too unrelated to serve as a cluster anchor.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43257", - "right": "issue:43931", - "accept": false, - "reason": "Both are Qwen3/VL-related loading bugs, but one is about MoE weight conversion under accelerate+deepspeed and the other is a weight-shape mismatch for Qwen3-VL-30B-A3B-Instruct. Different failure modes and code paths." - }, - { - "left": "issue:29159", - "right": "issue:34567", - "accept": false, - "reason": "Tokenizer inconsistency vs TrainerState token-count tracking. Same broad area only at a high level, but unrelated bugs." - }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Qwen3Moe fine-grained FP8 failure vs MobileLLM key error during model loading. Different models and different loading/quantization failures." + "reason": "Different failures: Qwen3Moe FineGrainedFP8 loading vs MobileLLM key error." }, { "left": "issue:44291", "right": "issue:45042", "accept": false, - "reason": "Init-empty-weights argument handling vs PIL backend requiring torchvision. Completely different subsystems." + "reason": "Unrelated code paths: init_empty_weights/_is_hf_initialized vs PIL image processor torchvision dependency." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "FP8 config failure vs AutoImageProcessor URL loading. Different functionality and error classes." + "reason": "Different subsystems: FP8 Qwen3Moe loading vs AutoImageProcessor URL loading." }, { "left": "issue:44849", "right": "issue:45701", "accept": false, - "reason": "Qwen3.5 output_hidden_states bug is a model-specific forward-pass regression, while 45701 is a broad tokenization/version-change complaint." + "reason": "Different bugs: Qwen3.5 hidden-states behavior vs general tokenization changes across versions." }, { "left": "issue:44568", "right": "issue:45488", "accept": false, - "reason": "mdeberta tokenizer add_special_tokens regression vs LlamaTokenizer overriding ByteLevel with Metaspace. Both tokenizer-related, but distinct tokenizer implementations and behaviors." + "reason": "Both tokenizer-related, but different model families and different failure mechanisms; not the same bug." }, { "left": "issue:44987", "right": "issue:45538", "accept": false, - "reason": "Loading a specific model fails vs CLIPTokenizer max length bug. Not the same issue." + "reason": "Unrelated: model loading failure for one repo vs CLIPTokenizer max_length bug." }, { "left": "issue:43452", "right": "issue:43582", "accept": false, - "reason": "gguf_file breaks pretrained loading APIs vs Apple Silicon TypeError in caching_allocator_warmup. Unrelated." + "reason": "Completely different areas: gguf/AutoModel loading vs Apple Silicon allocator warmup TypeError." }, { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Qwen3 auto device map cuda assert vs Qwen3OmniMoe video-input ValueError. Same family name, different concrete failure." + "reason": "Different Qwen issues and different symptoms: auto device-map CUDA assert vs video-input ValueError." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "RTDetrV2 output mismatch in v5 vs Qwen3.5 save_pretrained saving wrong visual encoder keys. Different model behavior and different lifecycle stage." + "reason": "Different regressions: RTDetrV2 output drift vs save_pretrained key serialization." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "DiaConfig token IDs lost on save/load vs MXFP4 quantization fallback. Different models and unrelated bugs." + "reason": "Different model-loading failures with different causes: config token ID loss vs MXFP4 fallback." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Qwen3Moe FP8 loading failure vs CLIPTokenizer model_max_length bug. No substantive overlap." + "reason": "No overlap beyond being Transformers issues; separate topics." }, { "left": "issue:44479", "right": "issue:45381", - "accept": false, - "reason": "Qwen2.5/Qwen3.5 video-input regression vs qwen2.5-vl vision_position_ids wrong. Related family, but one is a broad multi-model regression and the other is a specific positional-ID bug." + "accept": true, + "reason": "Both report the same Qwen VL video-input regression; 45381 is a narrower manifestation of the broader v5.3.0 breakage in 44479." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "MacOS bus error with community CLIP model vs TrainerState token counter bug. Different components and symptoms." + "reason": "Unrelated: macOS CLIP bus error vs TrainerState token-count tracking." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Qwen3OmniMoe video processing ValueError vs broader v5.3 video-input regression across several model families. Overlapping theme, but not the same concrete bug." + "reason": "Both concern video input, but different model families and symptoms; not enough evidence of one underlying bug." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "TensorFlow missing from torch install vs import error for Qwen2_5_VLForConditionalGeneration. Dependency packaging vs symbol export/import issue." + "reason": "Different problems: packaging extras missing TensorFlow vs missing Qwen2.5-VL symbol import." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Different issues: Qwen3 MoE weight conversion vs Smollm3 RoPE-layer mismatch." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Qwen3OmniMoe video ValueError vs Qwen3-vl-embedding StopIteration in 5.3.0. Different model and different exception path." + "reason": "Similar area, but separate Qwen video-input bugs with different failures and model families." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "device_map=auto dequantized model offload failure vs CLIPTokenizer max length bug. Different code paths." + "reason": "Unrelated: device_map/offload loading bug vs CLIPTokenizer max_length." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "AddedToken special-argument collision vs tokenizer loading failure for est-roberta. Same area, but not the same bug." + "reason": "Both tokenizer loading failures, but the reports do not clearly point to the same underlying AddedToken bug." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Different API regressions: model loading for a repo vs list_repo_files not forwarding revision." }, { "left": "issue:24643", "right": "issue:29159", "accept": false, - "reason": "DeepSpeed training runtime error about weight dimensionality vs tokenizer slow/fast inconsistency. Unrelated." + "reason": "Distinct bugs: DeepSpeed training RuntimeError vs tokenizer slow/fast inconsistency." }, { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error vs PEFT minimum version bump. Completely unrelated." + "reason": "Unrelated: ZeRO3 deque error vs PEFT version bump." }, { - "left": "issue:44987", - "right": "issue:45907", + "left": "issue:43577", + "right": "issue:45072", "accept": false, - "reason": "Model loading failure for physical-intelligence/fast vs list_repo_files not forwarding revision. Different APIs and failures." + "reason": "Both dtype-related, but different models and different failure modes; not the same code-path problem." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "Missing PyTorch ImportError for AutoModel vs Llama 3.2 90B Vision model loading regression. Both are load-time errors, but the causes are different." - }, - { - "left": "issue:43577", - "right": "issue:45072", - "accept": false, - "reason": "Blip2 dtype staying float32 vs SwitchTransformers/TimmWrapperModel dtype mismatches in bf16 inference. Both dtype-related, but different models and symptoms." + "reason": "Different issues: missing PyTorch import vs a specific model-loading regression." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "MobileLLM key error while loading a model vs init_empty_weights TypeError from unexpected internal argument. Not the same bug." + "reason": "Distinct loading bugs with different stack traces and causes." }, { "left": "issue:43257", "right": "issue:43756", "accept": false, - "reason": "Qwen3 MoE weight conversion/loading issue vs Smollm3 RoPE layer count mismatch. Different model families and different bug types." + "reason": "Separate model-specific bugs; no shared code-path evidence." }, { - "left": "issue:28282", - "right": "issue:29127", + "left": "issue:42175", + "right": "issue:44488", "accept": false, - "reason": "PyTorch missing ImportError vs LayoutLMv3 box-information error message clarity. No duplicate signal." + "reason": "Packaging/backend dependency issue vs model loading failure; not duplicates." }, { - "left": "issue:42175", - "right": "issue:44488", + "left": "issue:28282", + "right": "issue:29127", "accept": false, - "reason": "Packaging dependency omission vs failure to load cjvt/sleng-bert. Different issues." + "reason": "Different errors and domains: missing torch import vs LayoutLMv3 box validation message." }, { - "left": "issue:42915", - "right": "issue:43749", + "left": "issue:43749", + "right": "issue:45405", "accept": false, - "reason": "Qwen3Moe FP8 failure vs FSDP_CPU_RAM_EFFICIENT_LOADING broken. Both involve advanced loading, but distinct mechanisms and models." + "reason": "Different issues: FSDP CPU RAM efficient loading regression vs unreleased PEFT version bump." }, { "left": "issue:43582", "right": "issue:43782", "accept": false, - "reason": "Apple Silicon warmup TypeError vs Qwen3VL weight_only=True error. Different backend and failure mode." + "reason": "Different failures: Apple Silicon allocator TypeError vs Qwen3VL weight_only loading error." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "Unrelated: XLNet device placement bug vs flash-attention integration crash." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "Unexpected zero probabilities in a specific model vs Perceiver non-default resolution failure. Separate model behaviors, not duplicates." + "reason": "Different model behaviors and failure modes; not the same underlying bug." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: it identifies one plausible duplicate-like pair (44479/45381) and treats the rest as separate issues. The soft-edge verdicts are generally conservative and do not overclaim duplication from the available titles/evidence." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -6070,8 +6067,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -6087,9 +6084,11 @@ "issue:43531", "issue:43540", "issue:43643", + "issue:43688", "issue:43697", "issue:43720", "issue:43742", + "issue:43749", "issue:43756", "issue:43906", "issue:43994", @@ -6099,7 +6098,6 @@ "issue:44514", "issue:44589", "issue:44625", - "issue:44737", "issue:44792", "issue:44821", "issue:44877", @@ -6116,7 +6114,6 @@ "issue:45381", "issue:45405", "issue:45479", - "issue:45588", "issue:45600", "issue:45823", "issue:45920", @@ -6125,21 +6122,20 @@ "issue:46202" ], "soft_pairs": [ - "issue:44737|issue:45588", "issue:43540|issue:45823", "issue:43404|issue:45987", "issue:43526|issue:45479", "issue:43906|issue:45381", "issue:43493|issue:43994", - "issue:34567|issue:36331", "issue:44291|issue:44821", + "issue:34567|issue:36331", "issue:43756|issue:45325", "issue:44913|issue:46121", "issue:44062|issue:45081", "issue:44792|issue:44964", "issue:45042|issue:45081", - "issue:45357|issue:45920", "issue:43742|issue:44589", + "issue:45357|issue:45920", "issue:43452|issue:44488", "issue:43531|issue:44821", "issue:43643|issue:45375", @@ -6150,198 +6146,199 @@ "issue:43697|issue:45216", "issue:43906|issue:46202", "issue:41762|issue:44936", - "issue:44514|issue:45381" + "issue:44514|issue:45381", + "issue:43688|issue:43749" ], - "prepared_review_unit_hash": "667db3c19375eec2524434462e643aecf3f81564776b50f3075753054ac4bd45", + "prepared_review_unit_hash": "84a95db6ec4e5576626fa6937da946ec8e85b4afcf78f66024583e17f04b359a", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 26, - "serialized_chars": 35941, - "estimated_input_tokens": 8986, - "estimated_eval_tokens": 18228 + "serialized_chars": 35835, + "estimated_input_tokens": 8959, + "estimated_eval_tokens": 18174 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated closed issues across different models/components. The only plausible duplicate pair is the SigLIP2 implementation mismatch vs. user-visible nonsensical outputs; everything else looks distinct and should stay separate.", - "confidence": 0.78, - "canonical_issue_reason": "No single canonical issue fits the whole cluster; if we need one anchor, issue 43493 is the best root-cause-style report for the only likely duplicate pair (SigLIP2).", + "summary": "The cluster is mostly a set of loosely related bug reports connected only by superficial model/subsystem similarity. I found two likely true duplicate pairs: SigLIP2 discrepancy reports, and Qwen2.5-VL video/position-id reports. The rest are distinct bugs and should not be merged.", + "confidence": 0.7, + "canonical_issue_reason": "issue:43493 is the strongest root-cause style report among the clear duplicate pairs: it describes the underlying SigLIP2 HF-vs-JAX implementation discrepancy, while the paired issue is just a concrete symptom on a specific checkpoint.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43493 is the clearest technical statement of the underlying bug, while 43994 reads more like a downstream symptom/reproduction on the same SigLIP2 path.", + "best_issue_reason": "issue:45920 is the best overall issue for the cluster as a whole because it is a broad, current regression report with multiple affected models and clear user impact, making it the most useful general anchor if one issue had to represent the set.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44737", - "right": "issue:45588", - "accept": false, - "reason": "Unrelated problems: XLNet CPU rope indexing vs flash-attention aux-state crash." - }, { "left": "issue:43540", "right": "issue:45823", "accept": false, - "reason": "Different models and failure modes: Qwen3OmniMoe video processing vs Gemma4 device-mapping mismatch." + "reason": "Different models and failures: Qwen3OmniMoe video processing vs Gemma4 device placement under auto device mapping." }, { "left": "issue:43404", "right": "issue:45987", "accept": false, - "reason": "Different subsystems: Mistral3 weight tying vs StaticCache seq-length return type." + "reason": "Unrelated bugs: missing lm_head tying in Mistral3 vs StaticCache sequence-length return type." }, { "left": "issue:43526", "right": "issue:45479", "accept": false, - "reason": "Unrelated image-label reduction bug vs sequence-classification loss degeneracy." + "reason": "Different layer of the stack: BeitImageProcessorFast label reduction bug vs sequence-classification zero-loss bug across models." }, { "left": "issue:43906", "right": "issue:45381", - "accept": false, - "reason": "Both involve Qwen/VL themes, but one is a reproduction of another issue and the other is a specific video position-id bug; not enough evidence of the same concrete defect." + "accept": true, + "reason": "Both point to the same Qwen2.5-VL video/position-id code path; 43906 is an isolated reproduction of the same underlying bug." }, { "left": "issue:43493", "right": "issue:43994", "accept": true, - "reason": "Both point to the same SigLIP2 model correctness problem: HF/JAX discrepancy and broken/nonsensical outputs on the same model family." + "reason": "Same SigLIP2 implementation problem: one report is the general HF-vs-JAX discrepancy, the other is a specific model/pipeline symptom." }, { - "left": "issue:34567", - "right": "issue:36331", + "left": "issue:44291", + "right": "issue:44821", "accept": false, - "reason": "Trainer state token counting and custom loss signature incompatibility are different bugs." + "reason": "Different failure modes: init_empty_weights/_is_hf_initialized argument handling vs loading AutoImageProcessor from a URL." }, { - "left": "issue:44291", - "right": "issue:44821", + "left": "issue:34567", + "right": "issue:36331", "accept": false, - "reason": "Different areas: init_empty_weights/_is_hf_initialized loading path vs AutoImageProcessor URL loading." + "reason": "Both trainer-related, but they describe different bugs in different trainer code paths." }, { "left": "issue:43756", "right": "issue:45325", "accept": false, - "reason": "Different models and different RoPE-related bugs." + "reason": "Both involve RoPE, but the concrete bugs are different models and different position/rope calculations." }, { "left": "issue:44913", "right": "issue:46121", "accept": false, - "reason": "Config reload defaulting and rope-validation list handling are separate issues." + "reason": "Both touch rope config handling, but one is silent config reload fallback and the other is a TypeError during JSON-loaded validation." }, { "left": "issue:44062", "right": "issue:45081", "accept": false, - "reason": "Tokenizer AddedToken argument collision is unrelated to Mistral regex patching." + "reason": "Distinct tokenizer bugs: AddedToken special-arg duplication vs Mistral regex patching against a tokenizer without backend_tokenizer." }, { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "Both are multimodal-related, but one is a failing test and the other is a model-loading regression for a different model." + "reason": "Different models and symptoms: janus test failure vs Phi-4 multimodal model loading failure." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Image processor torchvision dependency issue is unrelated to tokenizer regex crash." + "reason": "PIL/image-processor torchvision dependency regression is unrelated to Mistral tokenizer regex patching." }, { - "left": "issue:45357", - "right": "issue:45920", + "left": "issue:43742", + "right": "issue:44589", "accept": false, - "reason": "Different regression classes: save_pretrained key serialization vs incorrect token IDs from AutoTokenizer." + "reason": "Different loading failures: MobileLLM key error vs Float8 storage lookup failure." }, { - "left": "issue:43742", - "right": "issue:44589", + "left": "issue:45357", + "right": "issue:45920", "accept": false, - "reason": "Different load failures: MobileLLM key error vs missing Float8 storage type." + "reason": "Both are save/load regressions, but they affect different APIs and different model families with no shared code path evidence." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "Both are loading issues, but for different models and likely different code paths." + "reason": "Both are loading issues, but one is gguf_file/from_pretrained handling and the other is a specific model load failure without clear shared cause." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Qwen3-MoE sliding window behavior and AutoImageProcessor URL loading are unrelated." + "reason": "Unrelated subsystems: Qwen3-MoE sliding-window behavior vs image-processor loading from URL." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both mention missing config fields, but the former is a generic trust_remote_code AutoConfig issue and the latter is a model-specific @strict field drop; not clearly the same bug." + "reason": "Both are config/strictness issues, but one is generic trust_remote_code field loss and the other is a specific Qwen vision config field being dropped." }, { "left": "issue:45405", "right": "issue:45600", "accept": false, - "reason": "Version-pin/release issue is unrelated to stale auto_mappings references." + "reason": "Different maintenance issues: PEFT version pin vs stale auto_mappings references breaking CI." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "num_labels propagation and strict config loading are different configuration bugs." + "reason": "Different config bugs: num_labels propagation vs strict config rejecting granite_speech." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed training RuntimeError and LayoutLMv3 error-message clarity are unrelated." + "reason": "Different problems and layers: DeepSpeed training RuntimeError vs LayoutLMv3 error-message clarity." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "BitNet packed-weight loading and Perceiver interpolation failure are different model-specific bugs." + "reason": "Different model/loading paths: BitNet packed-weight unpacking vs Perceiver non-default-resolution handling." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "RTDetrV2 output drift and Qwen3.5 save_pretrained checkpoint corruption are unrelated regressions." + "reason": "Different regressions: RTDetrV2 output drift vs Qwen3.5 save_pretrained checkpoint corruption." }, { "left": "issue:43906", "right": "issue:46202", "accept": false, - "reason": "No clear shared underlying defect; one is a reproduction of an older issue, the other is a Gemma pipeline roundtrip bug." + "reason": "Both mention model/tokenization behavior, but one is a Qwen2.5-VL reproduction and the other is a Gemma pipeline roundtrip issue." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "Gemma3 ZeRO-3 load failure and trainer.evaluate-after-train are unrelated." + "reason": "Different bugs: Gemma3 ZeRO-3 load failure vs trainer.evaluate() after trainer.train()." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Same model family, but batched chat-template padding crash is a different code path from wrong video vision_position_ids." + "reason": "Both are Qwen2.5-VL issues, but one is batched chat-template padding behavior and the other is video vision_position_ids; different code paths." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Different training/runtime issues: auxiliary-loss normalization vs FSDP CPU RAM efficient loading." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall: the packet is heterogeneous, and the soft-edge rejections are mostly conservative given the available titles. The only accepted pair, issue:43493 / issue:43994, is a reasonable likely duplicate on the same SigLIP2 path based on the titles alone. Minor caveat: the cluster summary says \"closed issues\" even though a few items are open, but that does not materially affect the verdict." + "feedback": "The summary is grounded in the packet and stays mostly conservative. The broad anchor choice for issue:45920 is defensible from the title, and the two accepted duplicate pairs are plausible given the shared model/topic signals. The rejected soft-edge pairs are also reasoned at a cautious level. No obvious overstatement that would justify rejecting." }, "evaluator_used": true, "retried": false, @@ -6357,8 +6354,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -6412,21 +6409,20 @@ "issue:46178" ], "soft_pairs": [ - "issue:43688|issue:43749", "issue:45372|issue:46170", "issue:43452|issue:43577", "issue:44466|issue:45005", "issue:44493|issue:46050", "issue:43819|issue:44811", "issue:30064|issue:33666", - "issue:43452|issue:44843", "issue:44936|issue:45823", + "issue:43452|issue:44843", "issue:45542|issue:46178", "issue:44484|issue:44704", "issue:43295|issue:45356", "issue:28282|issue:32090", - "issue:43653|issue:44361", "issue:44361|issue:44442", + "issue:43653|issue:44361", "issue:28282|issue:33357", "issue:44514|issue:45290", "issue:44464|issue:44743", @@ -6437,227 +6433,221 @@ "issue:43577|issue:43697", "issue:43688|issue:45704", "issue:44843|issue:44991", - "issue:43646|issue:43749", "issue:43531|issue:43742", + "issue:43646|issue:43749", "issue:29159|issue:33357", "issue:43688|issue:45440", - "issue:43688|issue:45701", - "issue:43653|issue:43927" + "issue:43653|issue:43927", + "issue:43688|issue:45701" ], - "prepared_review_unit_hash": "413774f79871653359b4834ac934b7e853611d94b3f9b3e1aadc669efa9c5f7e", + "prepared_review_unit_hash": "0467617e4ab7314dc8a06a8b4d17c820d42d89c309421fbc0c4e6489e2b1ae14", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 37143, - "estimated_input_tokens": 9286, - "estimated_eval_tokens": 18828 + "soft_pair_count": 30, + "serialized_chars": 36842, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct bugs across tokenization, loading, training, serialization, and backend detection. I would not merge any of the proposed soft edges as duplicates.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 45701 is the broadest umbrella in the set, but the cluster is still too mixed to make it a safe duplicate canonical.", + "summary": "This cluster is broadly heterogeneous: it mixes unrelated tokenizer, model-loading, training, generation, backend, and platform-specific regressions. None of the soft-similarity pairs look like the same concrete bug/change, so no duplicates should be merged here.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the items span many unrelated failure modes, so there is no representative issue that safely subsumes the rest.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45701 is the best available anchor because it is broad, open, and about tokenizer/version regressions, but it still does not represent the unrelated training/loading/model-init failures in the rest of the set.", + "best_issue_reason": "No global best issue either; the cluster is not one duplicate family, so any single issue would be misleading as a representative.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43688", - "right": "issue:43749", - "accept": false, - "reason": "Auxiliary-loss normalization and FSDP CPU RAM loading are different bugs in different code paths." - }, { "left": "issue:45372", "right": "issue:46170", "accept": false, - "reason": "Both are import/dependency failures, but they involve different packages, platforms, and root causes." + "reason": "Different import/dependency failures: Gemma processor loading vs Windows AMD ROCm torch.distributed import/type errors." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "gguf_file loading and Blip2 dtype propagation are unrelated failures." + "reason": "Different bugs: gguf_file breaks tokenizer/model loading, while this is a BLIP2 dtype preservation issue." }, { "left": "issue:44466", "right": "issue:45005", "accept": false, - "reason": "Both mention tied weights, but one is device-dependent serialization and the other is a broader translation-model behavior report; not the same concrete bug." + "reason": "Both mention tied weights, but one is serialization/device-dependent lm_head.weight behavior and the other is a broader translation-model tied-weights issue." }, { "left": "issue:44493", "right": "issue:46050", "accept": false, - "reason": "Unexpected position-id keys and sinq quantization save/load breakage are different issues." + "reason": "Different persistence problems: unexpected position-id keys vs quantized model save/load breakage." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC latent mismatch and Whisper batch_decode skip_special_tokens are unrelated code paths." + "reason": "Different code paths: DAC latent reconstruction mismatch vs Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:30064", "right": "issue:33666", "accept": false, - "reason": "Void segmentation map processing and Qwen2-VL multi-GPU training are different problems." + "reason": "Unrelated areas: void segmentation-map processing vs Qwen2-VL multi-GPU training." }, { - "left": "issue:43452", - "right": "issue:44843", + "left": "issue:44936", + "right": "issue:45823", "accept": false, - "reason": "gguf_file loading and offline mistral-regex patch behavior are separate tokenizer/model-loading bugs." + "reason": "Different failures: trainer lifecycle regression vs Gemma4 device-map mismatch during forward." }, { - "left": "issue:44936", - "right": "issue:45823", + "left": "issue:43452", + "right": "issue:44843", "accept": false, - "reason": "trainer.evaluate() after train and Gemma4 device-map mismatch are unrelated training/runtime issues." + "reason": "Both involve tokenizer loading, but one is gguf_file handling and the other is an offline model_info / regex patch regression." }, { "left": "issue:45542", "right": "issue:46178", "accept": false, - "reason": "These are different backend-detection/import problems involving different dependencies." + "reason": "Both are backend dependency issues, but they involve different backends and different failure modes (tensorboard/TF vs tensorflow_text)." }, { "left": "issue:44484", "right": "issue:44704", "accept": false, - "reason": "max_shard_size documentation/behavior and cached_file kwargs forwarding are not the same bug." + "reason": "Different APIs: save_pretrained shard-size behavior vs AutoProcessor kwargs propagation to cached_file." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "Processor/tokenizer regression and Kimi-K2.5 codec handling are distinct tokenizer issues." + "reason": "Both are tokenizer/processor regressions, but they affect different models and different root causes." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "Missing PyTorch import and NoneType broadcast errors are unrelated failures." + "reason": "Missing PyTorch ImportError vs a Trainer distributed broadcast TypeError; not the same underlying bug." }, { - "left": "issue:43653", - "right": "issue:44361", + "left": "issue:44361", + "right": "issue:44442", "accept": false, - "reason": "BigBird special-token registration and MLukeTokenizer AttributeError are different tokenizer bugs." + "reason": "Different tokenizer failures: MLuke task AttributeError vs AutoTokenizer load failure for FastSpeech2ConformerTokenizer." }, { - "left": "issue:44361", - "right": "issue:44442", + "left": "issue:43653", + "right": "issue:44361", "accept": false, - "reason": "These are different tokenizer-loading failures for different models." + "reason": "Different tokenizers and symptoms: BigBird special-token registration vs MLuke tokenizer AttributeError." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "PyTorch import error and Mac bus error on CLIP are unrelated." + "reason": "Different environments and failure types: missing PyTorch vs MacOS bus error on a CLIP model." }, { "left": "issue:44514", "right": "issue:45290", "accept": false, - "reason": "Both are chat-template crashes, but they hit different inputs and failure modes, so they are not clearly the same bug." + "reason": "Same general API family, but different crashes and inputs in apply_chat_template; not one concrete bug." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Compiled chunked generation inconsistency and Qwen3 modular recurrent-state reset are different generation bugs." + "reason": "Both touch generation/cache behavior, but on different models and distinct code-path bugs." }, { "left": "issue:38175", "right": "issue:43295", "accept": false, - "reason": "Siglip2 zero probabilities and processor.tokenizer regression are unrelated." + "reason": "Unrelated: zero probabilities in SigLIP2 vs processor.tokenizer regression." }, { "left": "issue:28282", "right": "issue:35141", "accept": false, - "reason": "Missing PyTorch import and output-embedding reinitialization after resize are different issues." + "reason": "Missing-library ImportError vs token-embedding resize/post_init reinitialization bug." }, { "left": "issue:41720", "right": "issue:43906", "accept": false, - "reason": "A device-map cudaErrorAssert and a reproduction issue are not the same concrete bug." + "reason": "The second issue is only an isolated reproduction of another report; there is no evidence it is the same Qwen3 device-map cudaErrorAssert bug." }, { "left": "issue:43653", "right": "issue:44568", "accept": false, - "reason": "Both involve special tokens, but they affect different tokenizers and different mechanics." + "reason": "Both involve special tokens, but they are different tokenizers and distinct behaviors." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "Blip2 dtype propagation and RTDetrV2 output divergence are unrelated." + "reason": "Different problems: BLIP2 dtype casting vs RTDetrV2 output divergence in v5." }, { "left": "issue:43688", "right": "issue:45704", "accept": false, - "reason": "Auxiliary-loss normalization and T5 apex RMSNorm memory leak are unrelated." + "reason": "Different model bugs: auxiliary-loss normalization vs an apex RMSNorm memory leak." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Offline model_info call in the mistral regex patch and tokenizer load failure for est-roberta have different root causes." + "reason": "Both are tokenizer-loading regressions, but one is offline/model_info behavior and the other is a specific model tokenizer load failure." }, { - "left": "issue:43646", - "right": "issue:43749", + "left": "issue:43531", + "right": "issue:43742", "accept": false, - "reason": "Custom model initialization breakage and FSDP CPU RAM loading are unrelated." + "reason": "Different model-loading/runtime bugs: Qwen3-MoE sliding_window vs MobileLLM key error." }, { - "left": "issue:43531", - "right": "issue:43742", + "left": "issue:43646", + "right": "issue:43749", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and MobileLLM-125M key errors are different model-specific bugs." + "reason": "Different regression classes: custom model initialization breakage vs FSDP CPU RAM-efficient loading." }, { "left": "issue:29159", "right": "issue:33357", "accept": false, - "reason": "Slow/fast tokenizer inconsistency and MacOS bus error are not the same failure." + "reason": "Tokenizer inconsistency vs MacOS bus error; unrelated symptoms and subsystems." }, { "left": "issue:43688", "right": "issue:45440", "accept": false, - "reason": "Auxiliary-loss normalization and DeepSeekV3 MoE divergence are unrelated model bugs." + "reason": "Both involve MoE models, but one is auxiliary-loss normalization and the other is a native-vs-remote implementation divergence." }, { - "left": "issue:43688", - "right": "issue:45701", + "left": "issue:43653", + "right": "issue:43927", "accept": false, - "reason": "A specific auxiliary-loss bug and a broad tokenization-regression report do not match the same underlying issue." + "reason": "Different tokenizer/config bugs: BigBird special-token registration vs DiaConfig losing custom token IDs." }, { - "left": "issue:43653", - "right": "issue:43927", + "left": "issue:43688", + "right": "issue:45701", "accept": false, - "reason": "BigBird special-token registration and DiaConfig save/load token-ID loss are different persistence/tokenizer issues." + "reason": "Unrelated: auxiliary-loss normalization vs a broad tokenization-version regression." } ] }, @@ -6676,8 +6666,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -6721,6 +6711,7 @@ "issue:44849", "issue:44913", "issue:44977", + "issue:45357", "issue:45491", "issue:45588", "issue:45823", @@ -6745,8 +6736,8 @@ "issue:43504|issue:43720", "issue:43504|issue:44610", "issue:32090|issue:33666", - "issue:44466|issue:46050", "issue:33666|issue:36010", + "issue:44466|issue:46050", "issue:43742|issue:44821", "issue:43720|issue:43927", "issue:45823|issue:46050", @@ -6762,234 +6753,241 @@ "issue:43756|issue:44913", "issue:44625|issue:46002", "issue:43493|issue:43825", - "issue:28218|issue:33357" + "issue:28218|issue:33357", + "issue:44849|issue:45357" ], - "prepared_review_unit_hash": "d9f3e1583510069040ed682b12a2b78559e6b25b19b19eb45e423a4ec2ebc44a", + "prepared_review_unit_hash": "2370325747955e6cb4f1eedc96ba6451d3cb3a7f270be2d83392001cbd7c1f02", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 37046, - "estimated_input_tokens": 9262, - "estimated_eval_tokens": 18780 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37932, + "estimated_input_tokens": 9483, + "estimated_eval_tokens": 19222 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a cohesive duplicate cluster: the issues span unrelated bugs across training, loading, config serialization, multimodal processors, attention kernels, and model-specific regressions. The soft-similarity pairs mostly share only broad keywords, not the same concrete code path.", - "confidence": 0.98, + "summary": "This cluster is heterogeneous: the soft pairs share generic Transformers terms (config, loading, training, device/dtype, tokenizer) but describe different concrete bugs and code paths. I would not merge any of the suggested pairs as duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43550 is the strongest standalone report here: it names a specific model, a concrete failure mode (torch.compile with SDPA), and a plausible single code path to fix.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45491", "right": "issue:46002", "accept": false, - "reason": "Both mention NaNs, but one is Gemma3 sliding-window attention on GPU and the other is classification-head NaNs after checkout/init; different symptoms and likely different causes." + "reason": "Different failures: Gemma3 mixed-length sliding-window attention NaNs vs classification-head NaNs on checkout/init." }, { "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "One is a Bamba torch.compile/SDPA failure; the other is a DiaConfig save/load field-loss bug. No shared code path." + "reason": "One is Bamba-9B-v2 torch.compile SDPA, the other DiaConfig save/load custom token IDs and generation IndexError." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Both affect generation, but one is chunked generation with compiled forward and the other is Qwen3.5 flash-attention inference. Too broad to merge." + "reason": "Both involve generation, but one is chunked generation with compiled forward; the other is Qwen3.5 flash-attention inference failure." }, { "left": "issue:43643", "right": "issue:44561", "accept": false, - "reason": "Both involve trust_remote_code-adjacent behavior, but one is missing fields in AutoConfig output and the other is a v5 API removal breaking remote-code models. Different fixes." + "reason": "Both are trust_remote_code-adjacent, but one is missing fields from AutoConfig.from_pretrained and the other is removal of is_torch_fx_available breaking remote-code models." }, { "left": "issue:43550", "right": "issue:44361", "accept": false, - "reason": "Completely different models and failures: SDPA/compile versus tokenizer AttributeError on tasks." + "reason": "Bamba torch.compile/SDPA bug is unrelated to MLukeTokenizer AttributeError on tasks." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Both are Qwen3.5-related, but one is a video-input regression and the other is num_labels propagation into text config." + "reason": "Video-input regression for several Qwen VL models is not the same as Qwen3.5 num_labels propagation." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "A CUDA assert in Qwen3 auto device mapping is unrelated to chunked generation output inconsistency under compiled forward." + "reason": "Qwen3 auto device mapping cudaErrorAssert is a different path from chunked generation inconsistency under compiled forward." }, { "left": "issue:44977", "right": "issue:46209", "accept": false, - "reason": "Both are inference bugs, but one is flash-attention behavior in Qwen3.5 and the other is FP8 inference on B200; different paths." + "reason": "Qwen3.5 flash-attention generation bug is unrelated to FP8 inference failure on B200." }, { "left": "issue:43577", "right": "issue:45588", "accept": false, - "reason": "One is dtype drift in BLIP2 loading, the other is a flash_attention integration crash on s_aux=None; unrelated." + "reason": "Blip2 dtype persistence and flash_attention.py s_aux=None crash are unrelated model/runtime bugs." }, { "left": "issue:32090", "right": "issue:34567", "accept": false, - "reason": "Trainer broadcast TypeError and TrainerState token-count tracking are separate training bugs." + "reason": "Trainer GPU broadcast NoneType error is different from TrainerState.num_input_tokens_seen not updating." }, { "left": "issue:43577", "right": "issue:44554", "accept": false, - "reason": "BLIP2 dtype mismatch and an MPS attention correctness issue are different subsystems and failure modes." + "reason": "Blip2 dtype mismatch is unrelated to the MPS attention value/query head-dim correctness issue." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "Beit legacy-field loading and Bamba compile/SDPA are unrelated model-specific bugs." + "reason": "Legacy field blocking BEiT semantic segmentation loading is not the same as Bamba SDPA compile failure." }, { "left": "issue:43504", "right": "issue:43720", "accept": false, - "reason": "Legacy config-field loading for Beit is not the same as BitNet packed-weight handling during accelerate loading." + "reason": "Legacy preset loading failure is unrelated to BitNet packed-weight unpacking during accelerate loading." }, { "left": "issue:43504", "right": "issue:44610", "accept": false, - "reason": "Beit pretrained-load failure and OmDet processor/model size mismatch are unrelated." + "reason": "BEiT legacy-field load bug is different from OmDet-Turbo processor/model input-size mismatch." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Both mention training/multi-GPU, but one is a broadcast TypeError and the other is a Qwen2-VL multi-GPU training setup issue." + "reason": "Trainer GPU broadcast NoneType error and Qwen2-VL multi-GPU training are different training issues." }, { - "left": "issue:44466", - "right": "issue:46050", + "left": "issue:33666", + "right": "issue:36010", "accept": false, - "reason": "Serialization of tied lm_head weights and sinq quantization save/load breakage are both persistence bugs, but not the same concrete bug." + "reason": "Qwen2-VL multi-GPU training is unrelated to GenerationMixin import failure." }, { - "left": "issue:33666", - "right": "issue:36010", + "left": "issue:44466", + "right": "issue:46050", "accept": false, - "reason": "Qwen2-VL multi-GPU training and a GenerationMixin import error are unrelated." + "reason": "lm_head.weight serialization inconsistency is different from sinq quantized model save/load breakage." }, { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "MobileLLM loading key error and AutoImageProcessor-from-URL failure are different loading paths." + "reason": "MobileLLM loading key error is unrelated to AutoImageProcessor loading from URL." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "BitNet packed-weight unpacking and DiaConfig custom-token-ID loss are unrelated save/load bugs." + "reason": "BitNet packed-weight handling during loading is unrelated to DiaConfig custom token ID persistence." }, { "left": "issue:45823", "right": "issue:46050", "accept": false, - "reason": "Gemma4 device-map mismatch during forward and sinq quantized model persistence are unrelated." + "reason": "Gemma4 device mismatch with auto device map is unrelated to sinq quantization save/load regression." }, { "left": "issue:43606", "right": "issue:43927", "accept": false, - "reason": "CPU-offload device mismatch in bark-small and DiaConfig field loss after save/load are not the same bug." + "reason": "CPU offload device mismatch for bark-small is unrelated to DiaConfig save/load token IDs." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "GGUF loading failures for AutoTokenizer/AutoModel and a specific model load failure are too different to merge." + "reason": "gguf_file loader breakage for AutoTokenizer/AutoModel is unrelated to ScandiBERT load failure." }, { "left": "issue:43262", "right": "issue:43295", "accept": false, - "reason": "Audio chat-template sample-rate defaulting and a processor.tokenizer/image-input regression are both processor issues but different code paths." + "reason": "Audio processor chat-template sample-rate default bug is unrelated to processor.tokenizer/image-passing regression." }, { "left": "issue:44849", "right": "issue:46209", "accept": false, - "reason": "Qwen3.5 hidden-states bug and FP8 inference failure on B200 are unrelated." + "reason": "Qwen3.5 output_hidden_states bug is unrelated to FP8 inference failure on B200." }, { "left": "issue:43278", "right": "issue:43688", "accept": false, - "reason": "Embedding dtype drift during evaluate and incorrect auxiliary-loss normalization are unrelated model-train bugs." + "reason": "Embedding dtype changing between train/eval is unrelated to auxiliary-loss normalization in OLMoE/GPT Oss." }, { "left": "issue:43994", "right": "issue:46192", "accept": false, - "reason": "SigLIP2 bad outputs and return_tensors being ignored when text_kwargs is passed are unrelated." + "reason": "SigLIP2 bad outputs from AutoModel/pipeline are unrelated to return_tensors being ignored when text_kwargs is passed." }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Both are config padding-token attribute issues, but for different model configs and different missing-field names; not a single concrete bug." + "reason": "Llama4Config missing pad_token_id and StableLmConfig missing pad_token_idx are similar in shape but distinct model config fields and fixes." }, { "left": "issue:33666", "right": "issue:36331", "accept": false, - "reason": "Qwen2-VL multi-GPU training and CustomTrainer.compute_loss signature breakage are different training issues." + "reason": "Qwen2-VL multi-GPU training is unrelated to CustomTrainer.compute_loss unexpected num_items_in_batch." }, { "left": "issue:43720", "right": "issue:44361", "accept": false, - "reason": "BitNet loading/unpacking and MLukeTokenizer task AttributeError are unrelated." + "reason": "BitNet loading/unpacking issue is unrelated to MLukeTokenizer task AttributeError." }, { "left": "issue:43756", "right": "issue:44913", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch and GPTNeoXConfig rotary_pct reload regression are both config-related, but not the same underlying bug." + "reason": "Smollm3 RoPE-layer count bug is unrelated to GPTNeoXConfig rotary_pct reload defaulting." }, { "left": "issue:44625", "right": "issue:46002", "accept": false, - "reason": "Qwen3.5 num_labels propagation and classification-head NaNs are unrelated." + "reason": "Qwen3.5 num_labels propagation is unrelated to NaNs in classification heads after checkout/init." }, { "left": "issue:43493", "right": "issue:43825", "accept": false, - "reason": "SigLIP2 implementation discrepancy and a pipeline translation error-message issue are not the same bug." + "reason": "SigLIP2 implementation discrepancy is unrelated to pipeline translation-task error messaging in v5." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer added-space behavior and a MacOS bus error in CLIP loading are completely unrelated." + "reason": "Tokenizer extra-space-after-added-token bug is unrelated to CLIP MacOS bus error." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3.5 hidden-states bug is unrelated to incorrect visual encoder keys saved by Qwen3.5." } ] }, @@ -7008,8 +7006,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -7033,6 +7031,7 @@ "issue:43756", "issue:43824", "issue:43844", + "issue:43873", "issue:43927", "issue:43994", "issue:44112", @@ -7048,29 +7047,27 @@ "issue:44734", "issue:44792", "issue:44843", - "issue:44849", "issue:44964", "issue:44977", "issue:45083", "issue:45200", "issue:45229", - "issue:45357", "issue:45405", "issue:45520", "issue:45584", "issue:45759", "issue:45812", - "issue:45823" + "issue:45823", + "issue:46050" ], "soft_pairs": [ "issue:44964|issue:45405", - "issue:44849|issue:45357", "issue:39692|issue:43493", "issue:44568|issue:45812", "issue:43540|issue:44977", "issue:43334|issue:43572", - "issue:44488|issue:44843", "issue:45200|issue:45229", + "issue:44488|issue:44843", "issue:43493|issue:43844", "issue:43278|issue:43844", "issue:43334|issue:45759", @@ -7080,228 +7077,229 @@ "issue:43646|issue:44466", "issue:43824|issue:44589", "issue:33357|issue:33666", - "issue:43504|issue:43606", "issue:45229|issue:45823", + "issue:43504|issue:43606", "issue:43927|issue:44361", "issue:44589|issue:45520", "issue:43572|issue:44291", "issue:44464|issue:45584", "issue:43278|issue:43701", - "issue:43606|issue:44112", "issue:43504|issue:43653", + "issue:43606|issue:44112", "issue:43575|issue:45405", "issue:44792|issue:45405", "issue:43994|issue:44493", - "issue:44423|issue:44734" + "issue:44423|issue:44734", + "issue:43873|issue:46050" ], - "prepared_review_unit_hash": "1a7faf2eb776ce19006563196a7bcfb8e73f4f804bf6c9603a4d8eeb6b27e605", + "prepared_review_unit_hash": "0832b321236872c2c0a1bd3480dc2e4810bf0e820a33d3a211c79d6840fa5b71", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36912, - "estimated_input_tokens": 9228, - "estimated_eval_tokens": 18712 + "serialized_chars": 36884, + "estimated_input_tokens": 9221, + "estimated_eval_tokens": 18698 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a loose, heterogeneous cluster of mostly Transformers v5 regression reports across unrelated model families (Qwen, Gemma, SigLIP, tokenizer/config loading, serving, and CI). Based on the titles alone, none of the soft-edge pairs look like the same concrete bug or change.", - "confidence": 0.69, - "canonical_issue_reason": "Issue 43646 is the broadest representative of the recurring \"v5 breaks model initialization/loading\" theme, but the cluster is too mixed to treat any single issue as a true canonical duplicate target.", + "summary": "Most pairs are related only at a broad subsystem level (tokenizers, multimodal models, quantization, device/offload, or v5 regressions) but point to different concrete failures, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "No single issue cleanly represents the rest as a duplicate cluster; if one anchor is needed, issue 43873 is the broadest and most actively discussed, but it is still a separate bug from the others.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43646 is the best global representative because it most directly captures the common regression pattern seen across several nearby reports, even though the cluster is not a single bug.", + "best_issue_reason": "Issue 43873 is the best triage anchor because it is current, active, and broadly framed, but it should not be used to collapse the other issues into one duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Different problems: one is a model-load failure for Phi-4 multimodal, the other is a dependency/version bump issue for PEFT." - }, - { - "left": "issue:44849", - "right": "issue:45357", - "accept": false, - "reason": "Both involve Qwen3.5, but one is about output_hidden_states behavior and the other about save_pretrained writing wrong visual encoder keys; not the same bug path." + "reason": "Different problems: model-loading failure for Phi-4 multimodal vs an unreleased PEFT version bump." }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "SigLIP2 docs/example errors are not the same as a model-vs-JAX implementation discrepancy." + "reason": "Both involve SigLIP2, but one is a docs/example error and the other is an implementation-vs-JAX discrepancy." }, { "left": "issue:44568", "right": "issue:45812", "accept": false, - "reason": "Both concern tokenizer special-token behavior, but one is about mdeberta add_special_tokens and the other is a Granite token-ID regression; too different to merge." + "reason": "Tokenizer regressions, but for different models and different failure modes (BOS/EOS omission vs wrong Granite token IDs)." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Both are Qwen3.5-family issues, but one is a video-input ValueError in Qwen3OmniMoe and the other is flash-attention generation quality; different concrete failures." + "reason": "Distinct Qwen multimodal bugs: video-input processing error vs flash-attention generation issue." }, { "left": "issue:43334", "right": "issue:43572", "accept": false, - "reason": "Both are config attribute-missing regressions, but they affect different configs and fields (pad_token_id vs pad_token_idx) with no evidence of one underlying bug." + "reason": "Both are missing pad-token fields, but they affect different configs/models and different attribute names." }, { - "left": "issue:44488", - "right": "issue:44843", + "left": "issue:45200", + "right": "issue:45229", "accept": false, - "reason": "sleng-bert load failure and HF_HUB_OFFLINE/Mistral regex patch are unrelated code paths." + "reason": "Gemma4 text-only token-type handling is unrelated to the Gemma4 multi-GPU CUDA OOM report." }, { - "left": "issue:45200", - "right": "issue:45229", + "left": "issue:44488", + "right": "issue:44843", "accept": false, - "reason": "One is a missing default for mm_token_type_ids, the other is a multi-GPU CUDA OOM; not the same bug." + "reason": "One is a model-specific load failure; the other is offline-mode breakage from an unconditional hub call." }, { "left": "issue:43493", "right": "issue:43844", "accept": false, - "reason": "SigLIP2 implementation mismatch and a ZeRO-3 gradient increase issue are unrelated." + "reason": "SigLIP2 implementation mismatch is unrelated to the ZeRO-3 gradient amplification bug." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Embedding dtype drift and gradient explosion under DeepSpeed are different training bugs." + "reason": "Embedding dtype drift in eval is a different issue from the HfDeepSpeedConfig/ZeRO-3 gradient problem." }, { "left": "issue:43334", "right": "issue:45759", "accept": false, - "reason": "Qwen3-VL pad_token_id missing and AutoModelForCausalLM not unwrapping text_config are distinct model-loading issues." + "reason": "Different multimodal-config regressions: missing pad_token_id vs not unwrapping text_config in AutoModelForCausalLM.from_config." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "SigLIP2 example/documentation errors do not match the processor.tokenizer regression in custom model code." + "reason": "Both are regressions around model usage, but one is a SigLIP2 docs example and the other is a processor/tokenizer API break." }, { "left": "issue:43756", "right": "issue:45083", "accept": false, - "reason": "RoPE-layer count mismatch and a helper-function length calculation issue in qwen3_omni_moe are not the same defect." + "reason": "Separate model-specific bugs: Smollm3 RoPE layer handling vs qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Both mention Qwen3 MoE loading, but one is about accelerate/deepspeed weight conversion and the other is a broader Qwen3VL MoE load failure; insufficient evidence of one code-path bug." + "reason": "Both mention Qwen3 MoE, but one is accelerate+deepspeed conversion and the other is a dev-version load regression for Qwen3VL MoE." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Both are v5-related, but one is custom model initialization and the other is lm_head.weight serialization depending on device; not the same change." + "reason": "Different v5 regressions: custom model initialization vs lm_head.weight serialization depending on device." }, { "left": "issue:43824", "right": "issue:44589", "accept": false, - "reason": "Qwen2.5-VL import error and Float8 storage lookup failure are unrelated." + "reason": "Import error for a missing class is unrelated to the Float8 storage-type lookup failure." }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "MacOS bus error in CLIP and Qwen2-VL multi-GPU training are completely different issues." + "reason": "Completely different models and failures: MacOS bus error on CLIP vs Qwen2-VL multi-GPU training." }, { - "left": "issue:43504", - "right": "issue:43606", + "left": "issue:45229", + "right": "issue:45823", "accept": false, - "reason": "BEiT legacy-field loading and bark-small CPU offload device mismatch are unrelated." + "reason": "Both are Gemma4, but one is inference OOM and the other is a forward device mismatch with device_map=\"auto\"." }, { - "left": "issue:45229", - "right": "issue:45823", + "left": "issue:43504", + "right": "issue:43606", "accept": false, - "reason": "Both are Gemma4 issues, but one is CUDA OOM during inference and the other is a PLE device mismatch during forward; different bugs." + "reason": "Legacy-field load failure for BEiT is unrelated to the CPU offload device mismatch in bark-small." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "DiaConfig losing token IDs after save/load and MLukeTokenizer task AttributeError are unrelated." + "reason": "Different tokenizer/config bugs: DiaConfig save-load token IDs vs MLukeTokenizer task-time AttributeError." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage-object lookup and flash_attn import KeyError are different failure modes." + "reason": "Float8 storage lookup failure is not the same as a flash_attn import-utils KeyError on Python 3.13." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "StableLmConfig missing pad_token_idx and init_empty_weights receiving an unexpected argument are separate compatibility regressions." + "reason": "Missing pad_token_idx after v5 update is separate from init_empty_weights passing an unexpected _is_hf_initialized argument." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Chunked generation with compiled forward and Whisper empty-transcription generation after align_special_tokens are unrelated." + "reason": "Different generation bugs: compiled-forward chunked generation mismatch vs empty-transcription failure after align_special_tokens." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "Embedding dtype drift in eval and resume_from_checkpoint key mismatch are different problems." + "reason": "Embedding dtype drift is unrelated to a resume_from_checkpoint key mismatch." }, { - "left": "issue:43606", - "right": "issue:44112", + "left": "issue:43504", + "right": "issue:43653", "accept": false, - "reason": "CPU offload device mismatch and a stale GraniteSpeech CI test are not the same underlying bug." + "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are distinct bugs." }, { - "left": "issue:43504", - "right": "issue:43653", + "left": "issue:43606", + "right": "issue:44112", "accept": false, - "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are unrelated." + "reason": "A real device mismatch bug and a stale GraniteSpeech CI test are not the same underlying issue." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "Qwen2 large-model TP OOM and a PEFT version bump issue do not share the same code path." + "reason": "TP OOM for Qwen2-57B is unrelated to the PEFT minimum-version bump." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "A janus test failure and a dependency version bump are unrelated." + "reason": "Janus image-generation test failure is unrelated to the PEFT versioning issue." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 wrong outputs with AutoModel/pipeline and unexpected position-id keys are different serialization/loading issues." + "reason": "SigLIP2 nonsensical outputs and unexpected position-id keys are different symptoms with different likely causes." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Serve crashes with multimodal models and KV-cache continuation tensor indexing are both serving bugs, but the concrete failure modes are distinct." + "reason": "Serve crashes differ: one is a multimodal '.to' AttributeError, the other is KV-cache continuation indexing." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Both concern quantization/offloading, but one is an offloading behavior bug and the other is save/load corruption after sinq quantization." } ] }, @@ -7320,8 +7318,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -7347,7 +7345,6 @@ "issue:43701", "issue:43824", "issue:43844", - "issue:43873", "issue:43957", "issue:44186", "issue:44521", @@ -7362,6 +7359,7 @@ "issue:45198", "issue:45200", "issue:45290", + "issue:45325", "issue:45356", "issue:45381", "issue:45406", @@ -7375,7 +7373,6 @@ "issue:46209" ], "soft_pairs": [ - "issue:43873|issue:46050", "issue:29942|issue:30333", "issue:45406|issue:46121", "issue:45588|issue:46050", @@ -7387,8 +7384,8 @@ "issue:43582|issue:44977", "issue:44554|issue:45446", "issue:45084|issue:45520", - "issue:15354|issue:38175", "issue:43540|issue:45200", + "issue:15354|issue:38175", "issue:45200|issue:45823", "issue:45072|issue:45198", "issue:37428|issue:43957", @@ -7400,213 +7397,214 @@ "issue:43295|issue:43450", "issue:43295|issue:44610", "issue:43572|issue:44987", - "issue:44554|issue:45356", "issue:43572|issue:43957", + "issue:44554|issue:45356", + "issue:39692|issue:46192", "issue:37428|issue:43824", - "issue:39692|issue:46192" + "issue:45200|issue:45325" ], - "prepared_review_unit_hash": "fb0168e1e8442db267d917e8de506a97b472f4317f0247e63831a13615ff8eb9", + "prepared_review_unit_hash": "d5607f5e065fb60fabd435c78876293c9501d54ac277218ed5019e9f7a316022", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36426, - "estimated_input_tokens": 9107, - "estimated_eval_tokens": 18470 + "serialized_chars": 36487, + "estimated_input_tokens": 9122, + "estimated_eval_tokens": 18500 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is noisy and mostly spans separate transformers bugs around tokenization/processor APIs, multimodal chat templating, generation, quantization, and model loading. None of the soft pairs look like true duplicates; the closest subtheme is apply_chat_template / multimodal tokenization.", - "confidence": 0.34, - "canonical_issue_reason": "issue:44521 is the best cluster anchor because it is a clearly reproduced, user-facing multimodal chat-template/tokenization bug and is the closest semantic match to another soft candidate (45290), while the other items are more peripheral or unrelated.", + "summary": "These items are mostly unrelated individual Transformers issues rather than a duplicate cluster. None of the soft edges look like the same underlying bug or change, so all are rejected.", + "confidence": 0.97, + "canonical_issue_reason": "If one issue must be chosen as a representative, issue 44521 is the clearest, concrete bug report with active discussion and inbound references; it is the most broadly representative item in this set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the strongest standalone representative of the cluster\u2019s recurring tokenizer/processor surface area, with direct impact, active discussion, and the broadest overlap to the nearby chat-template regression reports.", + "best_issue_reason": "Issue 44521 is the strongest standalone issue for triage: it has a precise repro, clear user impact, and external references, making it the best representative issue here.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43873", - "right": "issue:46050", - "accept": false, - "reason": "Both involve quantization, but one is about offloading behavior and the other about save/load corruption; not the same bug." - }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "Unrelated failures: Flash Attention 2 tests vs MLFlow reporting after a failed training job." + "reason": "Flash Attention 2 test failures and MLflow reporting on job failure are unrelated subsystems and failure modes." }, { "left": "issue:45406", "right": "issue:46121", "accept": false, - "reason": "Different code paths: Gemma4Processor/_tokenizer regression vs rope-validation TypeError." + "reason": "Gemma4 processor _tokenizer access and rope-parameter JSON list handling are different code paths with different errors." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "AttributeError in flash_attention sinkless models is unrelated to quantized model save/load breakage." + "reason": "A flash_attention aux-argument crash is unrelated to sinq quantized model save/load corruption." }, { "left": "issue:42490", "right": "issue:43696", "accept": false, - "reason": "Local-path import warning and CUDA OOM are unrelated." + "reason": "Local-path import warning and GPT-oss CUDA OOM are clearly different issues." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "Different problems: ZeRO-3 gradient growth vs a model loading failure for Phi-4-multimodal-instruct." + "reason": "ZeRO-3 gradient behavior and Phi-4 multimodal loading are unrelated bugs." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Tokenizer spacing bug and TrainerState token count tracking are different features." + "reason": "Tokenizer spacing after added tokens does not match TrainerState token-count tracking." }, { "left": "issue:29159", "right": "issue:32090", "accept": false, - "reason": "Tokenizer consistency issue vs trainer broadcast NoneType error are unrelated." + "reason": "Slow-vs-fast tokenizer inconsistency is unrelated to NoneType broadcast errors in Trainer." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Same general apply_chat_template area, but different failure modes: assistant masks vs crash on tool-call messages with no content." + "reason": "Both involve apply_chat_template, but one is zeroed assistant masks for multimodal inputs while the other is a crash on tool-call messages without content; not the same concrete bug." }, { "left": "issue:43582", "right": "issue:44977", "accept": false, - "reason": "Apple Silicon allocator TypeError and Qwen3.5 flash-attention generation bug are not the same bug." + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation problems are unrelated." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "MPS attention correctness vs PyTorch version gating for AuxRequest import are different issues." + "reason": "An MPS attention correctness issue and a PyTorch version check bug in flex_attention are different problems." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Template compilation TypeError and flash_attn import KeyError are unrelated." + "reason": "Template-node compilation failure and flash_attn import KeyError on Python 3.13 are unrelated." }, { - "left": "issue:15354", - "right": "issue:38175", + "left": "issue:43540", + "right": "issue:45200", "accept": false, - "reason": "JIT scripting/export limitation and zero-probabilities on SigLIP2 are different bugs." + "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to Gemma4 mm_token_type_ids defaults." }, { - "left": "issue:43540", - "right": "issue:45200", + "left": "issue:15354", + "right": "issue:38175", "accept": false, - "reason": "Video-input processing error in Qwen3OmniMoe vs Gemma4 mm_token_type_ids defaulting issue." + "reason": "torch.jit.script GeneratorExp export failure and unexpected zero probabilities in SigLIP2 are unrelated." }, { "left": "issue:45200", "right": "issue:45823", "accept": false, - "reason": "Both are Gemma4-related, but one is missing token-type defaults and the other is a device mismatch in forward." + "reason": "Missing mm_token_type_ids defaults and Gemma4 device-map mismatch are different forward-path bugs." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "Different subsystems and symptoms: dtype mismatch in inference vs save_pretrained/tokenization failure." + "reason": "bfloat16 dtype mismatches in inference and Wav2Vec2 save/tokenization failures do not match." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Flash-attention import error and meta-device loading breakage are separate runtime issues." + "reason": "A missing flash attention helper import and torch.device(\"meta\") model-loading breakage are different compatibility issues." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "Generation kwargs update after sync_gpus is unrelated to the Qwen3.5 output_hidden_states bug." + "reason": "Generation kwargs sync after sync_gpus and Qwen3.5 output_hidden_states behavior are unrelated generation bugs." }, { "left": "issue:44186", "right": "issue:46192", "accept": false, - "reason": "LayoutLMv2 tokenizer crash on NER/padding is not the same as return_tensors being ignored when text_kwargs is passed." + "reason": "LayoutLMv2 tokenizer crashes on NER/padding are not the same as return_tensors being ignored when text_kwargs is set." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch do not share the same code path." + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." }, { "left": "issue:45381", "right": "issue:46209", "accept": false, - "reason": "Wrong video vision_position_ids in Qwen2.5-VL is unrelated to FP8 inference failure on B200." + "reason": "Qwen2.5-VL video position-id bug and FP8 inference failure on B200 are different hardware/model issues." }, { "left": "issue:45823", "right": "issue:46192", "accept": false, - "reason": "Gemma4 device placement mismatch and return_tensors precedence bug are different issues." + "reason": "Gemma4 device-map placement bug and return_tensors precedence bug are not the same failure." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Processor.tokenizer regression/passing images to tokenizer and batched video shape errors are distinct multimodal bugs." + "reason": "Processor.tokenizer regression with image inputs is unrelated to video processor output shape for batched inputs." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "These are both processor/model mismatches, but they affect different models and different failure modes." + "reason": "Custom processor.tokenizer regression and OmDet-Turbo image-size mismatch are different model-specific issues." }, { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "Missing pad_token_idx in config and a model-loading regression for physical-intelligence/fast are unrelated." + "reason": "StableLmConfig pad_token_idx regression and loading physical-intelligence/fast on transformers>=5.1.0 are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Missing StableLmConfig pad_token_idx and meta-device loading failures affect different mechanisms." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "MPS attention correctness issue and Kimi-K2.5 tokenizer regression are not the same bug." + "reason": "MPS attention correctness with mismatched head dims and Kimi-K2.5 tokenizer codec regression are unrelated." }, { - "left": "issue:43572", - "right": "issue:43957", + "left": "issue:39692", + "right": "issue:46192", "accept": false, - "reason": "Config default regression and meta-device loading failures are different problems." + "reason": "SigLIP2 documentation errors and return_tensors precedence are not the same bug." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "flash_attention_utils import error and Qwen2_5_VL import error are separate import regressions." + "reason": "Missing flash-support import and missing Qwen2.5-VL symbol export are different import errors." }, { - "left": "issue:39692", - "right": "issue:46192", + "left": "issue:45200", + "right": "issue:45325", "accept": false, - "reason": "SigLIP2 docs/example errors and a generic return_tensors precedence bug are not duplicates." + "reason": "Gemma4 multimodal token-type defaults and Qwen2.5-VL rope-index scaling are unrelated model bugs." } ] }, @@ -7625,8 +7623,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -7666,10 +7664,9 @@ "issue:44792", "issue:44857", "issue:44898", + "issue:45083", "issue:45127", "issue:45198", - "issue:45200", - "issue:45325", "issue:45412", "issue:45520", "issue:45563", @@ -7680,11 +7677,10 @@ "issue:46209" ], "soft_pairs": [ - "issue:45200|issue:45325", - "issue:45520|issue:46178", "issue:43295|issue:43873", - "issue:43550|issue:43844", + "issue:45520|issue:46178", "issue:44898|issue:45412", + "issue:43550|issue:43844", "issue:44466|issue:45127", "issue:43334|issue:43525", "issue:43450|issue:43575", @@ -7697,203 +7693,222 @@ "issue:43278|issue:43856", "issue:28218|issue:35141", "issue:43653|issue:44112", - "issue:38175|issue:43493", "issue:43526|issue:46192", - "issue:43975|issue:45520", + "issue:38175|issue:43493", "issue:44442|issue:45198", + "issue:43975|issue:45520", "issue:38175|issue:45412", "issue:44514|issue:44521", - "issue:45520|issue:46170", "issue:43262|issue:43450", + "issue:45520|issue:46170", "issue:44206|issue:45584", "issue:43819|issue:44112", + "issue:44610|issue:44857", "issue:43827|issue:43901", - "issue:44610|issue:44857" + "issue:43819|issue:45083" ], - "prepared_review_unit_hash": "3df6e8d5b52176db8fbe94b0e645a12a82145efa868c4954d5a7cf2a6e1fc256", + "prepared_review_unit_hash": "81b1cf32396df91c8872ddf470374cac6fc897b59501e9223e17815a733918a6", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36620, - "estimated_input_tokens": 9155, - "estimated_eval_tokens": 18566 + "serialized_chars": 36003, + "estimated_input_tokens": 9001, + "estimated_eval_tokens": 18258 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous single-issue reports rather than a true duplicate set. None of the soft pairs look like the same concrete bug/change, so I would not merge any of them.", - "confidence": 0.69, - "canonical_issue_reason": "Issue 44521 is the most representative issue here: it has the strongest activity/attention and is the closest to the only nearby Qwen2.5-VL processor/chat-template candidate.", + "summary": "All soft pairs are false positives: they share superficial themes (tokenizers, vision models, imports, docs, etc.) but describe different bugs, models, or code paths. No pull requests are present.", + "confidence": 0.92, + "canonical_issue_reason": "No pair looks like a true duplicate cluster. If one issue must serve as a representative, issue:45520 is the broadest core-runtime import bug and is the least model-specific of the set.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 is the best overall representative of the cluster because it is active, user-facing, and sits in the clearest adjacent area to one other item in the set (`apply_chat_template`/multimodal processor behavior).", + "best_issue_reason": "Issue:45520 is the best standalone representative because it concerns shared import infrastructure and has broader applicability than the model-specific or docs-only reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:45200", - "right": "issue:45325", + "left": "issue:43295", + "right": "issue:43873", "accept": false, - "reason": "Different bugs: Gemma 4 multimodal token-type IDs vs Qwen2.5-VL still-image rope/position scaling." + "reason": "Different failures: processor.tokenizer / batched image handling regression vs quantization offloading behavior." }, { "left": "issue:45520", "right": "issue:46178", "accept": false, - "reason": "Unrelated import/backend failures in different code paths: `flash_attn` on Python 3.13 vs `tensorflow_text` backend mapping." + "reason": "Both are import-related, but one is a flash_attn KeyError on Python 3.13 and the other is a BACKENDS_MAPPING error for tensorflow_text; different code paths and causes." }, { - "left": "issue:43295", - "right": "issue:43873", + "left": "issue:44898", + "right": "issue:45412", "accept": false, - "reason": "Processor regression around `processor.tokenizer`/image passing is unrelated to quantization offloading behavior." + "reason": "Perceiver resolution/interpolation failure is unrelated to RT-DETR memory not being released after deletion." }, { "left": "issue:43550", "right": "issue:43844", "accept": false, - "reason": "Different failure modes: `torch.compile` + SDPA on Bamba vs DeepSpeed ZeRO-3 gradient growth." - }, - { - "left": "issue:44898", - "right": "issue:45412", - "accept": false, - "reason": "One is a Perceiver resolution/interpolation bug; the other is a memory release/GC issue for RT-DETR." + "reason": "torch.compile + SDPA crash in Bamba is not the same as gradient explosion with DeepSpeed ZeRO-3." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Both involve tied weights, but one is device-dependent serialization of `lm_head.weight`; the other is LoRA merge collapse with extended vocab." + "reason": "Both mention tied weights, but one is device-dependent serialization of lm_head.weight and the other is LoRA merge collapse with extended vocabulary." }, { "left": "issue:43334", "right": "issue:43525", "accept": false, - "reason": "Same symptom class (`pad_token_id` missing) but different models/config paths; not the same concrete bug." + "reason": "Same missing-attribute symptom, but different configs/models (Qwen3-VL vs Llama4) and likely separate fixes." }, { "left": "issue:43450", "right": "issue:43575", "accept": false, - "reason": "Batched video processor shape bug is unrelated to Qwen2-57B tensor-parallel OOM." + "reason": "Video batched-shape bug and Qwen2-57B tensor-parallel OOM are unrelated." }, { "left": "issue:44792", "right": "issue:45563", "accept": false, - "reason": "Different subsystems: Janus image-generation test failure vs stale paged-generate warning." + "reason": "A failing janus generate_images test is not the same as a stale warning about num_return_sequences in paged generate()." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "ImportError for Qwen2.5-VL class is unrelated to DeepSeek detokenization regression." + "reason": "ImportError for Qwen2_5_VLForConditionalGeneration and DeepSeek-Coder detokenization regression are different bugs." }, { "left": "issue:41720", "right": "issue:46209", "accept": false, - "reason": "Both are hardware-related inference failures, but they affect different models and different failure mechanisms." + "reason": "Different models, hardware, and symptoms: Qwen3 auto device mapping cudaErrorAssert vs FP8 inference failure on B200." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Qwen3Moe FP8 failure is unrelated to the missing Qwen2.5-VL import." + "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to a Qwen2.5-VL import error." }, { "left": "issue:43388", "right": "issue:43526", "accept": false, - "reason": "Both are label-processing issues, but one is `gather_for_metrics` dropping tuple labels and the other is `reduce_labels` returning only one label." + "reason": "Both involve labels, but one drops elements in gather_for_metrics and the other reduces Beit image labels incorrectly; different layers and behavior." }, { "left": "issue:42490", "right": "issue:43493", "accept": false, - "reason": "Trailing-slash import warning and SigLIP2 implementation discrepancy are unrelated." + "reason": "Trailing-slash local model loading warning is unrelated to the SigLIP2 HF-vs-JAX discrepancy." }, { "left": "issue:43278", "right": "issue:43856", "accept": false, - "reason": "Embedding dtype regression in evaluation is not the same as MoE training memory inefficiency." + "reason": "Embedding dtype drift during evaluate and Qwen3 MoE memory usage are unrelated issues." }, { "left": "issue:28218", "right": "issue:35141", "accept": false, - "reason": "Tokenizer spacing after added token and embedding reinitialization after resize are different bugs." + "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization when resizing embeddings with tie_word_embedding=False." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug is unrelated to a GraniteSpeech CI device-override test failure." + "reason": "BigBirdTokenizer special-token registration bug and GraniteSpeech CI test flake are different problems." + }, + { + "left": "issue:43526", + "right": "issue:46192", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to return_tensors being ignored when text_kwargs is passed." }, { "left": "issue:38175", "right": "issue:43493", "accept": false, - "reason": "Both mention SigLIP2, but one reports zero probabilities and the other a broader HF-vs-JAX discrepancy; not enough to treat as the same bug." + "reason": "Both mention SigLIP2, but one reports zero probabilities while the other is a broader HF-vs-JAX implementation mismatch; not enough to treat as the same concrete bug." }, { - "left": "issue:43526", - "right": "issue:46192", + "left": "issue:44442", + "right": "issue:45198", "accept": false, - "reason": "Different API bugs: `reduce_labels` truncation vs `return_tensors` being ignored when `text_kwargs` is passed." + "reason": "FastSpeech2Conformer tokenizer loading failure and Wav2Vec2 save_pretrained/tokenization failure are separate audio-tokenizer bugs." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Detokenization regression and `flash_attn` import error are unrelated." + "reason": "Detokenization regression in DeepSeek-Coder v5 is unrelated to a flash_attn import KeyError." }, { - "left": "issue:44442", - "right": "issue:45198", + "left": "issue:38175", + "right": "issue:45412", "accept": false, - "reason": "Different speech-model failures: tokenizer loading vs save_pretrained/tokenization failure." + "reason": "SigLIP2 zero probabilities and RT-DETR memory leak are entirely different bugs." }, { "left": "issue:44514", "right": "issue:44521", "accept": false, - "reason": "Both are Qwen2.5-VL `apply_chat_template` issues, but one is a batching/padding crash and the other is incorrect all-zero assistant masks; not the same concrete bug." + "reason": "Both are Qwen2.5-VL chat-template issues, but one crashes on batched padding=False input while the other returns all-zero assistant masks; different symptoms and likely different fixes." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Audio chat-template sample-rate default bug is unrelated to video processor batched-shape output." + "reason": "Audio chat-template sampling-rate default bug is unrelated to batched video processor shape output." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Both are import failures, but one is flash_attn on Python 3.13 and the other is a Windows/ROCm torch.distributed dependency issue." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Unsupported center arg in LasrFeatureExtractor is unrelated to Whisper failing on empty transcription after align_special_tokens." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "DAC latent/forward mismatch is unrelated to the GraniteSpeech CI test flake." + "reason": "DAC.from_latents mismatch and a GraniteSpeech stale CI test have no shared underlying code path." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Processor output size mismatch for OmDet-Turbo is unrelated to an AMP/CUDA crash in LwDetrImageLoss." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are docs-related, but one is about removed `pipeline()` references and the other about `return_all_scores`; different content." + "reason": "Both are docs issues, but they concern different pipeline deprecations and different pages; not the same bug/change." }, { - "left": "issue:44610", - "right": "issue:44857", + "left": "issue:43819", + "right": "issue:45083", "accept": false, - "reason": "Processor output resolution mismatch and AMP/CUDA loss crash are different bugs." + "reason": "DAC STE mismatch and qwen3_omni_moe feature-extract length helper behavior are unrelated." } ] }, @@ -7912,8 +7927,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -7967,7 +7982,6 @@ "issue:46192" ], "soft_pairs": [ - "issue:43819|issue:45083", "issue:37428|issue:43425", "issue:43335|issue:43550", "issue:43606|issue:45083", @@ -7975,19 +7989,19 @@ "issue:29942|issue:32090", "issue:44448|issue:44743", "issue:45127|issue:45245", - "issue:37428|issue:44351", "issue:45290|issue:45381", "issue:45200|issue:45588", + "issue:37428|issue:44351", "issue:43335|issue:43653", "issue:43611|issue:43931", "issue:44206|issue:45823", "issue:43550|issue:44112", "issue:43819|issue:44493", - "issue:43278|issue:43825", "issue:44855|issue:45083", "issue:43335|issue:43504", "issue:45588|issue:45859", "issue:29127|issue:32090", + "issue:43278|issue:43825", "issue:43335|issue:45083", "issue:39692|issue:43526", "issue:39692|issue:44493", @@ -7996,250 +8010,244 @@ "issue:43335|issue:44112", "issue:43335|issue:45072", "issue:44448|issue:44779", - "issue:44360|issue:44561", "issue:42490|issue:43854", + "issue:44360|issue:44561", "issue:45920|issue:46192", "issue:43606|issue:44857", "issue:34689|issue:35141" ], - "prepared_review_unit_hash": "48cf2da9cc5e96143b088fcb8becdaee0e06116e9e243412dd54a9c3cd249034", + "prepared_review_unit_hash": "7ab05a267208bec123d92f9f7ff4e48545c6428d178a6d738d406c63274802ad", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 34, - "serialized_chars": 38091, - "estimated_input_tokens": 9523, - "estimated_eval_tokens": 19302 + "soft_pair_count": 33, + "serialized_chars": 37806, + "estimated_input_tokens": 9452, + "estimated_eval_tokens": 19160 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated issues; the only clear duplicate pair is the Qwen3.5 MoE from_pretrained regression. I would keep the earlier report as canonical.", - "confidence": 0.95, - "canonical_issue_reason": "Issue 45310 is the earlier report of the same Qwen3.5 MoE from_pretrained failure as 45478, with identical title and the same concrete regression.", + "summary": "Most items are unrelated regression reports across different models/subsystems. The only clear duplicate-like pair is the two near-identical Qwen3.5 MoE from_pretrained failures.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45310 is the clearest canonical issue: it is the older, more specific report for the Qwen3.5 MoE from_pretrained regression, and issue:45478 reads like a follow-up/duplicate of the same failure.", "canonical_pr_reason": null, - "best_issue_reason": "45310 is the best issue to keep as the canonical representative because it is the first report of the duplicated Qwen3.5 MoE loading bug and already captures the core failure mode.", + "best_issue_reason": "issue:45310 is the best representative issue overall because it captures a concrete, reproducible model-loading regression and has a strong duplicate counterpart, making it the most suitable anchor for triage.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43819", - "right": "issue:45083", - "accept": false, - "reason": "Different models and failures: DAC latent decoding/STE mismatch vs a Qwen3-Omni helper length computation issue." - }, { "left": "issue:37428", "right": "issue:43425", "accept": false, - "reason": "One is an import error for a flash-attention helper; the other is a broad Torch 2.10 compatibility issue." + "reason": "Different failures: flash-attention helper import error vs Torch 2.10 compatibility break." }, { "left": "issue:43335", "right": "issue:43550", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug and Bamba torch.compile/SDPA failure are unrelated code paths." + "reason": "Unrelated bugs: sparse-layer config creation vs torch.compile/SDPA failure in Bamba." }, { "left": "issue:43606", "right": "issue:45083", "accept": false, - "reason": "CPU offload device mismatch in Bark-small is unrelated to the Qwen3-Omni output-length helper bug." + "reason": "Different code paths: CPU offload device mismatch vs qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "Float16 AMP crash in LwDetrImageLoss is not the same bug as D-FINE skipping auxiliary losses." + "reason": "Different model/components: AMP float16 crash in LwDetrImageLoss vs missing auxiliary losses in D-FINE." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Flash Attention 2 test failures and a Trainer NoneType broadcast error are different failures." + "reason": "Flash Attention 2 test failures are not the same as the Trainer _gpu_broadcast_one NoneType error." }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Pegasus v4/v5 output differences and Qwen3 recurrent-state reset are separate model-specific regressions." + "reason": "Different regressions: Pegasus output drift between v4/v5 vs Qwen recurrent-state reset with cache." }, { "left": "issue:45127", "right": "issue:45245", "accept": false, - "reason": "LoRA/vocab merge collapse and category-count overflow are unrelated problems." - }, - { - "left": "issue:37428", - "right": "issue:44351", - "accept": false, - "reason": "Both are import-related, but they concern different missing symbols and different breakages." + "reason": "LoRA/tied-embedding collapse is unrelated to the category-count limit RuntimeError." }, { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Chat-template tool-call crash and Qwen2.5-VL vision position-id bug are unrelated." + "reason": "Chat template tool-call crash is unrelated to Qwen2.5-VL video vision_position_ids behavior." }, { "left": "issue:45200", "right": "issue:45588", "accept": false, - "reason": "Gemma4 text-only mm_token_type_ids handling and flash_attention sink-less s_aux handling are different code paths." + "reason": "Gemma4 text-only token-type handling is unrelated to the flash_attention s_aux=None crash." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Different import failures: flash-attention utility symbol missing vs HybridCache import error." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "SwitchTransformers sparse-layer config and BigBirdTokenizer special-token registration are unrelated." + "reason": "Sparse-layer config bug and BigBirdTokenizer special-token registration are unrelated." }, { "left": "issue:43611", "right": "issue:43931", "accept": false, - "reason": "base_model_prefix loading breakage and Qwen3-VL weight-shape mismatch are different model-loading issues." + "reason": "base_model_prefix loading breakage is not the same as Qwen3-VL weight-shape mismatch." }, { "left": "issue:44206", "right": "issue:45823", "accept": false, - "reason": "Unsupported center arg in an audio feature extractor is unrelated to Gemma4 PLE device mismatch." + "reason": "Unsupported feature-extractor arg and Gemma4 PLE device mismatch are different problems." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "Bamba torch.compile/SDPA failure and GraniteSpeech stale device-override test are not the same bug." + "reason": "Torch.compile SDPA bug in Bamba is unrelated to a stale CI device-override test." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "DAC STE mismatch and unexpected position-id keys are different symptoms and code paths." - }, - { - "left": "issue:43278", - "right": "issue:43825", - "accept": false, - "reason": "Embedding dtype drift in evaluate and a pipeline translation-task warning are unrelated regressions." + "reason": "DAC from_latents/STE mismatch is unrelated to unexpected position-id keys across models." }, { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "DebertaV2 Python 3.13 import parsing error and Qwen3-Omni helper behavior are unrelated." + "reason": "Python 3.13 IndentationError on import is unrelated to qwen3_omni_moe output-length behavior." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "SwitchTransformers sparse-layer creation bug and BEiT legacy-field loading bug are not the same issue." + "reason": "Sparse-layer creation bug and Beit legacy-field loading failure do not share the same code-path." }, { "left": "issue:45588", "right": "issue:45859", "accept": false, - "reason": "Both touch offload/attention-adjacent logic, but they fail in different modules with different root causes." + "reason": "Both involve Qwen3.5/MoE-ish flash/offload areas, but one is an AttributeError on sink-less flash attention while the other is CPU-offload incompatibility in rotary embedding." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "LayoutLMv3 error-message clarity and Trainer broadcast TypeError are unrelated." + "reason": "LayoutLMv3 error-message clarity issue is unrelated to the Trainer broadcast NoneType crash." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype drift during evaluate is unrelated to the pipeline translation-task error message regression." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "SwitchTransformers sparse-layer config and Qwen3-Omni feature-length helper are unrelated." + "reason": "SwitchTransformers sparse-layer config bug is unrelated to qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "SigLIP2 doc/example problems and BeitImageProcessorFast label reduction are unrelated." + "reason": "SigLIP2 docs example errors and BeitImageProcessorFast reduce_labels bug are different issues." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "Documentation/model-processor mismatch and unexpected position-id keys are different issues." + "reason": "Documentation example failures are unrelated to position-id key regressions." }, { "left": "issue:45310", "right": "issue:45478", "accept": true, - "reason": "Same title, same model family, same from_pretrained regression, and both describe the same loading failure." + "reason": "Near-identical Qwen3.5 MoE from_pretrained failures with the same regression surface; these look like the same underlying bug." }, { "left": "issue:43720", "right": "issue:44112", "accept": false, - "reason": "BitNet packed-weight loading and GraniteSpeech CI device-override failures are unrelated." + "reason": "Packed-weight unpacking during accelerate loading is unrelated to the stale GraniteSpeech CI device override test." }, { "left": "issue:43335", "right": "issue:44112", "accept": false, - "reason": "SwitchTransformers config bug and GraniteSpeech CI failure do not share a concrete root cause." + "reason": "Sparse-layer config creation and GraniteSpeech CI device-override flakiness are unrelated." }, { "left": "issue:43335", "right": "issue:45072", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug and bfloat16 dtype mismatch in inference are unrelated." + "reason": "Sparse-layer creation bug is unrelated to bfloat16 dtype mismatches in inference." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "Pegasus output change and DeepSeek tokenizer regression are different model-specific v5 regressions." + "reason": "Pegasus v4/v5 output drift is unrelated to DeepSeek tokenizer regression." }, { - "left": "issue:44360", - "right": "issue:44561", + "left": "issue:42490", + "right": "issue:43854", "accept": false, - "reason": "DSA indexer activation discussion and removal of is_torch_fx_available breakage are not the same bug." + "reason": "Local-path trailing-slash import warning is unrelated to GLM-4.7-Flash model loading in tests." }, { - "left": "issue:42490", - "right": "issue:43854", + "left": "issue:44360", + "right": "issue:44561", "accept": false, - "reason": "Local-path trailing-slash import warning and GLM-4.7-Flash unit-test loading failure are unrelated." + "reason": "DSA indexer missing ReLU is unrelated to removal of is_torch_fx_available affecting trust_remote_code models." }, { "left": "issue:45920", "right": "issue:46192", "accept": false, - "reason": "Tokenizer ID regressions across many models and return_tensors being ignored with explicit text_kwargs are different issues." + "reason": "AutoTokenizer wrong token IDs regression is not the same as silently ignoring return_tensors when text_kwargs is passed." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "CPU offload device mismatch and float16 AMP loss crash are different failure modes." + "reason": "CPU offload device mismatch in Bark is unrelated to LwDetrImageLoss AMP crashes." }, { "left": "issue:34689", "right": "issue:35141", "accept": false, - "reason": "Llama 3.2 Vision model-loading breakage and post_init reinitializing tied embeddings are unrelated." + "reason": "Llama 3.2 Vision model-loading regression is unrelated to embedding reinitialization after resizing token embeddings." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: the only clearly supported duplicate pair is issue:45310 and issue:45478, which share the same title and the same Qwen3.5 MoE from_pretrained regression. The soft-edge verdicts are conservative and stay within the evidence provided, mostly rejecting pairs as unrelated without overclaiming. No PR-pair mergeability concerns apply here." + "feedback": "The summary is well grounded in the packet and stays conservative overall. The duplicate-like call on issue:45310 / issue:45478 is supported by the identical titles and high similarity, and the rejected soft pairs are consistent with the packet. No PR-mergeability concerns apply here." }, "evaluator_used": true, "retried": false, @@ -8255,8 +8263,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -8317,8 +8325,8 @@ "issue:40444|issue:43526", "issue:28218|issue:32090", "issue:43701|issue:45083", - "issue:32090|issue:33357", "issue:29942|issue:35141", + "issue:32090|issue:33357", "issue:43550|issue:45083", "issue:43723|issue:44554", "issue:37428|issue:44987", @@ -8340,214 +8348,214 @@ "issue:43278|issue:43381", "issue:45588|issue:45823" ], - "prepared_review_unit_hash": "3501a7cbe037eb3adae7942ff13c9ceddf5677c85bb5ba052415ef00b46ff028", + "prepared_review_unit_hash": "e02635811b9a4fc6577f8c5dc5472aeb7d4e0f8eaca48df451d8e8c15697dbd3", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36251, - "estimated_input_tokens": 9063, - "estimated_eval_tokens": 18382 + "serialized_chars": 36245, + "estimated_input_tokens": 9062, + "estimated_eval_tokens": 18380 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The soft pairs are mostly broad semantic matches rather than true duplicate bugs; I would reject all of them. The strongest standalone issue in the set looks like 43873 because it is open and has the most inbound discussion, while 45698 is the clearest self-contained bug report.", - "confidence": 0.84, - "canonical_issue_reason": "issue:43873 is the best canonical anchor here: it is open, has the highest discussion/inbound-reference signal in the set, and describes a broadly relevant user-facing bug that can absorb related reports.", + "summary": "The soft links are largely topical false positives: these issues span unrelated tokenizer, model-loading, attention, dtype, and runtime bugs. I don't see any pair that is clearly the same underlying defect; the only near-adjacent case is the Gemma4 processor-loading subgroup, but even there the failures differ.", + "confidence": 0.94, + "canonical_issue_reason": "No strong canonical issue exists for this cluster because the items are heterogeneous; issue 45372 is only the closest local representative of the Gemma4 processor-loading subgroup.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45698 is the most clearly written standalone bug report with a concrete failure mode in from_pretrained/save_pretrained, so it is the best single issue for a reader to understand and act on quickly.", + "best_issue_reason": "45372 is the best issue to anchor the nearest related subgroup because it has the clearest, most specific failure report, but it is not a good global duplicate hub for the full set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "Both mention incorrect model outputs, but one is siglip2 zero probabilities and the other is a position-id key warning; different symptoms and likely different code paths." + "reason": "Different symptoms and domains: zero probabilities in SigLIP2 vs unexpected position-id keys; not the same bug." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Flash-attention generation failure for Qwen3.5 and multi-GPU CUDA OOM for Gemma4 are unrelated bugs." + "reason": "One is flash-attention generation behavior, the other is multi-GPU CUDA OOM; different failure modes and code paths." }, { "left": "issue:44857", "right": "issue:45072", "accept": false, - "reason": "AMP/CUDA crash in a loss module and bfloat16 dtype mismatches in inference are distinct issues." + "reason": "Both involve numeric precision, but they hit different models/components and distinct crashes/correctness issues." }, { "left": "issue:43746", "right": "issue:46209", "accept": false, - "reason": "PEFT/local checkpoint loading for GraniteSpeech and FP8 inference failure on B200 are different problems." + "reason": "PEFT/local-checkpoint loading for GraniteSpeech is unrelated to FP8 inference failure on B200." }, { "left": "issue:44448", "right": "issue:44977", "accept": false, - "reason": "Version-to-version output drift for Pegasus and a flash-attention generation failure for Qwen3.5 are not the same underlying bug." + "reason": "Version output differences for Pegasus are not the same as Qwen flash-attention generation failures." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "IterableDataset multi-image finetuning failure and BeitImageProcessorFast label reduction are unrelated data-processing bugs." + "reason": "Different subsystems and defects: multimodal iterable-dataset handling vs Beit fast processor label reduction." }, { "left": "issue:28218", "right": "issue:32090", "accept": false, - "reason": "Tokenizer spacing after added tokens has no overlap with Trainer broadcast TypeError." + "reason": "Tokenizer spacing after added tokens is unrelated to Trainer/_gpu_broadcast_one NoneType errors." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "Checkpoint resume key mismatch and an omni-moe feature-length helper bug are different code paths." + "reason": "Checkpoint key mismatch and a helper-function length calculation bug are unrelated." }, { - "left": "issue:32090", - "right": "issue:33357", + "left": "issue:29942", + "right": "issue:35141", "accept": false, - "reason": "Trainer NoneType broadcast error and MacOS bus error with a CLIP model do not describe the same failure." + "reason": "Flash Attention test failures and embedding reinitialization on resize are different bugs." }, { - "left": "issue:29942", - "right": "issue:35141", + "left": "issue:32090", + "right": "issue:33357", "accept": false, - "reason": "Flash Attention 2 test failures are unrelated to embedding reinitialization after resizing token embeddings." + "reason": "Different runtime failures: NoneType broadcast in Trainer vs MacOS bus error with CLIP." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "torch.compile SDPA failure in Bamba and a helper-function logic bug in qwen3_omni_moe are different." + "reason": "Bamba torch.compile/SDPA failure is unrelated to qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:43723", "right": "issue:44554", "accept": false, - "reason": "AutoTokenizer v5 loading issue and MPS attention correctness on value/query head mismatch are unrelated." + "reason": "Tokenizer loading regression in v5 is not the same as the MPS attention correctness issue." }, { "left": "issue:37428", "right": "issue:44987", "accept": false, - "reason": "Missing flash-attention utility import and a model load failure for physical-intelligence/fast are both import-related at a high level, but not the same concrete bug." + "reason": "Missing flash-attention symbol import and loading a specific model in v5.1 are different problems." }, { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "Removal of is_torch_fx_available breaking trust_remote_code models and Deepseek tokenizer regression in v5 are distinct issues." + "reason": "Both are v5 regressions, but one is trust_remote_code FX removal and the other is Deepseek tokenizer correctness." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Quantization offloading behavior and multi-GPU CUDA OOM are different root causes and would not be merged into one fix." + "reason": "Quantization/offloading behavior is not the same as Gemma4 multi-GPU OOM." }, { "left": "issue:43927", "right": "issue:45698", "accept": false, - "reason": "Both involve save/load flows, but one is custom token IDs being lost and the other is the wrong custom module being loaded; separate bugs." + "reason": "Save/load config token IDs and loading the wrong custom module after save_pretrained are distinct bugs." }, { "left": "issue:44462", "right": "issue:44561", "accept": false, - "reason": "Tokenizer.json being ignored and trust_remote_code breaking after a removed helper are unrelated." + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to removal of is_torch_fx_available." }, { "left": "issue:43278", "right": "issue:43329", "accept": false, - "reason": "BF16-to-FP32 dtype drift in evaluate is unrelated to undefined variables in the video token-count helper." + "reason": "Embedding dtype drift and an undefined-variable bug in multimodal token counting are different issues." }, { "left": "issue:42490", "right": "issue:43575", "accept": false, - "reason": "Trailing-slash local path import warning and tensor-parallel OOM loading Qwen2-57B are unrelated." + "reason": "Local-path import warning is unrelated to tensor-parallel OOM while loading Qwen2-57B-A14B-Instruct." }, { "left": "issue:10000", "right": "issue:15354", "accept": false, - "reason": "German DistilBertModel issue and torch.jit.script generator-expression export failure are different historical bugs." + "reason": "German DistilBertModel issue and a torch.jit GeneratorExp export problem are unrelated." }, { "left": "issue:41720", "right": "issue:45229", "accept": false, - "reason": "cudaErrorAssert with auto device mapping and multi-GPU OOM are separate inference failures." + "reason": "Different models and failures: auto device mapping cudaErrorAssert vs multi-GPU CUDA OOM." }, { "left": "issue:44977", "right": "issue:45563", "accept": false, - "reason": "Flash-attention generation correctness and a stale warning for num_return_sequences are different problems." + "reason": "Flash-attention generation failure and a stale warning about num_return_sequences are not the same bug." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma 4 processor loading import failure and flash_attention crash on s_aux=None are unrelated." + "reason": "Related Gemma4/flash-attention area, but one is an import error for ReasoningEffort and the other is an AttributeError on s_aux=None; different root causes." }, { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Tokenizer special-token insertion regression and Gemma3 NaN embeddings are different failure modes." + "reason": "Tokenizer add_special_tokens behavior and Gemma3 NaN embeddings are unrelated defects." }, { "left": "issue:44857", "right": "issue:45588", "accept": false, - "reason": "AMP/CUDA loss crash and flash_attention s_aux=None AttributeError are not the same code-path issue." + "reason": "AMP/CUDA loss crash is not the same as flash_attention.py crashing on sink-less models." }, { "left": "issue:43412", "right": "issue:45588", "accept": false, - "reason": "Executorch export issues in Gemma3n and flash_attention sink-less-model crash are unrelated." + "reason": "Executorch export failures and flash_attention.py AttributeError are different code paths and bugs." }, { "left": "issue:45198", "right": "issue:45803", "accept": false, - "reason": "Wav2Vec2 save/tokenization failure and a bare-except bug in Fuyu tensor conversion are different bugs." + "reason": "Wav2Vec2 save/tokenization failures are unrelated to FuyuBatchFeature's bare-except problem." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "ReasoningEffort import failure for Gemma 4 processors and missing _tokenizer in Gemma4Processor are distinct load-time issues." + "reason": "Both involve Gemma4 processor loading, but one is an import dependency failure and the other is a serve-time missing-_tokenizer AttributeError; not the same concrete bug." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Embedding dtype changes and gradient checkpointing in eval mode are unrelated." + "reason": "Embedding dtype drift and gradient checkpointing being unusable in eval mode are different problems." }, { "left": "issue:45588", "right": "issue:45823", "accept": false, - "reason": "flash_attention.py crashing on s_aux=None and Gemma4 PLE device mismatch under device_map=auto are different bugs." + "reason": "Flash_attention.py crashing on s_aux=None is unrelated to Gemma4 PLE device mismatch under device_map=\"auto\"." } ] }, @@ -8566,8 +8574,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -8623,18 +8631,18 @@ "soft_pairs": [ "issue:43976|issue:45440", "issue:44448|issue:44466", - "issue:45459|issue:45712", "issue:43653|issue:44743", "issue:25251|issue:32090", + "issue:45459|issue:45712", "issue:45245|issue:45563", "issue:44206|issue:44479", "issue:43065|issue:43994", - "issue:45372|issue:45459", "issue:44186|issue:45198", "issue:43065|issue:45412", "issue:43611|issue:46121", - "issue:44871|issue:44977", + "issue:45372|issue:45459", "issue:44361|issue:45198", + "issue:44871|issue:44977", "issue:32090|issue:33290", "issue:42490|issue:43873", "issue:33357|issue:39692", @@ -8651,208 +8659,202 @@ "issue:43653|issue:44442", "issue:44610|issue:45412" ], - "prepared_review_unit_hash": "220f151e3a5ead8f5c97ba880bb847846d0202f4cfbe3ff74eab79b21290783c", + "prepared_review_unit_hash": "67263a04103ca719749ee6c04d942947bd4d62028cf7da15e784993e7ecea5f1", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36409, - "estimated_input_tokens": 9103, - "estimated_eval_tokens": 18462 + "serialized_chars": 36406, + "estimated_input_tokens": 9102, + "estimated_eval_tokens": 18460 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is not a true duplicate cluster: the issues span unrelated areas (tokenizers, training, quantization, vision processors, model loading, and CI). No single canonical issue or PR stands out, and all soft edges should be rejected.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "This cluster is noisy and mostly spans unrelated Transformers bug reports across tokenizers, model loading, generation, quantization, and processors. I did not find any true duplicate pair among the soft candidates; all should stay separate.", + "confidence": 0.79, + "canonical_issue_reason": "If forced to pick one issue as the cluster anchor, issue 43066 is the broadest user-facing Transformers v5 regression in the set and is closest to the recurring tokenizer/loading theme, though the cluster is not a true duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Issue 43066 is the best representative of the cluster because it describes a general v5 tokenizer regression rather than a highly model-specific edge case.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Different failures: Python version incompatibility vs DeepseekV3MoE implementation divergence." + "reason": "Different problems: a Python version/import compatibility issue versus a model implementation divergence." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "Different bugs: generation output drift vs inconsistent lm_head.weight serialization." - }, - { - "left": "issue:45459", - "right": "issue:45712", - "accept": false, - "reason": "Unrelated tokenizer error masking vs leftover dummy class cleanup." + "reason": "One is a generation output regression for Pegasus; the other is serialization inconsistency for lm_head.weight. Different code paths." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "BigBirdTokenizer special-token registration is unrelated to Qwen3.5 recurrent-state reset behavior." + "reason": "Tokenizer special-token decode bug versus Qwen recurrent-state/cache reset bug. Not the same underlying issue." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Nested pipeline output shape change is unrelated to Trainer _gpu_broadcast_one NoneType crash." + "reason": "Pipeline top_k nesting change and Trainer NoneType broadcast failure are unrelated." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer error masking on missing protobuf versus cleanup of leftover dummy classes. Different fixes." }, { "left": "issue:45245", "right": "issue:45563", "accept": false, - "reason": "Category cardinality/runtime error is unrelated to stale generate() warning behavior." + "reason": "Category cardinality runtime error and a stale warning in paged generate() are distinct bugs." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Unsupported center argument in LasrFeatureExtractor is unrelated to video-input regression in Qwen models." + "reason": "Unsupported center argument in a feature extractor versus a video-input regression in Qwen models. Different failures." }, { "left": "issue:43065", "right": "issue:43994", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d is unrelated to SigLIP2 AutoModel/pipeline nonsense outputs." - }, - { - "left": "issue:45372", - "right": "issue:45459", - "accept": false, - "reason": "Gemma 4 processor import failure from mistral_common is unrelated to protobuf error masking in tokenizers." + "reason": "SAM3 dummy Conv2d placeholder issue versus SigLIP2 incorrect outputs. Not the same bug." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes on NER/padding are unrelated to Wav2Vec2 save_pretrained/tokenization failures." + "reason": "LayoutLMv2Tokenizer crash on NER/padding versus Wav2Vec2 save/tokenization failure; unrelated tokenizer paths." }, { "left": "issue:43065", "right": "issue:45412", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder is unrelated to RT-DETR memory release problems." + "reason": "Dummy layer placeholder cleanup and RT-DETR memory release are unrelated." }, { "left": "issue:43611", "right": "issue:46121", "accept": false, - "reason": "base_model_prefix loading breakage is unrelated to rope-parameter JSON list handling." + "reason": "base_model_prefix loading regression and rope-parameter validation TypeError are different model/config issues." }, { - "left": "issue:44871", - "right": "issue:44977", + "left": "issue:45372", + "right": "issue:45459", "accept": false, - "reason": "Gemma-3 eos_token_id mismatch is unrelated to Qwen3.5 flash-attention generation issues." + "reason": "Gemma processor import dependency breakage versus hidden tokenizer decode errors; different root causes." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "MLukeTokenizer AttributeError is unrelated to Wav2Vec2 save/tokenization failure." + "reason": "MLukeTokenizer task AttributeError and Wav2Vec2 save/tokenization failure do not share a concrete code path." }, { - "left": "issue:32090", - "right": "issue:33290", + "left": "issue:44871", + "right": "issue:44977", "accept": false, - "reason": "Trainer NoneType broadcast crash is unrelated to Adafactor/DeepSpeed OOM." + "reason": "Gemma eos_token configuration mismatch versus Qwen flash-attention generation regression; unrelated." }, { "left": "issue:42490", "right": "issue:43873", "accept": false, - "reason": "Trailing-slash import warning is unrelated to offloading-with-quantization behavior." + "reason": "Local-path import warning and quantization/offloading behavior are different issues." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "MacOS bus error with CLIP is unrelated to SigLIP2 doc example mistakes." + "reason": "MacOS bus error on a community CLIP model is not the same as documentation example errors for SigLIP2." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 test failures are unrelated to void segmentation map processing." + "reason": "Flash Attention 2 CI failures versus void segmentation map processing failure; unrelated." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 error-message clarity is unrelated to TrainerState token counter not updating." + "reason": "LayoutLMv3 error-message clarity issue and TrainerState token counter not updating are different bugs." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "FineGrainedFP8Config failure for Qwen3Moe is unrelated to quantized model saving." + "reason": "Both are quantization-related, but one fails on load with FineGrainedFP8Config and the other fails while saving quantized models; distinct phases and likely distinct causes." }, { "left": "issue:43994", "right": "issue:45412", "accept": false, - "reason": "SigLIP2 inference mismatch is unrelated to RT-DETR memory leaks." + "reason": "SigLIP2 nonsensical outputs versus RT-DETR memory leak; unrelated." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "torch.export/torch_compilable_check failure is unrelated to Wav2Vec2 save/tokenization issues." + "reason": "torch.export failure on torch_compilable_check and Wav2Vec2 tokenization/save failure are unrelated." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Wrong tokenizer decoder type in v5 is unrelated to AutoTokenizer ignoring tokenizer.json." + "reason": "Both involve tokenizers, but one is wrong decoder selection in v5 and the other is AutoTokenizer ignoring tokenizer.json; different underlying failures." }, { "left": "issue:43278", "right": "issue:43575", "accept": false, - "reason": "Embedding dtype drift during eval is unrelated to tp loading OOM for Qwen2-57B-A14B-Instruct." + "reason": "BF16-to-FP32 dtype change in eval and tensor-parallel OOM are unrelated." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch is unrelated to DeepseekV3MoE implementation divergence." + "reason": "RoPE layer count bug in Smollm3 and DeepSeekV3 implementation divergence are separate model-specific issues." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Runtime special-token post-processor updates are unrelated to Gemma3 NaN embeddings with mixed-length batching." + "reason": "Runtime post-processor update behavior and Gemma3 NaN embeddings are different code paths." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient checkpointing in eval mode is unrelated to resume_from_checkpoint key mismatch." + "reason": "Gradient checkpointing in eval mode versus resume_from_checkpoint key mismatch are unrelated training issues." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "BigBirdTokenizer mask-token registration is unrelated to AutoTokenizer failing to load FastSpeech2ConformerTokenizer." + "reason": "BigBirdTokenizer special-token decode issue and AutoTokenizer load failure for FastSpeech2ConformerTokenizer are different tokenizer bugs." }, { "left": "issue:44610", "right": "issue:45412", "accept": false, - "reason": "OmDet-Turbo processor size mismatch is unrelated to RT-DETR memory reclamation." + "reason": "Processor image-size mismatch and RT-DETR memory leak are unrelated." } ] }, @@ -8871,8 +8873,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -8939,8 +8941,8 @@ "issue:43334|issue:43531", "issue:44442|issue:45072", "issue:43575|issue:43701", - "issue:42915|issue:43931", "issue:43526|issue:44857", + "issue:42915|issue:43931", "issue:42490|issue:43381", "issue:44977|issue:45245", "issue:43575|issue:44387", @@ -8954,20 +8956,20 @@ "issue:33290|issue:34689", "issue:44442|issue:44898" ], - "prepared_review_unit_hash": "49e63cdcda54e50d7465058e56f34643575ab7606f127b38863fd9c07b392e7f", + "prepared_review_unit_hash": "18cb09e631fb8215bb0725c3b4860ce76bd30a3e031e5ab1dcd6121c05d2a120", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 35765, + "serialized_chars": 35767, "estimated_input_tokens": 8942, "estimated_eval_tokens": 18140 }, @@ -8976,174 +8978,174 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items do not form a true duplicate cluster; they span many unrelated Transformers bugs across different models, loaders, processors, and runtime paths. All proposed soft edges are reject-worthy on the available evidence.", - "confidence": 0.77, - "canonical_issue_reason": "If a representative must be chosen, issue 44279 is the broadest/most umbrella-like title, but it is still too generic to be a strong canonical for this set.", + "summary": "This cluster is highly heterogeneous: the proposed edges connect unrelated regressions, model-specific issues, and API/import problems rather than the same underlying bug. I would reject all soft edges and treat the set as noise rather than a duplicate cluster.", + "confidence": 0.96, + "canonical_issue_reason": "No true canonical issue stands out because the items are unrelated; if a single anchor is required, issue 45372 is the least-bad choice since it is open, well-described, and has inbound references, but it is not a duplicate representative.", "canonical_pr_reason": null, - "best_issue_reason": "44279 is the least specific issue here, but the collection is too heterogeneous for any issue to be a reliable duplicate anchor.", + "best_issue_reason": "Issue 45372 is the best cluster anchor only in the loose sense that it is open and already referenced; however, it should not be treated as the canonical duplicate because the cluster does not share one underlying bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43824", "right": "issue:44279", "accept": false, - "reason": "Specific Qwen2.5-VL import failure vs generic dependency issue; not the same bug." + "reason": "Different problems: Qwen2.5-VL import failure vs a broad dependency issue; no shared code-path." }, { "left": "issue:29127", "right": "issue:33357", "accept": false, - "reason": "LayoutLMv3 error-message clarity vs CLIP MacOS bus error; unrelated symptoms and code paths." + "reason": "LayoutLMv3 error-message clarity vs a MacOS CLIP bus error; unrelated symptoms and subsystems." }, { "left": "issue:29942", "right": "issue:36010", "accept": false, - "reason": "Flash Attention 2 test failures vs GenerationMixin import regression; different failures." + "reason": "Flash Attention 2 test failures are unrelated to the GenerationMixin import error." }, { "left": "issue:45440", "right": "issue:45593", "accept": false, - "reason": "DeepSeekV3MoE implementation divergence vs D-FINE auxiliary-loss omission; unrelated model bugs." + "reason": "Different model families and failure modes: DeepseekV3MoE divergence vs D-FINE auxiliary-loss behavior." }, { "left": "issue:43761", "right": "issue:45859", "accept": false, - "reason": "CLIPVisionModel hidden_states regression vs Qwen3.5Moe CPU-offload incompatibility; different components." + "reason": "Both involve Qwen-family components, but one is a CLIPVisionModel hidden-state regression and the other is CPU-offload rotary embedding compatibility." }, { "left": "issue:43526", "right": "issue:45834", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug vs Kosmos2.5 long-OCR index error; not the same issue." + "reason": "BeitImageProcessorFast label reduction bug is unrelated to Kosmos2.5 OCR index errors." }, { "left": "issue:43526", "right": "issue:44186", "accept": false, - "reason": "Image processor label handling vs LayoutLMv2Tokenizer NER/padding crash; unrelated." + "reason": "Both are image/tokenization issues, but they affect different processors and different failure mechanisms." }, { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "auto_docstring annotations crash vs mistral_common import error in Gemma processor loading; different root causes." + "reason": "auto_docstring annotation handling is unrelated to the mistral_common ReasoningEffort import failure." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "Error-message clarity issue vs post_init reinitialization bug; not the same code path." + "reason": "LayoutLMv3 box validation messaging and token-embedding reinitialization are different code paths." }, { "left": "issue:44625", "right": "issue:45584", "accept": false, - "reason": "Config propagation bug vs Whisper empty-transcription generation failure; unrelated." + "reason": "Qwen3.5 num_labels propagation and Whisper empty-transcription generation are unrelated." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Missing pad_token_id for Qwen3-VL text config vs sliding_window issue in Qwen3-MoE; different model families." + "reason": "Qwen3-VL missing pad_token_id and Qwen3-MoE sliding_window are separate configuration bugs." }, { "left": "issue:44442", "right": "issue:45072", "accept": false, - "reason": "Tokenizer loading failure vs bfloat16 dtype mismatch during inference; different concrete problems." + "reason": "FastSpeech2ConformerTokenizer loading failure is unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." }, { "left": "issue:43575", "right": "issue:43701", "accept": false, - "reason": "Tensor-parallel OOM on model load vs checkpoint key mismatch; unrelated failures." + "reason": "Tensor-parallel OOM on Qwen2-57B is not the same as resume_from_checkpoint key mismatch." }, { - "left": "issue:42915", - "right": "issue:43931", + "left": "issue:43526", + "right": "issue:44857", "accept": false, - "reason": "FineGrainedFP8Config failure vs Qwen3-VL weight-shape mismatch; different issues." + "reason": "BeitImageProcessorFast label handling and LwDetrImageLoss AMP crashes do not share a concrete implementation bug." }, { - "left": "issue:43526", - "right": "issue:44857", + "left": "issue:42915", + "right": "issue:43931", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug vs LwDetrImageLoss AMP/CUDA crash; unrelated." + "reason": "FineGrainedFP8Config failure for Qwen3Moe and Qwen3-VL weight-shape mismatch are distinct model-loading issues." }, { "left": "issue:42490", "right": "issue:43381", "accept": false, - "reason": "Path import warning vs gradient-checkpointing-in-eval restriction; not the same bug." + "reason": "Trailing-slash import warning and gradient-checkpointing eval-mode restriction are unrelated." }, { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Qwen3.5 flash-attention generation bug vs category-count limit error; unrelated." + "reason": "Qwen3.5 flash-attention generation bug is not the same as the 2^24 category limit error." }, { "left": "issue:43575", "right": "issue:44387", "accept": false, - "reason": "Both are OOM-related, but one is TP load OOM and the other is int4 memory growth; not the same concrete problem." + "reason": "Both mention memory, but one is load-time TP OOM and the other is increased CUDA reserved memory under int4 quantization; not the same bug." }, { "left": "issue:44464", "right": "issue:45491", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency vs Gemma3 NaN embeddings from sliding-window padding; different bugs." + "reason": "Chunked generation with compiled forward and Gemma3 mixed-length NaN embeddings are different generation/attention failures." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Chat-template sample-rate default vs feature-extractor center-arg crash; unrelated." + "reason": "Audio processor chat-template sampling-rate default and LasrFeatureExtractor unsupported center arg are unrelated audio preprocessing bugs." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Generic custom-init regression vs patchtsmixer post_init policy issue; not a shared concrete fix." + "reason": "Custom model initialization breakage in Transformers 5.0.0 is not the same as patchtsmixer optional post_init handling." }, { "left": "issue:44743", "right": "issue:44811", "accept": false, - "reason": "Recurrent-state reset bug vs Whisper batch_decode skip_special_tokens behavior; unrelated." + "reason": "Qwen3.5 recurrent-state reset bug and Whisper batch_decode skip_special_tokens issue affect different logic paths." }, { "left": "issue:43452", "right": "issue:45588", "accept": false, - "reason": "gguf_file loading regression vs flash_attention.py s_aux=None crash; different subsystems and failures." + "reason": "gguf_file loading breakage and flash_attention.py s_aux=None crash are separate loader/integration bugs." }, { "left": "issue:44062", "right": "issue:45907", "accept": false, - "reason": "AddedToken keyword collision vs list_repo_files revision forwarding; unrelated API bugs." + "reason": "AddedToken duplicate special-argument error is unrelated to list_repo_files missing revision forwarding." }, { "left": "issue:43761", "right": "issue:44079", "accept": false, - "reason": "CLIP hidden_states regression vs ModelOutput key assignment bug; different code paths." + "reason": "CLIPVisionModel hidden_states regression and ModelOutput key assignment bug are different abstractions." }, { "left": "issue:33290", "right": "issue:34689", "accept": false, - "reason": "Adafactor+DeepSpeed OOM vs Llama 3.2 vision model loading breakage; unrelated." + "reason": "AdaFactor+DeepSpeed OOM and Llama 3.2 Vision model-loading breakage are different failures." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Tokenizer load failure vs Perceiver non-default resolution failure; not the same bug." + "reason": "Tokenizer loading and Perceiver non-default-resolution image classification failures are unrelated." } ] }, @@ -9162,8 +9164,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -9221,19 +9223,19 @@ "issue:45003|issue:45440", "issue:43723|issue:45479", "issue:43381|issue:43873", - "issue:43232|issue:43388", "issue:43316|issue:43525", + "issue:43232|issue:43388", "issue:42490|issue:44743", "issue:43761|issue:44188", - "issue:34689|issue:43611", "issue:44479|issue:45290", + "issue:34689|issue:43611", "issue:43720|issue:44442", "issue:44987|issue:46178", "issue:44442|issue:45803", "issue:44077|issue:44479", "issue:43931|issue:45216", - "issue:43493|issue:43827", "issue:34689|issue:36010", + "issue:43493|issue:43827", "issue:33290|issue:36331", "issue:43316|issue:45759", "issue:42907|issue:45698", @@ -9245,14 +9247,14 @@ "issue:44190|issue:44496", "issue:44855|issue:45137" ], - "prepared_review_unit_hash": "a38092142cb8957161dfccba6207c799108865976f3a1ec35a6c8f85cab5209a", + "prepared_review_unit_hash": "563af6ed5ffaec350ac1c20a5e9ff5c8f988e1e8fe1f5d4a13bd710303dd8b4b", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 47, @@ -9267,180 +9269,180 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair appears to describe the same underlying bug or change. The set is a grab bag of unrelated issues spanning loading, generation, training, quantization, and docs regressions, so there is no canonical duplicate to select.", - "confidence": 0.95, - "canonical_issue_reason": null, + "summary": "The cluster is highly heterogeneous: most pairs are only weak title-similarity matches and do not look like the same underlying bug or change. No soft edges should be merged.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45698 is the broadest and most representative open bug in the set: a loading/persistence problem around `from_pretrained`/`save_pretrained`, with inbound references and discussion. It is the least specific single issue to anchor the cluster, though the cluster itself is not cohesive.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45698 is the best global representative because it is open, broadly phrased, and centered on a common model-loading workflow that is more likely to recur than the very narrow edge cases elsewhere in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:33290", "right": "issue:34567", "accept": false, - "reason": "OOM with Adafactor/DeepSpeed is unrelated to TrainerState.num_input_tokens_seen not updating." + "reason": "Different failures: DeepSpeed Adafactor OOM vs TrainerState token counter not updating. Same general training area, but not the same bug or code path." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "Different failure modes: optimizer memory blowup vs embedding reinitialization after resize." + "reason": "Unrelated: optimizer memory blow-up vs embedding resize/post_init reinitialization bug." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "sys.modules access bug is unrelated to DeepSeekV3MoE implementation divergence." + "reason": "Both involve model loading internals, but one is unsafe `sys.modules` access and the other is DeepSeekV3 implementation divergence; not the same underlying issue." }, { "left": "issue:43723", "right": "issue:45479", "accept": false, - "reason": "Tokenizer loading regression and zero-loss sequence-classification bug are unrelated." + "reason": "Tokenizer loading failure vs sequence-classification loss degeneracy; different subsystems and symptoms." }, { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Gradient checkpointing in eval mode and offloading/quantization are different issues." + "reason": "Gradient checkpointing eval-mode behavior vs quantization/offloading bug. Similar high-level area, but not one concrete defect." }, { - "left": "issue:43232", - "right": "issue:43388", + "left": "issue:43316", + "right": "issue:43525", "accept": false, - "reason": "Generation kwargs after sync_gpus does not match metric gathering dropping tuple labels." + "reason": "Both are config/API discrepancies, but they affect different models and fields (`Gemma3TextConfig` vs `Llama4Config.pad_token_id`)." }, { - "left": "issue:43316", - "right": "issue:43525", + "left": "issue:43232", + "right": "issue:43388", "accept": false, - "reason": "Different config API problems: Gemma3TextConfig discrepancy vs missing pad_token_id on Llama4Config." + "reason": "Generation kwargs ordering bug vs metric gathering dropping tuple labels. Different code paths." }, { "left": "issue:42490", "right": "issue:44743", "accept": false, - "reason": "Trailing-slash path import warning is unrelated to Qwen recurrent-state reset with cache." + "reason": "Local-path import warning vs recurrent-state reset in Qwen3.5 modular model; not the same bug." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "CLIP hidden_states regression is unrelated to attention-kernel divergence under torch.compile." + "reason": "Hidden states regression in CLIP vision forward vs attention-kernel divergence under `torch.compile`." }, { - "left": "issue:34689", - "right": "issue:43611", + "left": "issue:44479", + "right": "issue:45290", "accept": false, - "reason": "Llama 3.2 model loading break and base_model_prefix loading regression are distinct." + "reason": "Video-input regression for several Qwen VL models vs chat template crash on tool-call assistant messages; unrelated input pipelines." }, { - "left": "issue:44479", - "right": "issue:45290", + "left": "issue:34689", + "right": "issue:43611", "accept": false, - "reason": "Video-input regression in Qwen VL models is unrelated to chat-template tool-call crashes." + "reason": "Model-loading breakage in one release vs base_model_prefix loading regression in another; same broad theme, but not a single concrete fix." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "BitNet packed-weight loading bug is unrelated to AutoTokenizer failing on FastSpeech2ConformerTokenizer." + "reason": "BitNet packed-weight unpacking during accelerate loading vs AutoTokenizer loading FastSpeech2ConformerTokenizer; different loading layers and model types." }, { "left": "issue:44987", "right": "issue:46178", "accept": false, - "reason": "Physical-intelligence/fast load failure is unrelated to missing tensorflow_text backend mapping." + "reason": "Loading a specific model failing in v5.1.0+ vs backend mapping error for `tensorflow_text`; not the same defect." }, { "left": "issue:44442", "right": "issue:45803", "accept": false, - "reason": "Tokenizer loading failure and FuyuBatchFeature exception handling are different code paths." + "reason": "Tokenizer loading failure vs Fuyu tensor conversion exception handling; unrelated code paths." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "patchtsmixer post_init policy issue is unrelated to Qwen video-input regression." + "reason": "Model post-init policy for `patchtsmixer` vs video-input regression in Qwen VL models." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Weight-shape mismatch on load is unrelated to a bad save_pretrained checkpoint for Qwen3.5." + "reason": "Shape mismatch loading a specific model vs incorrect checkpoint saving for Qwen3.5; both are serialization-related but not the same bug." }, { - "left": "issue:43493", - "right": "issue:43827", + "left": "issue:34689", + "right": "issue:36010", "accept": false, - "reason": "SigLIP2 implementation mismatch is unrelated to docs still mentioning pipeline()." + "reason": "Model loading regression vs missing `GenerationMixin` import; different failure modes and fixes." }, { - "left": "issue:34689", - "right": "issue:36010", + "left": "issue:43493", + "right": "issue:43827", "accept": false, - "reason": "Model-loading failure and GenerationMixin import error are separate regressions." + "reason": "SigLIP2 implementation mismatch vs docs still referencing removed pipeline APIs; one is model behavior, the other documentation cleanup." }, { "left": "issue:33290", "right": "issue:36331", "accept": false, - "reason": "DeepSpeed Adafactor OOM is unrelated to CustomTrainer.compute_loss receiving a new kwarg." + "reason": "DeepSpeed/Adafactor OOM vs custom trainer loss signature mismatch. No shared underlying bug." }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Gemma3TextConfig API discrepancy is unrelated to composite Qwen multimodal from_config unwrapping." + "reason": "Config API discrepancy for Gemma3TextConfig vs `from_config` text_config unwrapping for Qwen multimodal configs; related to configs but distinct defects." }, { "left": "issue:42907", "right": "issue:45698", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals is unrelated to loading the wrong custom module after save_pretrained." + "reason": "Both are save/load related, but one is dequantized Ministral saving and the other is wrong custom module reloading after save_pretrained; not the same code-path problem." }, { "left": "issue:45290", "right": "issue:46209", "accept": false, - "reason": "Chat-template tool-call crash is unrelated to FP8 inference failure on B200." + "reason": "Chat template crash on tool-call assistant messages vs FP8 inference failure on B200; unrelated." }, { "left": "issue:43873", "right": "issue:44871", "accept": false, - "reason": "Quantization/offloading behavior does not match Gemma eos_token_id configuration mismatch." + "reason": "Quantization/offloading issue vs Gemma eos token config mismatch; both model/runtime issues but not the same bug." }, { "left": "issue:43749", "right": "issue:43931", "accept": false, - "reason": "FSDP CPU RAM efficient loading and Qwen3-VL shape mismatch are different loading bugs." + "reason": "FSDP CPU RAM efficient loading vs Qwen3-VL weight-shape mismatch. Both loading-related, but not the same defect." }, { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "trust_remote_code returning missing fields is unrelated to GPTNeoX rotary_pct resetting on reload." + "reason": "Missing fields with `trust_remote_code=True` vs GPTNeoX rotary_pct not persisting on reload; both config serialization issues but distinct." }, { "left": "issue:43650", "right": "issue:46202", "accept": false, - "reason": "The vague 'ADD THE DATA' issue is not the same as Gemma pipeline roundtrip failure." + "reason": "Generic/low-signal issue vs Gemma pipeline roundtrip failure; no evidence they share the same bug." }, { "left": "issue:44190", "right": "issue:44496", "accept": false, - "reason": "Local dataset loading script issue is unrelated to unrecognized model config loading." + "reason": "Local dataset loading in a training script vs unrecognized model/config.json error for a specific repo; different surfaces." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "PyTorch 3.13 decorator parsing/import error is unrelated to DeepSpeed ZeRO3 deque underflow." + "reason": "Python 3.13 TorchScript parsing/import error vs DeepSpeed ZeRO3 deque underflow; unrelated." } ] }, @@ -9459,8 +9461,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -9518,18 +9520,18 @@ "issue:43065|issue:44617", "issue:43295|issue:43825", "issue:43550|issue:43701", - "issue:43606|issue:43701", "issue:43316|issue:44877", + "issue:43606|issue:43701", "issue:43316|issue:43716", - "issue:42617|issue:43575", "issue:43540|issue:44514", + "issue:42617|issue:43575", "issue:44077|issue:44534", "issue:44162|issue:44792", - "issue:16998|issue:22355", "issue:44871|issue:45584", + "issue:16998|issue:22355", "issue:43575|issue:43742", - "issue:43404|issue:43441", "issue:45803|issue:46170", + "issue:43404|issue:43441", "issue:43295|issue:44871", "issue:43673|issue:44743", "issue:43761|issue:44479", @@ -9540,198 +9542,205 @@ "issue:44871|issue:46050", "issue:45127|issue:45200", "issue:43572|issue:46178", - "issue:32090|issue:35141" + "issue:32090|issue:35141", + "issue:44617|issue:45584" ], - "prepared_review_unit_hash": "865123dcf87d23c2abc168a7458f99fbbec2be29425cfc2de3e5a58e0caf8204", + "prepared_review_unit_hash": "2ecea2b83c198bde38cffbbe3afa47bf843cacafbc7b6321cf1079783a0aa570", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 27, - "serialized_chars": 35721, - "estimated_input_tokens": 8931, - "estimated_eval_tokens": 18118 + "soft_pair_count": 28, + "serialized_chars": 36021, + "estimated_input_tokens": 9006, + "estimated_eval_tokens": 18268 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated transformer issues that only share broad subsystem or release-regression language; none look like true duplicates, so there is no strong canonical item.", - "confidence": 0.18, - "canonical_issue_reason": "No clear canonical duplicate exists; the set spans many different bug classes (pipelines, configs, model loading, training, multimodal/video, distributed, and checkpoints).", + "summary": "The items are a heterogeneous mix of unrelated Transformers bugs and regressions. The soft-paired issues share only broad subsystem keywords (e.g. loading, config, video, cache, v5 regressions) but not the same concrete underlying failure, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the broadest umbrella-style regression report in the set, but it still does not subsume the others and should not be treated as a duplicate target.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:25251", "right": "issue:28282", "accept": false, - "reason": "Pipeline output nesting bug vs missing PyTorch import are unrelated failure modes." + "reason": "Pipeline top_k output-shape bug and missing `transformers.onnx` import are unrelated failures." }, { "left": "issue:43065", "right": "issue:44617", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder and Sam3Video OOM are different SAM3 problems." + "reason": "One is about a dummy Conv2d in Sam3PixelDecoder, the other is CUDA OOM in Sam3Video; different code paths and symptoms." }, { "left": "issue:43295", "right": "issue:43825", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to a pipeline error message about translation tasks." + "reason": "Processor/tokenizer regression and a misleading pipeline error message are separate problems." }, { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "torch.compile/SDPA failure is unrelated to a checkpoint key mismatch." + "reason": "torch.compile/SDPA failure and resume_from_checkpoint key mismatch are unrelated." }, { - "left": "issue:43606", - "right": "issue:43701", + "left": "issue:43316", + "right": "issue:44877", "accept": false, - "reason": "CPU offload device mismatch and resume_from_checkpoint key mismatch are different bugs." + "reason": "Gemma3TextConfig API mismatch and granite_speech strict-config loading are different config issues." }, { - "left": "issue:43316", - "right": "issue:44877", + "left": "issue:43606", + "right": "issue:43701", "accept": false, - "reason": "Gemma3TextConfig API discrepancy is not the same as strict config loading failure for granite_speech." + "reason": "CPU offload device mismatch has nothing to do with checkpoint resume key mapping." }, { "left": "issue:43316", "right": "issue:43716", "accept": false, - "reason": "Config API discrepancy and image preprocessor dtype mismatch are separate issues." + "reason": "Config API discrepancy and image preprocessor/model dtype mismatch are distinct bugs." }, { - "left": "issue:42617", - "right": "issue:43575", + "left": "issue:43540", + "right": "issue:44514", "accept": false, - "reason": "3d_parallel.py startup failure and tp-related OOM are unrelated." + "reason": "Both involve Qwen video/text processing, but one is a video-input validation error and the other is batched apply_chat_template padding behavior." }, { - "left": "issue:43540", - "right": "issue:44514", + "left": "issue:42617", + "right": "issue:43575", "accept": false, - "reason": "Both are multimodal/video-related, but one is a Qwen3OmniMoe video input validation error and the other is a Qwen2_5_VL batching crash." + "reason": "3d_parallel.py runtime failure and tensor-parallel OOM are different distributed-inference problems." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "Optional post_init policy for patchtsmixer is unrelated to non-persistent buffer junk initialization." + "reason": "Optional post_init enforcement and non-persistent buffer initialization are separate v5 concerns." }, { "left": "issue:44162", "right": "issue:44792", "accept": false, - "reason": "ESM2 training breakage and janus image-generation test failure do not share the same bug." + "reason": "ESM2 model breakage and a janus test failure do not describe the same bug." }, { - "left": "issue:16998", - "right": "issue:22355", + "left": "issue:44871", + "right": "issue:45584", "accept": false, - "reason": "model_max_length question for DeBERTa-V3 is unrelated to missing transformers.onnx module import." + "reason": "Gemma-3 eos_token_id config inconsistency and Whisper empty-transcription generation failure are unrelated." }, { - "left": "issue:44871", - "right": "issue:45584", + "left": "issue:16998", + "right": "issue:22355", "accept": false, - "reason": "Gemma eos_token_id config inconsistency is not the same as Whisper empty transcription failure." + "reason": "model_max_length question for DeBERTa-V3 and missing `transformers.onnx` import are unrelated." }, { "left": "issue:43575", "right": "issue:43742", "accept": false, - "reason": "OOM on large Qwen2 model loading is unrelated to a KeyError while loading MobileLLM-125M." + "reason": "Tensor-parallel OOM on a large Qwen model and a key error loading MobileLLM are different load-time failures." }, { - "left": "issue:43404", - "right": "issue:43441", + "left": "issue:45803", + "right": "issue:46170", "accept": false, - "reason": "Mistral3 lm_head tying bug is different from a FlashAttention failure in Ministral-3." + "reason": "Bare except swallowing interrupts in FuyuBatchFeature is unrelated to Windows/ROCm distributed import issues." }, { - "left": "issue:45803", - "right": "issue:46170", + "left": "issue:43404", + "right": "issue:43441", "accept": false, - "reason": "Bare except swallowing errors in tensor conversion is unrelated to Windows/ROCm torch.distributed import issues." + "reason": "Missing lm_head tying in Mistral3ForConditionalGeneration and FlashAttention failure in Ministral-3 are different code-path bugs." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "Processor regression and Gemma eos_token_id config mismatch are different code paths." + "reason": "Processor/tokenizer regression and Gemma-3 eos_token_id config mismatch are separate issues." }, { "left": "issue:43673", "right": "issue:44743", "accept": false, - "reason": "Missing GenerationMixin cache during chunked_prefill is not the same as recurrent state reset in modular_qwen3_5.py." + "reason": "Missing GenerationMixin cache support during chunked_prefill and recurrent state reset in modular_qwen3_5 are not the same defect." }, { "left": "issue:43761", "right": "issue:44479", "accept": false, - "reason": "CLIPVisionModel hidden_states regression and Qwen video input regression are unrelated." + "reason": "CLIPVisionModel hidden_states regression and Qwen VL video-input regression affect different models and behaviors." }, { "left": "issue:41950", "right": "issue:43475", "accept": false, - "reason": "Video-classification pipeline selecting image processors is different from a missing fpn_position_embeddings attribute in Sam3 video." + "reason": "Video-classification pipeline picking image processors and Sam3VisionEncoderOutput missing an attribute are unrelated." }, { "left": "issue:43388", "right": "issue:43688", "accept": false, - "reason": "gather_for_metrics label truncation is unrelated to auxiliary loss normalization in OLMoE/GPT OSS." + "reason": "Metric gathering drops tuple labels, while the other issue is about auxiliary-loss normalization." }, { "left": "issue:43452", "right": "issue:44496", "accept": false, - "reason": "gguf loading breaks for tokenizer/model APIs, while the other is an unrecognized model config error." + "reason": "`gguf_file` argument handling and missing `model_type` in config.json are different model-loading failures." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Generic custom-model initialization breakage is not the same as GPTNeoX rotary_pct reload persistence." + "reason": "Broad v5 custom model initialization breakage is not the same as GPTNeoX rotary_pct persistence on reload." }, { "left": "issue:44871", "right": "issue:46050", "accept": false, - "reason": "Gemma eos_token_id mismatch is unrelated to sinq quantization save/load breakage." + "reason": "Gemma-3 eos_token_id mismatch and sinq quantized save/load breakage are separate serialization/config bugs." }, { "left": "issue:45127", "right": "issue:45200", "accept": false, - "reason": "LoRA merge collapse with tied embeddings is a different bug from missing mm_token_type_ids defaults." + "reason": "LoRA merge collapse with tied embeddings and Gemma 4 mm_token_type_ids defaults are unrelated." }, { "left": "issue:43572", "right": "issue:46178", "accept": false, - "reason": "StableLmConfig pad_token_idx regression is unrelated to tensorflow_text backend mapping errors." + "reason": "Missing pad_token_idx in StableLmConfig and tensorflow_text backend registration are different issues." }, { "left": "issue:32090", "right": "issue:35141", "accept": false, - "reason": "NoneType GPU broadcast failure and embedding reinitialization after resize are different training/model-init bugs." + "reason": "Trainer GPU broadcast NoneType error and output embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Sam3Video CUDA OOM and Whisper empty-transcription failure are not the same underlying bug." } ] }, @@ -9750,8 +9759,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -9773,6 +9782,7 @@ "issue:43994", "issue:44060", "issue:44077", + "issue:44112", "issue:44190", "issue:44242", "issue:44464", @@ -9780,7 +9790,6 @@ "issue:44521", "issue:44534", "issue:44610", - "issue:44617", "issue:44734", "issue:44737", "issue:44792", @@ -9806,17 +9815,16 @@ ], "soft_pairs": [ "issue:43504|issue:45698", - "issue:44617|issue:45584", "issue:44734|issue:45406", - "issue:45563|issue:45584", "issue:43525|issue:43531", + "issue:45563|issue:45584", "issue:43761|issue:44242", "issue:44792|issue:44977", "issue:43931|issue:44060", "issue:44479|issue:44734", "issue:44534|issue:44913", - "issue:44190|issue:44964", "issue:44792|issue:45600", + "issue:44190|issue:44964", "issue:44464|issue:45563", "issue:43827|issue:43994", "issue:44855|issue:45588", @@ -9833,210 +9841,211 @@ "issue:43550|issue:44190", "issue:43540|issue:44734", "issue:45127|issue:45357", - "issue:44737|issue:45803" + "issue:44737|issue:45803", + "issue:44112|issue:45600" ], - "prepared_review_unit_hash": "bf7d97444aec167c24983f713d74e949ec1daa3b58a0249f09638cb4cd3f7706", + "prepared_review_unit_hash": "0a82fef3c1b2cd654196228609d6206454b1d7a238ca96a8383b18530eb3028b", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36387, - "estimated_input_tokens": 9097, - "estimated_eval_tokens": 18450 + "serialized_chars": 36419, + "estimated_input_tokens": 9105, + "estimated_eval_tokens": 18466 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The soft pairs are largely heterogeneous and do not describe the same underlying bug or change. They span unrelated model-loading, multimodal preprocessing, generation, config, and docs/runtime regression issues, so none of the proposed duplicate links should be accepted.", - "confidence": 0.93, - "canonical_issue_reason": "issue:44521 is the most active/open issue in the set and has the strongest cross-referencing signal, so it is the best representative if one must be chosen; however, the overall set is not a true duplicate cluster.", + "summary": "The set is mostly unrelated one-off bugs across different models and subsystems. I do not see any soft pair that is clearly the same underlying issue, so I reject all duplicate candidates. The best representative issue is the broad serialization/init regression around non-persistent buffers.", + "confidence": 0.92, + "canonical_issue_reason": "issue:44534 is the cleanest broad statement of the non-persistent-buffer junk-initialization regression and is more generic than the version-specific wording in the older report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the strongest global issue candidate by discussion and inbound references, but it should be treated as representative only, not as a merge target for the other unrelated issues.", + "best_issue_reason": "issue:44534 is the most actionable and generally applicable issue in the set; it describes a concrete core bug rather than a narrow model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43504", "right": "issue:45698", "accept": false, - "reason": "BEiT legacy-field preset loading vs custom-module resolution after save/load; different subsystems and failure modes." - }, - { - "left": "issue:44617", - "right": "issue:45584", - "accept": false, - "reason": "Sam3Video CUDA OOM vs Whisper empty-transcription failure; unrelated code paths and symptoms." + "reason": "Different failures: Beit semantic-segmentation preset loading with a legacy field vs wrong custom module restored after save_pretrained." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "KV-cache continuation tensor indexing in serve vs Gemma4Processor missing _tokenizer; distinct serve bugs." + "reason": "Different code paths: KV-cache continuation indexing in transformers serve vs Gemma4Processor missing _tokenizer." }, { - "left": "issue:45563", - "right": "issue:45584", + "left": "issue:43525", + "right": "issue:43531", "accept": false, - "reason": "Paged-generate warning regression vs empty transcription after align_special_tokens; not the same bug." + "reason": "Different bugs: Llama4Config missing pad_token_id vs Qwen3-MoE sliding_window behavior." }, { - "left": "issue:43525", - "right": "issue:43531", + "left": "issue:45563", + "right": "issue:45584", "accept": false, - "reason": "Missing pad_token_id in Llama4Config vs sliding_window handling in Qwen3-MoE; unrelated config issues." + "reason": "Different symptoms and APIs: stale num_return_sequences warning vs Whisper empty-transcription failure after align_special_tokens." }, { "left": "issue:43761", "right": "issue:44242", "accept": false, - "reason": "CLIP hidden_states regression vs MoE load-balancing loss gating; separate model behaviors." + "reason": "Different model behaviors: CLIPVisionModel hidden_states regression vs load-balancing loss omission when output_router_logits=False." }, { "left": "issue:44792", "right": "issue:44977", "accept": false, - "reason": "janus image-generation test failure vs Qwen3.5 flash-attention generation bug; different failure surfaces." + "reason": "Different model families and failure modes: janus image-generation test failure vs Qwen3.5 flash-attention generation issue." }, { "left": "issue:43931", "right": "issue:44060", "accept": false, - "reason": "Qwen3-VL weight-shape load mismatch vs incorrect tied-weights warning; not the same underlying defect." + "reason": "One is a weight-shape mismatch loading Qwen3-VL; the other is an incorrect tied-weights warning in Qwen3-Next." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Video-input regression in Qwen VL models vs serve KV-cache indexing crash; unrelated paths." + "reason": "Both mention multimodal/serve areas, but one is a video-input regression in Qwen VL models and the other is a serve KV-cache indexing crash." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "Non-persistent buffer junk regression vs GPTNeoX rotary_pct reload defaulting; different config/runtime problems." + "reason": "Both are serialization-related, but one is non-persistent buffers being junk-filled and the other is GPTNeoX rotary_pct not surviving reload." }, { - "left": "issue:44190", - "right": "issue:44964", + "left": "issue:44792", + "right": "issue:45600", "accept": false, - "reason": "Example script dataset loading failure vs Phi-4 multimodal model loading error; not duplicates." + "reason": "Unrelated: janus image-generation test failure vs stale auto_mappings references to removed Sam3LiteText configs." }, { - "left": "issue:44792", - "right": "issue:45600", + "left": "issue:44190", + "right": "issue:44964", "accept": false, - "reason": "Janus test failure vs removed Sam3LiteText config references in auto_mappings; unrelated CI/test issues." + "reason": "Different scope: local dataset loading in an example script vs a model-loading regression for microsoft/Phi-4-multimodal-instruct." }, { "left": "issue:44464", "right": "issue:45563", "accept": false, - "reason": "Chunked generation inconsistency with compiled forward vs stale num_return_sequences warning; different generation bugs." + "reason": "Different problems: compiled-forward chunked generation inconsistency vs a stale warning about num_return_sequences." }, { "left": "issue:43827", "right": "issue:43994", "accept": false, - "reason": "Docs still referencing pipeline() vs SigLIP2 nonsensical runtime results; documentation and model behavior are separate." + "reason": "One is docs still referencing removed pipeline() APIs; the other is a SigLIP2 inference correctness issue." }, { "left": "issue:44855", "right": "issue:45588", "accept": false, - "reason": "Python 3.13 import parsing error vs flash_attention s_aux=None crash; unrelated breakages." + "reason": "Different failures: Python 3.13 IndentationError in DebertaV2Model import vs flash_attention crash on s_aux=None." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "Both are multimodal-preprocessing bugs, but assistant_masks being zero and vision_position_ids being wrong are different code paths and fixes." + "reason": "Both are multimodal-template regressions, but they affect different outputs: assistant_masks vs vision_position_ids." }, { "left": "issue:35141", "right": "issue:42490", "accept": false, - "reason": "Embedding resize reinitialization bug vs trailing-slash local-path warning; no shared underlying issue." + "reason": "Completely different issues: token-embedding resize reinit bug vs import warning when local model path ends with a slash." }, { "left": "issue:41720", "right": "issue:45245", "accept": false, - "reason": "Qwen3 auto device mapping cuda assert vs category-count limit error; distinct runtime failures." + "reason": "No shared underlying bug: Qwen3 auto device mapping cuda assert vs category-count overflow at 2^24." }, { "left": "issue:45803", "right": "issue:45987", "accept": false, - "reason": "Bare except swallowing errors vs StaticCache length contract mismatch; unrelated utilities." + "reason": "Different concerns: bare except swallowing errors in FuyuBatchFeature vs StaticCache.get_seq_length returning the wrong type." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder vs OmDet-Turbo processor/model size mismatch; different model-specific problems." + "reason": "Different model components: dummy Conv2d in Sam3PixelDecoder vs OmDet-Turbo processor/model input-size mismatch." }, { "left": "issue:44933", "right": "issue:45362", "accept": false, - "reason": "Missing import from image_utils vs Qwen3.5 chat crash; unrelated defects." + "reason": "Unrelated: nonexistent image_utils import vs Qwen3.5-35B crash in transformers chat." }, { "left": "issue:43335", "right": "issue:44855", "accept": false, - "reason": "SwitchTransformers sparse-layer creation bug vs Python 3.13 import parsing error; different areas." + "reason": "Different failures: SwitchTransformersConfig creates an unwanted sparse layer vs a Python 3.13 parser/import error." }, { "left": "issue:43644", "right": "issue:44913", "accept": false, - "reason": "Both involve config/regression symptoms, but non-persistent buffer junk and rotary_pct reload behavior are not the same bug." + "reason": "Related only at a high level (config/serialization), but not the same bug: non-persistent buffer junk vs rotary_pct resetting on reload." }, { "left": "issue:44077", "right": "issue:44877", "accept": false, - "reason": "Optional post_init allowance vs strict config blocking granite_speech; related theme, but different concrete change and failure." + "reason": "Both involve config handling, but one is optional post_init on patchtsmixer and the other is strict config blocking granite_speech loading." }, { "left": "issue:43976", "right": "issue:44855", "accept": false, - "reason": "Python-version compatibility issues, but one is version support failure and the other is a parser/indentation import error; not a duplicate." + "reason": "Both mention Python compatibility, but they are different failures: broad version incompatibility vs a specific IndentationError in one model file." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "Bamba torch.compile/SDPA bug vs local dataset loading in a training script; unrelated." + "reason": "Different areas entirely: Bamba torch.compile/SDPA failure vs local dataset loading in an image-classification training script." }, { "left": "issue:43540", "right": "issue:44734", "accept": false, - "reason": "Qwen3OmniMoe video-input ValueError vs serve KV-cache indexing crash; different subsystems." + "reason": "Different multimodal bugs: video input ValueError in Qwen3OmniMoe vs wrong tensor indexing during serve KV-cache continuation." }, { "left": "issue:45127", "right": "issue:45357", "accept": false, - "reason": "LoRA merge with extended vocab collapsing model vs incorrect visual encoder keys saved in Qwen3.5; separate save/merge issues." + "reason": "Different save/load regressions: LoRA merge with extended vocabulary on tied embeddings vs incorrect visual-encoder keys in save_pretrained." }, { "left": "issue:44737", "right": "issue:45803", "accept": false, - "reason": "XLNet CPU arange bug vs bare except swallowing errors; not the same code path or symptom." + "reason": "Different code paths: XLNet CPU-side relative position computation vs a bare except in FuyuBatchFeature.convert_to_tensors()." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "Unrelated CI issues: stale GraniteSpeech device override test vs auto_mappings references to removed Sam3LiteText configs." } ] }, @@ -10055,8 +10064,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -10110,23 +10119,22 @@ "issue:46121" ], "soft_pairs": [ - "issue:44112|issue:45600", "issue:44462|issue:45907", "issue:43335|issue:44190", "issue:46082|issue:46121", "issue:43116|issue:44190", - "issue:44938|issue:45542", "issue:45276|issue:45584", + "issue:44938|issue:45542", "issue:43295|issue:45276", "issue:43479|issue:45375", "issue:36246|issue:41628", "issue:45245|issue:45584", "issue:43650|issue:44220", - "issue:36246|issue:38617", "issue:42175|issue:44279", "issue:44186|issue:44734", - "issue:41950|issue:44016", "issue:44246|issue:45003", + "issue:41950|issue:44016", + "issue:36246|issue:38617", "issue:44683|issue:45446", "issue:43404|issue:45216", "issue:38175|issue:43065", @@ -10139,223 +10147,217 @@ "issue:44561|issue:45632", "issue:43976|issue:44987", "issue:43531|issue:45600", - "issue:42907|issue:43441", - "issue:43295|issue:45335" + "issue:43295|issue:45335", + "issue:42907|issue:43441" ], - "prepared_review_unit_hash": "20689046d7b1c35d566095ae2528820d5218de2f717cc8a13794be4c26d4ecf5", + "prepared_review_unit_hash": "9a359d109b37b18ad304ec8a362180d7fb9b91d628c8471acd629a52b47974e6", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 36948, - "estimated_input_tokens": 9237, - "estimated_eval_tokens": 18730 + "soft_pair_count": 30, + "serialized_chars": 36651, + "estimated_input_tokens": 9163, + "estimated_eval_tokens": 18582 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is very heterogeneous: it mixes import/export regressions, model-specific configuration bugs, backend/dependency problems, and notebook/script failures. None of the soft edges look like true duplicates of the same underlying bug or change.", - "confidence": 0.91, - "canonical_issue_reason": "issue:41628 is the broadest, clearest representative of the small import-regression subset in this set, but the overall cluster is too diverse to have a strong true canonical issue.", + "summary": "This set is not a tight duplicate cluster; it mixes unrelated import regressions, model/config bugs, pipeline issues, and environment-specific failures. I would not merge any of the soft-edge pairs.", + "confidence": 0.79, + "canonical_issue_reason": "If one issue must serve as canonical, issue 41628 is the broadest user-facing import regression in the set and best represents the general class of API/export breakages.", "canonical_pr_reason": null, - "best_issue_reason": "issue:41628 is the best global anchor by title clarity and breadth of symptom, though it still only represents one slice of the cluster rather than the whole.", + "best_issue_reason": "Issue 41628 is the strongest standalone report: clear failure mode, broad impact, and closest to the import/API regression theme seen in a few neighboring issues.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44112", - "right": "issue:45600", - "accept": false, - "reason": "Both are CI-related, but they involve different models, different files, and different failure modes." - }, { "left": "issue:44462", "right": "issue:45907", "accept": false, - "reason": "One is AutoTokenizer ignoring tokenizer.json; the other is list_repo_files not forwarding revision. Related area, different bug." + "reason": "Different bugs: AutoTokenizer ignoring tokenizer.json is a loading precedence issue, while list_repo_files forwarding revision is an Hub API parameter bug." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "SwitchTransformers config initialization is unrelated to a local-dataset loading failure in an example script." + "reason": "Unrelated subsystems: SwitchTransformers config construction vs a local dataset loading failure in an example script." }, { "left": "issue:46082", "right": "issue:46121", "accept": false, - "reason": "LlamaConfig head_dim validation and rope-param JSON list handling are separate config-validation bugs." + "reason": "Different validation paths: LlamaConfig head_dim handling is not the same as rope-parameter JSON list coercion." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Multi-label classification output logic is unrelated to dataset loading in a different example script." + "reason": "Different example-script failures: empty multi-label outputs vs inability to load a local dataset." }, { - "left": "issue:44938", - "right": "issue:45542", + "left": "issue:45276", + "right": "issue:45584", "accept": false, - "reason": "Python 3.14 import/load failure and missing TensorFlow backend wiring are different root causes." + "reason": "Unrelated model behaviors: Gemma4 embedding resizing vs Whisper empty-transcription generation after token alignment." }, { - "left": "issue:45276", - "right": "issue:45584", + "left": "issue:44938", + "right": "issue:45542", "accept": false, - "reason": "Gemma4 embedding-resize behavior is unrelated to Whisper empty-transcription generation." + "reason": "Separate environment issues: Python 3.14 import/load failure vs TensorFlow backend installation confusion with tensorboard." }, { "left": "issue:43295", "right": "issue:45276", "accept": false, - "reason": "Processor.tokenizer regression is a different code path from Gemma4 token-embedding resizing." + "reason": "Different regression surfaces and code paths: processor/tokenizer image handling vs Gemma4 resize_token_embeddings behavior." }, { "left": "issue:43479", "right": "issue:45375", "accept": false, - "reason": "Phi4Multimodal default sub-config initialization and missing deepstack_visual_indexes are different config bugs." + "reason": "Different multimodal config bugs affecting different models and fields; not the same underlying defect." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but one is a model-module symbol and the other is a top-level AutoImageProcessor export." + "reason": "Both are import errors, but for different symbols and namespaces; they do not point to the same missing export." }, { "left": "issue:45245", "right": "issue:45584", "accept": false, - "reason": "Category cardinality limits and Whisper special-token alignment are unrelated." + "reason": "Completely unrelated: category count runtime limit vs Whisper transcription edge case." }, { "left": "issue:43650", "right": "issue:44220", "accept": false, - "reason": "Generic content and fbank feature extraction are not the same bug." - }, - { - "left": "issue:36246", - "right": "issue:38617", - "accept": false, - "reason": "Different missing symbols from different modules; same symptom class, not the same defect." + "reason": "No shared bug: an issue about missing data vs a feature-extraction helper crash." }, { "left": "issue:42175", "right": "issue:44279", "accept": false, - "reason": "Pip extra/backend dependency packaging and a vague dependency issue are not concrete duplicates." + "reason": "Broad dependency complaints, but no shared concrete code-path or fix." }, { "left": "issue:44186", "right": "issue:44734", "accept": false, - "reason": "Tokenizer padding/truncation crash is unrelated to serve-time KV-cache continuation indexing." + "reason": "Tokenizer/NER padding crash is unrelated to serve-time KV-cache continuation indexing." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import performance and unsafe sys.modules access are different problems, even if both touch startup/import behavior." }, { "left": "issue:41950", "right": "issue:44016", "accept": false, - "reason": "Pipeline processor lookup and a notebook syntax error are completely different issues." + "reason": "Different topics entirely: pipeline processor lookup vs a notebook syntax error." }, { - "left": "issue:44246", - "right": "issue:45003", + "left": "issue:36246", + "right": "issue:38617", "accept": false, - "reason": "Import slowness and unsafe sys.modules access may be adjacent, but the reports describe different failures." + "reason": "Both are import errors, but they concern different missing names and different modules." }, { "left": "issue:44683", "right": "issue:45446", "accept": false, - "reason": "Compiled flex_attention failure on new torch and a version-check bug for AuxRequest import are not the same concrete problem." + "reason": "Same subsystem name, but different failures: compiled flex_attention on new torch versions vs a wrong version gate for AuxRequest import." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Mistral3 lm_head tying and Qwen3.5 save_pretrained regression affect different model behaviors." + "reason": "Different model-specific regressions: untied lm_head weights vs incorrect checkpoint saving." }, { "left": "issue:38175", "right": "issue:43065", "accept": false, - "reason": "Zero probabilities in SigLIP2 is unrelated to a dummy Conv2d in Sam3PixelDecoder." + "reason": "Unrelated model bugs in different architectures and code paths." }, { "left": "issue:44016", "right": "issue:44220", "accept": false, - "reason": "Notebook syntax error and audio feature extraction are unrelated." + "reason": "Notebook syntax error and audio feature extraction failure are not the same issue." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "Missing fields from trust_remote_code config loading and unrecognized-model config.json validation are different config-loader failures." + "reason": "Different config-loading failures: missing fields with trust_remote_code vs an unrecognized-model/config.json problem." }, { "left": "issue:44964", "right": "issue:45600", "accept": false, - "reason": "Phi-4 multimodal load failure and stale auto_mappings references to removed configs are different issues." + "reason": "Different breakages: loading a specific multimodal model vs stale autogenerated auto-mappings for removed configs." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "GraniteSpeech CI device override and Phi-4 multimodal model loading are unrelated." + "reason": "CI-only GraniteSpeech test failure is unrelated to Phi-4 multimodal loading." }, { "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "Both involve model loading, but they are different models with different failure contexts." + "reason": "Different model families and failure modes; these are not the same bug." }, { "left": "issue:44734", "right": "issue:45632", "accept": false, - "reason": "Serve KV-cache indexing and trust_remote_code cache-path collisions are unrelated." + "reason": "Serve-time tensor indexing bug is unrelated to local trust_remote_code cache-path collisions." }, { "left": "issue:44561", "right": "issue:45632", "accept": false, - "reason": "Removal of is_torch_fx_available breaks remote-code models; cache-path collisions are a different trust_remote_code bug." + "reason": "One is a removed helper breaking trust_remote_code models; the other is a cache-path naming collision." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Python version support and loading a specific model with transformers>=5.1.0 are different problems." + "reason": "A Python-version compatibility report is not the same as a specific model-load failure on a version range." }, { "left": "issue:43531", "right": "issue:45600", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior has nothing to do with stale auto_mappings references." + "reason": "Qwen3-MoE sliding-window behavior and auto_mappings CI breakage are unrelated." }, { - "left": "issue:42907", - "right": "issue:43441", + "left": "issue:43295", + "right": "issue:45335", "accept": false, - "reason": "Saving dequantized models and FlashAttention failure in Ministral-3 are distinct code paths." + "reason": "Both mention embedding/tokenizer mutation, but they affect different models and different internal components." }, { - "left": "issue:43295", - "right": "issue:45335", + "left": "issue:42907", + "right": "issue:43441", "accept": false, - "reason": "Processor.tokenizer regression and t5gemma embed_tokens resizing are separate model/API bugs." + "reason": "Different Ministral issues: saving dequantized models vs FlashAttention compatibility in v5 RC." } ] }, @@ -10374,8 +10376,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -10436,8 +10438,8 @@ "issue:43673|issue:44336", "issue:43335|issue:45593", "issue:44016|issue:44062", - "issue:43824|issue:44933", "issue:44075|issue:44368", + "issue:43824|issue:44933", "issue:43299|issue:45216", "issue:43381|issue:45593", "issue:43901|issue:44496", @@ -10451,198 +10453,198 @@ "issue:43299|issue:45357", "issue:38175|issue:44617", "issue:43577|issue:45216", - "issue:44351|issue:45278", "issue:41950|issue:45020", + "issue:44351|issue:45278", "issue:44279|issue:45405", "issue:44779|issue:45812", "issue:41950|issue:45823", "issue:39692|issue:43901", "issue:43295|issue:44779" ], - "prepared_review_unit_hash": "8e4efc9ba432e1338c1b9d1d3f64a23f07997d88b2b6cf470c37fdd10bda8290", + "prepared_review_unit_hash": "177608cc55e117776954ebf8a20a1e2b4ca8bfc27fec96aa554cf8476e38ed91", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36383, - "estimated_input_tokens": 9096, - "estimated_eval_tokens": 18448 + "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": "These items are mostly unrelated single-issue reports spanning different model families, pipelines, and regressions; I found no strong duplicate/merge candidates among the soft pairs.", - "confidence": 0.93, - "canonical_issue_reason": "issue:45278 is the broadest representative regression report in the set, covering multiple import failures after an upgrade and plausibly subsuming narrower import-surface breakages.", + "summary": "These items are mostly unrelated issue reports clustered by superficial text similarity rather than true duplicates. The only near-umbrella report is the broad import-regression issue, but even that does not safely subsume the others.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45278 is the broadest representative in the set: it describes a general import-breakage regression after upgrading transformers, which is the closest thing here to an umbrella issue.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the best global representative because it has the clearest high-impact upgrade-regression framing and broad user reach.", + "best_issue_reason": "issue:45278 is the best global representative because it is the most general user-facing regression report and could plausibly absorb some specific import-failure manifestations.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42490", "right": "issue:45632", "accept": false, - "reason": "Trailing-slash local-path import warning and trust_remote_code cache-path collision are different path-handling bugs." + "reason": "Trailing-slash local-path warning is a different bug from trust_remote_code cache collisions on shared leaf names." }, { "left": "issue:43526", "right": "issue:43901", "accept": false, - "reason": "A processor output bug and a documentation mismatch about return_all_scores are not the same issue." + "reason": "BeitImageProcessorFast reduce_labels corruption is unrelated to a TextClassificationPipeline docs mismatch." }, { "left": "issue:43441", "right": "issue:45588", "accept": false, - "reason": "Both mention flash attention, but one is model-specific Ministral-3 loading and the other is a generic sink-less s_aux=None crash." + "reason": "Both touch flash_attention, but one is a Ministral-3 RC failure and the other is an s_aux=None crash in sink-less models." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Unrecognized model config loading and PIL image processor torchvision dependency are unrelated." + "reason": "Unrecognized model/config loading is unrelated to PIL backend processors incorrectly requiring torchvision." }, { "left": "issue:43673", "right": "issue:44336", "accept": false, - "reason": "Generation cache lookup during chunked_prefill and ANSI color output in loading_report are unrelated." + "reason": "GenerationMixin cache missing in chunked_prefill has nothing to do with ANSI codes in loading_report." }, { "left": "issue:43335", "right": "issue:45593", "accept": false, - "reason": "Sparse-layer construction with num_sparse_encoder_layers=0 is different from D-FINE auxiliary-loss handling." + "reason": "Sparse-layer creation in SwitchTransformers and missing auxiliary losses in D-FINE are different model bugs." }, { "left": "issue:44016", "right": "issue:44062", "accept": false, - "reason": "A notebook syntax error and an AddedToken keyword conflict are separate bugs." + "reason": "Notebook syntax error and AddedToken special-argument TypeError are unrelated failures." }, { - "left": "issue:43824", - "right": "issue:44933", + "left": "issue:44075", + "right": "issue:44368", "accept": false, - "reason": "These are different import failures in different modules/symbols, not the same concrete bug." + "reason": "SGD optimizer args not applied is unrelated to the Qwen3.5 tie_word_embeddings warning." }, { - "left": "issue:44075", - "right": "issue:44368", + "left": "issue:43824", + "right": "issue:44933", "accept": false, - "reason": "Unused SGD args and a tie_word_embeddings warning in ms-swift are unrelated." + "reason": "Missing Qwen2_5_VLForConditionalGeneration export is not the same as a nonexistent image_utils import." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "One is a Qwen3VL model loading failure; the other is an incorrect save_pretrained checkpoint regression." + "reason": "Qwen3VL loading regression and Qwen3.5 save_pretrained regression affect different code paths." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient-checkpointing eval-mode behavior and D-FINE auxiliary-loss omission are different code paths." + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "A pipeline docs mismatch and a model-import/config error are not the same issue." + "reason": "A docs/API wording issue is not the same as a model-type loading failure." }, { "left": "issue:43976", "right": "issue:44279", "accept": false, - "reason": "Python-version compatibility breakage and a vague dependency issue do not point to one concrete bug." + "reason": "Python version incompatibility and a generic dependency issue are too broad and not the same bug." }, { "left": "issue:43650", "right": "issue:46209", "accept": false, - "reason": "Placeholder/non-technical issue text and an FP8 B200 inference failure are unrelated." + "reason": "Placeholder text and FP8 inference failure on B200 are unrelated." }, { "left": "issue:44496", "right": "issue:45759", "accept": false, - "reason": "These are different Qwen-related loading problems with different failure modes and code paths." + "reason": "Unrecognized model config loading is different from failing to unwrap text_config in composite Qwen multimodal configs." }, { "left": "issue:45335", "right": "issue:45584", "accept": false, - "reason": "Embedding-resize propagation and Whisper empty-transcription failure after align_special_tokens are unrelated." + "reason": "Decoder embed resize bug and Whisper empty-transcription failure are different generation/model issues." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Both are config issues, but they affect different model classes and different defaults/attributes." + "reason": "Default multimodal config initialization and missing pad_token_id on Llama4Config are unrelated attribute/config bugs." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "Missing load-balancing loss and mm_token_type_ids defaulting are separate training semantics bugs." + "reason": "Router load-balancing loss gating and mm_token_type_ids defaulting are different training-path issues." }, { "left": "issue:44521", "right": "issue:45325", "accept": false, - "reason": "All-zero assistant masks and rope-index temporal scaling are different multimodal preprocessing bugs." + "reason": "All-zero assistant masks and still-image rope-index scaling are separate multimodal processing bugs." }, { "left": "issue:43299", "right": "issue:45357", "accept": false, - "reason": "These are distinct save/load regressions for Qwen3.5 rather than one shared bug." + "reason": "Model loading failure and incorrect visual-encoder keys on save_pretrained are not the same concrete regression." }, { "left": "issue:38175", "right": "issue:44617", "accept": false, - "reason": "SigLIP2 zero probabilities and Sam3Video OOM are unrelated." + "reason": "Zero probabilities in SigLIP2 is unrelated to Sam3Video CUDA OOM." }, { "left": "issue:43577", "right": "issue:45216", "accept": false, - "reason": "BLIP2 dtype propagation and Qwen3.5 checkpoint saving are unrelated issues." + "reason": "Blip2 dtype handling and Qwen3.5 checkpoint saving are different code paths and symptoms." }, { - "left": "issue:44351", - "right": "issue:45278", + "left": "issue:41950", + "right": "issue:45020", "accept": false, - "reason": "HybridCache import failure may be a symptom in the broader import-regression report, but it is not clearly the same concrete bug." + "reason": "Pipeline processor lookup bug is unrelated to remote_code model breakage." }, { - "left": "issue:41950", - "right": "issue:45020", + "left": "issue:44351", + "right": "issue:45278", "accept": false, - "reason": "Video-classification processor lookup and remote_code breakage are different failures." + "reason": "Missing HybridCache import could be one symptom in a broader import regression, but the reports are not the same concrete bug." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "A generic dependency issue and an unreleased MIN_PEFT_VERSION bump are not the same bug." + "reason": "A generic dependency issue is not the same as bumping MIN_PEFT_VERSION to an unreleased release." }, { "left": "issue:44779", "right": "issue:45812", "accept": false, - "reason": "Deepseek tokenizer regression and Granite tokenizer regression affect different model families and behaviors." + "reason": "Deepseek tokenizer regression and Granite tokenizer regression affect different model families and likely different fixes." }, { "left": "issue:41950", @@ -10654,13 +10656,19 @@ "left": "issue:39692", "right": "issue:43901", "accept": false, - "reason": "These are both docs-related, but one is a SigLIP2 example error report and the other is a pipeline docs mismatch." + "reason": "SigLIP2 doc-example errors are not the same as a TextClassificationPipeline docs mismatch." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Custom processor/tokenizer regression and Deepseek tokenization regression are different failures." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Custom processor.tokenizer/images regression and Deepseek tokenizer incorrect results are different issues." + "reason": "Custom processor/tokenizer regression and Deepseek tokenization regression are different failures." } ] }, @@ -10679,8 +10687,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -10741,8 +10749,8 @@ "issue:44462|issue:45081", "issue:43450|issue:44279", "issue:44297|issue:44462", - "issue:44016|issue:44291", "issue:45081|issue:45907", + "issue:44016|issue:44291", "issue:44466|issue:45276", "issue:43441|issue:45834", "issue:45600|issue:45712", @@ -10753,9 +10761,9 @@ "issue:44623|issue:44749", "issue:43065|issue:44898", "issue:44261|issue:44263", + "issue:44016|issue:45823", "issue:44361|issue:45593", "issue:43441|issue:44683", - "issue:44016|issue:45823", "issue:44671|issue:44779", "issue:43723|issue:44162", "issue:42175|issue:43976", @@ -10763,202 +10771,202 @@ "issue:43901|issue:43994", "issue:43673|issue:43761" ], - "prepared_review_unit_hash": "50718ffb6bae847f8c2d1a390a43c4feae0a17f69f3c3d106cf9d0c606192ef1", + "prepared_review_unit_hash": "d52ebadb998f684ef4e5b14f0c39484b86734b44b4331f93b14817b0daa69c4b", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36010, - "estimated_input_tokens": 9003, - "estimated_eval_tokens": 18262 + "serialized_chars": 36015, + "estimated_input_tokens": 9004, + "estimated_eval_tokens": 18264 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set looks heterogeneous rather than one true duplicate cluster; most soft pairs are only broad topical matches (often v5/tokenizer/load/save regressions) and should be rejected. If a single issue must anchor the cluster, the tokenizer-loading regression around AutoTokenizer is the closest fit.", - "confidence": 0.43, - "canonical_issue_reason": "issue:44462 is the best provisional anchor because it is a concrete v5 tokenizer-loading regression and is the closest match to the other tokenizer-related near-duplicates in the set.", + "summary": "No soft pair looks like the same underlying bug/change; the set is a grab bag of unrelated Transformers regressions, packaging issues, and model-specific failures.", + "confidence": 0.93, + "canonical_issue_reason": "issue:44521 has the strongest signal in this mixed set: it is open, has substantial discussion/inbound references, and describes a concrete regression in a core API (`apply_chat_template`) affecting multimodal workflows.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44462 is the most representative of the cluster\u2019s strongest theme (tokenizer loading/serialization regressions in v5), even though the overall cluster is not truly cohesive.", + "best_issue_reason": "issue:44521 is the most actionable and representative issue here because it is well-scoped, actively discussed, and touches a central user-facing API.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "Compiled flex_attention failure and TensorFlow backend installation are unrelated bugs." + "reason": "Unrelated failures: compiled flex_attention on torch 2.9 vs TensorFlow backend missing from a torch install." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Both involve model loading, but they are different model-specific failures with different underlying code paths." + "reason": "Both are model-loading regressions, but they affect different models and appear to be distinct code-path failures." }, { "left": "issue:43976", "right": "issue:45405", "accept": false, - "reason": "Python version compatibility and an unreleased PEFT minimum version are separate release/dependency issues." + "reason": "Different dependency/version problems: Python support breakage vs an unreleased PEFT minimum version bump." }, { "left": "issue:43650", "right": "issue:46132", "accept": false, - "reason": "A placeholder issue and an AttentionInterface behavior change do not describe the same change or bug." + "reason": "Completely unrelated: a generic data issue vs an AttentionInterface.register behavior change." }, { "left": "issue:44462", "right": "issue:45081", "accept": false, - "reason": "Both are tokenizer-loading regressions, but one ignores tokenizer.json while the other crashes in Mistral regex patching." + "reason": "Both involve tokenizers, but one is AutoTokenizer ignoring repo tokenizer.json while the other is a Mistral regex patch AttributeError." }, { "left": "issue:43450", "right": "issue:44279", "accept": false, - "reason": "Video processor shape handling and a generic dependency issue are unrelated." + "reason": "No substantive overlap; batched video processor shape bug vs generic transformers dependency issue." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Tokenizer class serialization mismatch is different from AutoTokenizer ignoring tokenizer.json on load." + "reason": "One is save_pretrained metadata mismatch; the other is tokenizer loading ignoring tokenizer.json. Related area, different bug." }, { - "left": "issue:44016", - "right": "issue:44291", + "left": "issue:45081", + "right": "issue:45907", "accept": false, - "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated failure modes." + "reason": "Different subsystems entirely: Mistral tokenizer regex patch vs repo file-listing revision forwarding." }, { - "left": "issue:45081", - "right": "issue:45907", + "left": "issue:44016", + "right": "issue:44291", "accept": false, - "reason": "Tokenizer regex patch crash and list_repo_files revision forwarding are different code paths." + "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "lm_head serialization inconsistency and Gemma4 embedding resize behavior are distinct model bugs." + "reason": "Both touch Gemma4 weights, but one is serialization inconsistency and the other is resize_token_embeddings propagation; not the same bug." }, { "left": "issue:43441", "right": "issue:45834", "accept": false, - "reason": "FlashAttention failure on Ministral and OCR index error in Kosmos2.5 are unrelated model issues." + "reason": "Different model-specific failures with no shared concrete code path." }, { "left": "issue:45600", "right": "issue:45712", "accept": false, - "reason": "Removed auto-mappings entries and leftover dummy classes are separate cleanup/CI problems." + "reason": "Both are CI/cleanup related, but they stem from different stale-reference files and distinct failures." }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Quantized model saving and dequantized model loading with offload are different quantization workflows." + "reason": "Quantization is the broad theme, but saving a quantized model and loading a dequantized model with offload are different problems." }, { "left": "issue:43653", "right": "issue:45593", "accept": false, - "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss handling are unrelated." + "reason": "Tokenizer special-token registration bug vs D-FINE auxiliary-loss logic; unrelated." }, { "left": "issue:33290", "right": "issue:45656", "accept": false, - "reason": "OOM from Adafactor in DeepSpeed and optimizer.step being called twice are different DeepSpeed bugs." + "reason": "Both mention DeepSpeed/optimizer, but one is OOM with Adafactor and the other is double optimizer steps; not the same underlying issue." }, { "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "Missing processor save files and Gemma-4 ZeRO3 from_pretrained are separate save/load issues." + "reason": "Processor save_pretrained missing files vs Gemma-4 zero3 from_pretrained load failure are different save/load bugs." }, { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "Processor saving missing files and a data-filtering slowdown are unrelated." + "reason": "Unrelated: processor serialization bug vs a performance regression in Chinese text filtering." }, { "left": "issue:43065", "right": "issue:44898", "accept": false, - "reason": "A dummy Conv2d in Sam3PixelDecoder and Perceiver non-default-resolution failure are different model bugs." + "reason": "Both are vision-model issues, but dummy Conv2d in Sam3PixelDecoder and Perceiver interpolation failure are different defects." }, { "left": "issue:44261", "right": "issue:44263", "accept": false, - "reason": "Missing rms_norm_eps precision issue and torch.split return semantics are unrelated." + "reason": "Different problems: missing layernorm config precision error vs torch.split return handling in a separate model indexer." }, { - "left": "issue:44361", - "right": "issue:45593", + "left": "issue:44016", + "right": "issue:45823", "accept": false, - "reason": "MLukeTokenizer AttributeError and D-FINE auxiliary-loss behavior are unrelated." + "reason": "Notebook syntax issue and Gemma4 device mismatch are unrelated." }, { - "left": "issue:43441", - "right": "issue:44683", + "left": "issue:44361", + "right": "issue:45593", "accept": false, - "reason": "FlashAttention breakage for a specific model and compiled flex_attention failure are different attention-stack issues." + "reason": "No overlap: MLukeTokenizer AttributeError vs D-FINE auxiliary-loss behavior." }, { - "left": "issue:44016", - "right": "issue:45823", + "left": "issue:43441", + "right": "issue:44683", "accept": false, - "reason": "Notebook syntax error and Gemma4 PLE device mismatch are unrelated." + "reason": "Both are attention-related, but FlashAttention model failure and compiled flex_attention torch-compat failure are distinct code paths." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "CamemBERT masked-LM regression and Deepseek tokenizer regression are different model/tokenizer bugs." + "reason": "Different regression surfaces: masked LM predictions for CamemBERT vs tokenizer output correctness for DeepSeek." }, { "left": "issue:43723", "right": "issue:44162", "accept": false, - "reason": "Tokenizer loading regression and ESM2 model breakage are not the same underlying defect." + "reason": "Tokenizer loading regression vs a broader ESM2 breakage report; not the same bug." }, { "left": "issue:42175", "right": "issue:43976", "accept": false, - "reason": "Missing TensorFlow backend after pip install and Python 3.9/3.10 compatibility are unrelated." + "reason": "Packaging/backend omission and Python-version incompatibility are unrelated." }, { "left": "issue:43262", "right": "issue:44521", "accept": false, - "reason": "Default audio sampling rate and multimodal assistant mask generation are different apply_chat_template bugs." + "reason": "Both involve apply_chat_template, but one is audio sampling-rate defaulting and the other is multimodal assistant masks; different bugs." }, { "left": "issue:43901", "right": "issue:43994", "accept": false, - "reason": "Documentation mismatch for return_all_scores and incorrect SigLIP2 model outputs are unrelated." + "reason": "Docs mismatch for return_all_scores vs an inference-quality bug on a specific model/pipeline." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Generation cache missing and CLIPVision hidden_states regression are different v5 regressions." + "reason": "Generation cache missing in chunked_prefill and CLIPVisionModel hidden_states omission are unrelated." } ] }, @@ -10977,8 +10985,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -11033,8 +11041,8 @@ ], "soft_pairs": [ "issue:43784|issue:45542", - "issue:43824|issue:43976", "issue:43450|issue:44617", + "issue:43824|issue:43976", "issue:44297|issue:45397", "issue:43525|issue:43881", "issue:43901|issue:44246", @@ -11042,8 +11050,8 @@ "issue:44188|issue:44336", "issue:41628|issue:43976", "issue:44589|issue:45362", - "issue:44336|issue:44373", "issue:42491|issue:43299", + "issue:44336|issue:44373", "issue:45561|issue:45563", "issue:44295|issue:45397", "issue:44246|issue:44496", @@ -11052,232 +11060,232 @@ "issue:41950|issue:44220", "issue:43526|issue:45003", "issue:42886|issue:46132", - "issue:36246|issue:42757", - "issue:36246|issue:44351", "issue:44246|issue:44871", "issue:44336|issue:44485", + "issue:36246|issue:42757", + "issue:36246|issue:44351", "issue:45003|issue:45561", - "issue:43673|issue:44561", "issue:44779|issue:45292", + "issue:43673|issue:44561", "issue:36246|issue:43596", "issue:44295|issue:44749", "issue:44242|issue:44857", "issue:43352|issue:45250" ], - "prepared_review_unit_hash": "185909242fa3092bcb1c4c9bf9da7f58c7b714394f2e28c277657cee525d59f3", + "prepared_review_unit_hash": "a611ace7a416df4677abf090768d0ad7d07550cf2f88da354bd3b6054cb7eb31", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36685, - "estimated_input_tokens": 9172, - "estimated_eval_tokens": 18600 + "serialized_chars": 36689, + "estimated_input_tokens": 9173, + "estimated_eval_tokens": 18602 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of distinct Transformers regressions (import/export failures, model-loading issues, pipeline/config bugs). The only pair that looks like the same underlying bug is the Qwen2.5-VL import/export failure family.", - "confidence": 0.78, - "canonical_issue_reason": "issue:43824 is the clearest and most central representative of the Qwen2.5-VL import/export regression, with higher discussion and inbound references than the older sibling report.", + "summary": "The cluster is mostly heterogeneous: many are model-specific import/loading regressions, but the soft pairs do not look like true duplicates of the same concrete bug/change.", + "confidence": 0.31, + "canonical_issue_reason": "Issue 41628 is the broadest user-facing import-regression report in the set and is the closest thing to a central API breakage; the other issues are mostly model-specific failures, config mismatches, docs, or unrelated runtime behavior.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43824 is the best standalone issue to anchor this cluster because it is specific, well-scoped, and most clearly describes a real duplicate-prone import failure.", + "best_issue_reason": "41628 is the best representative issue for this noisy cluster because it captures a generic transformers import failure, but it is still not a strong duplicate anchor for the rest of the items.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43784", "right": "issue:45542", "accept": false, - "reason": "Both are import-time failures, but they involve different missing backends/variables and different packages; not the same bug." + "reason": "Both are import/backend-related, but one is a sentence-transformers import NameError and the other is a TensorBoard-without-TensorFlow backend issue; different code paths." }, { - "left": "issue:43824", - "right": "issue:43976", + "left": "issue:43450", + "right": "issue:44617", "accept": false, - "reason": "Qwen2.5-VL import failure vs Python version compatibility; unrelated problems." + "reason": "Both involve video, but one is a batched video-processor shape bug and the other is a CUDA OOM in Sam3Video; not the same defect." }, { - "left": "issue:43450", - "right": "issue:44617", + "left": "issue:43824", + "right": "issue:43976", "accept": false, - "reason": "Video shape bug vs CUDA OOM; different failure modes and code paths." + "reason": "Qwen2.5-VL import failure vs Python version compatibility breakage; unrelated underlying causes." }, { "left": "issue:44297", "right": "issue:45397", "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch vs ZeRO3 from_pretrained loading failure; not the same issue." + "reason": "Tokenizer save metadata mismatch is a serialization bug; gemma-4 zero3 from_pretrained is a model-loading/config issue." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Missing pad_token_id on one model config vs a different model loading failure; too model-specific to be duplicates." + "reason": "Missing pad_token_id on Llama4Config and GLM-4V loading failure are different model/config problems." }, { "left": "issue:43901", "right": "issue:44246", "accept": false, - "reason": "Docs mismatch for return_all_scores vs import slowdown; unrelated." + "reason": "Docs mismatch for return_all_scores vs intermittent slow import are unrelated." }, { "left": "issue:43525", "right": "issue:43643", "accept": false, - "reason": "Both involve config fields, but one is a specific missing attribute and the other is a trust_remote_code config hydration problem; not clearly the same bug." + "reason": "Both mention missing config fields, but one is a specific Llama4 attribute error and the other is a trust_remote_code AutoConfig fields regression; not the same bug." }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "Attention-kernel divergence vs ANSI codes in a loading report; unrelated subsystems." + "reason": "Attention-kernel divergence under torch.compile is unrelated to ANSI codes emitted by loading_report." }, { "left": "issue:41628", "right": "issue:43976", "accept": false, - "reason": "AutoImageProcessor import error vs Python 3.9/3.10 compatibility; different root causes." + "reason": "Generic import-name failure vs Python version support issue; different causes and fixes." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Storage-object lookup failure vs chat-model crash; no concrete shared code path." + "reason": "Float8 storage lookup failure and Qwen3.5 chat crash are unrelated runtime errors." }, { - "left": "issue:44336", - "right": "issue:44373", + "left": "issue:42491", + "right": "issue:43299", "accept": false, - "reason": "Terminal formatting bug vs docstring issue; not duplicates." + "reason": "Both are Qwen3 MoE loading regressions, but one is LoRA compatibility across hf4.x/hf5.x and the other is Qwen3VL MoE loading on dev v5; similar area, not clearly the same code-path bug." }, { - "left": "issue:42491", - "right": "issue:43299", + "left": "issue:44336", + "right": "issue:44373", "accept": false, - "reason": "Both mention Qwen3 MoE, but one is LoRA compatibility on hf4.x/hf5.x and the other is Qwen3VL MoE loading; too different to merge as one PR." + "reason": "ANSI output in loading_report vs a wrong docstring for position_ids are not duplicates." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "pytest-xdist file race vs a stale generate warning; unrelated." + "reason": "pytest-xdist file race and stale warning in paged generate() touch different subsystems." }, { "left": "issue:44295", "right": "issue:45397", "accept": false, - "reason": "position_ids buffer access bug vs ZeRO3 from_pretrained failure; different issue." + "reason": "position_ids buffer access bug vs gemma-4 zero3 loading bug are unrelated." }, { "left": "issue:44246", - "right": "issue:44496", + "right": "issue:44871", "accept": false, - "reason": "Import slowdown vs unrecognized model/config error; not the same bug." + "reason": "Import slowness and eos_token_id configuration inconsistency are different issues." }, { "left": "issue:44242", - "right": "issue:44871", + "right": "issue:45593", "accept": false, - "reason": "Router-logits load-balancing loss bug vs Gemma eos_token_id config inconsistency; different model behaviors." + "reason": "Missing load-balancing loss when router logits are off vs missing auxiliary losses for D-FINE denoising are distinct training-loss bugs." }, { "left": "issue:43531", "right": "issue:45362", "accept": false, - "reason": "Sliding-window bug in Qwen3-MoE vs chat crash in Qwen3.5-35B; too broad and not the same concrete failure." + "reason": "Qwen3-MoE sliding_window behavior and Qwen3.5 chat crash are separate model-specific failures." }, { "left": "issue:41950", "right": "issue:44220", "accept": false, - "reason": "Video-classification image-processor lookup vs fbank feature extraction; different pipelines." + "reason": "Video-classification pipeline looking for image processors and _torch_extract_fbank_features() are different processing bugs." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "BeitImageProcessorFast reduce_labels bug vs unsafe sys.modules access in modeling_utils; unrelated." + "reason": "BEiT fast image processor reduce_labels bug vs modeling_utils sys.modules access bug are unrelated." }, { "left": "issue:42886", "right": "issue:46132", "accept": false, - "reason": "Offline cache loading bug vs AttentionInterface.register behavior change; unrelated." + "reason": "Offline cache loading failure and AttentionInterface.register behavior change are different concerns." }, { - "left": "issue:36246", - "right": "issue:42757", + "left": "issue:44246", + "right": "issue:44871", "accept": false, - "reason": "Qwen2.5-VL import error vs huggingface_hub offline-mode import error; both are imports, but different modules and failures." + "reason": "Repeated pair: import slowness vs Gemma eos_token_id inconsistency are unrelated." }, { - "left": "issue:36246", - "right": "issue:44351", + "left": "issue:44336", + "right": "issue:44485", "accept": false, - "reason": "Different missing symbols from different namespaces; not the same regression." + "reason": "ANSI code emission in a utility vs GLM-5 RoPE implementation discussion are not the same bug." }, { - "left": "issue:44246", - "right": "issue:44871", + "left": "issue:36246", + "right": "issue:42757", "accept": false, - "reason": "Import performance issue vs eos_token_id inconsistency; unrelated." + "reason": "Both are import errors, but one is Qwen2.5-VLImageProcessor and the other is is_offline_mode from huggingface_hub; different symbols and packages." }, { - "left": "issue:44336", - "right": "issue:44485", + "left": "issue:36246", + "right": "issue:44351", "accept": false, - "reason": "ANSI terminal formatting bug vs GLM-5 RoPE discussion; not the same underlying problem." + "reason": "Different missing exports: Qwen2.5-VLImageProcessor vs HybridCache." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "sys.modules access bug vs xdist file race; different subsystems." + "reason": "Unsafe sys.modules access in modeling_utils and xdist captured_info.txt race are unrelated." }, { - "left": "issue:43673", - "right": "issue:44561", + "left": "issue:44779", + "right": "issue:45292", "accept": false, - "reason": "chunked_prefill cache issue vs trust_remote_code breakage after utility removal; different concrete failures." + "reason": "DeepSeek tokenizer regression and resize_token_embeddings not updating output_embeddings are different tokenization/model-head bugs." }, { - "left": "issue:44779", - "right": "issue:45292", + "left": "issue:43673", + "right": "issue:44561", "accept": false, - "reason": "DeepSeek tokenizer regression vs embedding resize mismatch; unrelated." + "reason": "Chunked-prefill cache regression and removal of is_torch_fx_available breaking trust_remote_code are different v5 regressions with different failure modes." }, { "left": "issue:36246", "right": "issue:43596", "accept": false, - "reason": "Qwen2.5-VL import failure vs DeepSpeed/BertModel index error; unrelated." + "reason": "Qwen2.5-VL import error and deepspeed zero3 BertModel IndexError are unrelated." }, { "left": "issue:44295", "right": "issue:44749", "accept": false, - "reason": "position_ids buffer issue vs performance regression after version upgrade; different symptoms and likely code paths." + "reason": "position_ids buffer read issue vs performance slowdown after upgrading transformers are not the same bug." }, { "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "Missing auxiliary loss vs AMP/CUDA crash; unrelated." + "reason": "Load-balancing loss omission and LwDetrImageLoss float16 AMP crash are separate training/runtime issues." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "Specific model not supporting Flash Attention 2 vs a general Flash Attention 2 topic; not the same concrete bug." + "reason": "Nemotron Flash Attention 2 unsupported and a generic Flash Attention 2.0 issue title do not provide enough evidence of the same concrete bug." } ] }, @@ -11296,8 +11304,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -11371,31 +11379,31 @@ "issue:44466|issue:45335", "issue:45397|issue:45656", "issue:42886|issue:45229", - "issue:44483|issue:46202", "issue:44060|issue:44077", - "issue:44188|issue:46202", + "issue:44483|issue:46202", "issue:44995|issue:45588", "issue:45137|issue:45600", - "issue:43653|issue:44162", + "issue:44188|issue:46202", "issue:44368|issue:44977", + "issue:43653|issue:44162", "issue:44877|issue:46170", "issue:44368|issue:45563", "issue:43931|issue:44368" ], - "prepared_review_unit_hash": "bcb8fc576507416b12cfe43829c63919cdc67ef9f0a374fe762f3b0e9032e7b0", + "prepared_review_unit_hash": "c4d29dd3877eb727f3c74b2ef2b7feed37a43862779dbf297bb8c5e8b76bf1a9", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36742, + "serialized_chars": 36744, "estimated_input_tokens": 9186, "estimated_eval_tokens": 18628 }, @@ -11404,192 +11412,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a grab-bag of unrelated issues across tokenizers, multimodal configs, generation, logging, and CI; there is no single underlying bug/change tying them together.", - "confidence": 0.98, - "canonical_issue_reason": "No canonical issue: the items cover many distinct subsystems and failure modes (tokenizer caching, model config loading, generation internals, multimodal/video handling, memory leaks, docs/CI), so they are not duplicates of one another.", + "summary": "This cluster is mostly a collection of unrelated model, config, tokenizer, and generation regressions. The soft links are largely broad topical similarity rather than the same concrete bug, so none should be merged.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45003 is the clearest broad core-library bug here: it targets `modeling_utils` and can affect many import/load paths, unlike the more model-specific reports in the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a good cluster representative because the cluster is heterogeneous rather than a duplicate set.", + "best_issue_reason": "issue:45003 is a strong representative issue because it describes a concrete infrastructure bug in a central code path, making it more reusable as the cluster anchor than the narrower model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44162", "right": "issue:45200", "accept": false, - "reason": "ESM2 model breakage vs Gemma 4 mm_token_type_ids defaulting; different models and failure mechanisms." + "reason": "Different failures: ESM2 runtime breakage vs Gemma 4 token-type ID defaults." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "Docs mismatch for TextClassificationPipeline vs unsafe sys.modules access in modeling_utils; unrelated subsystems." + "reason": "Docs mismatch in pipeline behavior vs unsafe `sys.modules` access in core modeling code." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Unrecognized model/config loading error vs sys.modules access bug; different code paths and symptoms." + "reason": "Unrecognized model/config loading error vs unrelated `sys.modules` access bug." }, { "left": "issue:45412", "right": "issue:46123", "accept": false, - "reason": "RT-DETR memory not released vs MaskGenerationPipeline dropping final partial batch; unrelated behavior." + "reason": "RT-DETR memory leak and MaskGenerationPipeline partial-batch handling are unrelated code paths." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Strict config loading for granite_speech vs missing deepstack_visual_indexes in Qwen config; different model families/config fields." + "reason": "Both involve config validation, but they are different models and distinct missing-field/loading failures." }, { "left": "issue:45292", "right": "issue:45584", "accept": false, - "reason": "resize_token_embeddings output embedding sync issue vs Whisper empty transcription after align_special_tokens; not the same bug." + "reason": "Token-embedding resize behavior vs Whisper empty-transcription generation failure are unrelated." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio processor chat-template sampling-rate default vs slow import of transformers; unrelated." + "reason": "Audio chat-template sampling-rate bug vs intermittent import slowness." }, { "left": "issue:45292", "right": "issue:45701", "accept": false, - "reason": "Embedding resize propagation vs tokenization differing by transformers version; distinct problems." + "reason": "Resize-token-embeddings bug vs version-dependent tokenization change." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed vs ZeRO3 deque pop error; different failure points." + "reason": "Qwen3 MOE weight conversion/loading issue vs DeepSpeed ZeRO3 deque underflow." }, { "left": "issue:45375", "right": "issue:45759", "accept": false, - "reason": "Missing config field in Qwen3_5MoeVisionConfig vs AutoModelForCausalLM.from_config unwrapping composite multimodal configs; related area but not same concrete bug." + "reason": "Missing config field dropped by strict validation vs failing to unwrap `text_config` in `from_config`." }, { "left": "issue:45987", "right": "issue:46082", "accept": false, - "reason": "StaticCache seq length contract issue vs LlamaConfig head_dim validation; unrelated APIs." + "reason": "Cache length return-type bug vs LlamaConfig `head_dim` validation error." }, { "left": "issue:43232", "right": "issue:44336", "accept": false, - "reason": "Generation kwargs after sync_gpus vs ANSI codes in loading_report; no overlap in bug or fix." + "reason": "Generation kwargs update bug vs ANSI output in loading report." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "Model loading failure on a specific repo/version vs tokenizer errors hidden by protobuf exception handling; different root causes." + "reason": "Model loading failure vs tokenizer error masking when protobuf is missing." }, { "left": "issue:44479", "right": "issue:46170", "accept": false, - "reason": "Video input regression in Qwen variants vs Windows/ROCm torch.distributed import issue; different environments and code paths." + "reason": "Video-input regression vs Windows/ROCm torch.distributed import dependency problem." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "Stale indexer cache on second forward pass vs Gemma 4 token type ids default; unrelated model internals." + "reason": "Stale indexer cache crash vs Gemma 4 multimodal token-type defaulting." }, { "left": "issue:43065", "right": "issue:43408", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder vs model type mismatch warning for sam3_video/sam3_tracker; not the same defect." + "reason": "Dummy Conv2d placeholder vs model-type mismatch warning." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "Missing fpn_position_embeddings on Sam3VisionEncoderOutput vs Qwen3OmniMoe video ValueError; separate model families and failures." + "reason": "SAM 3 missing attribute vs Qwen3OmniMoe video-input ValueError." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "v5 tied-weight serialization inconsistency vs t5gemma resize_token_embeddings not affecting decoder.embed_tokens; different layers and symptoms." + "reason": "Tied-weight serialization inconsistency vs token-embedding resize not propagating to decoder embeddings." }, { "left": "issue:45397", "right": "issue:45656", "accept": false, - "reason": "Gemma-4 zero3 from_pretrained bug vs optimizer step called twice with deepspeed; unrelated training/loading issues." + "reason": "Gemma-4 ZeRO3 loading bug vs optimizer step being called twice under deepspeed." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "HF_HUB_OFFLINE tokenizer cache load issue vs Gemma4 multi-GPU CUDA OOM; completely different problems." - }, - { - "left": "issue:44483", - "right": "issue:46202", - "accept": false, - "reason": "Chat completions request rejection vs Gemma pipeline roundtrip failure on special character; not the same underlying bug." + "reason": "Offline tokenizer cache loading bug vs Gemma 4 multi-GPU OOM." }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "Wrong tied-weights warning in Qwen3-Next vs patchtsmixer post_init allowance; unrelated model/config issues." + "reason": "Incorrect tied-weights warning vs patchtsmixer `post_init` policy change." }, { - "left": "issue:44188", + "left": "issue:44483", "right": "issue:46202", "accept": false, - "reason": "torch.compile attention-kernel divergence vs Gemma pipeline token roundtrip failure; no common code path." + "reason": "Chat/completions request rejection vs Gemma pipeline roundtrip token issue." }, { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "Stale cache on second forward pass vs flash_attention crash on s_aux=None; different bug surfaces." + "reason": "Stale indexer cache crash vs flash-attention `s_aux=None` AttributeError." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error vs stale Sam3LiteText references in auto_mappings; one is runtime, the other CI/import metadata." + "reason": "DeepSpeed ZeRO3 deque error vs stale auto-mappings reference to removed configs." }, { - "left": "issue:43653", - "right": "issue:44162", + "left": "issue:44188", + "right": "issue:46202", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug vs ESM2 breakage; different tokenizer/model areas." + "reason": "torch.compile attention-kernel divergence vs Gemma roundtrip failure on \u2581." }, { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "Tie_word_embeddings warning in Qwen3.5 LoRA fine-tuning vs flash-attention generation failure; not the same issue." + "reason": "Tie-word-embeddings warning vs flash-attention generation regression." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token bug vs ESM2 breakage." }, { "left": "issue:44877", "right": "issue:46170", "accept": false, - "reason": "granite_speech config strictness vs Windows ROCm torch.distributed import/type error; unrelated." + "reason": "Config strictness regression vs Windows ROCm torch.distributed import issue." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "Tie-word-embeddings warning vs stale num_return_sequences warning in paged generate; different warnings and code paths." + "reason": "Different stale warnings: tied-weights config warning vs num_return_sequences warning." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch vs Qwen3.5 tie_word_embeddings warning; distinct loading vs config warning problems." + "reason": "Qwen3-VL weight-shape mismatch vs tied-word-embeddings warning." } ] }, @@ -11608,8 +11616,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -11631,8 +11639,8 @@ "issue:43873", "issue:44038", "issue:44162", + "issue:44188", "issue:44297", - "issue:44351", "issue:44361", "issue:44530", "issue:44623", @@ -11666,8 +11674,8 @@ "issue:43673|issue:44734", "issue:44530|issue:46170", "issue:43065|issue:46123", - "issue:44933|issue:45459", "issue:44297|issue:44623", + "issue:44933|issue:45459", "issue:44361|issue:44749", "issue:44869|issue:45356", "issue:43646|issue:44162", @@ -11679,10 +11687,10 @@ "issue:44871|issue:46132", "issue:45230|issue:45397", "issue:42994|issue:43482", - "issue:43475|issue:45381", "issue:43475|issue:43643", - "issue:28282|issue:30990", + "issue:43475|issue:45381", "issue:44038|issue:44623", + "issue:28282|issue:30990", "issue:46077|issue:46123", "issue:45406|issue:46202", "issue:42371|issue:45092", @@ -11693,222 +11701,222 @@ "issue:45081|issue:45103", "issue:45310|issue:45375", "issue:43873|issue:44945", - "issue:41628|issue:44351" + "issue:43232|issue:44188" ], - "prepared_review_unit_hash": "d080db09b54f121d0600490426d03eb76ac6a0a720da37f8ea7a1d28551094fb", + "prepared_review_unit_hash": "041bd16f579a6bd46df2fdcf8130b016a7fce8aa61aff62f0f595eb909c5d990", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36838, - "estimated_input_tokens": 9210, - "estimated_eval_tokens": 18676 + "serialized_chars": 36874, + "estimated_input_tokens": 9219, + "estimated_eval_tokens": 18694 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is mostly heterogeneous issue reports. A few are loosely related Transformers v5 loading/custom-model regressions, but none of the soft pairs are close enough to be treated as the same underlying bug or change, so I reject all soft edges. No pull requests are present. The broadest issue is the generic v5 custom-model initialization regression report.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43646 is the best umbrella candidate because it describes a broad Transformers 5.0 custom-model initialization regression that can plausibly encompass several narrower loading/meta-init failures in the cluster.", + "summary": "This is a loose collection of mostly unrelated Transformers issues. Several share broad themes like generation, tokenizers, or model-family regressions, but none of the soft pairs look like the same concrete bug/change, so I reject all soft edges.", + "confidence": 0.86, + "canonical_issue_reason": "issue 43873 is the strongest representative issue here: it is detailed, active, has inbound references, and describes a concrete reproducible bug with broader impact.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the most generally applicable user-facing regression report among the issues here; the others are mostly model-specific or unrelated subsystems.", + "best_issue_reason": "issue 43873 is the best standalone triage candidate in this set because it is well-documented, still open, and actionable.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43673", "right": "issue:44734", "accept": false, - "reason": "Both involve generation/cache behavior, but they describe different concrete failures and code paths." + "reason": "Both involve generation/KV-cache behavior, but they fail in different code paths: chunked-prefill cache handling vs serve continuation tensor indexing." }, { "left": "issue:44530", "right": "issue:46170", "accept": false, - "reason": "PagedAttention/Qwen3.5 cache failure is unrelated to a Windows ROCm import/dependency problem." + "reason": "Completely different problems: Qwen3.5 paged attention group parsing versus Windows/ROCm import and torch.distributed dependency issues." }, { "left": "issue:43065", "right": "issue:46123", "accept": false, - "reason": "SAM3 pixel decoder dummy layer and mask pipeline batch handling are unrelated bugs." + "reason": "Different subsystems and symptoms: SAM3 pixel decoder dummy layer versus pipeline batching dropping final results." }, { - "left": "issue:44933", - "right": "issue:45459", + "left": "issue:44297", + "right": "issue:44623", "accept": false, - "reason": "One is a missing import/export, the other is tokenizer error handling when protobuf is absent; not the same bug." + "reason": "Both are save_pretrained-related, but one is tokenizer metadata/class mismatch and the other is processor file emission; not the same bug." }, { - "left": "issue:44297", - "right": "issue:44623", + "left": "issue:44933", + "right": "issue:45459", "accept": false, - "reason": "Both are save_pretrained regressions, but one concerns tokenizer config content while the other concerns processor file output." + "reason": "Different issues: a missing/nonexistent image_utils import versus tokenizer error handling that hides protobuf-related exceptions." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "MLukeTokenizer task AttributeError and a post-upgrade performance slowdown are different issues." + "reason": "Unrelated failures: MLukeTokenizer task AttributeError versus a performance regression after upgrading Transformers." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Both mention tokenizer/codec problems, but they affect different models and failure modes." + "reason": "Both mention token decoding/codec behavior, but they are different model-specific regressions with different failure modes." }, { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "General custom-model init regression and ESM2 breakage are too different to merge as one bug." + "reason": "Different areas entirely: trust_remote_code config field loss versus an ESM2 model breakage report." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Both touch generation internals, but one is sync_gpus kwargs handling and the other is chunked_prefill cache state." + "reason": "Both are generation-related, but one is after sync_gpus bookkeeping and the other is missing generation cache in chunked prefill." }, { "left": "issue:44623", "right": "issue:45656", "accept": false, - "reason": "Processor save output files and optimizer stepping under deepspeed are unrelated." + "reason": "No shared bug: missing processor save files versus optimizer stepping twice under deepspeed." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "Training slowdown after upgrade and duplicate optimizer stepping are unrelated." + "reason": "A data-filtering slowdown regression is unrelated to the deepspeed double-step optimizer bug." }, { "left": "issue:43701", "right": "issue:44749", "accept": false, - "reason": "Checkpoint key mismatch and data filtering slowdown are unrelated." + "reason": "Checkpoint key mismatch and upgrade-related filtering slowdown are different failure modes." }, { "left": "issue:43673", "right": "issue:43827", "accept": false, - "reason": "Generation cache regression and docs still referencing pipeline() are not the same underlying change." + "reason": "Generation cache bug versus documentation still referencing removed pipeline APIs; one is runtime, the other docs." }, { "left": "issue:44871", "right": "issue:46132", "accept": false, - "reason": "Gemma eos_token_id inconsistency and AttentionInterface.register behavior change are unrelated." + "reason": "Gemma eos_token_id inconsistency is unrelated to AttentionInterface.register changing callback behavior." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "The generic bug report has no evidence tying it to gemma-4 zero3 from_pretrained." + "reason": "Issue 45230 is too vague to match, and 45397 is a specific gemma-4 ZeRO3 from_pretrained failure." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model saving and GGUF loading are different code paths." + "reason": "Related to quantized models in a broad sense, but saving failure and GGUF loading failure are different concrete bugs." }, { "left": "issue:43475", - "right": "issue:45381", + "right": "issue:43643", "accept": false, - "reason": "Sam3 vision encoder output missing field and Qwen2.5-VL video position ids are different model bugs." + "reason": "Different objects and code paths: SAM3 vision encoder output missing an attribute versus AutoConfig trust_remote_code dropping fields." }, { "left": "issue:43475", - "right": "issue:43643", + "right": "issue:45381", "accept": false, - "reason": "A missing attribute on Sam3 output and missing fields from trust_remote_code config are not the same defect." + "reason": "Both are vision/video-related, but SAM3 output attributes and Qwen2.5-VL vision_position_ids are unrelated bugs." }, { - "left": "issue:28282", - "right": "issue:30990", + "left": "issue:44038", + "right": "issue:44623", "accept": false, - "reason": "PyTorch missing import error and sentence-transformers loading hang are unrelated." + "reason": "Qwen3-VL-Moe loading failure and processor save_pretrained missing files do not describe the same issue." }, { - "left": "issue:44038", - "right": "issue:44623", + "left": "issue:28282", + "right": "issue:30990", "accept": false, - "reason": "Qwen3-VL-Moe v5 bug and processor save_pretrained file omission are different issues." + "reason": "An immediate PyTorch ImportError is not the same as Sentence Transformers getting stuck during loading." }, { "left": "issue:46077", "right": "issue:46123", "accept": false, - "reason": "Gemma4 attention-mask construction and MaskGenerationPipeline final-batch handling are unrelated." + "reason": "Attention mask construction for Gemma4 is unrelated to MaskGenerationPipeline's final-batch handling." }, { "left": "issue:45406", "right": "issue:46202", "accept": false, - "reason": "Gemma4 processor missing _tokenizer and Gemma pipeline roundtrip on \u2581 are different bugs." + "reason": "Gemma4 processor initialization failure and Gemma pipeline roundtrip tokenization failure are different bugs." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 settings guidance and InternVL2 meta-init incompatibility are unrelated." + "reason": "TF32 API guidance is unrelated to remote-code checkpoint incompatibility with v5 meta initialization." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Both are import errors, but they concern different symbols/modules and different failure mechanisms." + "reason": "Both are import-related, but they concern different symbols and different failure surfaces; not the same underlying export bug." }, { "left": "issue:43475", "right": "issue:44661", "accept": false, - "reason": "Sam3VisionEncoderOutput and add-new-model-like TOKENIZER_MAPPING_NAMES are unrelated." + "reason": "A missing SAM3 output attribute is unrelated to add-new-model-like failing on TOKENIZER_MAPPING_NAMES membership." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized model saving and Llama4Config missing pad_token_id are different bugs." + "reason": "Quantized saving failure and missing pad_token_id on Llama4Config are different model/config bugs." }, { "left": "issue:45375", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3.5 MoE, but one is a missing strict field and the other is a from_pretrained regression; not enough to treat as the same bug." + "reason": "Same model family, but one is a specific missing config field and the other is a generic from_pretrained error; too little evidence they are the same bug." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Mistral regex patch tokenizer crash and auto_docstring future-annotations crash are unrelated." + "reason": "Both produce AttributeError, but they are in unrelated components: tokenizer regex patching versus auto_docstring processing." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "Both are Qwen3.5 MoE regressions, but one is a generic loading error and the other is a specific vision-config field drop." + "reason": "Both involve Qwen3.5-Moe, but one is a generic from_pretrained failure while the other is a specific vision-config field omission." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization offloading and pipeline-parallelism output correctness are different problems." + "reason": "Quantization/offloading behavior and pipeline-parallel output correctness are different code paths and symptoms." }, { - "left": "issue:41628", - "right": "issue:44351", + "left": "issue:43232", + "right": "issue:44188", "accept": false, - "reason": "Different missing-import issues affecting different symbols; not the same underlying fix." + "reason": "Generation kwargs bookkeeping and attention kernel divergence under torch.compile are not the same bug." } ] }, @@ -11927,8 +11935,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -11961,11 +11969,11 @@ "issue:44188", "issue:44222", "issue:44242", + "issue:44351", "issue:44496", "issue:44568", "issue:44623", "issue:44734", - "issue:44779", "issue:44792", "issue:44869", "issue:44933", @@ -11982,17 +11990,17 @@ "issue:45812" ], "soft_pairs": [ - "issue:43232|issue:44188", + "issue:41628|issue:44351", "issue:42673|issue:43704", "issue:44964|issue:45137", - "issue:44945|issue:45753", - "issue:36010|issue:41628", "issue:42994|issue:43824", "issue:42831|issue:44222", + "issue:44945|issue:45753", + "issue:36010|issue:41628", "issue:44222|issue:45310", "issue:43232|issue:44792", - "issue:43673|issue:44188", "issue:44038|issue:45310", + "issue:43673|issue:44188", "issue:42175|issue:45712", "issue:43526|issue:45561", "issue:44496|issue:45561", @@ -12012,235 +12020,228 @@ "issue:43784|issue:43824", "issue:44734|issue:45103", "issue:43723|issue:44568", - "issue:42898|issue:44779", "issue:42898|issue:45356" ], - "prepared_review_unit_hash": "1099455bee87f0e63b283dca567f174fa7ffd646704ba91baa63ff9370639d9f", + "prepared_review_unit_hash": "bb784333617437e71ac4740af354522633a7e49c3c399bedd0e2a0d9dbfc7f2f", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 37221, - "estimated_input_tokens": 9306, - "estimated_eval_tokens": 18868 + "soft_pair_count": 31, + "serialized_chars": 36897, + "estimated_input_tokens": 9225, + "estimated_eval_tokens": 18706 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of unrelated Transformers regressions (tokenizers, imports, model loading, FP8, generation, multimodal, etc.). Only the Qwen3ForCausalLM VRAM-leak reports look like a true duplicate pair; the rest are too different in code path or symptom to merge.", + "summary": "This cluster is mostly a grab-bag of unrelated Transformers bugs. The only clear duplicate pair is the repeated Qwen3ForCausalLM VRAM leak report; the rest are similar in subsystem at best, but describe different concrete failures.", "confidence": 0.87, - "canonical_issue_reason": "issue:42673 is the earlier, fuller report of the same Qwen3ForCausalLM multi-dataloader VRAM leak; issue:43704 is a duplicate with the same title/symptom.", + "canonical_issue_reason": "issue:42673 is the earliest and fuller report of the exact same Qwen3ForCausalLM VRAM leak problem; issue:43704 repeats the same title later with less context.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the strongest canonical issue candidate in this set because it has the clearest one-to-one duplicate match and the richer original bug report.", + "best_issue_reason": "issue:42673 is the strongest representative because it is the original, detailed report for the only clear duplicate issue in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43232", - "right": "issue:44188", + "left": "issue:41628", + "right": "issue:44351", "accept": false, - "reason": "Both are generation/runtime bugs, but they hit different mechanisms: sync_gpus kwargs update vs torch.compile attention-kernel branching." + "reason": "Both are import errors, but they reference different missing symbols (`AutoImageProcessor` vs `HybridCache`) and likely different export gaps." }, { "left": "issue:42673", "right": "issue:43704", "accept": true, - "reason": "Same title, same Qwen3ForCausalLM VRAM leak symptom, and the later report is clearly the same bug." + "reason": "Same title and same underlying bug: Qwen3ForCausalLM leaks VRAM when used across multiple dataloader threads." }, { "left": "issue:44964", "right": "issue:45137", "accept": false, - "reason": "One is a Phi-4 multimodal load failure; the other is a DeepSpeed ZeRO3 deque IndexError. Different bugs." + "reason": "Model loading failure vs DeepSpeed ZeRO3 deque error; these are unrelated code paths." }, { - "left": "issue:44945", - "right": "issue:45753", + "left": "issue:42994", + "right": "issue:43824", "accept": false, - "reason": "Pipeline-parallelism output corruption and Qwen3_5 image infinite loop are different model/workflow failures." + "reason": "Quantized model saving failure and a Qwen2.5-VL import error are different problems." }, { - "left": "issue:36010", - "right": "issue:41628", + "left": "issue:42831", + "right": "issue:44222", "accept": false, - "reason": "Both are import errors, but for different symbols and likely different fixes; not the same underlying bug." + "reason": "Both mention FP8, but one is an accuracy regression and the other is a save_pretrained/MoE serialization issue." }, { - "left": "issue:42994", - "right": "issue:43824", + "left": "issue:44945", + "right": "issue:45753", "accept": false, - "reason": "Quantized save failure and missing Qwen2.5-VL class import are unrelated." + "reason": "Pipeline parallelism output corruption and a Qwen3.5 image infinite loop are different bugs." }, { - "left": "issue:42831", - "right": "issue:44222", + "left": "issue:36010", + "right": "issue:41628", "accept": false, - "reason": "FP8 accuracy drift and FP8 save_pretrained/moe serialization are different concrete issues." + "reason": "Both are import failures, but for different missing APIs and likely different root causes." }, { "left": "issue:44222", "right": "issue:45310", "accept": false, - "reason": "MoE save_pretrained failure vs Qwen3.5-MoE from_pretrained load error are not the same code path." + "reason": "Different symptoms and model families: FP8 MoE save_pretrained vs Qwen3.5 MoE from_pretrained error." }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation kwargs/cache update bug and a janus test failure do not describe the same defect." + "reason": "Generation kwargs/cache handling and a janus image-generation test failure do not describe the same bug." }, { - "left": "issue:43673", - "right": "issue:44188", + "left": "issue:44038", + "right": "issue:45310", "accept": false, - "reason": "Chunked-prefill GenerationMixin cache issue and torch.compile attention divergence are separate generation bugs." + "reason": "Related model family naming, but the reported failures are different and not the same concrete issue." }, { - "left": "issue:44038", - "right": "issue:45310", + "left": "issue:43673", + "right": "issue:44188", "accept": false, - "reason": "Both mention Qwen3-MoE, but one is a broader v5.0 compatibility issue and the other is a specific v5.4.0 from_pretrained regression." + "reason": "Chunked prefill cache behavior and torch.compile attention-kernel branching are unrelated." }, { "left": "issue:42175", "right": "issue:45712", "accept": false, - "reason": "TensorFlow backend packaging and leftover dummy classes in dir(transformers) are unrelated maintenance issues." + "reason": "Packaging/backend dependency issues and leftover dummy classes are different maintenance problems." }, { "left": "issue:43526", "right": "issue:45561", "accept": false, - "reason": "Beit label reduction bug and pytest-xdist file-race are unrelated." + "reason": "BeitImageProcessorFast label reduction and a pytest-xdist file race are unrelated." }, { "left": "issue:44496", "right": "issue:45561", "accept": false, - "reason": "Unrecognized model/config loading and captured_info.txt race are different problems." + "reason": "Model config loading failure vs xdist race; no shared underlying bug." }, { "left": "issue:43504", "right": "issue:45561", "accept": false, - "reason": "Legacy-field pretrained loading failure and xdist file race are unrelated." + "reason": "Legacy BEiT preset loading and xdist capture-file race are different issues." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Packaging dependency mismatch and a missing image_utils import are unrelated." + "reason": "Torch extra/backend packaging and a missing image_utils import are not the same defect." }, { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "Padding-mask mismatch in MIMI encoder vs missing load-balancing loss when router logits are off are distinct bugs." + "reason": "Missing padding-mask support in MIMI encoder and missing load-balancing loss when router logits are off are different code paths." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Tokenizer cleanup-space behavior change and missing BOS/EOS insertion are different tokenizer regressions." + "reason": "Tokenizer cleanup behavior change and BOS/EOS insertion regression are distinct tokenizer bugs." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "Broken vitpose model loading and Gemma 4 mm_token_type_ids defaulting are unrelated model/config issues." + "reason": "VitPose breakage and Gemma 4 multimodal token type defaults are unrelated." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "Python version compatibility and nonexistent import from image_utils are unrelated." + "reason": "Python version compatibility and missing image_utils import are unrelated." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "Sentence-transformers loading hang and embedding reinitialization on resize are unrelated." + "reason": "Sentence-Transformers loading hang and embedding resize reinitialization are different issues." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "Dtype drift between train/eval and a ZeRO3 deque IndexError are different failures." + "reason": "Embedding dtype drift in evaluation and a ZeRO3 deque IndexError are unrelated." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Default multimodal config initialization and Whisper word-timestamp decode crash are unrelated." + "reason": "Phi4 multimodal config defaults and Whisper timestamp decode crash are different bugs." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Default config construction and ModelOutput key assignment are different bugs." + "reason": "Config initialization bug and ModelOutput key assignment bug are separate problems." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Missing processor save files and Qwen3.5-MoE from_pretrained loading error are unrelated." + "reason": "Processor save_pretrained missing files and Qwen3.5 MoE from_pretrained error are different failures." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLflow reporting on training failure and Adafactor OOM in DeepSpeed are unrelated." + "reason": "MLFlow reporting after training failure and Adafactor OOM in DeepSpeed are unrelated." }, { "left": "issue:42898", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but one is cleanup-space behavior and the other is wrong token IDs for Granite; different underlying issues." + "reason": "Both are tokenizer regressions, but they affect different behaviors and different model families." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Sentence-transformers import NameError and missing Qwen2.5-VL class import are unrelated import failures." + "reason": "Sentence-Transformers import NameError and Qwen2.5-VL import error are unrelated import issues." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "KV-cache continuation crash in serve and auto_docstring AttributeError under future annotations are unrelated." + "reason": "KV-cache continuation indexing in serve and auto_docstring annotation handling are unrelated." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "AutoTokenizer loading regression and add_special_tokens BOS/EOS regression are different tokenizer bugs." - }, - { - "left": "issue:42898", - "right": "issue:44779", - "accept": false, - "reason": "Different tokenizer regressions affecting different models/symptoms; not the same bug." + "reason": "Tokenizer loading in v5 and BOS/EOS token insertion regression are distinct tokenizer issues." }, { "left": "issue:42898", "right": "issue:45356", "accept": false, - "reason": "Clean-up-space change and Kimi-K2.5 codec/regex regression are unrelated tokenizer issues." + "reason": "v5 cleanup behavior and Kimi-K2.5 codec handling are different tokenizer regressions." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The only accepted soft pair, issue:42673 vs issue:43704, is strongly supported by identical titles and matching symptom. The remaining verdicts are appropriately cautious and do not overstate duplicate evidence." + "feedback": "The summary is well grounded in the packet: the only accepted duplicate pair is indeed issue:42673 / issue:43704, with 42673 being earlier and more detailed. The other soft-edge rejections are conservative and consistent with the titles and descriptions, and there are no PR mergeability concerns here." }, "evaluator_used": true, "retried": false, @@ -12256,8 +12257,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -12285,7 +12286,6 @@ "issue:44263", "issue:44295", "issue:44297", - "issue:44360", "issue:44373", "issue:44448", "issue:44483", @@ -12293,6 +12293,7 @@ "issue:44485", "issue:44514", "issue:44623", + "issue:44779", "issue:44933", "issue:44977", "issue:45071", @@ -12311,10 +12312,11 @@ "issue:46209" ], "soft_pairs": [ + "issue:42898|issue:44779", "issue:43874|issue:45092", "issue:44060|issue:45125", - "issue:45405|issue:45712", "issue:42898|issue:44448", + "issue:45405|issue:45712", "issue:44295|issue:45656", "issue:44448|issue:46132", "issue:44933|issue:45712", @@ -12343,97 +12345,102 @@ "issue:44295|issue:44623", "issue:44373|issue:44485", "issue:44483|issue:45406", - "issue:44514|issue:45103", - "issue:44263|issue:44360" + "issue:44514|issue:45103" ], - "prepared_review_unit_hash": "c01abc47d7ff4f427c61f70e9fcbcee59cf06a36e584dd8e4deaa521d56bfefd", + "prepared_review_unit_hash": "6483c58bd454d63b558fbe46a04d83e6b4295b01ff7ee44825e202d348e814f2", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37624, - "estimated_input_tokens": 9406, - "estimated_eval_tokens": 19068 + "serialized_chars": 37646, + "estimated_input_tokens": 9412, + "estimated_eval_tokens": 19080 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly unrelated issue reports that only share superficial keyword overlap (imports, v5 regressions, save/load, multimodal, FP8, trainers). None of the soft pairs looks like the same underlying bug or change, so all soft edges should be rejected.", - "confidence": 0.95, - "canonical_issue_reason": "No single canonical issue is safe here: the items span many distinct bugs and regressions rather than one duplicate thread.", + "summary": "These items are not a duplicate set; they span unrelated bugs and regressions across tokenization, model loading, multimodal processors, config validation, FP8, and tooling. The soft pairs mostly share only broad subsystem similarity, not the same underlying defect.", + "confidence": 0.17, + "canonical_issue_reason": "issue:42898 is the closest broad v5 tokenizer-regression candidate, but it still only represents one narrow symptom and does not unify the rest of the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "No issue is a strong global representative; the set is heterogeneous and the apparent matches are only lexical, not the same code-path failure.", + "best_issue_reason": "issue:42898 is the best single representative by breadth of nearby v5/tokenization reports, though the cluster is too heterogeneous to treat it as a real canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 tokenizer regressions, but one is clean_up_tokenization_spaces behavior and the other is Deepseek output drift; different code paths." + }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Different failures: missing image patch method in GLM46V vs meta-init incompatibility in old InternVL2 checkpoints." + "reason": "Missing image-patch method in an image processor is unrelated to remote-code checkpoint meta-initialization incompatibility." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Both mention Qwen MoE, but one is a tied-weights warning bug and the other is a missing tensor-parallel plan." + "reason": "Tied-weights warning in Qwen3-Next is a different defect from missing _tp_plan in Qwen3_5Moe." }, { - "left": "issue:45405", - "right": "issue:45712", + "left": "issue:42898", + "right": "issue:44448", "accept": false, - "reason": "PEFT version bump mismatch and leftover dummy classes are unrelated packaging/check-repo issues." + "reason": "Tokenization cleanup behavior change is not the same as Pegasus output differences across v4/v5." }, { - "left": "issue:42898", - "right": "issue:44448", + "left": "issue:45405", + "right": "issue:45712", "accept": false, - "reason": "Tokenization-space cleanup behavior and Pegasus output differences are separate tokenizer/model-output regressions." + "reason": "A version pin bump for PEFT is unrelated to leftover dummy class cleanup and dir() leakage." }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "Position_ids buffer reading error is unrelated to the deepspeed optimizer-step duplication bug." + "reason": "Reading a registered position_ids buffer is unrelated to deepspeed causing optimizer.step to run twice." }, { "left": "issue:44448", "right": "issue:46132", "accept": false, - "reason": "Pegasus v4/v5 output differences do not match an AttentionInterface registration behavior change." + "reason": "Pegasus output drift and AttentionInterface.register behavior changes are separate code paths." }, { "left": "issue:44933", "right": "issue:45712", "accept": false, - "reason": "An image_utils import problem is not the same as stale dummy classes leaking into dir(transformers)." + "reason": "Nonexistent image_utils import and dummy_pt_objects cleanup are different import-layer issues." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "FP8 save_pretrained behavior for MoE models is unrelated to tokenizer_class mismatch on save_pretrained." + "reason": "FP8 save_pretrained for MoE is not the same as tokenizer_class mismatch in tokenizer_config.json." }, { "left": "issue:44297", "right": "issue:45656", "accept": false, - "reason": "Tokenizer save metadata corruption and duplicated optimizer steps are unrelated bugs." + "reason": "Tokenizer save metadata mismatch and deepspeed double-step are unrelated failures." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Phi4Multimodal config defaults and non-persistent buffer junk initialization are different initialization paths." + "reason": "Defaulting multimodal configs when None is unrelated to non-persistent buffers being filled with junk." }, { "left": "issue:44190", @@ -12445,139 +12452,133 @@ "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "MoE FP8 save_pretrained and processor.save_pretrained missing files are different save-path issues." + "reason": "FP8 save_pretrained on MoE and processor.save_pretrained missing files are different serialization bugs." }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "MLFlow reporting on failed training jobs is unrelated to a Llama 3.2 Vision model-loading regression." + "reason": "MLFlow job status reporting and Llama 3.2 Vision model-loading breakage do not share the same bug." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "FineGrainedFP8 accuracy and processor save_pretrained missing files are not the same bug." + "reason": "FineGrainedFP8 accuracy issues are unrelated to missing processor save_pretrained artifacts." }, { "left": "issue:43122", "right": "issue:46209", "accept": false, - "reason": "Tokenization differences across versions and FP8 inference failure on B200 are unrelated." + "reason": "Tokenizer behavior drift from 4.57.3 to 5.0 is unrelated to FP8 inference failure on B200." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "MLFlow job-state reporting and CustomTrainer compute_loss signature mismatch are separate trainer issues." + "reason": "MLFlow reporting after failed training is not the same as compute_loss signature breakage." }, { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Missing transformers.onnx module and missing PyTorch dependency are distinct import errors." + "reason": "Missing transformers.onnx and missing torch for AutoModel are both import issues, but they affect different modules and causes." }, { "left": "issue:28282", "right": "issue:30333", "accept": false, - "reason": "PyTorch import requirement failure is unrelated to MLFlow reporting behavior." + "reason": "PyTorch-not-found ImportError and MLFlow job reporting are unrelated." }, { "left": "issue:45245", "right": "issue:46132", "accept": false, - "reason": "Category-count runtime error and AttentionInterface registration behavior are unrelated APIs." + "reason": "Category-count overflow is unrelated to AttentionInterface.register changing registered function behavior." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "Failed training job reporting and embedding resize reinitialization are unrelated trainer/model-init bugs." + "reason": "MLFlow failure reporting and embedding reinitialization in post_init are different training/model-init bugs." }, { "left": "issue:45071", "right": "issue:46082", "accept": false, - "reason": "PretrainedConfig type checking regression is unrelated to LlamaConfig head_dim validation." + "reason": "PretrainedConfig type-checking regression is unrelated to LlamaConfig head_dim rejection." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "AutoImageProcessor import failure and is_offline_mode import failure are different missing-symbol issues." + "reason": "AutoImageProcessor import failure and is_offline_mode import failure come from different packages and symbols." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "Unsupported center argument in a feature extractor is unrelated to the auto-docstring AttributeError with future annotations." + "reason": "Unsupported center arg in a feature extractor is unrelated to auto_docstring crashing on future annotations." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer save metadata mismatch and Qwen3.5 MoE from_pretrained error are different failure modes." + "reason": "Tokenizer save metadata mismatch and Qwen3.5 MoE from_pretrained error are separate serialization/load failures." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "Tokenizer loading in v5 and Qwen3.5 flash-attention generation problems are unrelated." + "reason": "Tokenizer loading regression and Qwen3.5 flash-attention generation failure are different runtime paths." }, { "left": "issue:43232", "right": "issue:45834", "accept": false, - "reason": "Generation kwargs update after sync_gpus is not the same as Kosmos2.5 OCR long-input index error." + "reason": "Generation kwargs update after sync_gpus is unrelated to Kosmos2.5 OCR long-input indexing." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "torch.split return values in a GLM MoE indexer are unrelated to old InternVL2 meta-init incompatibility." + "reason": "torch.split indexing semantics in GlmMoeDsaIndexer do not match remote-code meta-init incompatibility." }, { "left": "issue:43761", "right": "issue:46082", "accept": false, - "reason": "CLIPVisionModel hidden_states regression is unrelated to LlamaConfig head_dim rejection." + "reason": "CLIPVision hidden_states regression and LlamaConfig head_dim validation are unrelated." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "FP8 MoE save_pretrained bug and max_shard_size default question are not the same issue." + "reason": "FP8 MoE save_pretrained and max_shard_size default policy are different concerns." }, { "left": "issue:44295", "right": "issue:44623", "accept": false, - "reason": "Position_ids buffer access error does not match missing processor save_pretrained files." + "reason": "position_ids buffer read errors are unrelated to processor.save_pretrained missing files." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "A wrong docstring and a GLM-5 RoPE implementation discussion are not duplicate bugs." + "reason": "A docstring mistake for position_ids is not the same as a GLM-5 RoPE implementation discussion." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "Chat completions endpoint acceptance failure is unrelated to Gemma4Processor missing _tokenizer." + "reason": "chat/completions endpoint rejection is unrelated to Gemma4Processor missing _tokenizer." }, { "left": "issue:44514", "right": "issue:45103", "accept": false, - "reason": "Batched apply_chat_template crash is unrelated to the auto-docstring future-annotations AttributeError." - }, - { - "left": "issue:44263", - "right": "issue:44360", - "accept": false, - "reason": "Both mention DSA/GLM, but one is a torch.split return-value bug and the other is a discussion about a missing ReLU; not the same underlying defect." + "reason": "Batched apply_chat_template crash and auto_docstring future-annotations crash are different processor/documentation code paths." } ] }, @@ -12596,8 +12597,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -12629,6 +12630,7 @@ "issue:44117", "issue:44261", "issue:44263", + "issue:44360", "issue:44373", "issue:44442", "issue:44521", @@ -12650,6 +12652,7 @@ "issue:46132" ], "soft_pairs": [ + "issue:44263|issue:44360", "issue:45092|issue:45357", "issue:43723|issue:44871", "issue:30990|issue:38175", @@ -12659,8 +12662,8 @@ "issue:44829|issue:46002", "issue:43723|issue:44442", "issue:43479|issue:44913", - "issue:41628|issue:43824", "issue:42831|issue:45310", + "issue:41628|issue:43824", "issue:44263|issue:44373", "issue:42831|issue:44038", "issue:39401|issue:44568", @@ -12669,8 +12672,8 @@ "issue:42886|issue:42994", "issue:42886|issue:43066", "issue:22355|issue:28218", - "issue:42831|issue:45753", "issue:43756|issue:44060", + "issue:42831|issue:45753", "issue:44117|issue:44704", "issue:44671|issue:46132", "issue:43576|issue:43784", @@ -12679,202 +12682,208 @@ "issue:42371|issue:44263", "issue:30990|issue:33357" ], - "prepared_review_unit_hash": "4aaa9316b424e5e6c526ff7f468f1361eac9a8a9d301fc14f6e54af5bf933265", + "prepared_review_unit_hash": "c898a99c8bc5cf522786cc95a8ea7193c11748fb8e4c6e65cd90486292feb846", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 28, - "serialized_chars": 35096, - "estimated_input_tokens": 8774, - "estimated_eval_tokens": 17804 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35928, + "estimated_input_tokens": 8982, + "estimated_eval_tokens": 18220 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The candidate edges are mostly broad topical similarities rather than true duplicates. They span unrelated tokenizer, import, quantization, configuration, and model-specific regressions, so I would reject all soft pairs as the same underlying bug/change.", - "confidence": 0.87, - "canonical_issue_reason": "issue:43723 is the broadest representative of the cluster\u2019s recurring v5 tokenizer-loading regressions; several other items are narrower or model-specific variants, but none are clear duplicates of it.", + "summary": "The set is mostly a mix of unrelated Transformers v5 regressions; none of the soft pairs look like true duplicates of the same underlying bug/change, so all soft edges are rejected. The most representative issue is the broad tokenizer-loading regression in issue 43723.", + "confidence": 0.62, + "canonical_issue_reason": "issue:43723 is the broadest, most central tokenizer-loading regression in this set and overlaps conceptually with several nearby tokenizer-related reports, making it the best canonical issue candidate.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43723 is the best overall issue representative because it captures the central failure mode repeated across multiple nearby reports: tokenizer loading breakage in Transformers v5.", + "best_issue_reason": "issue:43723 best captures the shared theme of v5 tokenizer/from_pretrained breakage without being too model-specific.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Same DSA/GLM-MoE area, but the titles indicate different defects: split return handling vs missing ReLU." + }, { "left": "issue:45092", "right": "issue:45357", "accept": false, - "reason": "Both are multimodal save/load regressions, but they affect different model families and different failure modes; not the same concrete bug." + "reason": "Both are multimodal checkpoint/config issues, but one is meta-init incompatibility and the other is incorrect visual-encoder save keys." }, { "left": "issue:43723", "right": "issue:44871", "accept": false, - "reason": "Both involve tokenizer/model loading in v5, but one is a generic AutoTokenizer loading failure and the other is an eos_token_id configuration inconsistency." + "reason": "Both involve tokenizer/model loading in v5, but one is an AutoTokenizer load failure and the other is an eos_token_id config mismatch." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Completely different symptoms and model areas: sentence-transformers loading hang vs zero probabilities in a SigLIP model." + "reason": "Unrelated model-loading stall vs zero-probability inference bug." }, { "left": "issue:42907", "right": "issue:46050", "accept": false, - "reason": "Both concern quantized model persistence, but they target different quantization flows and different save/load failure paths." + "reason": "Both touch quantized save/load, but the concrete failure modes differ: dequantized Ministrals/Devstrals saving vs sinq round-trip corruption." }, { "left": "issue:44261", "right": "issue:45092", "accept": false, - "reason": "Unrelated bugs: MLA layernorm precision/config issue vs remote-code meta-initialization incompatibility." + "reason": "Different problems: RMS norm epsilon precision mismatch vs remote-code checkpoint meta-init incompatibility." }, { "left": "issue:43723", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer loading reports, but they concern different models and likely different tokenizer registration/mapping causes." + "reason": "Both are tokenizer-load regressions, but the affected code paths and model-specific failures are not shown to be the same." }, { "left": "issue:44829", "right": "issue:46002", "accept": false, - "reason": "Different problems: flash_attention_3 training degeneration vs NaNs during classification-head init/checkout." + "reason": "Different training pathologies: flash_attention_3 degeneration vs NaNs in classification heads after checkout/init." }, { "left": "issue:43723", "right": "issue:44442", "accept": false, - "reason": "Both are AutoTokenizer loading failures, but for different tokenizers and likely different missing-registration issues." + "reason": "Both are AutoTokenizer load issues, but they appear model-specific and there is no evidence they share the same root cause." }, { "left": "issue:43479", "right": "issue:44913", "accept": false, - "reason": "Different config serialization bugs in different model families; not the same code path." + "reason": "Different configuration bugs: multimodal sub-config initialization vs rotary_pct not persisting on reload." }, { - "left": "issue:41628", - "right": "issue:43824", + "left": "issue:42831", + "right": "issue:45310", "accept": false, - "reason": "Both are import errors, but for different missing exports and different classes; too broad to merge." + "reason": "Accuracy problem in FineGrainedFP8 vs a Qwen3.5 MoE from_pretrained error; unrelated failures." }, { - "left": "issue:42831", - "right": "issue:45310", + "left": "issue:41628", + "right": "issue:43824", "accept": false, - "reason": "Different issues: FineGrainedFP8 accuracy regression vs Qwen3.5 MoE from_pretrained error." + "reason": "Both are import errors, but for different missing symbols and likely different export paths." }, { "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "A torch.split behavior report and a docstring issue are not the same underlying bug." + "reason": "Different issue types: runtime tensor indexing behavior vs a docstring correction." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "Different model/regression reports with no shared concrete failure mode." + "reason": "Separate bugs in different areas: FP8 accuracy vs Qwen3-VL-Moe loading." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer behavior bugs, but one is offset_mapping and the other is add_special_tokens/BOS-EOS handling; different code paths." + "reason": "Different tokenizer regressions: offset_mapping vs add_special_tokens/BOS-EOS handling." }, { "left": "issue:42757", "right": "issue:45278", "accept": false, - "reason": "The second is a broad collection of import errors, not the same single bug as the specific is_offline_mode import failure." + "reason": "Related only at a very broad 'import error' level; the specific missing symbol and scope are different." }, { "left": "issue:43232", "right": "issue:45593", "accept": false, - "reason": "Unrelated generation-state bug vs D-FINE auxiliary-loss behavior." + "reason": "Generation-state update bug vs loss-computation bug in D-FINE; unrelated." }, { "left": "issue:42886", "right": "issue:42994", "accept": false, - "reason": "Different persistence/loading failures: offline tokenizer cache loading vs quantized model save failure." + "reason": "Cache/offline tokenizer loading bug vs quantized model saving failure." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Tokenizer cache offline loading and wrong tokenizer decoder type are distinct failures." + "reason": "Both tokenizer-related, but one is offline cache loading and the other is a decoder-type mismatch in v5." }, { "left": "issue:22355", "right": "issue:28218", "accept": false, - "reason": "No shared concrete bug: missing module import vs tokenizer adding an extra space." + "reason": "Completely different areas: missing onnx module import vs tokenizer spacing behavior." }, { - "left": "issue:42831", - "right": "issue:45753", + "left": "issue:43756", + "right": "issue:44060", "accept": false, - "reason": "Accuracy regression in FP8 training is unrelated to an infinite-loop multimodal inference bug." + "reason": "Different model families and different logic errors: RoPE layer dropout vs tied-weights warning behavior." }, { - "left": "issue:43756", - "right": "issue:44060", + "left": "issue:42831", + "right": "issue:45753", "accept": false, - "reason": "Different model-specific implementation bugs: RoPE-layer dropping vs incorrect tied-weights warning behavior." + "reason": "Unrelated: FP8 accuracy issue vs Qwen3_5 infinite-loop generation on a specific image." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "Both are loading-path issues, but one is tokenizer mapping fallback and the other is AutoProcessor kwargs forwarding." + "reason": "Both are from_pretrained plumbing issues, but one is tokenizer mapping resolution and the other is cached_file kwargs forwarding." }, { "left": "issue:44671", "right": "issue:46132", "accept": false, - "reason": "Masked LM prediction regression and AttentionInterface.register behavior change are unrelated." + "reason": "CamemBERT MLM regression and AttentionInterface.register behavior change are different bugs." }, { "left": "issue:43576", "right": "issue:43784", "accept": false, - "reason": "A CLI/env command breakage and an import NameError in sentence-transformers are different issues." + "reason": "Different failures: env CLI breakage vs import-time NameError in sentence-transformers integration." }, { "left": "issue:43646", "right": "issue:46002", "accept": false, - "reason": "Different initialization regressions; no evidence they stem from the same underlying change." + "reason": "Custom model initialization breakage is not the same as NaNs in classification heads." }, { "left": "issue:43122", "right": "issue:44521", "accept": false, - "reason": "Different areas: tokenization drift across versions vs multimodal assistant_mask computation." + "reason": "Tokenizer output drift vs multimodal assistant-mask failure; not the same code path." }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "TF32 settings guidance and a torch.split return-value report are unrelated." + "reason": "TF32 settings warning/behavior issue vs a tensor split/indexer bug." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Different symptoms and systems: loading hang vs macOS bus error on a CLIP model." + "reason": "Different model/runtime failures: sentence-transformers load hang vs MacOS bus error with a CLIP model." } ] }, @@ -12893,8 +12902,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -12952,28 +12961,28 @@ "issue:44704|issue:45290", "issue:44188|issue:44373", "issue:42907|issue:44945", - "issue:41628|issue:45278", "issue:43232|issue:44743", "issue:43784|issue:45600", "issue:44261|issue:44373", "issue:39401|issue:44488", + "issue:41628|issue:45278", "issue:43976|issue:44945", "issue:43122|issue:44514", "issue:43408|issue:43475", "issue:42907|issue:43937", "issue:39401|issue:44521", - "issue:44625|issue:44829", "issue:43116|issue:45200", - "issue:44360|issue:44485", + "issue:44625|issue:44829", "issue:42907|issue:44964", + "issue:44360|issue:44485", "issue:44038|issue:44297", "issue:44038|issue:44295", "issue:28218|issue:30990", "issue:43122|issue:44779", - "issue:42907|issue:43452", "issue:36296|issue:43116", - "issue:43824|issue:44351", + "issue:42907|issue:43452", "issue:44038|issue:44222", + "issue:43824|issue:44351", "issue:43122|issue:43906", "issue:45216|issue:45478", "issue:43784|issue:45446", @@ -12981,220 +12990,226 @@ "issue:22355|issue:29159", "issue:43278|issue:44945" ], - "prepared_review_unit_hash": "4041b766c39146bfa13c1ed876411b810ce551139676803211307530ec2d3514", + "prepared_review_unit_hash": "1fba7439c5fa4cd99654846ff334152de6aee76fb043526134c75173af5fb137", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 36912, - "estimated_input_tokens": 9228, - "estimated_eval_tokens": 18712 + "serialized_chars": 36902, + "estimated_input_tokens": 9226, + "estimated_eval_tokens": 18708 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated transformer issues with only superficial lexical similarity. I found no soft pair that clearly refers to the same underlying bug or change, so all candidate links are rejected.", - "confidence": 0.91, - "canonical_issue_reason": null, + "summary": "No duplicate cluster emerges here: the items span unrelated tokenizer, import, model-loading, and training bugs. I rejected every soft edge as a true duplicate match; the only strong standalone issue is the open multimodal assistant_mask bug.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 44521 is the clearest current bug report: open, detailed, reproducible, and already attracting references, so it is the best anchor issue among this otherwise unrelated set.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Issue 44521 is the most actionable and representative standalone issue in the group; the others are either narrower, already closed, or clearly on different code paths.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44295", "right": "issue:44297", "accept": false, - "reason": "One is about reading a registered buffer; the other is about tokenizer_class metadata on save. Different failure modes and code paths." + "reason": "Different bugs: one is about reading a bufferized position_ids tensor, the other about tokenizer_config metadata written by save_pretrained." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "AutoProcessor kwargs plumbing to cached_file is unrelated to apply_chat_template crashing on tool-call messages." + "reason": "Different code paths: cached_file kwargs passthrough vs apply_chat_template crashing on tool-call assistant messages." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "torch.compile attention divergence and a wrong docstring for position_ids are unrelated." + "reason": "One is an attention-kernel branching issue under torch.compile; the other is just a docstring problem for position_ids." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals and wrong pipeline-parallel outputs are different bugs." - }, - { - "left": "issue:41628", - "right": "issue:45278", - "accept": false, - "reason": "Both mention import errors, but they concern different symbols and likely different regressions." + "reason": "Unrelated failures: dequantized model saving vs incorrect output under pipeline parallelism." }, { "left": "issue:43232", "right": "issue:44743", "accept": false, - "reason": "Both are generation-related, but one is sync_gpus kwargs handling and the other is cache-state reset logic." + "reason": "Both are generation-state bugs, but the symptoms and affected logic differ: sync_gpus kwargs handling vs recurrent state reset in modular_qwen3_5." }, { "left": "issue:43784", "right": "issue:45600", "accept": false, - "reason": "Sentence-transformers import NameError and auto_mappings references to removed configs are not the same issue." + "reason": "Different regressions: a missing nn import in sentence-transformers integration vs stale auto_mappings references to removed configs." }, { "left": "issue:44261", "right": "issue:44373", "accept": false, - "reason": "A precision bug in MLA layernorm config is unrelated to a docstring issue for position_ids." + "reason": "Model numerical precision/config issue vs a docstring mismatch; not the same underlying bug." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Wrong offset_mapping for Qwen3 tokenizer and failure to load cjvt/sleng-bert are separate tokenizer problems." + "reason": "Both touch tokenization, but they concern different models and different failures (offset_mapping vs load failure)." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "The second is a broad upgrade regression with many import errors; the first is a specific missing export. Too broad to treat as the same bug." }, { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Python version compatibility and pipeline parallel output correctness are unrelated." + "reason": "Different problem domains: Python-version compatibility vs incorrect outputs from pipeline parallelism." }, { "left": "issue:43122", "right": "issue:44514", "accept": false, - "reason": "General tokenizer regression across versions and a batched apply_chat_template crash are different code paths." + "reason": "Both involve tokenization/chat handling, but one is a version-to-version tokenization regression and the other is an apply_chat_template crash on batched multimodal input." }, { "left": "issue:43408", "right": "issue:43475", "accept": false, - "reason": "Wrong model-class warning and a missing attribute on the SAM3 vision encoder output are not the same bug." + "reason": "Same model family, but different concrete bugs: model-type warning vs missing fpn_position_embeddings attribute." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Dequantized save_pretrained failure and invalid GenerationConfig for GLM-5 are unrelated." + "reason": "Different Qwen/Mistral-related failures: saving dequantized models vs invalid GenerationConfig for GLM-5." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Offset mapping errors and zero assistant masks in multimodal apply_chat_template are distinct tokenizer/template issues." + "reason": "Both are tokenizer/template issues, but one is Qwen3 offset_mapping and the other is assistant_masks being all-zero for multimodal inputs." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Different tasks and symptoms: run_classification.py multi-label inference vs Gemma 4 fine-tuning token type defaults." }, { "left": "issue:44625", "right": "issue:44829", "accept": false, - "reason": "num_labels propagation in Qwen3.5 and flash_attention_3 training degeneration are unrelated." + "reason": "Different bugs: num_labels propagation in Qwen3.5 config vs degenerate training with flash_attention_3." }, { - "left": "issue:43116", - "right": "issue:45200", + "left": "issue:42907", + "right": "issue:44964", "accept": false, - "reason": "Empty multi-label example output and mm_token_type_ids defaults in Gemma 4 affect different features." + "reason": "Both are model-loading failures, but for different models and different causes; not the same code-path problem." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "A missing ReLU in DSA indexer is unrelated to GLM-5 RoPE implementation details." + "reason": "Completely different topics: DSA indexer activation vs GLM-5 RoPE implementation." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Qwen3-VL-Moe loading and tokenizer_class mismatch on save are different problems." + "reason": "One is a Qwen3-VL-Moe bug; the other is tokenizer_config metadata mismatch. Different root causes." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Model loading for Qwen3-VL-Moe does not match a position_ids buffer access error." + "reason": "Different failures: Qwen3-VL-Moe bug vs reading position_ids after buffer registration." }, { "left": "issue:28218", "right": "issue:30990", "accept": false, - "reason": "Tokenizer spacing after added token and Sentence Transformers hanging on load are unrelated." + "reason": "Tokenizer spacing issue vs Sentence Transformers loading hang; no shared underlying bug." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and behaviors." - }, - { - "left": "issue:42907", - "right": "issue:43452", - "accept": false, - "reason": "Saving dequantized models and gguf_file breakage in from_pretrained are different loading/saving paths." + "reason": "Both are tokenizer regressions, but they affect different models and different behaviors (general tokenization drift vs DeepSeek incorrect results)." }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Tensor parallel training bugs and empty multi-label example outputs are unrelated." + "reason": "Different training bugs in unrelated workflows: tensor parallel training vs example-script multi-label classification output." }, { - "left": "issue:43824", - "right": "issue:44351", + "left": "issue:42907", + "right": "issue:43452", "accept": false, - "reason": "Missing Qwen2_5_VLForConditionalGeneration and missing HybridCache are different missing-export/import errors." + "reason": "Different loading/saving paths and different model APIs; not the same concrete bug." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Qwen3-VL-Moe loading and FP8 save_pretrained for MoE are not the same concrete bug." + "reason": "Different MoE-related issues: Qwen3-VL-Moe behavior vs FP8 save_pretrained handling." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols from different packages and likely different fixes." }, { "left": "issue:43122", "right": "issue:43906", "accept": false, - "reason": "A version-to-version tokenization regression is not the same as an isolated reproduction of another issue." + "reason": "The second is just an isolated reproduction of another issue; the titles do not indicate the same concrete failure as the tokenization regression." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both involve Qwen3.5 and save/load regressions, but one is about save_pretrained checkpoints and the other about from_pretrained loading; too different to merge." + "reason": "Related Qwen3.5 family, but one is save_pretrained checkpoint correctness and the other is from_pretrained loading error; too different to merge as one bug." }, { "left": "issue:43784", "right": "issue:45446", "accept": false, - "reason": "A missing nn import in sentence-transformers is unrelated to a PyTorch version check in flex_attention." + "reason": "Sentence-transformers import failure vs a PyTorch version-check bug in flex_attention; unrelated code paths." }, { "left": "issue:42757", "right": "issue:44351", "accept": false, - "reason": "Missing is_offline_mode from huggingface_hub and missing HybridCache from transformers are separate import regressions." + "reason": "Both are import errors, but one is from huggingface_hub and the other is a transformers symbol export; different dependencies and fixes." }, { "left": "issue:22355", "right": "issue:29159", "accept": false, - "reason": "An absent transformers.onnx module and slow/fast tokenizer inconsistency are unrelated." + "reason": "Module import failure for transformers.onnx vs slow/fast tokenizer inconsistency; unrelated issues." }, { "left": "issue:43278", "right": "issue:44945", "accept": false, - "reason": "Embedding dtype drift in evaluate is unrelated to incorrect outputs under pipeline parallelism." + "reason": "Different runtime problems: dtype drift between train/eval and incorrect outputs under pipeline parallelism." } ] }, @@ -13213,8 +13228,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -13270,11 +13285,11 @@ "soft_pairs": [ "issue:42175|issue:44938", "issue:42673|issue:45072", + "issue:43992|issue:44704", "issue:43493|issue:45753", "issue:43066|issue:44779", - "issue:43992|issue:44704", - "issue:38617|issue:45278", "issue:22355|issue:29127", + "issue:38617|issue:45278", "issue:43611|issue:46002", "issue:43937|issue:44485", "issue:42222|issue:43441", @@ -13282,8 +13297,8 @@ "issue:43116|issue:43493", "issue:42371|issue:43874", "issue:44075|issue:44936", - "issue:45357|issue:45478", "issue:44038|issue:45478", + "issue:45357|issue:45478", "issue:43064|issue:43673", "issue:43116|issue:43825", "issue:45083|issue:45446", @@ -13298,208 +13313,208 @@ "issue:43450|issue:43986", "issue:43937|issue:44977" ], - "prepared_review_unit_hash": "8344b9a73ae8b746df9dc4f88fed3006f03efd0956eb2f8f13237cc0fa8a8e99", + "prepared_review_unit_hash": "3f164a69579ba898b28382b9a8803a66a051bc6ffe672328c413546302636b0b", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36033, - "estimated_input_tokens": 9009, - "estimated_eval_tokens": 18274 + "serialized_chars": 36040, + "estimated_input_tokens": 9010, + "estimated_eval_tokens": 18276 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated Transformers issues that only share broad themes like v5 regressions, tokenizers, or multimodal models. The soft-edge candidates are superficially similar but do not describe the same concrete bug/change, so none should be merged. No PRs are present.", - "confidence": 0.9, - "canonical_issue_reason": null, + "summary": "This cluster is a grab-bag of Transformers v5 regression reports across imports, tokenizers, model-specific bugs, docs, and training/runtime issues. The soft pairs mostly share broad vocabulary but not the same concrete code path, so I would reject all as duplicates; the broadest umbrella issue is the v5 upgrade/import-regression report.", + "confidence": 0.24, + "canonical_issue_reason": "issue:45278 is the broadest umbrella in the set: it describes multiple import failures after upgrading to Transformers 5.5.0, which is the closest match to the recurring v5 breakage pattern across the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the broadest user-facing regression report in the set (many import errors after upgrading from 4.57.0 to 5.5.0), so it's the most representative issue if one must be chosen.", + "best_issue_reason": "issue:45278 is the best representative because it is high-level and can subsume several v5 import/load regressions, unlike the narrower model- or tokenizer-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Packaging/backend dependency mismatch vs a Python 3.14 import/load failure; different failure modes and code paths." + "reason": "Different failures: missing TensorFlow backend in an install extra vs Python 3.14 load failure. Same release context, not the same bug." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM threading vs dtype mismatches in inference; unrelated models and bugs." + "reason": "VRAM leak in Qwen3ForCausalLM multi-threaded dataloading is unrelated to bfloat16 dtype mismatches in CI." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Missing embed_tokens.weight during UMT5 loading is a model-specific checkpoint issue, not the same as AutoProcessor kwargs forwarding." }, { "left": "issue:43493", "right": "issue:45753", "accept": false, - "reason": "SigLIP2 implementation discrepancy vs Qwen3.5 image-triggered infinite loop; both multimodal, but not the same defect." + "reason": "Both concern Qwen-related behavior, but one is a SigLIP2 implementation discrepancy and the other is a Qwen3_5 generation loop on a specific image." }, { "left": "issue:43066", "right": "issue:44779", "accept": false, - "reason": "Tokenizer decoder type regression vs Deepseek tokenizer output regression; same broad area, but distinct underlying problems." + "reason": "Wrong tokenizer decoder type in v5 is not the same as Deepseek tokenizer output regressions; different tokenizer code paths." }, { - "left": "issue:43992", - "right": "issue:44704", + "left": "issue:22355", + "right": "issue:29127", "accept": false, - "reason": "Missing encoder weights during load vs AutoProcessor kwargs forwarding; different components and symptoms." + "reason": "Import error for transformers.onnx is unrelated to a LayoutLMv3 box-validation error message." }, { "left": "issue:38617", "right": "issue:45278", "accept": false, - "reason": "Single missing import symbol vs broad upgrade-caused import failures; the latter is an umbrella report, not the same concrete bug." - }, - { - "left": "issue:22355", - "right": "issue:29127", - "accept": false, - "reason": "ONNX import failure vs a LayoutLMv3 error-message clarity issue; not the same bug." + "reason": "The former is a specific missing symbol import; the latter is a broad multi-import regression report. Not the same concrete failure." }, { "left": "issue:43611", "right": "issue:46002", "accept": false, - "reason": "base_model_prefix loading regression vs NaNs in classification heads; different code paths and outcomes." + "reason": "Base_model_prefix loading breakage is a model-loading compatibility bug, while NaNs in classification heads are an initialization/numerics issue." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "GenerationConfig validation failure vs GLM-5 RoPE implementation discussion; related model family, but not the same change or bug." + "reason": "GenerationConfig validation for GLM-5 is not the same as a RoPE implementation discussion." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "VitPose breakage vs Ministral-3 FlashAttention failure; separate models and distinct failure mechanisms." + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are distinct model-specific bugs." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python version compatibility break vs PyTorch version check in flex_attention; both version-related, but not the same issue." + "reason": "Python version compatibility regression is unrelated to a flex_attention version-check bug." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Example-script multi-label behavior vs SigLIP2 implementation discrepancy; unrelated functionality." + "reason": "Classification script returning empty multi-label results is unrelated to SigLIP2 HF/JAX discrepancy." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 API messaging vs missing image-patch helper in GLM46V; completely different subsystems." + "reason": "TF32 API warning changes and missing get_number_of_image_patches are different issues in different subsystems." }, { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "Optimizer arguments ignored vs train/evaluate interaction failure; different training bugs." + "reason": "SGD optimizer arguments not used is not the same as trainer.evaluate() failing after train()." }, { - "left": "issue:45357", + "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "save_pretrained saves wrong visual encoder keys vs from_pretrained error for Qwen3.5 Moe; same family, but different concrete paths." + "reason": "Qwen3-VL-Moe compatibility and Qwen3.5 Moe from_pretrained error are related only by model family, not the same code-path bug." }, { - "left": "issue:44038", + "left": "issue:45357", "right": "issue:45478", "accept": false, - "reason": "Qwen3-VL-Moe breakage vs Qwen3.5 Moe load error; too different in model/version and symptom." + "reason": "Incorrect visual encoder keys on save_pretrained is a save bug; from_pretrained loading failure is a separate load bug." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption vs missing GenerationMixin cache during chunked prefill; unrelated training vs generation bugs." + "reason": "Wrong optimizer state/params in FSDP2+PEFT training is unrelated to missing GenerationMixin cache in chunked prefill." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Classification example output bug vs pipeline() translation-support messaging; different user flows and defects." + "reason": "Empty multi-label classification results are not the same as the pipeline() translation-support error message problem." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Qwen3 Omni-MoE feature-length helper behavior vs flex_attention version gating; unrelated code paths." + "reason": "Feature-extract output length behavior in qwen3_omni_moe is unrelated to the flex_attention PyTorch version check." }, { "left": "issue:44380", "right": "issue:46077", "accept": false, - "reason": "GPT2 attention scaling ignored in SDPA/FlashAttention vs Gemma4 bidirectional mask construction; both attention-related but not the same bug." + "reason": "GPT2 attention scaling under SDPA/FlashAttention and Gemma4 bidirectional attention masking are different attention bugs." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "Invalid GenerationConfig vs stale warning for num_return_sequences; different generation-layer issues." + "reason": "GLM-5 generation config invalidity is not the same as a paged generate warning for num_return_sequences." }, { "left": "issue:44117", "right": "issue:45588", "accept": false, - "reason": "Tokenizer mapping fallback bug vs flash_attention AttributeError on s_aux=None; unrelated components." + "reason": "TOKENIZER_MAPPING_NAMES returning None is unrelated to flash_attention.py crashing on s_aux=None." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Generic tokenizer loading regression vs Qwen3 offset_mapping bug; the latter is much more specific." + "reason": "Tokenizer loading regression and Qwen3 offset_mapping bug are different tokenizer failures with different symptoms." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Qwen3 offset mapping bug vs Deepseek tokenizer regression; same broad area, different model-specific issues." + "reason": "Both are tokenizer bugs, but offset_mapping corruption and Deepseek incorrect results are not the same underlying defect." }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs still referencing pipeline() vs PyTorch version check in flex_attention; docs cleanup is not the same as a runtime version bug." + "reason": "Docs still referencing pipeline() and a flex_attention version check are unrelated." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "VRAM leak report vs isolated reproduction of another issue; not enough evidence they are the same underlying bug." + "reason": "Multi-thread VRAM leak and an isolated reproduction of another issue are not the same bug." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "init_empty_weights TypeError vs PretrainedConfig type-checking regression; both v5 load-time issues, but different failure points and causes." + "reason": "Unexpected _is_hf_initialized argument during init_empty_weights is unrelated to PretrainedConfig type-checking breakage." }, { "left": "issue:43450", "right": "issue:43986", "accept": false, - "reason": "Batched video processor shape bug vs missing-torchvision crash when loading video model; distinct defects." + "reason": "Batched video processor shape bug and missing torchvision crash on AutoProcessor loading are different failures." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "GenerationConfig invalid vs Qwen3.5 flash-attention generation failure; not the same bug." + "reason": "GenerationConfig validation for GLM-5 is unrelated to Qwen3.5 flash-attention generation issues." } ] }, @@ -13518,8 +13533,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -13578,45 +13593,45 @@ "issue:44117|issue:45081", "issue:45341|issue:45561", "issue:42222|issue:46178", - "issue:38617|issue:41628", "issue:44410|issue:45310", - "issue:42617|issue:45341", + "issue:38617|issue:41628", "issue:44829|issue:45216", + "issue:42617|issue:45341", "issue:42673|issue:43582", "issue:39401|issue:43421", "issue:42490|issue:42673", "issue:43408|issue:43742", "issue:42371|issue:44261", - "issue:44933|issue:45278", "issue:38617|issue:44220", "issue:44484|issue:45310", + "issue:44933|issue:45278", "issue:43937|issue:44496", - "issue:43066|issue:43122", "issue:42222|issue:45542", - "issue:44483|issue:45464", + "issue:43066|issue:43122", "issue:42371|issue:42831", + "issue:44483|issue:45464", "issue:42175|issue:42222", "issue:44075|issue:45656", - "issue:44410|issue:45478", "issue:43012|issue:43577", + "issue:44410|issue:45478", "issue:43673|issue:45341", "issue:43097|issue:44493", "issue:41628|issue:45071" ], - "prepared_review_unit_hash": "19149ae030576a3d84ed43d3e25e9c918c46fa463e232a17140962f27e8c9433", + "prepared_review_unit_hash": "a8b87e2b8d342371f640ea493fa1750468c62184483b4456d5789d26a8c231d2", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36142, + "serialized_chars": 36144, "estimated_input_tokens": 9036, "estimated_eval_tokens": 18328 }, @@ -13625,196 +13640,193 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated closed issues (imports, tokenizers, model loading, optimizer behavior, backend/version checks). Only one soft edge looks like a plausible duplicate: a specific import failure fitting under a broader post-upgrade import-regression report.", - "confidence": 0.39, - "canonical_issue_reason": "issue:45278 is the best umbrella representative because it explicitly describes the broader import-regression pattern after upgrading from 4.57.0 to 5.5.0.", + "summary": "This cluster is a loose collection of Transformers v5 regressions and API/import/model-loading complaints, but most items are not duplicates of one another. The soft edges are mostly false positives from broad lexical similarity rather than the same concrete bug.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster\u2019s recurring API/import regression theme, making it the best single issue anchor among otherwise heterogeneous reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most central and reusable summary of the cluster\u2019s closest shared theme: broad import errors after a version upgrade.", + "best_issue_reason": "issue:45278 best fits as a cluster representative because it captures the widest user-facing regression pattern (post-upgrade import failures) seen across several nearby issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44466", "right": "issue:46002", "accept": false, - "reason": "Different bugs: tied-weight serialization vs NaNs in classification heads." + "reason": "Different bugs: lm_head serialization inconsistency vs NaNs in classification heads after checkout/init." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "Separate dependency/version issues with different affected code paths." + "reason": "Unrelated: a PEFT version bump issue vs a PyTorch version check bug in flex_attention." }, { "left": "issue:44117", "right": "issue:45081", "accept": false, - "reason": "Both involve tokenizers, but one is mapping-name handling and the other is a Mistral regex/backend attribute crash." + "reason": "Both tokenizer-related, but one is a TOKENIZER_MAPPING_NAMES None-handling bug and the other is a Mistral regex patch crash." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "The second is a specific xdist race in testing_utils; the first is too vague to treat as the same bug." + "reason": "Both mention testing utils, but one is a vague general bug and the other is a specific pytest-xdist race on captured_info.txt." }, { "left": "issue:42222", "right": "issue:46178", "accept": false, - "reason": "VitPose model breakage and a tensorflow_text backend-mapping error are unrelated." - }, - { - "left": "issue:38617", - "right": "issue:41628", - "accept": false, - "reason": "Both are import errors, but for different symbols and different breakages." + "reason": "Different model/backend failures: vitpose breakage vs missing tensorflow_text in BACKENDS_MAPPING." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Different Qwen model families and different failures." + "reason": "Different Qwen variants and different failure modes: missing projections vs from_pretrained error." }, { - "left": "issue:42617", - "right": "issue:45341", + "left": "issue:38617", + "right": "issue:41628", "accept": false, - "reason": "Unrelated issues: a script runtime failure vs a testing_utils bug." + "reason": "Both are import errors, but they concern different missing symbols and different code paths." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "Training degeneration with flash_attention_3 is not the same as a save_pretrained regression." + "reason": "Training degeneration under flash_attention_3 is not the same as incorrect save_pretrained checkpoint serialization." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "Unrelated: a runtime script failure vs a testing_utils bug." }, { "left": "issue:42673", "right": "issue:43582", "accept": false, - "reason": "Both mention memory-ish symptoms, but one is a VRAM leak and the other is an Apple Silicon TypeError." + "reason": "Different issues entirely: VRAM leak in dataloader threads vs Apple Silicon TypeError in allocator warmup." }, { "left": "issue:39401", "right": "issue:43421", "accept": false, - "reason": "Offset-mapping bug and runtime special-token post-processor update are different tokenizer issues." + "reason": "Both tokenizer-related, but offset_mapping is a different problem from post-processor updates after runtime special-token changes." }, { "left": "issue:42490", "right": "issue:42673", "accept": false, - "reason": "Local-path import warning is unrelated to a Qwen3 VRAM leak." + "reason": "Local-path import warning is unrelated to the Qwen3ForCausalLM VRAM leak." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "Model-type warning and MobileLLM key error are different loading problems." + "reason": "Different model-loading problems: wrong model type warning vs a key error for MobileLLM." }, { "left": "issue:42371", "right": "issue:44261", "accept": false, - "reason": "TF32 API warning and missing rms_norm_eps are unrelated." - }, - { - "left": "issue:44933", - "right": "issue:45278", - "accept": true, - "reason": "45278 is a broad umbrella for post-upgrade import errors, and 44933 reads like one specific missing import under that same regression." + "reason": "TF32 behavior settings and RMS norm epsilon precision are unrelated issues." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "Different import/path vs feature-function failure; not the same underlying bug." + "reason": "Missing configuration_utils import and _torch_extract_fbank_features are different code-path failures." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "Different API/model-loading failures." + "reason": "A shard-size question is unrelated to the Qwen3.5 MoE from_pretrained error." }, { - "left": "issue:43937", - "right": "issue:44496", + "left": "issue:44933", + "right": "issue:45278", "accept": false, - "reason": "GenerationConfig validation and unrecognized-model config errors are distinct." + "reason": "Both concern imports, but one is a specific missing image_utils export and the other is a broad upgrade-wide import regression; not enough to treat as the same bug." }, { - "left": "issue:43066", - "right": "issue:43122", + "left": "issue:43937", + "right": "issue:44496", "accept": false, - "reason": "Both are v5 tokenizer regressions, but they describe different concrete symptoms and code paths." + "reason": "Different model/config failures: invalid GenerationConfig vs missing model_type in config.json." }, { "left": "issue:42222", "right": "issue:45542", "accept": false, - "reason": "VitPose model breakage is unrelated to missing TensorFlow in a backend install." + "reason": "Vitpose breakage is unrelated to the missing TensorFlow backend when only tensorboard is installed." }, { - "left": "issue:44483", - "right": "issue:45464", + "left": "issue:43066", + "right": "issue:43122", "accept": false, - "reason": "Both mention chat/completions, but one is a general request rejection and the other is a Qwen3.5 streaming failure." + "reason": "Related tokenization regression area, but one is decoder type selection and the other is broader tokenization differences across versions." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 behavior control and FineGrainedFP8 accuracy issues are unrelated." + "reason": "TF32 API usage and FineGrainedFP8 accuracy are distinct issues." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Different API failures: generic /v1/chat/completions rejection vs a Qwen3.5 streaming inference failure." }, { "left": "issue:42175", "right": "issue:42222", "accept": false, - "reason": "Backend-install packaging issue and VitPose model breakage are unrelated." + "reason": "Backend installation packaging issue vs vitpose model breakage are unrelated." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "Different optimizer bugs: SGD args ignored vs double optimizer step under deepspeed." + "reason": "Optimizer args being ignored is different from optimizer.step being called twice with deepspeed." }, { - "left": "issue:44410", - "right": "issue:45478", + "left": "issue:43012", + "right": "issue:43577", "accept": false, - "reason": "Different Qwen model variants and different failure modes." + "reason": "Both dtype-related, but one is a compile-time warning and the other is incorrect dtype propagation on load." }, { - "left": "issue:43012", - "right": "issue:43577", + "left": "issue:44410", + "right": "issue:45478", "accept": false, - "reason": "Compile-time warning about bfloat16 precision is not the same as Blip2 dtype fields remaining float32 on load." + "reason": "Both are Qwen-related, but they describe different model families and different from_pretrained failures." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "Generation cache/chunked_prefill bug is unrelated to a testing_utils issue." + "reason": "Chunked_prefill GenerationMixin cache regression is unrelated to the testing_utils bug." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Removal of a config field and unexpected position-id keys are different issues." + "reason": "Removed tie_embeddings_and_encoder_decoder and unexpected position-id keys are different regressions." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "Importing AutoImageProcessor and PreTrainedConfig type-checking are unrelated." + "reason": "Different regressions: missing AutoImageProcessor export vs PretrainedConfig type-checking breakage." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The cluster does look heterogeneous, and the one accepted soft edge (44933 \u2194 45278) is a plausible umbrella-duplicate relationship based on the issue titles. The other pair verdicts are cautious and supported by the visible symptoms in the packet." - }, - "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 }, @@ -13826,8 +13838,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -13897,21 +13909,21 @@ "issue:45003|issue:45341", "issue:43986|issue:45341", "issue:44117|issue:44987", + "issue:43986|issue:44484", "issue:43874|issue:44263", "issue:43408|issue:44368", - "issue:43986|issue:44484", - "issue:43531|issue:45341", "issue:43986|issue:45372", + "issue:43531|issue:45341", "issue:45440|issue:46172", "issue:44261|issue:44485", - "issue:38617|issue:42757", "issue:43097|issue:46050", - "issue:43097|issue:45588", "issue:43064|issue:44315", + "issue:43097|issue:45588", + "issue:38617|issue:42757", "issue:44062|issue:44117", "issue:44410|issue:45375", - "issue:38617|issue:44351", "issue:44038|issue:44410", + "issue:38617|issue:44351", "issue:43502|issue:43992", "issue:43097|issue:43873", "issue:44704|issue:45103", @@ -13919,250 +13931,256 @@ "issue:43824|issue:43986", "issue:43066|issue:43906" ], - "prepared_review_unit_hash": "c848b961e48e457497a59c0eddde0976c45db59e8c078cec8e0c03ac6cab2cda", + "prepared_review_unit_hash": "23b0b8c37900f10c7d24cca0aef44a14d3b005b23c4a9753ec0a824f772aa960", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 37, - "serialized_chars": 38785, - "estimated_input_tokens": 9697, - "estimated_eval_tokens": 19650 + "serialized_chars": 38784, + "estimated_input_tokens": 9696, + "estimated_eval_tokens": 19648 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a true duplicate cluster; it mixes unrelated Transformers regressions, import/export breakages, model-family-specific bugs, and warning/typing issues. I would not merge any of the soft edges.", - "confidence": 0.94, - "canonical_issue_reason": "Issue 43986 is the most self-contained, user-facing regression in the set and the clearest anchor for a loading-path bug, but the surrounding items are too heterogeneous to treat this as a real duplicate cluster.", + "summary": "Mostly unrelated transformer issues with superficial overlap in model names, loading paths, or dependency/import errors. None of the soft-edge pairs look like the same underlying bug or fix, so I would not merge any of them; issue 45071 is the best representative because it is a broad core-regression report.", + "confidence": 0.95, + "canonical_issue_reason": "Broad, cross-cutting regression in core config/type checking; more central than the model-specific and dependency-specific reports in this set.", "canonical_pr_reason": null, - "best_issue_reason": "43986 is the best representative only in the sense that it is concrete and actionable; it is still not a good umbrella for the rest of the set because most neighbors are unrelated.", + "best_issue_reason": "Impacts a core code path and is likely to be the most generally useful anchor for triage among these mostly narrow issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43874", "right": "issue:44261", "accept": false, - "reason": "Different failures in different GLM-related paths: image-patch counting vs RMS norm precision config." + "reason": "Different models and failures: missing image-patch method vs LayerNorm epsilon precision issue." }, { "left": "issue:44263", "right": "issue:44485", "accept": false, - "reason": "Same broad GLM area, but distinct bugs: DSA indexer split behavior vs RoPE implementation." + "reason": "Unrelated GLM-MoE indexing behavior vs GLM-5 RoPE implementation." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "Unrelated regressions: PretrainedConfig type-checking vs TensorFlow backend detection when only tensorboard is installed." + "reason": "Core config type-checking regression vs TensorFlow backend detection error." }, { "left": "issue:44263", "right": "issue:46172", "accept": false, - "reason": "Different model families and code paths; DSA indexer split handling is unrelated to DeepSeek transpose logic." + "reason": "Different bugs in different model families and tensor math paths." }, { "left": "issue:44360", "right": "issue:46172", "accept": false, - "reason": "Distinct model/code-path bugs: missing ReLU in DSA indexer vs DeepSeek tensor conversion/transpose." + "reason": "DSA ReLU complaint is unrelated to deepseek-v4 transpose behavior." }, { "left": "issue:43012", "right": "issue:43408", "accept": false, - "reason": "Both are warnings, but they come from different models and different mechanisms." + "reason": "Precision warning during bfloat16 compilation vs tokenizer/model type mismatch warning." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "Both mention FlashAttention, but one is a loading failure and the other is degenerate training; not the same bug." + "reason": "FlashAttention inference failure for Ministral-3 vs flash_attention_3 training degeneration." }, { "left": "issue:44484", "right": "issue:45478", "accept": false, - "reason": "A save_pretrained shard-size question is unrelated to Qwen3.5 MoE from_pretrained errors." + "reason": "save_pretrained shard-size question is unrelated to Qwen3.5-Moe loading failure." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Different Qwen model problems with different symptoms and affected layers; not one concrete bug." + "reason": "Qwen3-VL weight-shape mismatch vs Qwen3next missing projections." }, { "left": "issue:43335", "right": "issue:45071", "accept": false, - "reason": "Sparse-layer creation in SwitchTransformers is unrelated to a v5 PretrainedConfig type-checking regression." + "reason": "SwitchTransformers config edge case vs global PreTrainedConfig type-check regression." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Both are Qwen-related loading issues, but they affect different models and failure modes." + "reason": "Different Qwen multimodal loading problems with different symptoms and code paths." }, { "left": "issue:43673", "right": "issue:46123", "accept": false, - "reason": "Generation cache handling during chunked prefill is unrelated to MaskGenerationPipeline batch-finalization behavior." + "reason": "Generation cache/chunked prefill issue vs pipeline partial-batch result dropping." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "Configuration removal warning vs gguf_file/from_pretrained loading breakage; different code paths." + "reason": "Removed tie_embeddings API warning is unrelated to gguf_file loading breakage." }, { "left": "issue:45003", "right": "issue:45341", "accept": false, - "reason": "sys.modules access bug is unrelated to a testing_utils bug." + "reason": "sys.modules access bug in modeling_utils vs a testing_utils bug." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "AutoProcessor crashing without torchvision is not the same issue as the testing_utils bug." + "reason": "Missing torchvision crash vs an unrelated testing utility bug." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Tokenizer mapping None handling is a different loading bug than the physical-intelligence/fast regression." + "reason": "Tokenizer mapping None handling vs loading failure for a specific model repo." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Video-model processor dependency crash is unrelated to save_pretrained shard sizing." }, { "left": "issue:43874", "right": "issue:44263", "accept": false, - "reason": "Different GLM subcomponents and different failure signatures." + "reason": "Different GLM-related failures: image processor method missing vs tensor split return handling." }, { "left": "issue:43408", "right": "issue:44368", "accept": false, - "reason": "Model-type warning for sam3 is unrelated to a tie_word_embeddings warning in a Qwen fine-tuning setup." + "reason": "Model-type warning mismatch vs tie_word_embeddings warning during Qwen3.5 LoRA." }, { "left": "issue:43986", - "right": "issue:44484", + "right": "issue:45372", "accept": false, - "reason": "Optional-dependency crash in AutoProcessor is unrelated to a save_pretrained shard-size question." + "reason": "Different missing-dependency/import issues affecting different processor paths." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to the testing_utils bug." - }, - { - "left": "issue:43986", - "right": "issue:45372", - "accept": false, - "reason": "Missing torchvision crash is unrelated to the mistral_common ReasoningEffort import failure." + "reason": "Qwen3-MoE sliding window behavior is unrelated to a testing utility bug." }, { "left": "issue:45440", "right": "issue:46172", "accept": false, - "reason": "DeepSeek implementation divergence is a different problem from the comb.to(dtype).transpose(-1, -2) bug." + "reason": "DeepSeekV3 MoE divergence vs deepseek-v4 comb transpose bug." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "Missing rms_norm_eps precision issue is unrelated to the GLM-5 RoPE implementation bug." + "reason": "LayerNorm epsilon precision issue is unrelated to RoPE implementation." }, { - "left": "issue:38617", - "right": "issue:42757", + "left": "issue:43097", + "right": "issue:46050", "accept": false, - "reason": "Both are import errors, but they concern different missing symbols in different modules and are not the same bug." + "reason": "Removed API warning is unrelated to quantized model save/load corruption." }, { - "left": "issue:43097", - "right": "issue:46050", + "left": "issue:43064", + "right": "issue:44315", "accept": false, - "reason": "Tie-embeddings removal and quantized save/load corruption are unrelated regressions." + "reason": "Distributed optimizer-state bug vs Liger Kernel not applied in model_init." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "Removed config flag warning and flash_attention s_aux=None crash are different code-path issues." + "reason": "Tie-embeddings deprecation/removal warning is unrelated to flash_attention.py crash on sink-less models." }, { - "left": "issue:43064", - "right": "issue:44315", + "left": "issue:38617", + "right": "issue:42757", "accept": false, - "reason": "Distributed optimizer-state corruption is unrelated to Liger Kernel not being applied with model_init." + "reason": "Both are import errors, but for different missing symbols from different modules." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "AddedToken constructor argument collision is unrelated to tokenizer mapping returning None." + "reason": "AddedToken keyword collision is unrelated to tokenizer mapping None handling." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Missing attention projections in qwen3next is unrelated to a missing deepstack_visual_indexes config field." + "reason": "Different Qwen-family issues: missing projections vs missing config field being dropped." }, { - "left": "issue:38617", - "right": "issue:44351", + "left": "issue:44038", + "right": "issue:44410", "accept": false, - "reason": "Different import failures: layer_type_validation vs HybridCache, with no shared code path." + "reason": "Qwen3-VL-Moe loading bug is unrelated to qwen3next missing attention projections." }, { - "left": "issue:44038", - "right": "issue:44410", + "left": "issue:38617", + "right": "issue:44351", "accept": false, - "reason": "Both are Qwen-family regressions, but they affect different models and different architectural components." + "reason": "Distinct import errors from different modules and symbols." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "Ignoring local_files_only is unrelated to a missing embed_tokens.weight during UMT5Encoder loading." + "reason": "local_files_only still making requests vs missing embed_tokens weight on UMT5Encoder." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "Config/removal behavior is unrelated to quantization offloading not working as expected." + "reason": "Embedding-tying removal warning is unrelated to quantization offloading behavior." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "kwargs propagation to cached_file is unrelated to the docstring annotation crash." + "reason": "AutoProcessor kwargs passthrough bug vs auto_docstring annotation crash." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "AutoProcessor without torchvision and TOKENIZER_MAPPING_NAMES returning None are separate loading problems." + "reason": "Missing torchvision processor crash vs tokenizer mapping assumption bug." }, { "left": "issue:43824", "right": "issue:43986", "accept": false, - "reason": "Missing Qwen2_5_VLForConditionalGeneration export and missing torchvision in AutoProcessor are different failures." + "reason": "Qwen2.5-VL import error is unrelated to video-model loading without torchvision." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Tokenizer decoder type issue is unrelated to the reproduced issue referenced by 38071." } ] }, @@ -14181,8 +14199,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -14243,8 +14261,8 @@ "issue:43012|issue:43550", "issue:43012|issue:44368", "issue:42907|issue:44655", - "issue:42907|issue:45753", "issue:43352|issue:44380", + "issue:42907|issue:45753", "issue:44162|issue:46172", "issue:43352|issue:43976", "issue:44380|issue:46132", @@ -14253,14 +14271,14 @@ "issue:43064|issue:44380", "issue:44483|issue:44514", "issue:43208|issue:43526", - "issue:36683|issue:43582", "issue:44222|issue:45478", + "issue:36683|issue:43582", "issue:43012|issue:44683", - "issue:42757|issue:43824", "issue:43618|issue:43761", "issue:42490|issue:43012", - "issue:45698|issue:46172", + "issue:42757|issue:43824", "issue:43618|issue:44380", + "issue:45698|issue:46172", "issue:38617|issue:43502", "issue:44038|issue:44230", "issue:44483|issue:45290", @@ -14269,226 +14287,226 @@ "issue:43299|issue:43867", "issue:36683|issue:43475" ], - "prepared_review_unit_hash": "058e0df70daac2a2ba0e06e81280cf75cb8f49d484fbcd355ede9089ebf02348", + "prepared_review_unit_hash": "629f81c964ea2c00b47db2b4ed31927c8aec16bfeedb159c16d59a7dad6f42f8", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 36948, - "estimated_input_tokens": 9237, - "estimated_eval_tokens": 18730 + "serialized_chars": 36944, + "estimated_input_tokens": 9236, + "estimated_eval_tokens": 18728 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is largely heterogeneous: the paired items share broad Transformers themes (loading, attention, FP8, chat templates, etc.) but not the same concrete defect. I would not merge any of the soft pairs as duplicates.", - "confidence": 0.86, - "canonical_issue_reason": null, + "summary": "This cluster is mostly a grab-bag of unrelated Transformers bug reports around v5 regressions, loading/saving, imports, and model-specific failures. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates.", + "confidence": 0.56, + "canonical_issue_reason": "issue:45698 is the broadest and most active open report here, centered on from_pretrained/save_pretrained custom-module loading behavior, which is the closest thing to a representative theme across the mostly serialization/loading-adjacent items in this set.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45698 is the best global issue candidate because it is open, has substantial discussion and inbound references, and covers a concrete user-facing loading/saving regression rather than a model-specific edge case.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44938", "right": "issue:46178", "accept": false, - "reason": "Different failures: Python 3.14 import/load issue vs invalid backend mapping for tensorflow_text." + "reason": "Python 3.14 load failure and invalid backend-name error are unrelated import/dispatch problems." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training bugs and Bamba SDPA/torch.compile failure are unrelated code paths." + "reason": "xLSTM training bugs and Bamba compile/SDPA failure affect different models and code paths." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing embed token weight on load is not the same as save_pretrained shard-size behavior." + "reason": "Missing embed_tokens.weight on from_pretrained is unrelated to a max_shard_size question." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Pipeline task warning text and saving Pipeline objects are distinct bugs." + "reason": "Incorrect pipeline task error messaging is different from saving Pipeline objects." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "BF16 compile warning is unrelated to Bamba SDPA compile failure." + "reason": "A bf16 compile warning is not the same as a model-specific torch.compile SDPA failure." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "PyTorch precision warning and Qwen3.5 LoRA warning come from different mechanisms." + "reason": "General bf16 compile warning and tie_word_embeddings warning are different symptoms with different causes." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Dequantized model save failure is not the same as generic Pipeline save_pretrained support." + "reason": "Both are save-related, but dequantized model saving and Pipeline.save_pretrained are distinct code paths." }, { - "left": "issue:42907", - "right": "issue:45753", + "left": "issue:43352", + "right": "issue:44380", "accept": false, - "reason": "Save/dequantization bug is unrelated to Qwen3.5 image-specific infinite loop." + "reason": "Nemotron Flash Attention 2 support and GPT2 attention scaling under SDPA/FA are different bugs." }, { - "left": "issue:43352", - "right": "issue:44380", + "left": "issue:42907", + "right": "issue:45753", "accept": false, - "reason": "FlashAttention support error for Nemotron is different from GPT2 attention scaling being ignored." + "reason": "A dequantized-save failure is unrelated to a Qwen3.5 image-induced infinite loop." }, { "left": "issue:44162", "right": "issue:46172", "accept": false, - "reason": "ESM2 breakage and deepseek-v4 tensor transpose issue are unrelated models and bugs." + "reason": "ESM2 loading breakage and a DeepSeek tensor-op bug are different model-specific issues." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "FlashAttention support error is unrelated to Python version compatibility." + "reason": "Flash Attention support warning and Python version incompatibility are unrelated." }, { "left": "issue:44380", "right": "issue:46132", "accept": false, - "reason": "Attention scaling regression and AttentionInterface.register behavior change are separate issues." + "reason": "Attention scaling being ignored is not the same as AttentionInterface.register changing registered function behavior." }, { "left": "issue:43792", "right": "issue:46123", "accept": false, - "reason": "Whisper runtime failure and MaskGenerationPipeline partial-batch dropping are different pipelines and symptoms." + "reason": "Whisper runtime failure and partial-batch pipeline result dropping are different pipeline bugs." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "Attention backend scaling bug is unrelated to zero-loss classification behavior." + "reason": "Attention backend scaling and degenerate zero loss classification are unrelated regressions." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP/PEFT optimizer-state corruption is not the same as attention scaling being ignored." + "reason": "FSDP2 optimizer-state corruption and GPT2 attention-scaling bugs do not share a concrete code path." }, { "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "Chat-completions API acceptance and Qwen2.5-VL chat-template crash are separate request-handling bugs." + "reason": "Chat-completions request rejection and chat-template batching crash are both chat-related but distinct failures." }, { "left": "issue:43208", "right": "issue:43526", "accept": false, - "reason": "xLSTM training bugs do not match BeitImageProcessorFast label reduction behavior." - }, - { - "left": "issue:36683", - "right": "issue:43582", - "accept": false, - "reason": "Gemma3Config missing vocab_size and Apple Silicon caching_allocator_warmup TypeError are unrelated." + "reason": "xLSTM training issues and BeitImageProcessorFast reduce_labels behavior are unrelated." }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "Both involve MoE/FP8, but one is save_pretrained and the other from_pretrained loading; not the same concrete bug." + "reason": "FP8 save_pretrained failure and Qwen3.5 MoE from_pretrained failure are different operations and symptoms." }, { - "left": "issue:43012", - "right": "issue:44683", + "left": "issue:36683", + "right": "issue:43582", "accept": false, - "reason": "BF16 warning is unrelated to compiled flex_attention failure on torch>=2.9." + "reason": "Missing Gemma3Config vocab_size and Apple Silicon TypeError are unrelated." }, { - "left": "issue:42757", - "right": "issue:43824", + "left": "issue:43012", + "right": "issue:44683", "accept": false, - "reason": "Missing is_offline_mode import and missing Qwen2.5-VL symbol are different import regressions." + "reason": "A precision warning is not the same as compiled flex_attention failing on torch >= 2.9." }, { "left": "issue:43618", "right": "issue:43761", "accept": false, - "reason": "CLIPOutput attentions assignment and CLIPVisionModel hidden_states regression are different CLIP bugs." + "reason": "Both affect CLIP, but one drops attentions while the other drops hidden_states; not the same bug." }, { "left": "issue:42490", "right": "issue:43012", "accept": false, - "reason": "Trailing-slash local model warning and bf16 compile warning are unrelated." + "reason": "Local-model-path slash warning and bf16 compile warning are unrelated." }, { - "left": "issue:45698", - "right": "issue:46172", + "left": "issue:42757", + "right": "issue:43824", "accept": false, - "reason": "Wrong custom module after save/load and deepseek-v4 tensor transpose are different serialization vs model-shape issues." + "reason": "Both are import errors, but they involve different missing symbols and different failure modes." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "CLIP attentions regression and GPT2 attention scaling regression affect different models and code paths." + "reason": "CLIP attentions regression and GPT2 attention-scaling regression are separate model-specific bugs." + }, + { + "left": "issue:45698", + "right": "issue:46172", + "accept": false, + "reason": "Custom module loading after save_pretrained and a DeepSeek tensor transform error are unrelated." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "Missing configuration_utils import and local_files_only network requests are unrelated." + "reason": "ImportError for layer_type_validation and API calls despite local_files_only=True are different problems." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Qwen3-VL-Moe loading bug and fp8 inference support issue are different failures." + "reason": "Qwen3-VL-Moe loading breakage and FP8 inference support are related only by model family, not by the same bug." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "Generic /v1/chat/completions compatibility issue and assistant tool-call templating crash are not the same bug." + "reason": "Chat API request handling and apply_chat_template tool-call crash are different layers of the stack." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "Smollm3 RoPE layer mismatch and GLM-5 RoPE implementation discussion are related only by topic, not by defect." + "reason": "Smollm3 RoPE-layer omission and GLM-5 RoPE implementation discussion are not the same underlying issue." }, { "left": "issue:43066", "right": "issue:46209", "accept": false, - "reason": "Tokenizer decoder type bug in v5 and FP8 inference failure on B200 are different subsystems." + "reason": "Wrong tokenizer decoder type and FP8 inference on B200 are unrelated." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Qwen3VL MoE loading regression and sorted-state_dict load error are different load-path failures." + "reason": "Qwen3VL-Moe loading regression and sorted-state_dict load error are distinct loading failures." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Gemma3Config vocab_size AttributeError and SAM3 vision encoder missing fpn_position_embeddings are unrelated model classes." + "reason": "Gemma3Config vocab_size missing and SAM3 output missing fpn_position_embeddings are unrelated attribute errors." } ] }, @@ -14507,8 +14525,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -14565,9 +14583,9 @@ "issue:45276|issue:46077", "issue:42994|issue:44655", "issue:43901|issue:44655", - "issue:36010|issue:42757", "issue:43299|issue:45753", "issue:44077|issue:44380", + "issue:36010|issue:42757", "issue:36683|issue:44077", "issue:44625|issue:46077", "issue:43611|issue:44655", @@ -14575,8 +14593,8 @@ "issue:43867|issue:45216", "issue:46077|issue:46082", "issue:43976|issue:44938", - "issue:43352|issue:45593", "issue:36683|issue:43525", + "issue:43352|issue:45593", "issue:43618|issue:45216", "issue:36683|issue:44242", "issue:43352|issue:43931", @@ -14598,20 +14616,20 @@ "issue:43010|issue:43976", "issue:36296|issue:43575" ], - "prepared_review_unit_hash": "c1f1c46e5772532056efcccff6ee6e44ddd86d2edb6e1408b3277bb24c490b33", + "prepared_review_unit_hash": "c6c7548030a3345045d0e95195321b04f619e71b7fc26e4592b9c6e0c4a20acf", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 37688, + "serialized_chars": 37687, "estimated_input_tokens": 9422, "estimated_eval_tokens": 19100 }, @@ -14620,240 +14638,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly heterogeneous, with only a few clear duplicate/near-duplicate issue pairs. The strongest matches are the Qwen3.5 Moe from_pretrained reports, the import-time slowdown reports, and the Qwen3.5 visual-config/save_pretrained serialization reports.", - "confidence": 0.71, - "canonical_issue_reason": "No single cluster-wide canonical issue fits this mixed set; the best umbrella issue is 44246 for the import-time slowdown subcluster, while 45310/45478 is the clearest exact-duplicate subcluster.", + "summary": "This cluster is mostly a loose collection of transformer v5 regression reports across import, config, loading, and save/pretrained paths. The only plausible duplicate edge is the import-time slowdown pair; the rest are distinct model- or bug-specific issues and should stay separate.", + "confidence": 0.64, + "canonical_issue_reason": "issue:45278 is the broadest and most representative issue in the cluster: it captures the general post-upgrade import-regression theme ('many import errors after update from 4.57.0 to 5.5.0') that several other items orbit around.", "canonical_pr_reason": null, - "best_issue_reason": "44246 is the best global issue anchor because it is a broad, user-visible regression with a plausible duplicate in 41669 and reads like the top-level symptom for that subcluster.", + "best_issue_reason": "issue:45278 is the best single issue to represent the cluster because it is broad, actionable, and closer to the shared version-regression theme than the many model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45276", "right": "issue:46077", "accept": false, - "reason": "Different Gemma4 bugs: token embedding resize propagation vs bidirectional attention mask construction." + "reason": "Different Gemma4 bugs: token embedding resize propagation vs causal attention-mask construction." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Both are save-related, but one is quantized model saving and the other is pipeline save_pretrained behavior/docs; not the same bug." + "reason": "One is quantized-model saving failure; the other is pipeline save_pretrained serialization." }, { "left": "issue:43901", "right": "issue:44655", "accept": false, - "reason": "Docs mismatch for TextClassificationPipeline is unrelated to pipeline object serialization failure." - }, - { - "left": "issue:36010", - "right": "issue:42757", - "accept": false, - "reason": "Both are import errors, but for different symbols and different upstream modules." + "reason": "Docs mismatch about return_all_scores is not the same bug as pipeline save_pretrained failing." }, { "left": "issue:43299", "right": "issue:45753", "accept": false, - "reason": "Qwen3VL model loading failure and Qwen3_5 infinite loop are different code-path problems." + "reason": "Different Qwen3VL issues: MoE model loading regression vs an image-triggered infinite loop." }, { "left": "issue:44077", "right": "issue:44380", "accept": false, - "reason": "Unrelated subsystems: patchtsmixer post_init validation vs GPT2 attention scaling under SDPA/FlashAttention." + "reason": "Unrelated: patchtsmixer post_init validation versus GPT2 attention scaling in SDPA/FlashAttention." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols from different packages and different fixes." }, { "left": "issue:36683", "right": "issue:44077", "accept": false, - "reason": "Missing config attribute in Gemma3 is unrelated to patchtsmixer post_init constraints." + "reason": "Gemma3 missing vocab_size and patchtsmixer post_init are unrelated config issues." }, { "left": "issue:44625", "right": "issue:46077", "accept": false, - "reason": "num_labels propagation bug is unrelated to Gemma4 bidirectional attention masks." + "reason": "Qwen3.5 num_labels propagation and Gemma4 attention-mask behavior are different model bugs." }, { "left": "issue:43611", "right": "issue:44655", "accept": false, - "reason": "Base_model_prefix loading regression is not the same as pipeline save_pretrained failure." + "reason": "Base_model_prefix loading regression and pipeline save_pretrained are different code paths." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "FlashAttention-2 support error and invalid GenerationConfig are unrelated failures." + "reason": "Flash Attention 2 support gate and invalid GenerationConfig are unrelated failures." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "Both involve model persistence/loading, but one is generic state_dict ordering and the other is a Qwen3.5 save_pretrained regression." + "reason": "State_dict ordering load error versus incorrect Qwen3.5 checkpoint saving are different serialization bugs." }, { "left": "issue:46077", "right": "issue:46082", "accept": false, - "reason": "Different Gemma4 issues: attention mask mode vs LlamaConfig head_dim validation." + "reason": "Gemma4 attention-mask bug and LlamaConfig head_dim validation are unrelated." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Both are Python-version compatibility issues, but they target different Python releases and different failures." + "reason": "Both mention Python compatibility, but they are different version-specific failures with different causes." }, { - "left": "issue:43352", - "right": "issue:45593", + "left": "issue:36683", + "right": "issue:43525", "accept": false, - "reason": "FlashAttention support and D-FINE auxiliary-loss behavior are unrelated." + "reason": "Different missing config attributes in different model families (Gemma3 vs Llama4)." }, { - "left": "issue:36683", - "right": "issue:43525", + "left": "issue:43352", + "right": "issue:45593", "accept": false, - "reason": "Different missing config attributes in different model families; not the same bug." + "reason": "Flash Attention support for Nemotron and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "CLIPOutput attentions regression is unrelated to Qwen3.5 checkpoint saving." + "reason": "CLIPOutput attentions assignment and Qwen3.5 save_pretrained corruption are separate bugs." }, { "left": "issue:36683", "right": "issue:44242", "accept": false, - "reason": "Gemma3 vocab_size attr error and load-balancing-loss gating are different model bugs." + "reason": "Missing Gemma3Config vocab_size is unrelated to MoE load-balancing loss gating." }, { "left": "issue:43352", "right": "issue:43931", "accept": false, - "reason": "Unsupported FlashAttention and Qwen3-VL weight-shape mismatch are unrelated." + "reason": "Nemotron Flash Attention support and Qwen3-VL weight-shape mismatch are different model-loading problems." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "FSDP/optimizer state corruption and a cache strategy typo do not share the same underlying bug." + "reason": "Distributed optimizer-state corruption and a cache-strategy typo are not the same issue." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "no_grad decoration for cache update is unrelated to broad import errors after upgrading to 5.5.0." + "reason": "Cache update no_grad behavior and import errors after upgrade are different parts of the stack." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "Different models and failure modes: UMT5 missing weight vs Qwen3.5 Moe from_pretrained regression." + "reason": "UMT5 missing embed_tokens.weight is unrelated to Qwen3.5 MoE from_pretrained failures." }, { "left": "issue:43867", "right": "issue:46202", "accept": false, - "reason": "Generic state_dict load error and Gemma pipeline roundtrip failure on a specific token are not the same bug." + "reason": "State_dict sorting load error and Gemma pipeline roundtrip on \u2581 are different serialization/pathology reports." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Compiled-file import errors and a Python 3.13 torch.jit parsing problem are different root causes." + "reason": "Compiled Python file execution errors are distinct from the Python 3.13 decorator/comment parsing regression." }, { "left": "issue:43867", "right": "issue:45375", "accept": false, - "reason": "Generic load failure via state_dict ordering is not the same as Qwen3_5 vision config field loss." + "reason": "State_dict load failure and missing Qwen3.5 vision config field are different issues." }, { "left": "issue:41669", "right": "issue:44246", "accept": true, - "reason": "These look like the same import-time slowdown regression: one is the concrete diagnosis, the other is the user-facing symptom." - }, - { - "left": "issue:36683", - "right": "issue:43064", - "accept": false, - "reason": "Missing Gemma3 config attribute is unrelated to the FSDP+PEFT optimizer-state bug." + "reason": "These look like the same import-time slowdown problem: 41669 identifies the import-star root cause and 44246 reports the same slow-import symptom." }, { "left": "issue:44938", "right": "issue:45071", "accept": false, - "reason": "Different breakages: Python 3.14 load failure vs PretrainedConfig type-checking regression." + "reason": "Python 3.14 load failure and PretrainedConfig type-checking breakage are different regressions." }, { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "Test flakiness from missing set_seed and pytest-xdist file races are not the same issue." + "reason": "SAM HQ test flakiness from missing set_seed and an xdist file race are different test issues." }, { "left": "issue:43010", "right": "issue:45850", "accept": false, - "reason": "Cache update no_grad bug is much more specific than the broad 'too many models' complaint." + "reason": "A cache update behavior change and a vague 'breaks too many models' complaint are not the same concrete bug." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "Generic load/state_dict issues are not the same as Qwen3.5 visual-encoder key serialization." + "reason": "State_dict sorting load error and Qwen3.5 visual-encoder save key corruption are different save/load regressions." }, { "left": "issue:42490", "right": "issue:44075", "accept": false, - "reason": "Local-path import warning and SGD arguments being ignored are unrelated." + "reason": "Local-path slash import warning and SGD args not being used are unrelated." }, { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Both are from_pretrained/load errors, but for different model families and different failure modes." + "reason": "UMT5Encoder weight-missing and Qwen3.5 MoE from_pretrained errors are different model-specific failures." }, { "left": "issue:43618", "right": "issue:44077", "accept": false, - "reason": "CLIPOutput attentions regression is unrelated to patchtsmixer post_init gating." + "reason": "CLIPOutput attentions regression and patchtsmixer post_init gating are unrelated." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "Cache no_grad behavior and Python version compatibility are unrelated." + "reason": "Cache no_grad behavior and Python-version compatibility failure are unrelated." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Tensor-parallel training bug and Qwen2-57B-A14B tp OOM are different issues despite both touching TP." - }, - { - "left": "issue:45357", - "right": "issue:45375", - "accept": true, - "reason": "Both point to the same Qwen3.5 vision/save_pretrained serialization problem: missing vision config fields causing incorrect saved keys." - }, - { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "Near-identical titles, same model family, same from_pretrained failure, and same version window; these are clearly the same bug." + "reason": "Both mention tensor parallelism, but one is a generic training bug and the other is a specific TP load OOM; not the same concrete issue." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary stays within the packet evidence and the soft-edge calls are conservative. The accepted duplicate pairs are well grounded by the titles/body themes, and the rejected pairs avoid overclaiming similarity across different model families or failure modes. The umbrella/canonical language is cautious enough for this mixed cluster." + "feedback": "The summary stays within the packet evidence: it frames the cluster conservatively as a mixed set of v5 regression reports and picks 45278 as the broadest representative issue without overclaiming. The only accepted soft edge, 41669\u219444246, is well-grounded by the shared import-time slowdown symptom. The rest of the pairwise rejections are cautious and supported by the titles, with no obvious overstatement." }, "evaluator_used": true, "retried": false, @@ -14869,8 +14869,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -14926,8 +14926,8 @@ "soft_pairs": [ "issue:28282|issue:33453", "issue:43618|issue:44857", - "issue:44360|issue:45440", "issue:43618|issue:44977", + "issue:44360|issue:45440", "issue:42890|issue:44964", "issue:43502|issue:44062", "issue:42890|issue:44112", @@ -14935,10 +14935,10 @@ "issue:43723|issue:44556", "issue:43576|issue:46132", "issue:44387|issue:44556", - "issue:42548|issue:44351", "issue:44222|issue:45125", - "issue:42890|issue:45341", "issue:43010|issue:43065", + "issue:42890|issue:45341", + "issue:42548|issue:44351", "issue:43502|issue:44485", "issue:44355|issue:44749", "issue:45125|issue:45478", @@ -14949,10 +14949,10 @@ "issue:43502|issue:44222", "issue:43299|issue:44556", "issue:42831|issue:43874", - "issue:43827|issue:44961", "issue:43502|issue:44961", - "issue:44355|issue:45071", + "issue:43827|issue:44961", "issue:30333|issue:36296", + "issue:44355|issue:45071", "issue:44315|issue:44360", "issue:44355|issue:44655", "issue:33453|issue:34689", @@ -14963,269 +14963,272 @@ "issue:43576|issue:43824", "issue:41628|issue:42548" ], - "prepared_review_unit_hash": "16a64a91c5aeefca2e53ee5ca5c5ad719362c1a3fac671b323e6b585bad220dd", + "prepared_review_unit_hash": "85e6b0ca235e677902de5001c0407305fe8c2d5f8e550c6447cf77e076b12955", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 38, - "serialized_chars": 38482, - "estimated_input_tokens": 9621, - "estimated_eval_tokens": 19498 + "serialized_chars": 38478, + "estimated_input_tokens": 9620, + "estimated_eval_tokens": 19496 }, "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 issues, many of them broad v5 regressions or loading/import problems. The soft pairs share superficial wording or subsystem overlap, but not the same concrete bug, so none should be merged as duplicates.", - "confidence": 0.88, - "canonical_issue_reason": "issue:33453 is the most representative standalone bug report in the set: it is a broad tokenizer-loading regression with high discussion activity and overlaps most with other loading-related reports.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers issues that only share broad v5-era breakage themes (imports, loading, CLI, quantization, tensor parallelism). I found one plausible duplicate edge in the Qwen3.5 MoE from_pretrained/tensor-parallel-plan reports; the rest are too coarse or describe different code paths.", + "confidence": 0.62, + "canonical_issue_reason": "issue:44556 is the clearest representative of the recurring v5 loading/regression theme: a concrete upgrade-related failure to reload checkpoints after moving from v4.57 to v5.2/5.3.", "canonical_pr_reason": null, - "best_issue_reason": "issue:33453 is the best single issue candidate to anchor the cluster because it is broad, concrete, and central to the loading-regression theme seen here.", + "best_issue_reason": "issue:44556 is the best single issue to stand in for the cluster because it is concrete, user-facing, and directly captures the v4\u2192v5 model-loading breakage pattern seen across several nearby reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "Both mention import/loading failures, but one is a missing PyTorch dependency error and the other is a tokenizer regression; different underlying bugs." + "reason": "Different failures: missing PyTorch dependency vs tokenizer-loading regression." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "Different failures in different model paths; one is CLIP attentions handling, the other is an AMP/CUDA crash in LwDetrImageLoss." + "reason": "Unrelated model/output bug vs FP16 AMP crash." }, { - "left": "issue:44360", - "right": "issue:45440", + "left": "issue:43618", + "right": "issue:44977", "accept": false, - "reason": "Unrelated topics: a discussion about a DSA indexer ReLU vs a DeepSeekV3MoE implementation divergence." + "reason": "Different code paths; CLIPOutput attention assignment is not the same as Qwen3.5 flash-attention generation." }, { - "left": "issue:43618", - "right": "issue:44977", + "left": "issue:44360", + "right": "issue:45440", "accept": false, - "reason": "Only broad generation/runtime context overlaps; CLIP attentions assignment and Qwen flash-attention generation are separate bugs." + "reason": "Different models and failures; DSA ReLU discussion is not the same as DeepSeekV3MoE divergence." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "One is a flaky test due to missing set_seed, the other is a Phi-4 multimodal model-loading regression." + "reason": "Different test stability issue vs Phi-4 multimodal load failure." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "Both involve loading-related code paths, but local_files_only network leakage and AddedToken argument duplication are different defects." + "reason": "Different bugs: offline/local-files behavior vs AddedToken constructor argument collision." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Both are test failures, but they concern different tests and causes: missing set_seed vs a stale device override assertion." + "reason": "CI flakiness in a test file is unrelated to GraniteSpeech device override behavior." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "A flaky SamHQ test and a CI break from removed auto_mappings entries are not the same bug." + "reason": "Different scope: seed-dependent test flakiness vs auto_mappings referencing removed configs." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Both are version-related loading complaints, but one is tokenizer loading and the other is checkpoint reload compatibility after upgrading to v5." + "reason": "Both are loading-related, but tokenizer.from_pretrained breakage is a different symptom and likely a different fix than checkpoint reload compatibility." }, { "left": "issue:43576", "right": "issue:46132", "accept": false, - "reason": "Both touch Transformers internals, but one is an env command regression and the other is changed behavior in AttentionInterface.register." + "reason": "CLI/environment command breakage is unrelated to AttentionInterface.register behavior change." }, { "left": "issue:44387", "right": "issue:44556", "accept": false, - "reason": "One is an int4 quantization memory regression; the other is a model reload compatibility issue." + "reason": "Memory/OOM under int4 quantization is a different failure than checkpoint reload compatibility." }, { - "left": "issue:42548", - "right": "issue:44351", + "left": "issue:44222", + "right": "issue:45125", "accept": false, - "reason": "Both are import errors, but they refer to different symbols and likely different API breakages." + "reason": "FP8 save_pretrained support and missing tensor-parallel plan are different problems." }, { - "left": "issue:44222", - "right": "issue:45125", + "left": "issue:43010", + "right": "issue:43065", "accept": false, - "reason": "Both mention FP8/MoE, but one is save_pretrained behavior and the other is missing tensor-parallel metadata." + "reason": "Cache/layer no_grad decoration and Sam3PixelDecoder dummy Conv2d are unrelated." }, { "left": "issue:42890", "right": "issue:45341", "accept": false, - "reason": "A flaky integration test and a small testing_utils bug are not the same underlying issue." + "reason": "Different test bug entirely." }, { - "left": "issue:43010", - "right": "issue:43065", + "left": "issue:42548", + "right": "issue:44351", "accept": false, - "reason": "Different code paths: cache/layer update decoration vs a dummy Conv2d in Sam3PixelDecoder." + "reason": "Both are import errors, but they concern different missing symbols and do not look like the same underlying bug." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "local_files_only network behavior is unrelated to the GLM-5 RoPE implementation discussion." + "reason": "API/local_files_only bug is unrelated to GLM-5 RoPE implementation." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "One concerns compiled Python file errors, the other is a data-filtering slowdown after upgrading Transformers." + "reason": "Compiled-Python-file errors are unrelated to the slowdown/filtering regression." }, { "left": "issue:45125", "right": "issue:45478", - "accept": false, - "reason": "Both are Qwen3.5/MoE-related, but one is tensor-parallel support metadata and the other is a from_pretrained loading error." + "accept": true, + "reason": "Same underlying Qwen3.5 MoE from_pretrained failure in v5.4.0; the missing _tp_plan tensor-parallel fix matches the concrete error report." }, { "left": "issue:30333", "right": "issue:33453", "accept": false, - "reason": "Training-job MLflow reporting and tokenizer loading regression are unrelated bugs." + "reason": "MLflow reporting after failed training and tokenizer loading are unrelated." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Both mention generation/cache internals, but they are distinct generation-state bugs." + "reason": "No_grad decoration for cache/layer updates is unrelated to the generation kwargs sync issue." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Different model metadata issues: tensor parallel plan vs missing vision config field silently dropped by strict validation." + "reason": "Tensor-parallel loading failure is not the same as missing deepstack_visual_indexes in config." }, { "left": "issue:30990", "right": "issue:31795", "accept": false, - "reason": "Sentence-transformers loading hang and documentation confusion around forward inputs are not the same issue." + "reason": "Sentence-transformers stuck loading and documentation confusion are unrelated." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Both are regressions, but one is unwanted API calls and the other is FP8 save_pretrained behavior." + "reason": "Local-files-only network calls and FP8 save_pretrained bugs are unrelated." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "A Qwen3VL MoE loading break and a generic v4-to-v5 checkpoint reload problem are different failures." + "reason": "Qwen3VL MoE loading and generic checkpoint reload breakage are different model-specific failures." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "One reports an FP8 accuracy issue; the other is a missing image-patch helper causing an AttributeError." + "reason": "FineGrainedFP8 accuracy and GLM46V image-patch counting are unrelated." }, { - "left": "issue:43827", + "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "Documentation referencing removed pipelines is unrelated to the 'racoon' issue." + "reason": "API-request leakage is unrelated to the placeholder issue titled racoon." }, { - "left": "issue:43502", + "left": "issue:43827", "right": "issue:44961", "accept": false, - "reason": "No real overlap beyond being bug reports; local_files_only network leakage and 'racoon' are unrelated." + "reason": "Docs still referencing pipeline() is unrelated to racoon." }, { - "left": "issue:44355", - "right": "issue:45071", + "left": "issue:30333", + "right": "issue:36296", "accept": false, - "reason": "Compiled-file errors and PretrainedConfig type-checking breakage are different regressions." + "reason": "MLflow failure reporting and tensor-parallel training are different bugs." }, { - "left": "issue:30333", - "right": "issue:36296", + "left": "issue:44355", + "right": "issue:45071", "accept": false, - "reason": "MLflow reporting on failed jobs and tensor-parallel training bugs are separate subsystems and failures." + "reason": "Compiled file execution errors are unrelated to PretrainedConfig type-checking regressions." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Liger Kernel application with model_init has nothing to do with a DSA indexer lacking ReLU." + "reason": "Liger Kernel application with model_init and DSA ReLU discussion are unrelated." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Compiled Python file errors and inability to save Pipeline objects are distinct issues." + "reason": "Compiled Python file errors and Pipeline.save_pretrained failures are different issues." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Both concern loading regressions, but one is tokenizer loading and the other is Llama 3.2 vision model loading under a specific version." + "reason": "Tokenizer-loading regression and Llama 3.2 vision model loading regression are different model/code paths." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "Both are v5-related API complaints, but one is a missing huggingface_hub import and the other is a broken env command." + "reason": "Missing is_offline_mode import and broken transformers env command are different failures." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "Unrelated: broken env command vs PretrainedConfig type-checking regression." + "reason": "CLI command breakage and PretrainedConfig type checking are unrelated." }, { "left": "issue:33453", "right": "issue:36296", "accept": false, - "reason": "Tokenizer loading regression and tensor-parallel training bug are not the same bug." + "reason": "Tokenizer-loading regression and tensor-parallel training bug are unrelated." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Both mention FP8/MoE, but one is a save_pretrained bug and the other is inference support." + "reason": "Both mention FP8/MoE, but one is save_pretrained and the other is inference support; not the same bug." }, { "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "Different import/loading failures: env command breakage versus missing Qwen2_5_VLForConditionalGeneration export." + "reason": "CLI/env command breakage and Qwen2_5_VL import failure are different issues." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "Both are import-name errors, but for different symbols and likely different package export regressions." + "reason": "Different missing imports: AutoImageProcessor vs PreTrainedModel." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall: the summary reflects the packet\u2019s mix of mostly unrelated v5-era issues, and the only accepted soft edge (issue:45125 \u2194 issue:45478) is cautiously stated and plausible from the titles. The other verdicts stay conservative rather than overreaching." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -15307,8 +15310,8 @@ "issue:44623|issue:45230", "issue:44273|issue:45542", "issue:44038|issue:45125", - "issue:44273|issue:45003", "issue:44230|issue:46095", + "issue:44273|issue:45003", "issue:44393|issue:45071", "issue:43576|issue:44556", "issue:43525|issue:45850", @@ -15324,248 +15327,230 @@ "issue:42890|issue:44908", "issue:43408|issue:44315", "issue:44704|issue:44908", - "issue:43976|issue:44393", "issue:44297|issue:45230", + "issue:43976|issue:44393", "issue:43352|issue:44393" ], - "prepared_review_unit_hash": "2716e78f647d024f2adb5180e32e8f99f5ac3967fd688d7e1f6cd5e7fd2270b2", + "prepared_review_unit_hash": "cd3e3c3de119bbcb89a4d53cff76182f0866b3cb0de7cb460ea8e672b816384b", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 37491, - "estimated_input_tokens": 9373, - "estimated_eval_tokens": 19002 + "serialized_chars": 37488, + "estimated_input_tokens": 9372, + "estimated_eval_tokens": 19000 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous, with several unrelated bug reports loosely tied together by Transformers v5 regressions, import issues, or model-specific failures. No PRs are present. I\u2019d use the broad v5 regression report as the best issue anchor, but I would not merge any of the soft pairs as duplicates.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45850 is the broadest umbrella report in the set and best matches the cluster\u2019s recurring theme of widespread v5 breakage, making it the most reasonable canonical anchor.", + "summary": "These items are largely separate bug reports/regressions across different models, APIs, and environments. None of the soft pairs look like the same underlying defect, so there is no safe duplicate merge among them.", + "confidence": 0.88, + "canonical_issue_reason": "issue:45850 is the broadest umbrella report ('Since 5.0 version it breaks too many models') and best matches the cluster\u2019s general v5-regression theme, even though the individual reports are not duplicates.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45850 has the most general scope, a clear regression-oriented title, and enough discussion to serve as the best triage hub for this mixed cluster.", + "best_issue_reason": "issue:45850 is the best global issue anchor because it is the most general, open-ended regression report covering many downstream breakages.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "Completely different bugs: a placeholder/title-only issue versus a Gemma-4 positional encoding problem." + "reason": "Unrelated issue content: a one-word/non-descriptive report vs a Gemma-4 positional-encoding bug." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "Documentation confusion in model.forward is unrelated to an ImportError for layer_type_validation." + "reason": "Documentation confusion about forward inputs vs an import error for layer_type_validation." }, { "left": "issue:44355", "right": "issue:46178", "accept": false, - "reason": "Compiled-Python-file errors and a tensorflow_text backend mapping error are different code paths." + "reason": "Compiled Python-file loading errors vs a backend mapping error for tensorflow_text; different failure modes." }, { "left": "issue:43576", "right": "issue:44938", "accept": false, - "reason": "Both mention version-related breakage, but one is the env command and the other is Python 3.14 import failure; not the same bug." + "reason": "Broken `transformers env` command vs Python 3.14 import/load failure; not the same code path." }, { "left": "issue:44315", "right": "issue:46209", "accept": false, - "reason": "Liger Kernel application via model_init and FP8 inference on B200 are unrelated model/runtime issues." + "reason": "Model-init/Liger-kernel behavior vs FP8 inference failure on B200; different problems." }, { "left": "issue:46143", "right": "issue:46144", "accept": false, - "reason": "Both involve RoFormer, but one is kwargs forwarding and the other is attention-interface implementation; different underlying defects." + "reason": "Passing `**kwargs` through RoFormer methods vs attention-interface usage; related subsystem but distinct bug." }, { "left": "issue:44908", "right": "issue:44961", "accept": false, - "reason": "Scheduler kwargs handling is unrelated to the racoon issue." + "reason": "Scheduler keyword handling vs an unrelated issue titled 'racoon'." }, { "left": "issue:44485", "right": "issue:44961", "accept": false, - "reason": "RoPE implementation discussion does not match the unrelated racoon issue." + "reason": "RoPE implementation discussion vs an unrelated issue; no shared bug." }, { "left": "issue:36296", "right": "issue:43901", "accept": false, - "reason": "Tensor parallel training bug and TextClassificationPipeline docs mismatch are unrelated." + "reason": "Tensor-parallel training bug vs docs mentioning `return_all_scores`; completely different concerns." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Chunked prefill cache regression and a typo in cache strategies are not the same bug." + "reason": "Chunked-prefill cache behavior vs a cache-strategies typo; not the same defect." }, { "left": "issue:36296", "right": "issue:43526", "accept": false, - "reason": "Tensor parallel training bug and BeitImageProcessorFast reduce_labels behavior are unrelated." + "reason": "Tensor-parallel training issue vs BeitImageProcessorFast label reduction bug; different code paths." }, { "left": "issue:45362", "right": "issue:45850", "accept": false, - "reason": "Qwen3.5 chat crash is a specific model failure; 45850 is only a broad umbrella regression report." + "reason": "Qwen3.5 chat crash is a specific model failure, while 45850 is a broad v5 regression umbrella." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "Missing processor save files and a generic bug report do not indicate the same underlying problem." + "reason": "`processor.save_pretrained` missing files vs a generic 'Bug report' with no concrete shared symptom." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading malfunction is unrelated to the TensorFlow/tensorboard backend error." + "reason": "Lazy loading malfunction vs undefined TensorFlow backend from tensorboard-only install; unrelated." }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "Qwen3-VL-Moe issue and missing _tp_plan for tensor parallelism are distinct model-support bugs." + "reason": "Qwen3-VL-Moe bug vs missing `_tp_plan` for tensor parallelism; same family, but not the same concrete issue." }, { - "left": "issue:44273", - "right": "issue:45003", + "left": "issue:44230", + "right": "issue:46095", "accept": false, - "reason": "Lazy loading is not the same as unsafe sys.modules access in modeling_utils." + "reason": "FP8 inference support request vs a weight-conversion-script question; not duplicates." }, { - "left": "issue:44230", - "right": "issue:46095", + "left": "issue:44273", + "right": "issue:45003", "accept": false, - "reason": "FP8 inference support for Qwen models is unrelated to a DeepSeekV4 weight-conversion question." + "reason": "Lazy loading problems vs unsafe `sys.modules` access in modeling_utils; different underlying bug." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Qwen3-VL bbox hallucination/error and PretrainedConfig type-checking regression are unrelated." + "reason": "Qwen3-VL bounding-box hallucination/error vs PretrainedConfig type-checking breakage; unrelated." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "Both are version-related, but one is the transformers env command and the other is checkpoint reload compatibility; different bugs." + "reason": "`transformers env` command breakage vs checkpoint reload incompatibility after upgrade; different symptoms and code paths." }, { "left": "issue:43525", "right": "issue:45850", "accept": false, - "reason": "Missing pad_token_id on Llama4Config is a specific config bug, not the same as the broad v5 regression report." - }, - { - "left": "issue:44393", - "right": "issue:45753", - "accept": false, - "reason": "Qwen3-VL bbox output failure and Qwen3_5 infinite-loop behavior are different failures." - }, - { - "left": "issue:41669", - "right": "issue:43673", - "accept": false, - "reason": "Removing import * usage for performance and a missing cache issue in chunked prefill are not the same change or bug." - }, - { - "left": "issue:44556", - "right": "issue:45092", - "accept": false, - "reason": "Checkpoint reload incompatibility and meta-initialization incompatibility are related to upgrades, but they are different concrete failures." + "reason": "Missing `pad_token_id` on Llama4Config is a specific config bug, not the broad v5 regression umbrella." }, { "left": "issue:43572", "right": "issue:45850", "accept": false, - "reason": "StableLmConfig pad_token_idx regression is a specific config issue, not the same as the broad v5 breakage report." + "reason": "Missing `pad_token_idx` in StableLmConfig is a specific model config issue, not the same as the generic v5-breakage report." }, { "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "TF32 API guidance and a cache-strategy typo are unrelated." + "reason": "TF32 behavior settings vs a typo in cache strategies; no common underlying bug." }, { "left": "issue:44492", "right": "issue:45850", "accept": false, - "reason": "A typo in cache strategies is not the same as the broad statement that v5 breaks too many models." + "reason": "A typo issue is not the same as a general claim that v5 breaks many models." }, { "left": "issue:45278", "right": "issue:45850", "accept": false, - "reason": "Many import errors after upgrade and broad v5 breakage overlap only at a high level; they are not the same concrete issue." + "reason": "Both are broad v5-regression complaints, but one is about import errors after upgrade and the other about many models; too general to treat as the same bug." }, { "left": "issue:41669", "right": "issue:46218", "accept": false, - "reason": "Import-* cleanup/performance work is unrelated to a specific missing Qwen3_5ForCausalLM module import." + "reason": "Import-star usage/performance issue vs a specific missing-module import error for Qwen3_5ForCausalLM." }, { "left": "issue:44246", "right": "issue:44273", "accept": false, - "reason": "Slow import sometimes and lazy loading not working properly may be adjacent, but there is not enough evidence they are the same underlying bug." + "reason": "Slow import timing vs lazy-loading malfunction; related theme, different defect." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "Missing set_seed in SAM-HQ tests is unrelated to inverse_sqrt scheduler kwargs handling." + "reason": "Missing `set_seed()` in tests vs scheduler kwarg handling; unrelated." }, { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "SAM3 tracker/model-type warning and Liger Kernel not applied via model_init are unrelated." + "reason": "sam3_video/sam3_tracker warning vs Liger kernel application with model_init; different bugs." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "AutoProcessor kwargs propagation to cached_file and scheduler kwargs handling are different APIs and bugs." + "reason": "`AutoProcessor.from_pretrained` kwargs forwarding vs inverse-sqrt scheduler kwargs; not the same component or bug." }, { - "left": "issue:43976", - "right": "issue:44393", + "left": "issue:44297", + "right": "issue:45230", "accept": false, - "reason": "Python version incompatibility and Qwen3-VL bounding-box errors are unrelated." + "reason": "Tokenizer class mismatch on save vs a vague bug report with no matching failure mode." }, { - "left": "issue:44297", - "right": "issue:45230", + "left": "issue:43976", + "right": "issue:44393", "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch is unrelated to a generic bug report." + "reason": "Python-version compatibility problem vs Qwen3-VL bounding-box output bug; unrelated." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Flash Attention 2 support for NemotronH and Qwen3-VL bbox errors are unrelated model-specific issues." + "reason": "Flash Attention 2 support gap vs Qwen3-VL output hallucination/error; different issues." } ] }, @@ -15595,10 +15580,10 @@ "issue:34689", "issue:41669", "issue:42371", + "issue:42757", "issue:42913", "issue:43066", "issue:43298", - "issue:43352", "issue:43502", "issue:43519", "issue:43576", @@ -15606,6 +15591,7 @@ "issue:43618", "issue:43704", "issue:43827", + "issue:43874", "issue:43976", "issue:44077", "issue:44261", @@ -15627,7 +15613,6 @@ "issue:44998", "issue:45071", "issue:45230", - "issue:45290", "issue:45310", "issue:45341", "issue:45468", @@ -15644,8 +15629,8 @@ "issue:41669|issue:44351", "issue:44908|issue:45071", "issue:41669|issue:45341", + "issue:43874|issue:45850", "issue:44492|issue:44945", - "issue:43976|issue:45850", "issue:43976|issue:46218", "issue:44077|issue:44861", "issue:44393|issue:45478", @@ -15665,311 +15650,303 @@ "issue:31795|issue:34689", "issue:31515|issue:31795", "issue:44485|issue:45468", - "issue:43704|issue:44485", "issue:43298|issue:46095", + "issue:43704|issue:44485", "issue:45850|issue:46218", - "issue:42371|issue:43704", "issue:43502|issue:43519", + "issue:42371|issue:43704", "issue:43519|issue:44623", "issue:43519|issue:43704", "issue:43606|issue:45468", "issue:44263|issue:45468", "issue:44261|issue:45468", "issue:44351|issue:46218", + "issue:42757|issue:46218", "issue:42913|issue:43066", - "issue:43352|issue:46218", "issue:44908|issue:44998", "issue:44492|issue:44998", - "issue:43618|issue:44998", - "issue:44861|issue:45290", - "issue:44998|issue:45850" + "issue:43618|issue:44998" ], - "prepared_review_unit_hash": "5312f4bf8e6a5e21b38ef562c928d5dcdd572babab22b84eeca15aaa65d0bae5", + "prepared_review_unit_hash": "9b1cb3bbf8a47bd46822211466a77931fcb573c72f797ca28cb41edb897a9787", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 44, - "serialized_chars": 40004, - "estimated_input_tokens": 10001, - "estimated_eval_tokens": 20258 + "soft_pair_count": 42, + "serialized_chars": 39418, + "estimated_input_tokens": 9855, + "estimated_eval_tokens": 19966 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous Transformers issues (tokenizer docs/regressions, model-loading bugs, training/runtime edge cases, and unrelated utilities). No listed soft-edge pair looks like the same concrete bug/change; the only obvious duplicate signal in the item set is the repeated Qwen3.5 MoE from_pretrained report (45310/45478).", - "confidence": 0.66, - "canonical_issue_reason": "issue:45310 is the earlier, concrete report of the same Qwen3.5 MoE from_pretrained failure as 45478, so it is the best canonical anchor among these items.", + "summary": "This cluster is heterogeneous, with many unrelated loading, tokenizer, docs, import, and model-specific issues. The clearest canonical thread is the Qwen3.5 MoE from_pretrained regression, while the soft-edge pairs mostly share only broad Transformers-v5 or loading-related wording and should stay separate.", + "confidence": 0.56, + "canonical_issue_reason": "issue:45310 is the most concrete, reproducible bug report in the cluster: it names a specific failing path (Qwen3.5 MoE + transformers>=5.4.0 + from_pretrained) and is representative of the loading-regression theme.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the most actionable representative: it names a specific model, a specific version range, and a concrete loading failure, unlike the broader or vaguer reports in the cluster.", + "best_issue_reason": "issue:45310 is the best overall representative because it is specific enough to anchor the cluster, describes an actionable failure mode, and is more precise than the surrounding broad/vague loading complaints.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "Specific buffer/position_ids bug vs a generic bug report; no shared code path." + "reason": "One is a position_ids buffer access bug; the other is only a generic bug report with no matching failure mode." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "CLI env command regression and scheduler kwargs bug are unrelated." + "reason": "Broken env command and scheduler kwargs bug are unrelated subsystems and concrete failures." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star cleanup/perf issue vs missing HybridCache import are different failures." + "reason": "Import-star performance issue and missing HybridCache export are different problems." }, { "left": "issue:44908", "right": "issue:45071", "accept": false, - "reason": "Scheduler kwargs handling and PretrainedConfig type checking are separate issues." + "reason": "Scheduler keyword handling and PretrainedConfig type-checking are distinct code paths." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Repository import cleanup has no overlap with a testing_utils typo." + "reason": "Model import cleanup/performance and a testing_utils typo are unrelated." }, { - "left": "issue:44492", - "right": "issue:44945", + "left": "issue:43874", + "right": "issue:45850", "accept": false, - "reason": "Cache-strategy typo and pipeline-parallelism output bug are unrelated." + "reason": "A missing image-processor method in one model is not the same as a broad complaint that many models break in v5." }, { - "left": "issue:43976", - "right": "issue:45850", + "left": "issue:44492", + "right": "issue:44945", "accept": false, - "reason": "Python-version compatibility problem vs broad v5 breakage complaint." + "reason": "Cache-strategy typo and pipeline-parallelism output corruption do not describe the same bug." }, { "left": "issue:43976", "right": "issue:46218", "accept": false, - "reason": "Interpreter support issue vs a specific missing import symbol." + "reason": "Python version compatibility and a missing Qwen3_5 module import are separate import/runtime issues." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Optional post_init policy and tied-weight key crash do not share a concrete bug." + "reason": "Optional post_init enforcement and tied-weights list handling are different model-init bugs." }, { "left": "issue:44393", "right": "issue:45478", "accept": false, - "reason": "Qwen3-VL bbox hallucination and Qwen3.5 MoE loading error are different model paths." + "reason": "Qwen3-VL bounding-box hallucination is unrelated to a Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Docs referencing removed pipeline APIs is unrelated to scheduler kwargs." + "reason": "Docs still referencing pipeline() and scheduler kwargs handling are unrelated." }, { "left": "issue:31795", "right": "issue:33453", "accept": false, - "reason": "Forward-doc confusion and tokenizer-loading regression are not the same underlying issue." + "reason": "Documentation confusion for model.forward and tokenizer-loading regression are not the same underlying problem." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Liger Kernel application in model_init and flash-attention training degeneration are different problems." + "reason": "Liger Kernel application in model_init and flash_attention_3 degenerate training are different training-path issues." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken env command and lazy loading malfunction are not the same bug." }, { "left": "issue:44655", "right": "issue:44861", "accept": false, - "reason": "Saving Pipeline objects and _tied_weights_keys list handling are unrelated code paths." + "reason": "Saving pipeline objects and crashing on _tied_weights_keys list are unrelated save/load behaviors." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer activation issue and Gemma4Audio positional encoding bug are unrelated." + "reason": "DSA indexer ReLU discussion and Gemma-4 audio positional encoding are different model internals." }, { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "Generic placeholder report vs a specific Qwen3.5 MoE from_pretrained failure." + "reason": "A generic bug report is not a duplicate of a specific from_pretrained error report." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy loading malfunction and inverse_sqrt scheduler kwargs are unrelated." + "reason": "Lazy loading failure and scheduler kwargs ignoring are unrelated." }, { "left": "issue:46018", "right": "issue:46172", "accept": false, - "reason": "DeepSeek-V4 gating logic and deepseek-v4 weight-transpose/conversion issue are different bugs." + "reason": "DeepSeek-V4 gating behavior and a comb transpose bug are different DeepSeek issues." }, { "left": "issue:43576", "right": "issue:46218", "accept": false, - "reason": "Broken env command and missing Qwen3_5 import are unrelated." + "reason": "Broken env command and missing Qwen3_5 import are unrelated commands/imports." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Timestamp calculation in Qwen3VL is unrelated to GLM-5 RoPE implementation." + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are different vision/text model bugs." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Processor timestamp bug and scheduler kwargs bug do not match." + "reason": "Timestamp math in a processor and scheduler kwargs handling are unrelated." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import-star performance issue vs VRAM leak in multithreaded loading are different." + "reason": "Import-star performance regression and VRAM leak in multi-threaded use are different issues." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Documentation confusion and Llama 3.2 model-loading regression are unrelated." + "reason": "Forward-doc confusion and Llama 3.2 model-loading breakage are not the same bug." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Slow checkpoint loading and forward-doc confusion are different issue types." + "reason": "Slow checkpoint loading and documentation confusion about forward inputs are unrelated." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "GLM-5 RoPE math and Gemma4Audio positional encoding are separate model-specific bugs." + "reason": "GLM-5 RoPE implementation and Gemma-4 audio positional encoding are different architecture bugs." }, { - "left": "issue:43704", - "right": "issue:44485", + "left": "issue:43298", + "right": "issue:46095", "accept": false, - "reason": "VRAM leak in multithreaded use is unrelated to RoPE implementation." + "reason": "Continuous batching for audio models and DeepSeek-V4 weight-conversion question are not the same underlying issue." }, { - "left": "issue:43298", - "right": "issue:46095", + "left": "issue:43704", + "right": "issue:44485", "accept": false, - "reason": "Continuous batching/audio-model support and weight conversion script availability are not the same bug." + "reason": "VRAM leak in threaded inference and GLM-5 RoPE implementation do not match." }, { "left": "issue:45850", "right": "issue:46218", "accept": false, - "reason": "Broad version-wide breakage complaint vs a specific import error." + "reason": "Broad complaint that v5 breaks many models and a specific missing module import are not duplicates." }, { - "left": "issue:42371", - "right": "issue:43704", + "left": "issue:43502", + "right": "issue:43519", "accept": false, - "reason": "TF32 configuration guidance and VRAM leak are unrelated." + "reason": "Offline file-loading network requests and Qwen3VL timestamp calculation are unrelated." }, { - "left": "issue:43502", - "right": "issue:43519", + "left": "issue:42371", + "right": "issue:43704", "accept": false, - "reason": "Unexpected network requests with local_files_only and timestamp calculation are different failures." + "reason": "TF32 config migration and VRAM leak in threaded inference are separate concerns." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Timestamp math bug and missing processor save files are unrelated." + "reason": "Timestamp calculation in a processor and missing processor save files are different processor bugs." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Processor timestamp bug and VRAM leak are unrelated." + "reason": "Processor timestamp math and multi-thread VRAM leak are unrelated." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch and Gemma4Audio positional encoding are unrelated." + "reason": "CPU offload device mismatch in bark-small is unrelated to Gemma-4 audio positional encoding." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split return handling and Gemma4Audio positional encoding are unrelated." + "reason": "GlmMoeDsaIndexer torch.split behavior and Gemma-4 audio positional encoding are different internals." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "MLA rms_norm_eps precision issue and Gemma4Audio positional encoding are unrelated." + "reason": "MLA q_a_layernorm epsilon precision issue and Gemma-4 audio positional encoding are not the same bug." }, { "left": "issue:44351", "right": "issue:46218", "accept": false, - "reason": "Different missing import symbols; no shared root cause is evident." + "reason": "Missing HybridCache export and missing Qwen3_5ForCausalLM import are separate import/export failures." }, { - "left": "issue:42913", - "right": "issue:43066", + "left": "issue:42757", + "right": "issue:46218", "accept": false, - "reason": "Both are tokenizer regressions in v5, but the symptoms and likely fixes differ." + "reason": "huggingface_hub is_offline_mode import breakage is unrelated to Qwen3_5 module import failure." }, { - "left": "issue:43352", - "right": "issue:46218", + "left": "issue:42913", + "right": "issue:43066", "accept": false, - "reason": "Flash Attention support limitation and missing Qwen3_5 import are different issues." + "reason": "Both mention tokenizer v5 behavior, but they describe different concrete tokenizer regressions." }, { - "left": "issue:44998", - "right": "issue:44908", + "left": "issue:44908", + "right": "issue:44998", "accept": false, - "reason": "Unemployment is unrelated noise, not a software bug match." + "reason": "Scheduler kwargs handling and an unrelated placeholder issue title are not the same bug." }, { "left": "issue:44492", "right": "issue:44998", "accept": false, - "reason": "A typo in cache strategies is unrelated to an off-topic issue." + "reason": "Cache-strategy typo and unrelated placeholder issue are not duplicates." }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions regression is unrelated to the off-topic item." - }, - { - "left": "issue:44861", - "right": "issue:45290", - "accept": false, - "reason": "Tied-weight-key crash and chat-template tool-call crash are different code paths." - }, - { - "left": "issue:44998", - "right": "issue:45850", - "accept": false, - "reason": "Off-topic item and broad version-breakage complaint are not duplicates." + "reason": "CLIPOutput attentions regression is unrelated to the placeholder issue." } ] }, @@ -15988,8 +15965,8 @@ "review_unit_count": 50, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-431", "nodes": [ @@ -16013,11 +15990,15 @@ "issue:44623", "issue:44861", "issue:44998", + "issue:45290", "issue:45310", + "issue:45850", "issue:46018", "issue:46144" ], "soft_pairs": [ + "issue:44861|issue:45290", + "issue:44998|issue:45850", "issue:44556|issue:44998", "issue:43525|issue:44861", "issue:43572|issue:44861", @@ -16041,166 +16022,178 @@ "issue:37934|issue:43668", "issue:42936|issue:43668" ], - "prepared_review_unit_hash": "0de79b46faa0c6f478fa9b0f53f9de609b7d016dbf7bce54291899ccd34eb6e6", + "prepared_review_unit_hash": "420eecf687ba8d7b1338d36000ace2d9034f9b4b1a15dbc9e64c5a98ea8c03bb", "component_budget": { "node_count": 431, "item_count": 431, "soft_pair_count": 1595, - "serialized_chars": 825263, - "estimated_input_tokens": 206316, - "estimated_eval_tokens": 412888 + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 }, "budget": { - "node_count": 23, - "item_count": 23, - "soft_pair_count": 22, - "serialized_chars": 19401, - "estimated_input_tokens": 4851, - "estimated_eval_tokens": 9958 + "node_count": 25, + "item_count": 25, + "soft_pair_count": 24, + "serialized_chars": 21130, + "estimated_input_tokens": 5283, + "estimated_eval_tokens": 10822 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a single duplicate cluster: the items span many unrelated bugs/features across loading, config, tokenizer, training, and model-specific behavior. All soft pairs look like loose topical similarity rather than the same underlying bug or change, so they should be rejected.", - "confidence": 0.9, - "canonical_issue_reason": "issue:44556 is the most concrete, actionable issue in the set: it describes a specific regression boundary (v4.57 to v5.x) and a clear failure mode (checkpoint reload breaks after upgrade).", + "summary": "This cluster is heterogeneous: most items are unrelated issues spanning loading regressions, config/API breakages, model-specific bugs, and a few generic feature requests. I do not see any true duplicate pair here; the soft-similarity matches are mostly surface-level topic overlap.", + "confidence": 0.91, + "canonical_issue_reason": "issue:44556 is the best representative because it is a concrete, user-facing version-regression report about checkpoint reloads, which is the closest substantive theme shared by several other items in this set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44556 is the best single issue to anchor triage because it is specific, reproducible, and framed as a versioned regression rather than a broad or ambiguous request.", + "best_issue_reason": "issue:44556 is the most actionable and broadly representative issue in the set: it names the failure mode, the affected upgrade path, and a clear regression scenario.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Different bugs in different code paths: tied-weight key extraction crash vs chat template crashes on tool-call assistant messages." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "Unemployment is not a technical bug report; the other is a broad model-regression complaint, so they are unrelated." + }, { "left": "issue:44556", "right": "issue:44998", "accept": false, - "reason": "Completely unrelated: checkpoint reload regression vs an unrelated low-signal title ('Unemployment')." + "reason": "A checkpoint reload regression is unrelated to the non-technical \"Unemployment\" issue." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Different bugs: missing pad_token_id on Llama4Config vs AttributeError in _get_tied_weight_keys for list handling." + "reason": "Missing pad_token_id on Llama4Config and an AttributeError in _get_tied_weight_keys are different failures in different subsystems." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "Different config regression (pad_token_idx missing) vs tied-weight key crash; no shared code path." + "reason": "StableLmConfig padding attribute regression is not the same bug as tied-weight key handling." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "One is slow from_pretrained checkpoint loading; the other is an ImportError for a missing symbol. Different failure modes." + "reason": "Slow from_pretrained checkpoint loading and an ImportError for layer_type_validation are distinct problems." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "Different areas: fp8 Qwen3-VL/Qwen3.5 inference support vs processor.save_pretrained missing files." + "reason": "FP8 inference support for Qwen models is unrelated to missing files in processor.save_pretrained()." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLflow reporting after failed training job vs slow checkpoint loading; unrelated subsystems and symptoms." + "reason": "MLflow reporting after failed training and slow checkpoint loading are different code paths and outcomes." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "Checkpoint load performance issue vs tensor parallel training bug; not the same underlying defect." + "reason": "from_pretrained performance and tensor-parallel training bugs are not the same underlying defect." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "Both mention Qwen3.5, but one is fp8 inference support and the other is a from_pretrained regression after version upgrade; too different to merge as one fix." + "reason": "FP8 support request and Qwen3.5 MoE from_pretrained error are different model-specific issues." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "Slow from_pretrained checkpoint load vs tokenizer loading regression; similar theme only at a very broad level." + "reason": "Checkpoint load slowness and tokenizer loading regression are both loading-related but not the same concrete bug." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching audio-model support vs fp8 inference support; different model classes and problem types." + "reason": "Continuous batching audio-model support is a different feature gap than FP8 inference support." }, { "left": "issue:43717", "right": "issue:46144", "accept": false, - "reason": "Mamba init_weights behavior vs RoFormer attention interface usage; unrelated model-specific implementation issues." + "reason": "Mamba init_weights behavior and RoFormer attention-interface usage are unrelated model implementation issues." }, { "left": "issue:42936", "right": "issue:46018", "accept": false, - "reason": "Mask2Former ignore_value handling vs DeepSeek-V4 shared expert gating; different models and different logic." + "reason": "Mask2Former ignore_value handling and DeepSeek-V4 shared-expert gating are unrelated model bugs." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "Incorrect type hint in ModernBERTConfig vs tokenizer regex warning regression; unrelated." + "reason": "A type-hint bug in ModernBERTConfig and a tokenizer regex warning regression are different issues." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Ignore_value not used in Mask2Former vs kwargs not forwarded to loss calculation; related only in the broadest sense of loss code." + "reason": "Ignored loss parameter handling and kwargs not forwarded to loss calculation are separate bugs, though both touch loss code." }, { "left": "issue:43240", "right": "issue:46144", "accept": false, - "reason": "Loss kwargs propagation vs attention interface implementation; no shared concrete bug." + "reason": "Loss kwargs forwarding and attention-interface usage are different subsystems with no shared concrete failure." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2Former ignore_value bug vs tokenizer regex warning after 4.57.3; different components and root causes." + "reason": "Mask2Former ignore_value handling does not match the tokenizer regex warning issue." }, { "left": "issue:42936", "right": "issue:46144", "accept": false, - "reason": "Mask2Former ignore_value handling vs RoFormer attention interface issue; not the same code path." + "reason": "Different model code paths: Mask2Former loss handling vs RoFormer attention implementation." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "Ignore_value usage bug vs a typo in quantization_operation text; unrelated." + "reason": "Mask2Former ignore_value handling and a typo in quantization_operation are unrelated." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Missing kwargs in loss calculation vs incorrect type hint; different severity and implementation areas." + "reason": "Loss kwargs forwarding bug and ModernBERTConfig type-hint bug are unrelated." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "Type hint error vs typo in an attribute name; only superficial similarity." + "reason": "A config type-hint issue and a typo fix are not the same underlying bug." }, { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "Question about whether Llama4TextL2Norm is really RMSNorm vs a type-hint correction; unrelated." + "reason": "A question about Llama4TextL2Norm being RMSNorm is unrelated to ModernBERTConfig norm_eps typing." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Mask2Former ignore_value bug vs ModernBERTConfig type hint issue; different models and bug classes." + "reason": "Mask2Former loss logic and ModernBERTConfig typing are unrelated." } ] }, @@ -16219,8 +16212,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10519-3", "nodes": [ @@ -16233,52 +16226,52 @@ "issue:10519|issue:11005", "issue:11005|issue:11307" ], - "prepared_review_unit_hash": "20c772d1f593ae0dd0f578ec9c37a2b967b86c8a0dda7ae658e3fa98f4fac2cb", + "prepared_review_unit_hash": "79bcc05d6892c55ae0f82f5045e5df4de37c28db7f56536151c44d798f1bac3d", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2878, - "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": 2878, - "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 about unrelated feature requests: tokenizer truncation behavior, learning-rate scheduling functionality, and Wav2Vec2 word time offsets. They do not appear to be duplicates or the same underlying bug/change.", + "summary": "The three issues are unrelated feature requests/questions: truncation behavior, ReduceLROnPlateau-like scheduling, and Wav2Vec2 word time offsets. They do not describe the same underlying bug or change, so this cluster should not be merged as duplicates.", "confidence": 0.97, - "canonical_issue_reason": "No true canonical issue for this cluster; the items cover different subsystems and different requested capabilities.", - "canonical_pr_reason": "No pull requests are present in this cluster.", - "best_issue_reason": "If one issue must serve as a representative, #10519 is the most concrete and narrowly scoped request, but it is not representative of the others as a duplicate target.", - "best_pr_reason": "No pull requests are present in this cluster.", + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a good canonical representative here because the topics are unrelated; if forced to pick a single issue, #10519 is the most broadly phrased feature request, but it is not a duplicate of the others.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:10519", "right": "issue:11307", "accept": false, - "reason": "Different domains and goals: tokenizer truncation option vs speech model word timestamp extraction." + "reason": "Different domains and requested behaviors: truncation from beginning vs Wav2Vec2 word boundary timestamps." }, { "left": "issue:10519", "right": "issue:11005", "accept": false, - "reason": "Unrelated functionality: text truncation behavior vs ReduceLROnPlateau-like scheduler behavior." + "reason": "One is about token truncation strategy, the other about learning-rate scheduling functionality; no shared underlying bug." }, { "left": "issue:11005", "right": "issue:11307", "accept": false, - "reason": "Different problems in different components: optimizer/scheduler feature request vs Wav2Vec2 alignment/time-offset request." + "reason": "Unrelated requests in different parts of the library: scheduler behavior versus Wav2Vec2 time offsets." } ] }, @@ -16297,8 +16290,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-13244-3", "nodes": [ @@ -16310,7 +16303,7 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "0b788c1aec8199c3ca715dea69ee7f9db99d3da602a02b70782ffb712d94fe6d", + "prepared_review_unit_hash": "7ad21377992da39338aa6ae60ff20a09973bd02e4a2cf2ce374067a67088fcc8", "component_budget": { "node_count": 3, "item_count": 3, @@ -16332,24 +16325,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates as a cluster: one issue is about Tapas tokenization, one about a private-model auth/error-message problem, and one about loading bert-base-uncased. Only the two model-loading issues are loosely related, but they still appear to describe different failure modes.", - "confidence": 0.77, - "canonical_issue_reason": "Issue 16618 is the broadest and most central model-loading report, but the cluster does not contain a true duplicate set.", + "summary": "These issues do not look like the same bug: one is about Tapas tokenization differing from TensorFlow code, one is about an unhelpful auth error when downloading a private model, and one is a generic model-loading failure for bert-base-uncased. The only loose commonality is model loading, but the concrete failure modes differ.", + "confidence": 0.88, + "canonical_issue_reason": "Issue 16618 is the broadest and most representative of the cluster because it concerns a core model-loading failure, whereas 15980 is specifically an authentication/error-message problem and 13244 is about tokenizer behavior.", "canonical_pr_reason": null, - "best_issue_reason": "16618 is the best representative because it is the most general user-facing model load failure among the three; the others are clearly about different topics.", + "best_issue_reason": "16618 is the best issue to keep as the cluster representative since it is the most general user-facing model-load problem; however, it should not be treated as a duplicate of the other two issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Different problems: Tapas tokenization vs. model loading for bert-base-uncased. No shared code-path or concrete failure mode." + "reason": "Tapas tokenization mismatch vs. bert-base-uncased loading failure are different subsystems and different failure modes." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "Both mention model download/load, but one is specifically a private-model login/error-message issue and the other is a generic bert-base-uncased load failure; too different to treat as the same bug." + "reason": "Private-model login/download error message is an auth/UX issue, while 16618 is a model-load failure for a public model; not the same underlying bug." } ] }, @@ -16368,8 +16361,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-26350-2", "nodes": [ @@ -16379,7 +16372,7 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "6b5dd876227352b905a3f199171ac178a6e7ce815e620d9af6bb8162a4f0de1c", + "prepared_review_unit_hash": "942127d98fb66aec749c5cdf0b97b35ed2e84f2a4a283b6f10921fd2ac58f293", "component_budget": { "node_count": 2, "item_count": 2, @@ -16401,18 +16394,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one issue requests Flash Attention 2 support for more architectures, while the other requests GGUF support for more architectures. They share a broad theme but describe different features and code paths.", - "confidence": 0.97, + "summary": "These are two different community-contribution feature requests: one for Flash Attention 2 support on more architectures, the other for GGUF support on more architectures. They share the same broad pattern but not the same underlying change or bug.", + "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "If a single representative is needed, issue 26350 is slightly better because it is older and more specific, but the two issues are heterogeneous and should remain separate.", + "best_issue_reason": null, "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 underlying changes; they are related only at a broad \"more architectures\" level." + "reason": "Different requested features and code paths: Flash Attention 2 architecture support vs GGUF architecture support. Similar wording and contribution framing are too broad to treat as duplicates." } ] }, @@ -16434,7 +16427,7 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-27088-7", + "cluster_id": "cluster-27088-8", "nodes": [ "issue:27088", "issue:28936", @@ -16442,7 +16435,8 @@ "issue:43834", "issue:44365", "issue:44370", - "issue:46222" + "issue:46222", + "issue:46226" ], "soft_pairs": [ "issue:27088|issue:44370", @@ -16459,161 +16453,196 @@ "issue:38435|issue:46222", "issue:38435|issue:44365", "issue:38435|issue:43834", + "issue:38435|issue:46226", "issue:43834|issue:44370", "issue:43834|issue:46222", "issue:43834|issue:44365", + "issue:43834|issue:46226", "issue:44365|issue:44370", "issue:44365|issue:46222", - "issue:44370|issue:46222" + "issue:44365|issue:46226", + "issue:44370|issue:46222", + "issue:44370|issue:46226", + "issue:46222|issue:46226" ], - "prepared_review_unit_hash": "996c6bf499666d136718c12050a3b3bab7cf69333d2125d87a979a1a28eb49db", + "prepared_review_unit_hash": "39ffaa3671b3302b6ae30b5b6f477b0dfefee359a99ddec76c620cb8c377b66c", "component_budget": { - "node_count": 7, - "item_count": 7, - "soft_pair_count": 20, - "serialized_chars": 11132, - "estimated_input_tokens": 2783, - "estimated_eval_tokens": 5822 + "node_count": 8, + "item_count": 8, + "soft_pair_count": 25, + "serialized_chars": 13678, + "estimated_input_tokens": 3420, + "estimated_eval_tokens": 7096 }, "budget": { - "node_count": 7, - "item_count": 7, - "soft_pair_count": 20, - "serialized_chars": 11132, - "estimated_input_tokens": 2783, - "estimated_eval_tokens": 5822 + "node_count": 8, + "item_count": 8, + "soft_pair_count": 25, + "serialized_chars": 13678, + "estimated_input_tokens": 3420, + "estimated_eval_tokens": 7096 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly separate documentation localization requests, not duplicates: Turkish, Spanish, and Romanian are clearly distinct, and the later placeholder-titled issues look like template-generated translation requests rather than the same underlying task.", - "confidence": 0.89, - "canonical_issue_reason": "Issue 27088 is the oldest concrete, language-specific request and is the clearest representative in the set, but it should not be treated as a true duplicate target for the other language-specific issues.", + "summary": "This cluster is mostly a set of i18n translation request issues for different languages. The titles/body templates are highly similar, but they represent separate translation tasks rather than the same underlying bug or change, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "No canonical issue: the items are separate translation requests for different languages, and the similarity is mostly from a shared issue template.", "canonical_pr_reason": null, - "best_issue_reason": "27088 is the best representative issue because it is specific and readable, while the later placeholder-titled issues are too generic to serve as a stable canonical duplicate.", + "best_issue_reason": "No single issue is a good global duplicate target; the closest matches are generic template issues, but they still do not represent the same underlying translation task.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Different translation requests; the right side is a generic template issue, not evidence of the same language/task as Turkish." + "reason": "Different language-request issues; same template, but no evidence they track the same translation task." }, { "left": "issue:27088", "right": "issue:46222", "accept": false, - "reason": "Different translation requests; generic placeholder title is too vague to equate with the concrete Turkish request." + "reason": "Different language-request issues; superficial template similarity only." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "Different translation requests; the placeholder issue could be a separate language-specific task, not a duplicate of Turkish." + "reason": "Turkish vs another translation request; not the same underlying issue." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Different translation requests; the placeholder issue is too underspecified to be the same underlying task as Turkish." + "reason": "Different language request; similar template does not make them duplicates." }, { "left": "issue:27088", "right": "issue:38435", "accept": false, - "reason": "Different target languages (Turkish vs Romanian) mean these are separate issues." + "reason": "Turkish vs Romanian translation requests are distinct tasks." }, { "left": "issue:27088", "right": "issue:28936", "accept": false, - "reason": "Different target languages (Turkish vs Spanish) mean these are separate issues." + "reason": "Turkish vs Spanish translation requests are separate issues." }, { "left": "issue:28936", "right": "issue:43834", "accept": false, - "reason": "Spanish translation request versus a generic placeholder issue; not enough evidence they are the same task." + "reason": "Spanish translation request vs generic template issue; no same-language/task evidence." }, { "left": "issue:28936", "right": "issue:44365", "accept": false, - "reason": "Spanish translation request versus a generic placeholder issue; template similarity alone is not enough." + "reason": "Same translation template, but likely different language-specific task." }, { "left": "issue:28936", "right": "issue:38435", "accept": false, - "reason": "Different target languages (Spanish vs Romanian) mean these are separate issues." + "reason": "Spanish and Romanian translation requests are distinct." }, { "left": "issue:28936", "right": "issue:44370", "accept": false, - "reason": "Spanish translation request versus a generic placeholder issue; no evidence of the same underlying request." + "reason": "Shared template only; no sign these are the same translation work item." }, { "left": "issue:38435", "right": "issue:44370", "accept": false, - "reason": "Romanian translation request versus a generic placeholder issue; not enough evidence they are duplicates." + "reason": "Different translation requests; similarity is only from the i18n template." }, { "left": "issue:38435", "right": "issue:46222", "accept": false, - "reason": "Romanian translation request versus a generic placeholder issue; generic similarity is insufficient." + "reason": "Different language-specific translation tasks." }, { "left": "issue:38435", "right": "issue:44365", "accept": false, - "reason": "Romanian translation request versus a generic placeholder issue; could be a different language/task." + "reason": "Romanian translation request vs another language request; not duplicates." }, { "left": "issue:38435", "right": "issue:43834", "accept": false, - "reason": "Romanian translation request versus a generic placeholder issue; no clear same-task evidence." + "reason": "Same i18n template, but not the same underlying request." + }, + { + "left": "issue:38435", + "right": "issue:46226", + "accept": false, + "reason": "Romanian translation request vs unrelated Thai text; clearly not duplicates." }, { "left": "issue:43834", "right": "issue:44370", "accept": false, - "reason": "Both are generic template issues, but there is no evidence they refer to the same language or same translation task." + "reason": "Both are generic translation templates, but there is no evidence they refer to the same language or task." }, { "left": "issue:43834", "right": "issue:46222", "accept": false, - "reason": "Both are generic template issues, but template similarity does not establish the same underlying request." + "reason": "Template similarity only; no explicit target-language match." }, { "left": "issue:43834", "right": "issue:44365", "accept": false, - "reason": "Both are generic template issues, but no concrete evidence they are the same issue." + "reason": "Separate translation issues despite near-identical wording." + }, + { + "left": "issue:43834", + "right": "issue:46226", + "accept": false, + "reason": "Generic i18n template vs unrelated Thai title; not the same issue." }, { "left": "issue:44365", "right": "issue:44370", "accept": false, - "reason": "Highly similar placeholder issues, but still no proof they are the same language/task." + "reason": "Near-identical placeholder titles, but no evidence they are the same language request." }, { "left": "issue:44365", "right": "issue:46222", "accept": false, - "reason": "Highly similar placeholder issues, but they may be separate translation requests created from the same template." + "reason": "Same template, but distinct translation tasks are plausible and should not be merged." + }, + { + "left": "issue:44365", + "right": "issue:46226", + "accept": false, + "reason": "Generic translation issue vs unrelated Thai text; not duplicates." }, { "left": "issue:44370", "right": "issue:46222", "accept": false, - "reason": "Exact placeholder match is not enough here; these could still be separate language-specific translation requests, so do not merge." + "reason": "Exact template match is not enough; could still be separate language-specific issues." + }, + { + "left": "issue:44370", + "right": "issue:46226", + "accept": false, + "reason": "Generic template issue vs unrelated Thai title; no duplicate signal." + }, + { + "left": "issue:46222", + "right": "issue:46226", + "accept": false, + "reason": "Generic i18n translation request vs unrelated Thai text; not the same underlying issue." } ] }, @@ -16632,8 +16661,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-36817-16", "nodes": [ @@ -16692,12 +16721,12 @@ "issue:43837|issue:45177", "issue:45322|issue:46133" ], - "prepared_review_unit_hash": "050753e8f3d5656ac78f6b5448008d0d644bb685b091e70b0f0c31a7e4cc95f6", + "prepared_review_unit_hash": "960c4a8c45a236dfad14c057b8df5f2898ebe5d7fda74da47fc03e15dd5c465d", "component_budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21658, + "serialized_chars": 21657, "estimated_input_tokens": 5415, "estimated_eval_tokens": 11086 }, @@ -16705,7 +16734,7 @@ "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21658, + "serialized_chars": 21657, "estimated_input_tokens": 5415, "estimated_eval_tokens": 11086 }, @@ -16714,228 +16743,228 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are a collection of separate model-support requests spanning different architectures and modalities (vision, OCR, ASR/TTS, audio, transformers). They are only superficially similar, so none of the soft edges look like true duplicates.", + "summary": "This cluster is a loose collection of model-support request issues, but the items are for different models and often different modalities/tasks. They are not duplicates of one another; at most they share the generic pattern of requesting new model additions.", "confidence": 0.97, - "canonical_issue_reason": "Issue 42738 ('BERT-like models with RoPE') is the broadest, most category-level request and is the closest thing to an umbrella representative for this noisy cluster, though it is not a duplicate of the others.", + "canonical_issue_reason": "No canonical issue: these are distinct model-request issues, not the same underlying bug/change.", "canonical_pr_reason": null, - "best_issue_reason": "42738 is the best single representative issue because it is the least model-specific and most general of the set; the others are distinct named-model requests that should stay separate.", + "best_issue_reason": "No single best issue stands out as a representative duplicate target; the requests span unrelated models, so a canonical merge point would be misleading.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "EuroBert config vs DEIMv2: different models and different request scopes." + "reason": "EuroBert config vs DEIMv2 request; different models and code paths." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert config vs Qwen3-TTS support: unrelated model families and modalities." + "reason": "BERT config request vs Qwen3-TTS support; different model families and tasks." }, { "left": "issue:36817", "right": "issue:36879", "accept": false, - "reason": "EuroBert vs RF-DETR: distinct models, no shared concrete change." + "reason": "EuroBert vs RF-DETR; unrelated architectures and modalities." }, { "left": "issue:36879", "right": "issue:41211", "accept": false, - "reason": "RF-DETR vs DEIMv2: different architectures and tasks." + "reason": "RF-DETR object detection vs DEIMv2; different concrete models." }, { "left": "issue:36879", "right": "issue:43909", "accept": false, - "reason": "RF-DETR detection request vs LFM2.5 Audio request are unrelated." + "reason": "RF-DETR vs LFM2.5 Audio; detection and audio are unrelated." }, { "left": "issue:36879", "right": "issue:44121", "accept": false, - "reason": "RF-DETR vs sparse transformer: different model types and code paths." + "reason": "RF-DETR vs weight-sparse transformer; no same underlying change." }, { "left": "issue:36879", "right": "issue:43837", "accept": false, - "reason": "RF-DETR detection request vs Qwen3-ASR support: different modalities." + "reason": "RF-DETR vs Qwen3-ASR; different models and tasks." }, { "left": "issue:36879", "right": "issue:43671", "accept": false, - "reason": "RF-DETR vs Qwen3-TTS support: different tasks and model families." + "reason": "RF-DETR vs Qwen3-TTS; different model families." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT vs Qwen3-ASR: unrelated model requests." + "reason": "NeoBERT config request vs Qwen3-ASR support; unrelated models." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT vs weight-sparse transformer: different concrete models." + "reason": "NeoBERT vs sparse transformer; different architecture requests." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT vs LFM2.5 Audio: unrelated architectures and modalities." + "reason": "NeoBERT vs LFM2.5 Audio; unrelated model types." }, { "left": "issue:36879", "right": "issue:37015", "accept": false, - "reason": "RF-DETR detection model vs NeoBERT language model: distinct requests." + "reason": "RF-DETR and NeoBERT are different models with different integration work." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT vs ModernVBERT: different model additions, not one bug/change." + "reason": "NeoBERT vs ModernVBERT; similar naming only, not the same model." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 vs Qwen3-ASR: different model support requests." + "reason": "MiniCPM3 architecture request vs Qwen3-ASR support; different targets." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 vs sparse transformer: unrelated concrete code changes." + "reason": "MiniCPM3 vs weight-sparse transformer; distinct model additions." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 vs ModernVBERT: separate model-specific requests." + "reason": "MiniCPM3 vs ModernVBERT; both model requests, but not duplicates." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 vs Voxtral TTS: different modalities and model families." + "reason": "MiniCPM3 vs Voxtral TTS; different model families and modalities." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 vs Qwen3-TTS: unrelated models and tasks." + "reason": "DEIMv2 vs Qwen3-TTS; unrelated model support requests." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT vs Qwen3-ASR: different support requests." + "reason": "ModernVBERT vs Qwen3-ASR; different models/tasks." }, { "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT vs Qwen3-TTS: different models and code paths." + "reason": "ModernVBERT vs Qwen3-TTS; no shared code-path problem." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT vs broad BERT-like RoPE request: related family, but not the same concrete change." + "reason": "Specific model request vs generic BERT-like/RoPE discussion; not duplicates." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "BERT-like models with RoPE vs Voxtral TTS: unrelated modalities." + "reason": "Generic BERT-like model request vs audio TTS model; unrelated." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS vs Qwen3-ASR: same vendor, different tasks and implementations." + "reason": "Qwen3-TTS and Qwen3-ASR are different tasks and likely separate implementations." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Qwen3-TTS vs Voxtral TTS: different model-specific support requests." + "reason": "Both are TTS-related, but they request different models and code paths." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "TTS support vs DeepSeek-OCR2: different modalities and code paths." + "reason": "Qwen3-TTS vs DeepSeek-OCR2; different model domains." }, { "left": "issue:43671", "right": "issue:44121", "accept": false, - "reason": "Qwen3-TTS vs weight-sparse transformer: unrelated model classes." + "reason": "Audio model support vs sparse transformer request; not the same change." }, { "left": "issue:43671", "right": "issue:43909", "accept": false, - "reason": "Qwen3-TTS vs LFM2.5 Audio: both audio-related but distinct model additions." + "reason": "Qwen3-TTS vs LFM2.5 Audio; both audio-related but different models." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR vs LFM2.5 Audio: different audio model support requests." + "reason": "Qwen3-ASR vs LFM2.5 Audio; different audio models and tasks." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR vs EUPE: ASR model request versus vision-language encoder request." + "reason": "ASR model request vs EUPE vision-language encoder; unrelated." }, { "left": "issue:43837", "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR vs sparse transformer: different concrete models." + "reason": "Qwen3-ASR support vs sparse transformer request; different change types." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "LFM2.5 Audio vs SAM3-LiteText: audio model versus vision-language model." + "reason": "Audio model request vs SAM3-LiteText; different modalities and code paths." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Sparse transformer request vs EUPE: different architectures and use cases." + "reason": "Weight-sparse transformer request vs EUPE encoder; not the same model or bug." }, { "left": "issue:43837", "right": "issue:45027", "accept": false, - "reason": "Qwen3-ASR vs Voxtral TTS: different tasks, not the same support change." + "reason": "Qwen3-ASR vs Voxtral TTS; both audio, but distinct tasks and models." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT vs Voxtral TTS: unrelated model families." + "reason": "NeoBERT config vs Voxtral TTS support; unrelated requests." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "ASR support vs OCR model support: different modalities." + "reason": "ASR support vs OCR2 model request; different model families." }, { "left": "issue:45322", "right": "issue:46133", "accept": false, - "reason": "EUPE vs TIPSv2: both vision-language related, but they are distinct model proposals with different changes." + "reason": "EUPE and TIPSv2 are different vision-language models; not the same underlying change." } ] }, @@ -16954,8 +16983,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-41084-2", "nodes": [ @@ -16965,7 +16994,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "34bc3063617364029e19ecad276c43ce05079dba3d4d7d3c60fb30cf00c4b6b3", + "prepared_review_unit_hash": "821b46ba1d6c91d6ac20e6b66383e8caaf38ae153b2457351055f8ccee9b5fba", "component_budget": { "node_count": 2, "item_count": 2, @@ -16987,18 +17016,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are not duplicates: #41084 is about block decoding behavior, while #43037 is a request for DeepSeek v3.2 support. They describe different underlying work with no clear shared code path.", - "confidence": 0.98, - "canonical_issue_reason": "No canonical issue: the titles and likely scopes are unrelated, so neither issue can stand in for the other.", + "summary": "The two issues are not duplicates: one is about enabling block decoding, the other is a request for DeepSeek v3.2 model support. They describe different features/problems with no shared targets or code path evidence.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #41084 is the more technical, implementation-oriented request and is the better representative of the cluster if one must be chosen, though it is not actually the same underlying problem as #43037.", "canonical_pr_reason": null, - "best_issue_reason": "No single best issue for deduplication; #43037 is more specific as a standalone feature request, but it does not overlap with #41084.", + "best_issue_reason": "Issue #41084 is the best single issue candidate because it is more concrete and likely to map to an actionable code change; #43037 is a separate model-support request.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different topics and likely different code paths: block decoding vs. model-specific support request." + "reason": "Different topics and likely different implementation paths: block decoding vs DeepSeek v3.2 support. No explicit links, shared targets, filenames, or bug signature indicate the same underlying issue." } ] }, @@ -17017,8 +17046,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-43598-2", "nodes": [ @@ -17028,7 +17057,7 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "89389ba2c07ccbc79da1b3829b4407b65fdf0ab09ed3b549a96c7ca986be465e", + "prepared_review_unit_hash": "c0e5508c1f4822971e9ede5aa5046d65b0dc84485d6215f7ccd51dfffbe3bf1f", "component_budget": { "node_count": 2, "item_count": 2, @@ -17050,18 +17079,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues are about different Trainer behaviors: one is about `remove_unused_column` and customizability, the other about FLOPS estimation for custom trainers. They are related to Trainer internals but do not describe the same bug or change.", - "confidence": 0.95, + "summary": "These are related only at a high level (custom trainers in Trainer), but they describe different problems: one is about `remove_unused_column` customizability, the other about FLOPS estimation handling custom trainers. Not duplicates.", + "confidence": 0.91, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No clear canonical issue for this cluster because the reports target different code paths; if forced to pick one, `#43598` is the more focused configuration-behavior report.", + "best_issue_reason": "Issue 43598 is the more central and broadly framed Trainer customization concern; issue 43600 is a narrower follow-on about FLOPS estimation rather than the core customization behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Different underlying problems: one concerns `remove_unused_column` customization in Trainer, the other concerns accounting for custom trainers in FLOPS estimation. Same subsystem, but not the same bug." + "reason": "They share the theme of custom trainers, but the concrete issues differ: one targets `remove_unused_column` behavior, the other targets FLOPS estimation logic. Not the same underlying bug or change." } ] }, @@ -17080,8 +17109,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-43684-2", "nodes": [ @@ -17091,7 +17120,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "1c6dafd3271d6c14d3ead1d5750bc33b17e7e25f763a6012ac3269595e389e34", + "prepared_review_unit_hash": "a0c78dc5e184c9f67fc294b2025bd85402428f14de21f09c6380b00864748cb7", "component_budget": { "node_count": 2, "item_count": 2, @@ -17113,18 +17142,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are separate feature requests, not duplicates: one asks for Qwen3-Omni registration on AutoModel/AutoModelForConditionalGeneration, while the other asks for AutoModelForSequenceClassification support for Qwen3.5 (Qwen3_5Config). They target different model families and different heads/code paths.", - "confidence": 0.95, - "canonical_issue_reason": null, + "summary": "These are not duplicates: one requests auto-class registration for Qwen3-Omni, while the other asks for sequence-classification support for Qwen3.5. They touch different model families and different AutoModel paths.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 43684 is the broader and more central request around auto-model registration, so it is the better representative issue if one must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "If one issue must represent the cluster, issue 43684 is the broader core-registration request, but it is still distinct from 44405 and should not be treated as a duplicate.", + "best_issue_reason": "Issue 43684 is the cleaner cluster representative because it concerns generic AutoModel/AutoModelForConditionalGeneration registration rather than a single task-specific head.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different model families and different AutoModel heads: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration vs Qwen3.5 sequence-classification support. Not the same underlying bug/change." + "reason": "Different underlying changes: Qwen3-Omni auto registration vs Qwen3.5 sequence-classification support. They are not the same bug or fix and would not be merged into one PR." } ] }, @@ -17143,8 +17172,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-43835-4", "nodes": [ @@ -17161,76 +17190,76 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "da62a0b1772c8bd2a65ee232dd22141c5491af492cfa4c133ef3af9b8bd19904", + "prepared_review_unit_hash": "249b3f80038f0322f7a1a52c9700b2d05986153c6b7f79c1e8a2cc11e8a20352", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4264, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4264, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Three issues titled \"huggingface\" look like the same duplicate report; the separate \"Fett\" issue does not. Among the duplicate-looking issues, #43859 is the best representative because it has the same title and at least some discussion activity.", - "confidence": 0.77, - "canonical_issue_reason": "Issue #43859 is the best canonical choice: it matches the repeated title and has the most activity among the duplicate-looking issues.", + "summary": "One issue (\u201cFett\u201d) appears unrelated. The three \u201chuggingface\u201d issues look like a small duplicate cluster based on identical titles and very similar creation timing/body size.", + "confidence": 0.83, + "canonical_issue_reason": "issue:43845 is the earliest of the near-identical \u201chuggingface\u201d issues, so it is the best representative for the duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": "#43859 is the strongest global issue candidate because it appears to be the same report as #43845/#43846 and has the most discussion.", + "best_issue_reason": "issue:43845 is the clearest cluster anchor: same title as the other likely duplicates, earliest timestamp, and no evidence that it differs materially from them.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Different titles and no body detail tying them to the same underlying issue." + "reason": "Different title; no evidence these describe the same underlying issue." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Different titles and no evidence that \"Fett\" refers to the same problem as the \"huggingface\" reports." + "reason": "Different title; no indication of the same bug or request." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "No substantive overlap beyond both being short issues; titles do not match." + "reason": "Different title and no shared concrete issue details." }, { "left": "issue:43845", "right": "issue:43846", "accept": true, - "reason": "Same title and highly similar sparse issue content, so they likely describe the same report." + "reason": "Identical title, nearly identical creation time, and matching body size strongly suggest the same underlying issue." }, { "left": "issue:43845", "right": "issue:43859", "accept": true, - "reason": "Same title and same apparent topic; no evidence of a distinct issue." + "reason": "Same title and clearly overlapping metadata; likely duplicate reports of the same issue." }, { "left": "issue:43846", "right": "issue:43859", "accept": true, - "reason": "Same title and same apparent topic; likely duplicate reports of the same issue." + "reason": "Identical title and strong similarity in metadata indicate the same underlying issue." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: #43845, #43846, and #43859 share the same title and strong soft-similarity signals, while #43835 has a different title and no stronger evidence tying it to the others. The representative choice of #43859 is conservative and supported by its slightly higher discussion activity. The pairwise verdicts are also suitably cautious given the sparse evidence." + "feedback": "The summary and pairwise verdicts are grounded in the packet-level metadata: issue:43835 has a different title, while issue:43845/43846/43859 share the same title and close timestamps, with body-length similarity supporting a conservative duplicate-cluster call. The canonical choice of issue:43845 as the earliest representative is also supported." }, "evaluator_used": true, "retried": false, @@ -17246,8 +17275,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-44371-5", "nodes": [ @@ -17268,88 +17297,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "f94db69eb35e276fff23d89b44f2f057e199a620a3aef1466def79e8f8528e37", + "prepared_review_unit_hash": "2383e2cb1af7c8fa6c8282c53b4f7d46792ded1accacdb6b7632a0d62a6d42ef", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5785, - "estimated_input_tokens": 1447, - "estimated_eval_tokens": 3150 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5785, - "estimated_input_tokens": 1447, - "estimated_eval_tokens": 3150 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster contains only vague/spam-like issues and no concrete bug report. The titles do not indicate the same underlying problem, so none of the soft pairs should be merged as duplicates.", - "confidence": 0.98, - "canonical_issue_reason": "No canonical issue: the items are unrelated or too content-free to establish a shared bug or request.", + "summary": "The cluster is heterogeneous: multiple spam placeholders, one generic documentation suggestion, and one unrelated company-name request. There is no shared concrete bug or change, so no duplicate canonicalization is justified.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No strong best issue exists; #44418 is only marginally more descriptive, but still too vague to represent the cluster.", + "best_issue_reason": "Issue 44371 is only marginally the best representative because it is an earliest, minimal placeholder, but the cluster is too mixed to treat any item as a true canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both are spam-like placeholders with no concrete problem description; no evidence they describe the same issue." + "reason": "Both are spam placeholders, but there is no concrete underlying bug or change to unify." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Both are spam-like placeholders with no concrete problem description; no evidence they describe the same issue." + "reason": "Same as above: identical placeholder titles do not establish the same issue." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "One is spam and the other is an extremely vague documentation suggestion; not the same underlying issue." + "reason": "Spam placeholder versus a documentation enhancement request; different topics." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "Completely different titles and no shared technical context; not duplicates." + "reason": "Spam placeholder versus a request about 'Google LLC'; no shared problem statement." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "A vague documentation request and a spam placeholder are not the same request." + "reason": "Generic docs enhancement and spam placeholder are unrelated." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "A vague documentation request and a spam placeholder are not the same request." + "reason": "Generic docs enhancement and spam placeholder are unrelated." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "Both are spam-like placeholders, but there is no content showing they refer to the same bug or change." + "reason": "Both are placeholders, but that is not enough to conclude the same underlying issue." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "No overlapping subject matter; one is spam-like, the other references a different named topic." + "reason": "Spam placeholder versus unrelated 'Google LLC' request." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "No overlapping subject matter; one is spam-like, the other references a different named topic." + "reason": "Spam placeholder versus unrelated 'Google LLC' request." } ] }, @@ -17368,8 +17397,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-45145-2", "nodes": [ @@ -17379,50 +17408,50 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "46b72ddb7de1090daa29d22822c72ee2ec0072229840af5126762d49935506cf", + "prepared_review_unit_hash": "b444870143764218620fedd0c6e661e8977edec4b9595175239ec07659d06f52", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1645, - "estimated_input_tokens": 412, - "estimated_eval_tokens": 1080 + "serialized_chars": 1644, + "estimated_input_tokens": 411, + "estimated_eval_tokens": 1078 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1645, - "estimated_input_tokens": 412, - "estimated_eval_tokens": 1080 + "serialized_chars": 1644, + "estimated_input_tokens": 411, + "estimated_eval_tokens": 1078 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two closed issues have identical titles and very similar body lengths, and #45145 was created first with an inbound reference, but the available packet does not include body text or other strong proof that they are the same report. The soft link is therefore too weak to accept confidently.", - "confidence": 0.81, - "canonical_issue_reason": "#45145 is the better canonical issue because it was created earlier and has an inbound reference, while #45151 appears to be the later, likely related report. The evidence is suggestive, but not strong enough to claim near-identical content.", + "summary": "The two closed issues look like likely duplicates: they have the same title and a strong text-similarity signal, with only limited metadata differences. The evidence is enough to cluster them together, but not enough to claim they are proven identical without seeing the full bodies.", + "confidence": 0.87, + "canonical_issue_reason": "Issue #45145 is the better canonical representative because it was created earlier and matches #45151 on title plus a strong similarity signal, making it the safer anchor for this likely duplicate pair.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is the strongest issue representative for the cluster due to earlier creation and the only inbound reference. It is the safest canonical choice given the limited evidence.", + "best_issue_reason": "#45145 is the best issue for the cluster because it is the earliest report and appears to be the same likely duplicate topic as #45151, based on the matching title and similarity signal.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", - "accept": false, - "reason": "Same title and close body lengths suggest similarity, but the packet lacks body text, explicit links, or other concrete evidence that these are the same underlying issue. That is not enough to accept the edge." + "accept": true, + "reason": "Accept as a likely duplicate: the titles are identical and the packet shows a solid soft-similarity signal, but the evidence is limited because the bodies are not shown and there is some distinguishing metadata." } ] }, "evaluator_result": { "accept": false, - "feedback": "The cluster looks plausibly duplicate, but the analyst overstates the evidence. The packet supports identical titles, close body lengths, and later creation for #45151, plus a soft-similarity signal; it does not provide body text or other proof that the content is \u2018nearly identical\u2019 or that the relationship is \u2018essentially exact.\u2019 The soft-edge accept is too strong for the available evidence, so this should be rejected as insufficiently conservative." + "feedback": "The clustering is plausible, but the writeup overstates the evidence. The packet only shows a soft_similarity signal (Jaccard ~0.63) plus matching titles and similar body lengths; it does not show the body contents, and there is distinguishing metadata such as different creation times and inbound references. Please soften claims like \"same underlying issue,\" \"fully equivalent,\" and \"no distinguishing metadata\" to reflect that this is only a likely duplicate, not a proven match." }, "evaluator_used": true, "retried": true, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -17488,8 +17517,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-45672-2", "nodes": [ @@ -17499,7 +17528,7 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "e86c262675e0651be4f58517e0204b7461c73f5e6d094baa4e04c89124b1ec02", + "prepared_review_unit_hash": "808fd1470b07eea820021d7b2bef0383affb059ecbd801d0ed9d1aae5d492310", "component_budget": { "node_count": 2, "item_count": 2, @@ -17521,18 +17550,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two issues describe different 4-bit inference failures in different code paths: one is a torch.finfo() TypeError triggered by uint8 weights in Gemma4 audio modules, the other is an AssertionError caused by llm_int8_skip_modules changing the default lm_head exclusion. They should not be deduplicated.", + "summary": "These are related only at a very broad level (4-bit inference / quantization), but they describe different bugs: one is a Gemma4 audio-module uint8/torch.finfo TypeError, the other is a BitsAndBytesConfig skip-module regression that drops the default lm_head exclusion and triggers an AssertionError.", "confidence": 0.97, - "canonical_issue_reason": null, + "canonical_issue_reason": "issue:45674 is the better anchor for the cluster because it is the more general configuration regression and has higher discussion/activity; issue:45672 is a model-specific Gemma4 failure with a different root cause.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45674 has the clearer, broader repro and is more likely to serve as the representative report for quantized-inference breakage, while issue:45672 is narrowly scoped to Gemma4 audio modules.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Different root causes and code paths: Gemma4 audio-module uint8/torch.finfo() failure vs. BitsAndBytesConfig skip-module handling that clears lm_head exclusion. Same broad 4-bit inference area, but not the same bug." + "reason": "Reject: both concern 4-bit inference, but the failure modes and code paths differ substantially (Gemma4 audio-module torch.finfo on uint8 vs. BitsAndBytesConfig llm_int8_skip_modules clearing lm_head exclusion)." } ] }, @@ -17551,8 +17580,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-45762-2", "nodes": [ @@ -17562,7 +17591,7 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "e106d292ca186bdaa15a5f5c40434590b993a0fe95f331b0cfec395ec46411a6", + "prepared_review_unit_hash": "9234982dd41d7ff6c7ad6313a07048f34b5f8f4e751d2b39f309b9d6356d90b8", "component_budget": { "node_count": 2, "item_count": 2, @@ -17584,18 +17613,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are unrelated entries with different personal names and dates, and contain no bug/report context. This is not a duplicate cluster.", + "summary": "These two closed issues do not appear to be duplicates: they have different names and dates, with no shared problem description or technical context.", "confidence": 0.98, - "canonical_issue_reason": null, + "canonical_issue_reason": "Neither issue is a clear canonical duplicate target; the titles refer to different people and contain no evidence of the same underlying issue.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Neither issue is suitable as a global best issue for this cluster because the items look unrelated rather than duplicate reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "Different names and birthdates; no shared technical problem, code context, or explicit linkage." + "reason": "Different names and dates, no common bug report or matching context; this looks unrelated rather than the same issue." } ] }, @@ -17614,8 +17643,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-45764-2", "nodes": [ @@ -17625,7 +17654,7 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "4b7a72edc099ab82626a2b51867a843381ac540c371993ab97a6e7d052d84ff3", + "prepared_review_unit_hash": "d4ba2403b946c1deec5746aa5a57f1cddd53625dc0301bc611c8cdaaa53e8788", "component_budget": { "node_count": 2, "item_count": 2, @@ -17647,18 +17676,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues have unrelated titles consisting of different personal names and dates, with no body content or shared problem evidence. This looks like accidental similarity rather than duplicate reports.", - "confidence": 0.98, - "canonical_issue_reason": "No clear canonical issue: neither issue describes a bug or task, and the titles reference different people/dates with no shared underlying problem.", + "summary": "The two issues are just separate name/date titles and do not describe the same bug, change, or request. They should not be clustered together.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a strong canonical candidate; if one must be chosen for bookkeeping, issue #45764 is marginally earlier, but there is no substantive basis to prefer it.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "Different names and dates, no body text, and no evidence of the same underlying issue." + "reason": "Different personal-name/date titles with no shared problem statement, target, or context; not the same underlying issue." } ] }, @@ -17677,8 +17706,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "cache", - "cache_hit": true, + "source": "llm", + "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-45869-5", "nodes": [ @@ -17698,12 +17727,12 @@ "issue:45871|issue:45873", "issue:45872|issue:45873" ], - "prepared_review_unit_hash": "74d8434447fc560323b09c2d48e3089d88333f33194de5e561a4128b812d2e00", + "prepared_review_unit_hash": "85f5d96ed41f20dcaef927d62d21a86f9a18172a4cf55d5fe1efc19c8fbb8b0f", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5728, + "serialized_chars": 5727, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -17711,7 +17740,7 @@ "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5728, + "serialized_chars": 5727, "estimated_input_tokens": 1432, "estimated_eval_tokens": 3120 }, @@ -17720,66 +17749,66 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This set splits into two separate exact-duplicate issue pairs: 45870/45871 and 45872/45873. Issue 45869 is a different dataset-promo item and is not a duplicate of the others.", - "confidence": 0.96, - "canonical_issue_reason": "45872 is the strongest representative of the substantive request thread: it is a concrete DataForge feature request and has an exact duplicate in 45873, while the 45870/45871 pair is a different example-oriented request and 45869 is unrelated.", + "summary": "This cluster contains two separate exact-duplicate pairs: 45870/45871 and 45872/45873. Issue 45869 is a different financial dataset promotion/sample request and does not match the DataForge example/service issues.", + "confidence": 0.98, + "canonical_issue_reason": "45870 is the earliest and cleanest representative of the exact-duplicate pair titled 'Example: Financial time series prediction with DataForge'.", "canonical_pr_reason": null, - "best_issue_reason": "45872 is the best single issue to anchor triage because it is the clearest, more specific request among the near-duplicates and most likely to capture the intended code-change discussion.", + "best_issue_reason": "45870 is the strongest single issue to keep as a canonical representative because it is an exact duplicate target, earliest created, and has a fully generic title for the repeated example issue.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45869", "right": "issue:45870", "accept": false, - "reason": "Different topics: a financial data bundle promotion vs a DataForge example request." + "reason": "Different topics: a dataset bundle promo/sample vs a DataForge example issue." }, { "left": "issue:45869", "right": "issue:45871", "accept": false, - "reason": "Different topics: a financial data bundle promotion vs a DataForge example request." + "reason": "Different topics: a dataset bundle promo/sample vs a DataForge example issue." }, { "left": "issue:45870", "right": "issue:45871", "accept": true, - "reason": "Exact duplicate titles/body length and same underlying example request." + "reason": "Same title and same underlying example issue; exact duplicate." }, { "left": "issue:45870", "right": "issue:45872", "accept": false, - "reason": "Related domain, but different underlying request: example prediction flow vs scraping-service feature." + "reason": "Related financial/DataForge theme, but this is a different requested feature/topic: example prediction vs scraping service." }, { "left": "issue:45870", "right": "issue:45873", "accept": false, - "reason": "Related domain, but different underlying request: example prediction flow vs scraping-service feature." + "reason": "Same as 45872: different underlying issue, not the same concrete request." }, { "left": "issue:45871", "right": "issue:45872", "accept": false, - "reason": "Related domain, but different underlying request: example prediction flow vs scraping-service feature." + "reason": "Different underlying requests; exact duplicate only with 45870, not with the scraping-service issue." }, { "left": "issue:45871", "right": "issue:45873", "accept": false, - "reason": "Related domain, but different underlying request: example prediction flow vs scraping-service feature." + "reason": "Different underlying requests; exact duplicate only with 45870, not with the scraping-service issue." }, { "left": "issue:45872", "right": "issue:45873", "accept": true, - "reason": "Exact duplicate titles/body length and same underlying scraping-service request." + "reason": "Same title and same underlying DataForge scraping-service request; exact duplicate." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet: 45870/45871 and 45872/45873 have identical titles and body lengths, while 45869 is clearly a different topic. The soft-edge verdicts are appropriately conservative overall, with the two accepted pairs supported by the strongest available evidence and the rejected cross-pairs remaining within distinct request themes." + "feedback": "The summary is grounded in the packet: 45870/45871 and 45872/45873 are well-supported exact-duplicate pairs by identical titles and body similarity, while 45869 is clearly a different dataset-promo request. The soft-edge rejections between the two topic clusters are conservative and consistent with the available evidence." }, "evaluator_used": true, "retried": false, @@ -17790,7 +17819,7 @@ { "label": "pull_request", "component_index": 1, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -17834,12 +17863,12 @@ "pull_request:43823|pull_request:45213", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "bb61d71e3e8c584c84a1795f4f681546c95fcdca68100e53dcfed976ee4c17c2", + "prepared_review_unit_hash": "45115f25b79dd681b8bc046b985fe23add2af7e5d3a7df5f868842205351f9e1", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 18, - "serialized_chars": 34776, + "serialized_chars": 34774, "estimated_input_tokens": 8694, "estimated_eval_tokens": 17644 }, @@ -17847,7 +17876,7 @@ "node_count": 13, "item_count": 13, "soft_pair_count": 18, - "serialized_chars": 34776, + "serialized_chars": 34774, "estimated_input_tokens": 8694, "estimated_eval_tokens": 17644 }, @@ -17856,126 +17885,126 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "One true duplicate-like pair exists: the two `fixed_cross_entropy` patches. The rest are unrelated model additions, refactors, or test/docs changes despite superficial similarity.", + "summary": "Only one pair appears to be a true duplicate/fork: the two fixed_cross_entropy PRs for issue 43240. The rest are unrelated PRs across different models or features and should not be merged.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the strongest representative because it directly targets the reported issue 43240 and applies the fix at the exact `fixed_cross_entropy` call site with a complete kwargs plumbing change.", + "canonical_pr_reason": "PR 43251 is the best canonical representative: it explicitly targets issue 43240 and contains the clearer direct fix to fixed_cross_entropy.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the best overall pick: it addresses a concrete bug, has an explicit issue target, and is the clearest standalone fix in the set.", + "best_pr_reason": "PR 43251 is the strongest standalone PR in this set for the same reason: explicit issue linkage and a concrete, minimal change to the shared loss path.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Unrelated PRs: new VideoPrism model integration vs an Executorch dynamic-shape test. No shared bug or mergeable code-path." + "reason": "Different models/features (VideoPrism docs/model add vs ExecuTorch dynamic-shape test); not the same underlying change." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Mistral tokenizer API additions and Executorch export tests are unrelated changes." + "reason": "Tokenizer API addition vs ExecuTorch export test; unrelated code paths." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Tokenizer method additions for MistralCommon do not overlap with adding a new MobileLLM model." + "reason": "Mistral tokenizer helper methods vs adding MobileLLM model files; different subsystems and goals." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor robustness fixes across existing models are not the same change as introducing MobileLLM." + "reason": "Processor robustness changes vs new MobileLLM model implementation; not the same bug or change." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Cross-entropy kwargs plumbing is unrelated to adding MobileLLM model files and tests." + "reason": "Cross-entropy kwargs fix vs adding a new model; unrelated." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Different bug domains: loss helper kwargs vs Mistral4 test/model adjustments." + "reason": "Loss utility fix vs Mistral4 test/model work; different concrete problems." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Executorch dynamic-shape testing is unrelated to Mistral4 test/model fixes." + "reason": "ExecuTorch dynamic-shape test vs Mistral4 test/model work; only broad testing theme overlaps." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "A dynamic export test and a new model integration are not the same underlying change." + "reason": "ExecuTorch export test vs MobileLLM model addition; unrelated." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New MobileLLM model support and config-resolution logic in `modeling_utils` are unrelated." + "reason": "New MobileLLM model addition vs config-resolution fix in modeling_utils; no shared code path." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM model addition is unrelated to Falcon output-interface refactoring." + "reason": "New MobileLLM model addition vs Falcon output-interface refactor; unrelated model implementations." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon model refactor is unrelated to the model-creation skill documentation/workflow PR." + "reason": "Falcon refactor vs model-creation skill/docs; no common underlying bug or change." }, { - "left": "pull_request:44827", - "right": "pull_request:45170", + "left": "pull_request:43995", + "right": "pull_request:44827", "accept": false, - "reason": "Mistral4 test fixes and `layrnorm` renaming are different concrete changes in different subsystems." + "reason": "Falcon output-collection refactor vs Mistral4 test fixes; different code paths." }, { "left": "pull_request:44827", - "right": "pull_request:46048", + "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test/model work is unrelated to adding a TDT loss kernel integration." + "reason": "Mistral4 test fixes vs layrnorm->layernorm renaming across CLIP-like models; unrelated fixes." }, { "left": "pull_request:44827", - "right": "pull_request:45655", + "right": "pull_request:46048", "accept": false, - "reason": "No shared underlying bug: Mistral4 test fixes versus config-class resolution ordering." + "reason": "Mistral4 test fixes vs TDT loss kernel integration; different functionality." }, { "left": "pull_request:44827", - "right": "pull_request:45213", + "right": "pull_request:45655", "accept": false, - "reason": "Model tests/fixes are not the same as the model creation skill PR." + "reason": "Mistral4 test fixes vs config resolution ordering fix; not the same issue." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Shared auto-model file aside, these are not the same concrete fix; one is Mistral4 test work, the other is a model-creation skill overhaul." + "reason": "Mistral4 test/model work vs model-creation skill repository changes; unrelated." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "New MobileLLM implementation and a workflow/skill PR are unrelated." + "reason": "New MobileLLM model addition vs model-creation skill; the latter is tooling/docs, not the model change itself." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both patch the same `fixed_cross_entropy` code path to forward supported kwargs to `nn.functional.cross_entropy`; they are near-identical fixes for the same underlying bug." + "reason": "Both PRs fix the same concrete loss-path issue in src/transformers/loss/loss_utils.py by extending fixed_cross_entropy to forward supported cross-entropy kwargs for issue 43240; they are plausible duplicates." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet, and the only accepted pair is a plausible duplicate/mergeable same-bugfix match (`fixed_cross_entropy` in the same file with the same issue target). The other pair judgments are conservative and consistent with the diffs shown." + "feedback": "The verdicts are grounded in the packet: the only mergeable same-bug pair is PR 43251/43254 on loss_utils.py for issue 43240, while the other pairs involve distinct model additions, tests, or tooling changes and are conservatively rejected." }, "evaluator_used": true, "retried": false, @@ -17986,7 +18015,7 @@ { "label": "pull_request", "component_index": 2, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18002,7 +18031,7 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "609edf3ab9010d3b524fe7db0fb7e7ca33e502e7709db36e266341b5cfb3a201", + "prepared_review_unit_hash": "b83a5fb80fef75147d1766147c0d5381ca4f18bb0027bff2c8fbae2039cc5baf", "component_budget": { "node_count": 2, "item_count": 2, @@ -18024,18 +18053,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Related VibeVoice work, but not the same change: PR 40546 introduces the base VibeVoice implementation, while PR 42781 adds a separate realtime/acoustic-tokenizer extension and extra generation plumbing. They share files because 42781 builds on the same family, but they are not duplicates.", - "confidence": 0.93, + "summary": "These are related VibeVoice PRs, but they are not duplicates: one is the base model implementation and the other is a later realtime/tokenizer extension.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR for dedupe: 40546 is the base VibeVoice implementation, while 42781 is a distinct realtime extension with additional tokenizer/feature-extraction and generation changes.", + "canonical_pr_reason": "PR 40546 is the foundational VibeVoice implementation and the broader, earlier change; PR 42781 is a later realtime-focused extension built on the same model family.", "best_issue_reason": null, - "best_pr_reason": "No single best duplicate PR: 42781 is broader, but it targets a different concrete feature scope than 40546 rather than the same bug/change.", + "best_pr_reason": "PR 40546 is the best representative for the cluster because it is the main initial integration, has substantial review activity, and is not a draft.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Reject: these are related VibeVoice-family PRs, but one implements the base model and the other adds realtime/acoustic-tokenizer support; not the same underlying change and not mergeable as one duplicate PR." + "reason": "Do not merge: 40546 adds VibeVoice support, while 42781 adds VibeVoice Realtime and the acoustic tokenizer. They share model-family files but target different functionality and would not plausibly be one PR." } ] }, @@ -18049,7 +18078,7 @@ { "label": "pull_request", "component_index": 3, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18065,7 +18094,7 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "76db3861fd7795142acb60d862798f8021c2a5f50120f3f79774095bb9022720", + "prepared_review_unit_hash": "9e53778aa5b28f61297df49e9f8de9f36fa44e8b73db64ca624425119445c80d", "component_budget": { "node_count": 2, "item_count": 2, @@ -18087,18 +18116,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are unrelated PRs: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. The shared \"add support\" phrasing is superficial.", + "summary": "These PRs are unrelated: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. They do not fix the same bug or implement the same change, so this is not a duplicate cluster.", "confidence": 0.99, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 41224 is the cleaner representative: it adds a single, concrete image-classification capability with explicit model/AutoModel wiring, docs, and tests. PR 40755 is a separate, broader forecasting feature in an unrelated model family.", + "canonical_pr_reason": "If a representative PR is needed, #41224 is the more focused and better-scoped change; #40755 is a separate TimesFM feature in a different code path and model family.", "best_issue_reason": null, - "best_pr_reason": "PR 41224 is the most self-contained and mergeable PR in this cluster; it has a narrowly scoped codepath and clear end-user behavior.", + "best_pr_reason": "#41224 is the stronger standalone PR to keep as the cluster representative because it is narrower and has substantial review activity, but it is not a duplicate of #40755.", "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different model families and tasks (TimesFM forecasting with covariates vs DINOv3 image classification); they do not fix the same concrete codepath and would not plausibly merge into one PR." + "reason": "Different models, different functionality, and different code paths: TimesFM covariate forecasting vs DINOv3 image classification support. They are not the same underlying change." } ] }, @@ -18112,7 +18141,7 @@ { "label": "pull_request", "component_index": 4, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18175,7 +18204,7 @@ { "label": "pull_request", "component_index": 5, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18193,7 +18222,7 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "c61376cf12bfb17897ff43931630286b0601abd2d794d726ed1d86b7c723f1c4", + "prepared_review_unit_hash": "cbf5c2858422f83689c1238420600c8a5592a66987670a0f7fa7499375570d39", "component_budget": { "node_count": 3, "item_count": 3, @@ -18215,24 +18244,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-addition PRs for different architectures (DeepSeek OCR, Isaac, Kimi2-6). They overlap in Transformers boilerplate, but they are not the same underlying change or bug fix.", - "confidence": 0.97, + "summary": "These are three separate model-addition PRs (DeepSeek OCR, Isaac, Kimi2-6). They share Transformers boilerplate and auto-registration changes, but they do not fix the same underlying bug or implement the same concrete model path, so they should not be clustered as duplicates.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45630 is the most representative cluster member: it has the highest review activity and a broad integration footprint across auto-mappings and model files, but it still corresponds to a distinct model addition rather than a duplicate of the others.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #45630 is the strongest standalone PR in the set because it appears most mature and comprehensive, with substantial framework integration and the most discussion, making it the best representative despite not being merge-equivalent to the others.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Both add different models (DeepSeek OCR vs Isaac). Shared docs/auto-registration changes are generic model-integration boilerplate, not the same code-path fix or a PR that could be merged into one." + "reason": "Different models and codepaths: DeepSeek OCR vs Isaac. Shared docs/auto-mapping scaffolding is generic model-onboarding work, not the same change." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "Isaac and Kimi2-6 are unrelated model additions. They touch similar Transformers plumbing, but the concrete implementation, model package, and tests are for different architectures, so they should not be merged together." + "reason": "Different model families and implementations: Isaac vs Kimi2-6. Overlap is limited to standard integration files, which is insufficient for duplicate PRs." } ] }, @@ -18246,7 +18275,7 @@ { "label": "pull_request", "component_index": 6, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18262,7 +18291,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "a5284344bf2caeb25acc01908e3fad2dda2286ccf053c6cb1d44151f035bc6b6", + "prepared_review_unit_hash": "d5520462a5e32a6c2b5270c31c7aba0fa12ea20d9c78fb2f10f3a62c814c8279", "component_budget": { "node_count": 2, "item_count": 2, @@ -18284,18 +18313,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related GitHub Actions maintenance PRs, but they change different action families and address different upgrade goals. PR 42975 is a broad Node 24 compatibility sweep for workflow actions, while PR 42976 is a separate upgrade of docker/build-push-action and a few other workflow actions to newer versions. Similar theme, not the same underlying change.", - "confidence": 0.95, + "summary": "These are related workflow-maintenance PRs, but they are not duplicates: one is a broad GitHub Actions upgrade for Node 24 compatibility, while the other is a narrower refresh of docker/build-push-action versions in docker image workflows.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 42975 is the broader, more clearly scoped representative of the cluster: it upgrades workflow actions specifically for Node 24 compatibility across many files.", + "canonical_pr_reason": "PR #42975 is the better canonical representative because it is the broader, more clearly scoped workflow-action compatibility upgrade across many files, centered on Node 24 support.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the best single representative because it covers the wider cross-workflow action compatibility update and has the clearest umbrella purpose.", + "best_pr_reason": "PR #42975 is the best cluster anchor: it captures the main infra-wide dependency modernization theme and covers the widest shared workflow surface, whereas #42976 is a separate docker-action version bump.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Both are workflow dependency bump PRs, but they target different actions and different concrete changes: 42975 upgrades checkout/cache/upload-artifact for Node 24 compatibility, while 42976 upgrades docker/build-push-action and related workflow versions. They are similar maintenance work, not the same fix and not something that would naturally merge into one PR." + "reason": "They touch overlapping GitHub Actions workflows, but the concrete changes differ: #42975 upgrades many actions for Node 24 compatibility, while #42976 mainly bumps docker/build-push-action in docker image workflows. They are adjacent maintenance PRs, not the same underlying change." } ] }, @@ -18309,7 +18338,7 @@ { "label": "pull_request", "component_index": 7, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18353,154 +18382,154 @@ "pull_request:46162|pull_request:46214", "pull_request:46162|pull_request:46225" ], - "prepared_review_unit_hash": "e357f60a7dfdb7dbc38df3c4f649d4d83294352cd58cf9d2d889e4e2553b5935", + "prepared_review_unit_hash": "6d3711fd08e70be02f7d202ca77275518449b905332eaafcbf11a1ed7f71f383", "component_budget": { "node_count": 11, "item_count": 11, "soft_pair_count": 20, - "serialized_chars": 34562, - "estimated_input_tokens": 8641, - "estimated_eval_tokens": 17538 + "serialized_chars": 34733, + "estimated_input_tokens": 8684, + "estimated_eval_tokens": 17624 }, "budget": { "node_count": 11, "item_count": 11, "soft_pair_count": 20, - "serialized_chars": 34562, - "estimated_input_tokens": 8641, - "estimated_eval_tokens": 17538 + "serialized_chars": 34733, + "estimated_input_tokens": 8684, + "estimated_eval_tokens": 17624 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a grab bag of unrelated PRs: several new model additions and several XPU/test expectation tweaks. They share broad tooling or subsystem areas, but not the same concrete bug/change, so none of the soft edges should be merged.", - "confidence": 0.96, + "summary": "This cluster is not a true duplicate set: it mixes several unrelated model-addition PRs and several separate test/expectation tweaks. The only thematic overlap is the small XPU/distributed-test expectation subgroup, but those still target different test suites and code paths.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43451 is the strongest representative artifact in the cluster: it is the most complete standalone model-addition PR, with broad implementation, processor/image/video support, tests, and heavy review activity.", + "canonical_pr_reason": "No single PR is a real duplicate canonical, but PR 46109 is the closest representative of the only tightly related subset because it is the broadest of the XPU/distributed-test expectation updates and connects to the other expectation PRs more than the model-addition PRs do.", "best_issue_reason": null, - "best_pr_reason": "PR 43451 is the best overall standalone PR in this mixed cluster because it has the broadest and most mature scope, making it the clearest exemplar even though it is not a duplicate of the others.", + "best_pr_reason": "PR 46109 is the best global representative of the cluster\u2019s only coherent slice: generic distributed trainer test expectations. It is still not a duplicate of the others, but it is the most central among the related test-only changes.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Both are new model-addition PRs, but one adds ViT NEPA and the other adds Molmo; different model families and different code paths." + "reason": "Different model additions (ViT NEPA vs Molmo); shared docs/auto wiring is incidental, not the same change." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "ViT NEPA and Molmo2 are distinct model implementations, not the same underlying change." + "reason": "Different model additions (ViT NEPA vs Molmo2); same subsystem, but distinct models and implementations." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "ViT NEPA versus CTSM: unrelated model additions with different implementations and targets." + "reason": "Different model additions (ViT NEPA vs CTSM) with unrelated architectures and files." }, { "left": "pull_request:42978", "right": "pull_request:46225", "accept": false, - "reason": "ViT NEPA and Rio2 are separate model additions; shared auto-registration patterns are too broad to imply duplication." + "reason": "Different model additions (ViT NEPA vs Rio2); not the same underlying feature." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are different models, not one bug/change; similarity is only at the generic model-onboarding level." + "reason": "Molmo and Molmo2 are separate model families/PRs, not one bug or one code change." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo and CTSM are unrelated model additions." + "reason": "Molmo vs CTSM are unrelated model additions despite shared auto/doc plumbing." }, { "left": "pull_request:43448", "right": "pull_request:46225", "accept": false, - "reason": "Molmo and Rio2 are different model families; no shared concrete fix." + "reason": "Molmo and Rio2 are distinct model integrations, not a duplicate fix." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are separate model onboarding PRs with different architectures and files." + "reason": "Molmo2 and CTSM are different model additions with no shared underlying change." }, { "left": "pull_request:43451", "right": "pull_request:46225", "accept": false, - "reason": "Molmo2 and Rio2 are distinct model additions; shared auto-mapping edits are incidental." + "reason": "Molmo2 and Rio2 are unrelated model additions." }, { "left": "pull_request:45487", "right": "pull_request:46151", "accept": false, - "reason": "AltCLIP/ChineseCLIP model-parallel fix versus token-classification XPU expectations are different code paths and different changes." + "reason": "AltCLIP/ChineseCLIP model-parallel fix is unrelated to pipeline token-classification test expectations." }, { "left": "pull_request:45487", "right": "pull_request:46214", "accept": false, - "reason": "AltCLIP/ChineseCLIP device-placement fix is unrelated to HRM text test expectations." + "reason": "AltCLIP/ChineseCLIP parallelism fix is a different code path from HRM text XPU expectations." }, { "left": "pull_request:45490", "right": "pull_request:46225", "accept": false, - "reason": "CTSM and Rio2 are different model additions; do not merge just because they touch auto-model mappings." + "reason": "CTSM and Rio2 are separate model-addition PRs; same auto-registration patterns do not make them duplicates." }, { "left": "pull_request:45490", "right": "pull_request:46162", "accept": false, - "reason": "CTSM model onboarding is unrelated to Florence2/LFM2-VL XPU expectation updates." + "reason": "Model addition for CTSM is unrelated to Florence2/LFM2-VL XPU test expectations." }, { "left": "pull_request:46109", - "right": "pull_request:46214", + "right": "pull_request:46151", "accept": false, - "reason": "Distributed trainer parity test adjustments are unrelated to HRM text expectations." + "reason": "Distributed trainer parity tests and token-classification pipeline expectations are different test suites and failure modes." }, { "left": "pull_request:46109", - "right": "pull_request:46151", + "right": "pull_request:46214", "accept": false, - "reason": "One changes distributed trainer tests, the other changes pipeline token-classification expectations; not the same bug." + "reason": "Both are test-related, but one is distributed trainer parity and the other is HRM text expectations; not the same change." }, { "left": "pull_request:46135", "right": "pull_request:46214", "accept": false, - "reason": "Kosmos2.5 XPU integration expectations and HRM text expectations affect different model tests and behaviors." + "reason": "Kosmos2.5 integration expectations and HRM text expectations are distinct model-specific test adjustments." }, { "left": "pull_request:46135", "right": "pull_request:46151", "accept": false, - "reason": "Both are test-only updates, but for different models and different behaviors; too broad to be duplicates." + "reason": "Kosmos2.5 integration test updates and pipeline token-classification expectations are unrelated." }, { "left": "pull_request:46151", "right": "pull_request:46214", "accept": false, - "reason": "Token-classification pipeline expectations versus HRM text expectations are separate test changes." + "reason": "Both add expectations, but for different components and failure signatures; not one underlying bug." }, { "left": "pull_request:46162", "right": "pull_request:46214", "accept": false, - "reason": "Florence2/LFM2-VL XPU expectations are not the same change as HRM text expectations." + "reason": "Florence2/LFM2-VL XPU expectations and HRM text XPU expectations are separate model-specific test changes." }, { "left": "pull_request:46162", "right": "pull_request:46225", "accept": false, - "reason": "FlFlorence2/LFM2-VL test expectation updates and Rio2 model addition are unrelated changes." + "reason": "Florence2/LFM2-VL XPU expectations are unrelated to adding the Rio2 model." } ] }, @@ -18514,7 +18543,7 @@ { "label": "pull_request", "component_index": 8, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18533,7 +18562,7 @@ "pull_request:43291|pull_request:44815", "pull_request:43665|pull_request:44815" ], - "prepared_review_unit_hash": "3dcb35036dd4ca58bb46951c14f730c86682c138306d2ffe68a742e60c3152a9", + "prepared_review_unit_hash": "a5a469b0458fe17c41e90ea5bac0d127df59cd6c2b0fc0209d30c8562032fb3d", "component_budget": { "node_count": 3, "item_count": 3, @@ -18555,30 +18584,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "No duplicate PR cluster here: the three PRs touch different problems, and the only apparent overlap is superficial test-only similarity. PR 44815 is the strongest standalone fix; 43291 is a narrow Whisper test/tokenizer change; 43665 looks like unrelated test noise.", - "confidence": 0.91, + "summary": "All three PRs are unrelated: one changes Whisper decoding logic, one is a test-only patch for CLIP/ViT, and one fixes FP8/dequant loading paths across several components. None should be clustered as the same underlying change.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44815 is the most substantive and coherent fix, centered on a concrete dequant/loading path bug with related implementation changes and tests. The other PRs are either test-only or unrelated.", + "canonical_pr_reason": "pull_request:44815 is the strongest standalone change: it has a concrete bug-fix scope across the FP8 loading/dequant path and includes the most substantive implementation work.", "best_issue_reason": null, - "best_pr_reason": "PR 44815 is the best representative of the cluster because it fixes a real code-path issue and has the clearest, broadest technical scope. PR 43291 is narrower and 43665 is not a meaningful bug fix.", + "best_pr_reason": "pull_request:44815 is the best representative PR in this set because it addresses a specific runtime correctness issue with real code-path changes, unlike the test-only or narrow Whisper patch.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Both are test-related, but they target different models and different behaviors; no shared bug or code path suggests they could be merged into one PR." + "reason": "Different models and different code paths: Whisper tokenizer decoding vs CLIP/ViT test edits. No shared underlying bug or mergeable common fix." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper tokenizer/test changes are unrelated to the FP8 dequant/model-loading fix; they do not address the same underlying defect." + "reason": "Both touch tokenization-related files, but they fix unrelated model-specific behaviors in different areas (Whisper decoding vs Mistral tokenization/FP8 loading). Not the same bug." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "The overlap is only superficial test churn. 43665 is not the same concrete bug/change as the FP8 loading/dequant fix in 44815." + "reason": "Test-only CLIP/ViT assertions are unrelated to the FP8/dequant loading fix. They do not target the same concrete code-path problem." } ] }, @@ -18592,7 +18621,7 @@ { "label": "pull_request", "component_index": 9, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18615,73 +18644,73 @@ "pull_request:45932|pull_request:46100", "pull_request:46100|pull_request:46114" ], - "prepared_review_unit_hash": "84a5da59e4484b364cdb567f759ccdc54e9ab1a0ec1cbf630d02b736ec9803b0", + "prepared_review_unit_hash": "c300c73a714a7274812a73fa76814a5de4f927f60250162a43b7aace2af6d53f", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 5, - "serialized_chars": 11342, - "estimated_input_tokens": 2836, - "estimated_eval_tokens": 5928 + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 5, - "serialized_chars": 11342, - "estimated_input_tokens": 2836, - "estimated_eval_tokens": 5928 + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The only convincing duplicate pair is the two near-identical ViT test-migration PRs (46100 and 46114). The broader grafana/CI PR (45932) is too expansive and unrelated to be merged as the same change.", - "confidence": 0.84, + "summary": "The strongest duplicate signal is the ViT test-migration pair 46100/46114; the other PRs are distinct. The packet\u2019s deterministic_accept flag on 45932/46100 is not supported by the visible diff, so it should not be treated as the same change.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46100 is the cleanest minimal representative of the exact test-only migration change; 46114 is essentially the same forked diff with a different title/body.", + "canonical_pr_reason": "PR 46100 is the cleanest representative of the ViT test-migration duplicate pair: it is the minimal diff showing the inserted failing tests, while 46114 is a forked copy with the same change but a more specialized title.", "best_issue_reason": null, - "best_pr_reason": "PR 46100 best represents the cluster because it is the smallest, most focused instance of the exact same ViT test-edit change.", + "best_pr_reason": "PR 46100 is the best global PR representative because it most clearly captures the underlying ViT test-migration change. 46114 is a near-copy of that same change, while 45932 and the other PRs touch different code paths or broader CI/modeling plumbing.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:46114", "accept": false, - "reason": "Different PR intents and different diffs; one is setup/version/dependency noise, the other is a test-migration fork." + "reason": "Both mention ViT tests, but 43488 also changes setup/versioning and dependency plumbing; 46114 is only a tiny test-migration PR, so they are not the same underlying change." }, { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "Both touch test_modeling_vit, but 43488 is a setup.py/dependency change while 43532 is a much broader CI/model/workflow PR." + "reason": "These are different dummy PRs: 43488 is packaging/version/dependency setup plus a test file, while 43532 rewrites workflow/comment-CI and related utilities." }, { "left": "pull_request:43532", "right": "pull_request:46114", "accept": false, - "reason": "Shared test file is not enough; 43532 has many unrelated workflow/model changes and does not look like the same concrete PR." + "reason": "Shared ViT test filename is not enough; 43532 is CI/workflow plumbing with broader unrelated edits, while 46114 is a minimal failing-test migration change." }, { "left": "pull_request:45932", "right": "pull_request:46100", "accept": false, - "reason": "45732/46100 are not the same underlying change: 45932 spans workflow secrets and multiple model files, while 46100 is only a tiny test migration." + "reason": "The deterministic_accept flag is inconsistent with the visible diffs: 45932 changes model code and CI secrets across several files, while 46100 only adds deliberately failing ViT tests, so they are not the same change." }, { "left": "pull_request:46100", "right": "pull_request:46114", "accept": true, - "reason": "These look like the same exact test-only change in the same file, with identical diff content and only title/body differences." + "reason": "These are effectively the same ViT test-migration change: same file, same inserted failing tests, and the same narrow purpose, so they look like duplicate copies of one PR." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The analysis is grounded in the packet and stays conservative. The duplicate call on 46100 and 46114 is supported by the identical diff preview and matching single-file test change. The rejections for 43488, 43532, and 45932 versus the ViT test-migration PRs are also reasonable because those PRs include broader or different changes beyond the shared test file. No evidence is overstated." + "accept": false, + "feedback": "The summary is not fully grounded in the packet. The packet marks pull_request:45932 | pull_request:46100 as deterministic_accept=true, but the analyst says the only true duplicate pair is 46100/46114. That overstates the evidence and ignores the strongest signal in the packet. Because of that, the global/canonical PR reasoning is not conservative enough. The 46100/46114 duplicate claim may be plausible from the diffs, but the writeup cannot assert exclusivity here." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -18689,7 +18718,7 @@ { "label": "pull_request", "component_index": 10, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18705,12 +18734,12 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "7111eb3ca68d3fb87965d9c454e1a9885113a096d92920c869923711ce6c82e7", + "prepared_review_unit_hash": "fc011bde4e24b815cfb08205bf207e3adb9e9aef912cd2dda1bcf2e9761266b8", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6083, + "serialized_chars": 6082, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -18718,7 +18747,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6083, + "serialized_chars": 6082, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -18727,18 +18756,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are related by shared pipeline/docs scaffolding, but they implement different features: promptable concept segmentation vs promptable visual segmentation. They should not be deduplicated.", + "summary": "These PRs are closely related in scope and infrastructure, but they add different pipelines for different tasks: promptable concept segmentation vs promptable visual segmentation. They should not be treated as duplicates.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No canonical PR: the two pull requests implement distinct pipeline features for different segmentation tasks, not the same underlying change.", "best_issue_reason": null, - "best_pr_reason": "43613 is the broader and more integration-heavy PR, but it still targets a different segmentation pipeline than 43612.", + "best_pr_reason": "Neither PR is a better duplicate canonical than the other; both are standalone feature additions, and the visual-segmentation PR is not a substitute for the concept-segmentation PR.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different tasks and code paths: 43612 adds PromptableConceptSegmentationPipeline, while 43613 adds PromptableVisualSegmentationPipeline with different model integrations and tests; shared docs/metadata updates are incidental." + "reason": "Different pipelines and different target tasks/code paths; they share framework/docs scaffolding but not the same underlying bug or change." } ] }, @@ -18752,7 +18781,7 @@ { "label": "pull_request", "component_index": 11, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18768,40 +18797,40 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "eda293ac9e53287437349d57c560afe3dcacbb8dc8998919e86ad6ddfaecaf53", + "prepared_review_unit_hash": "5590ee7fff176c2be8c71ba3686a579581d26c7559945374b9e69b6390053049", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4291, - "estimated_input_tokens": 1073, - "estimated_eval_tokens": 2402 + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4291, - "estimated_input_tokens": 1073, - "estimated_eval_tokens": 2402 + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two independent `trainer.py` extensibility PRs: one adds `_loss_is_scaled_for_ga` for gradient-accumulation loss scaling, the other adds `_get_signature_columns` for column filtering. They do not appear to fix the same bug or implement the same change.", - "confidence": 0.96, + "summary": "Two open PRs both touch `trainer.py`, but they address different extension points in `Trainer`: one adds `_loss_is_scaled_for_ga` for gradient-accumulation loss scaling, the other adds `_get_signature_columns` for column filtering. They are related only by file and broad subsystem, not by the same underlying bug or change.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR as a duplicate target: the PRs touch the same file but solve different trainer code paths and should be triaged separately.", + "canonical_pr_reason": "No clear canonical PR: the patches solve different problems in `Trainer` and should not be merged as duplicates.", "best_issue_reason": null, - "best_pr_reason": "PR 43651 is the more concrete bug-fix style change, with a tight code-path change in `training_step` and a clear user-facing extensibility need.", + "best_pr_reason": "Neither PR is a duplicate of the other; if forced to pick a representative, `#43651` is slightly more self-contained and narrowly scoped, but `#43663` is equally valid for its separate feature.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Both modify `trainer.py`, but one addresses gradient-accumulation loss scaling while the other adds a hook for signature-column filtering; different underlying behavior and not plausibly one merged PR for the same bug." + "reason": "Reject: both modify `trainer.py`, but one changes gradient-accumulation loss scaling and the other refactors signature-column inspection for custom trainers. They do not fix the same concrete code-path problem and would not plausibly be merged into one PR as duplicates." } ] }, @@ -18815,7 +18844,7 @@ { "label": "pull_request", "component_index": 12, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18831,12 +18860,12 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "fa675c3f9f53355aa700bdbcff68a01f5be8cb7d59d60d1268e2a2cb86b7633a", + "prepared_review_unit_hash": "8f0762a004e04e6f03f10485ebb193df9fbacc08950a8041d221f4964fc4e5c9", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3663, + "serialized_chars": 3664, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -18844,7 +18873,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3663, + "serialized_chars": 3664, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -18853,18 +18882,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs touch the same CLI file and both orbit TypeAdapter annotation/import handling, but they appear to address different failure modes rather than the same exact fix.", - "confidence": 0.72, + "summary": "Two open PRs touch `src/transformers/cli/serve.py`, but they address different runtime/type-hint problems. The stronger canonical candidate is the narrower `pydantic`-related NameError fix; the other is a separate typing/annotation cleanup. Soft match rejected.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43842 is the cleaner representative of the cluster: its title and diff directly target the TypeAdapter NameError when pydantic is absent, with a narrow fix in src/transformers/cli/serve.py.", + "canonical_pr_reason": "PR #43842 is the more specific and self-contained fix: it guards the `TypeAdapter` import behind `TYPE_CHECKING` and quotes the annotation so the CLI does not require `pydantic` at runtime.", "best_issue_reason": null, - "best_pr_reason": "PR #43842 is the best overall fit because it states the concrete failure mode clearly and applies a focused change to that code path, making it the most useful representative for triage.", + "best_pr_reason": "PR #43842 is the best representative of a concrete bug fix because it directly targets the runtime failure mode described in its title and diff, with minimal unrelated changes.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": false, - "reason": "Same file and similar TypeAdapter-related cleanup, but the fixes are not the same underlying bug: #43656 adds future annotations and broader typing adjustments, while #43842 specifically guards the TypeAdapter import/annotation for the pydantic-not-installed case. Different explicit issues also suggest distinct reports." + "reason": "Both touch the same file, but they fix different underlying issues: #43656 is a broader annotations/`__future__` typing change, while #43842 specifically fixes `TypeAdapter` being unavailable when `pydantic` is not installed. They are not the same code-path bug and would not be merged as one PR." } ] }, @@ -18878,7 +18907,7 @@ { "label": "pull_request", "component_index": 13, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -18944,7 +18973,7 @@ { "label": "pull_request", "component_index": 14, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 2, "status": "reviewed", @@ -19023,8 +19052,8 @@ "pull_request:44073|pull_request:44101", "pull_request:44059|pull_request:44071", "pull_request:44018|pull_request:44059", - "pull_request:44056|pull_request:44101", "pull_request:44073|pull_request:44161", + "pull_request:44056|pull_request:44101", "pull_request:44059|pull_request:44068", "pull_request:44101|pull_request:44161", "pull_request:44017|pull_request:44018", @@ -19056,352 +19085,418 @@ "pull_request:44044|pull_request:44129", "pull_request:44044|pull_request:44154" ], - "prepared_review_unit_hash": "40a424985269f5509b579b3a14884a95781ab69820fd7d1635e853f5101e83f1", + "prepared_review_unit_hash": "2f74dcab86a770e0a07b14c38c49704cc87a4a01b7b56b113443136807633a5a", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84890, - "estimated_input_tokens": 21223, - "estimated_eval_tokens": 42702 + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 }, "budget": { "node_count": 27, "item_count": 27, "soft_pair_count": 72, - "serialized_chars": 75087, - "estimated_input_tokens": 18772, - "estimated_eval_tokens": 37800 + "serialized_chars": 75091, + "estimated_input_tokens": 18773, + "estimated_eval_tokens": 37802 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a set of model-specific PRs to standardize output tracing / capture_outputs across different Transformer architectures. Most are related at the theme level, but only a few are true duplicates within the same concrete code path (notably GPT-J/CodeGen and GPT-Neo variants).", - "confidence": 0.89, + "summary": "This cluster is mostly a set of independent model-by-model refactors to the new output-tracing/capture_outputs interface. The only clear duplicate/overlapping PRs are the GPT-J/CodeGen changes; the DeBERTa, SpeechT5, ViLT, Vits, RegNet/ResNet, and similar PRs are related by theme but not the same concrete code-path change.", + "confidence": 0.86, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44066 is a strong representative of the core change: it targets the shared GPT-J/CodeGen output-tracing refactor, has an explicit issue link, and covers the same concrete files that later variants touch.", + "canonical_pr_reason": "PR 44722 is the best canonical representative for the duplicate GPT-J/CodeGen refactor: it targets the same concrete output-tracing change as 44066/44085, explicitly ties to #43979, and appears to cover the shared code path most completely.", "best_issue_reason": null, - "best_pr_reason": "PR 44066 is the best single representative for this cluster because it captures the main standardized output-tracing change in the most central GPT-J/CodeGen code path and overlaps with the closest duplicate variants.", + "best_pr_reason": "44722 is the strongest global PR candidate because it is the clearest concrete implementation of the shared GPT-J/CodeGen output-tracing refactor and the most plausible merge target among the near-duplicates.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families and code paths (CVT/FNet vs GPT-J/CodeGen); only the issue target is shared." + "reason": "Different models and different code paths (FNet/CVT vs GPT-J); same theme only, not the same underlying change." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Same broad refactor theme, but different architectures and implementation paths." + "reason": "FNet/CVT output-tracing refactor is unrelated to DeBERTa's attention/output changes." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Same GPT-J output-tracing refactor in the same file; 44085 looks like a narrower variant of the 44066 change." + "reason": "Both are GPT-J output-tracing refactors and touch the same underlying model code path; they look like overlapping implementations of the same change." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "Different models (RegNet/ResNet/RT-DETR vs EfficientNet) and different code paths." + "reason": "RegNet/ResNet output-tracing migrations are separate model implementations, not one duplicate bug or change." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "EfficientNet vs GPT-J/CodeGen; same theme, but not the same concrete bug/change." + "reason": "EfficientNet output tracing and GPT-J/CodeGen output tracing are different model-specific refactors." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen output tracing vs EfficientNet output tracing; different model implementations." + "reason": "GPT-J/CodeGen refactor vs EfficientNet refactor; broad standardization theme only." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTa-v2 are unrelated code paths despite sharing the refactor theme." + "reason": "MobileNetV2 and DeBERTa are unrelated code paths, so this is not a duplicate." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen refactor vs MGP-STR refactor; different model code." + "reason": "GPT-J/CodeGen refactor does not match MGP-STR's model-specific output handling." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen vs MPT; same architectural cleanup, but not the same underlying change." + "reason": "GPT-J/CodeGen and MPT are separate model refactors, not the same concrete bug/change." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both are GPT-Neo output-tracing refactors over the same model file set; they look like near-duplicate variants." + "reason": "Both target GPT-Neo output-tracing standardization in the same file/code path and are plausibly duplicate implementations of the same refactor." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo vs GPT-J/CodeGen; similar pattern, but different model code paths." + "reason": "GPT-Neo and GPT-J/CodeGen are different model code paths, so this is only a thematic match." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen and GPT-Neo are different implementations, so these are not the same concrete change." + "reason": "GPT-J/CodeGen vs GPT-Neo are distinct model refactors." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet output tracing and GPT-J/CodeGen output tracing are unrelated code paths." + "reason": "ResNet output-tracing refactor and GPT-J/CodeGen refactor are unrelated model-specific changes." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPT-J/CodeGen are separate model implementations." - }, - { - "left": "pull_request:44019", - "right": "pull_request:44085", - "accept": false, - "reason": "ResNet and GPT-J are different model families and fixes." + "reason": "MGP-STR output tracing is a different model and code path from GPT-J/CodeGen." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo and ResNet changes are unrelated beyond the generic refactor pattern." + "reason": "GPT-Neo and ResNet are different implementations; same refactor theme is not enough." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet vs MPT; different code paths." + "reason": "ResNet and MPT are unrelated model changes." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo output tracing vs Flaubert output tracing; not the same underlying change." + "reason": "GPT-Neo and Flaubert are separate model refactors, not duplicates." }, { "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo and Flaubert are different model families with separate implementations." + "reason": "GPT-Neo and Flaubert are unrelated code paths." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo vs GPT-J/CodeGen; same pattern, but not the same code-path problem." + "reason": "GPT-Neo and GPT-J/CodeGen are different model code paths." }, { "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "MPT and GPT-J/CodeGen are different model implementations." + "reason": "MPT and GPT-J/CodeGen are distinct model-specific changes." }, { "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo and GPT-J are different code paths." + "reason": "GPT-Neo refactor is separate from GPT-J/CodeGen refactor." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo vs MGP-STR; unrelated concrete fixes." + "reason": "GPT-Neo and MGP-STR are unrelated changes." }, { "left": "pull_request:44068", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are separate model changes." + "reason": "GPT-Neo and MGP-STR do not share the same concrete code path." }, { "left": "pull_request:44071", "right": "pull_request:44085", "accept": false, - "reason": "MPT and GPT-J are different architectures." + "reason": "MPT and GPT-J/CodeGen are different model-specific refactors." }, { "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT-2 and MGP-STR are unrelated implementations." + "reason": "GPT2 output tracing and MGP-STR output tracing are separate model implementations." }, { "left": "pull_request:44085", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J and MGP-STR are different model code paths." + "reason": "GPT-J and MGP-STR are different code paths; only the general output-tracing theme overlaps." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT and Flaubert/XLM are separate model families." + "reason": "ViLT and XLM/Flaubert are distinct model refactors, not the same underlying change." }, { "left": "pull_request:44059", - "right": "pull_request:44722", + "right": "pull_request:44071", "accept": false, - "reason": "GPT-2 vs GPT-J/CodeGen; same refactor theme, but not the same bug/change." + "reason": "GPT2 and MPT are different model code paths." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer and GPT-J/CodeGen are unrelated model implementations." + "reason": "SegFormer output-tracing refactor is unrelated to GPT-J/CodeGen." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and LongT5 touch different code paths." + "reason": "SegFormer and LongT5 are separate model-specific refactors." }, { "left": "pull_request:44017", "right": "pull_request:44073", "accept": false, - "reason": "SegFormer and VisualBert are different model families." + "reason": "SegFormer and VisualBert are unrelated model changes." }, { "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo vs MPT; same output-tracing motif, but different concrete changes." + "reason": "GPT-Neo and MPT are different model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44086", + "accept": false, + "reason": "MPT and MGP-STR are unrelated model-specific refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet output tracing and VisualBert output tracing are different code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44071", + "accept": false, + "reason": "SegFormer and MPT are separate model refactors." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44101", + "accept": false, + "reason": "VisualBert and XLM/Flaubert are not the same implementation change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT2 and GPT-Neo are different model code paths." + }, + { + "left": "pull_request:44101", + "right": "pull_request:44161", + "accept": false, + "reason": "Flaubert/XLM and LongT5 are unrelated refactors." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44018", + "accept": false, + "reason": "SegFormer and GPT-Neo are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are separate model refactors." }, { "left": "pull_request:44071", "right": "pull_request:44076", "accept": false, - "reason": "MPT and ImageGPT are separate implementations." + "reason": "MPT and ImageGPT are different model code paths." }, { "left": "pull_request:44074", "right": "pull_request:44101", "accept": false, - "reason": "TextNet and Flaubert/XLM are unrelated model changes." + "reason": "TextNet and Flaubert/XLM are unrelated model-specific changes." }, { "left": "pull_request:43997", "right": "pull_request:44018", "accept": false, - "reason": "RegNet and GPT-Neo are different code paths." + "reason": "RegNet/ResNet and GPT-Neo are not the same underlying change." }, { "left": "pull_request:44056", - "right": "pull_request:44101", + "right": "pull_request:44161", "accept": false, - "reason": "MPNet and Flaubert/XLM are separate model families." + "reason": "MPNet and LongT5 are different model refactors." }, { "left": "pull_request:43997", "right": "pull_request:44161", "accept": false, - "reason": "RegNet and LongT5 are unrelated implementations." + "reason": "RegNet/ResNet and LongT5 are unrelated code paths." }, { "left": "pull_request:44074", "right": "pull_request:44161", "accept": false, - "reason": "TextNet and LongT5 are different model code paths." + "reason": "TextNet and LongT5 are separate model changes." }, { "left": "pull_request:44018", + "right": "pull_request:44161", + "accept": false, + "reason": "GPT-Neo and LongT5 are different model code paths." + }, + { + "left": "pull_request:43997", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are different models; not mergeable as one PR." + "reason": "RegNet/ResNet and MPT are unrelated refactors." }, { "left": "pull_request:44010", - "right": "pull_request:44073", + "right": "pull_request:44071", "accept": false, - "reason": "SqueezeBert and VisualBert are separate code paths." + "reason": "SqueezeBert and MPT are different model-specific changes." }, { - "left": "pull_request:43997", - "right": "pull_request:44074", + "left": "pull_request:44018", + "right": "pull_request:44129", "accept": false, - "reason": "RegNet and TextNet are unrelated model implementations." + "reason": "GPT-Neo and SpeechT5 are distinct model code paths." }, { - "left": "pull_request:44056", - "right": "pull_request:44161", + "left": "pull_request:44010", + "right": "pull_request:44073", "accept": false, - "reason": "MPNet and LongT5 are different model families and fixes." + "reason": "SqueezeBert and VisualBert are unrelated model refactors." }, { - "left": "pull_request:44010", + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet/ResNet and TextNet are separate model-specific changes." + }, + { + "left": "pull_request:44018", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBert and DeBERTa-v2 are different code paths." + "reason": "GPT-Neo and DeBERTa are different model code paths." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa-v2 and TextNet are unrelated changes." + "reason": "DeBERTa and TextNet are unrelated changes." }, { "left": "pull_request:44066", "right": "pull_request:44722", "accept": true, - "reason": "Both touch the same GPT-J/CodeGen standardized output-tracing refactor and overlap on the same files." + "reason": "Same GPT-J/CodeGen output-tracing refactor with overlapping file changes; these are the closest duplicate pair in the cluster." }, { "left": "pull_request:44085", "right": "pull_request:44722", "accept": true, - "reason": "Both target the same GPT-J output-tracing change in modeling_gptj.py; one is a narrower variant of the other." + "reason": "Both are GPT-J output-tracing refactors that touch the same underlying implementation and look mergeable into one change." }, { "left": "pull_request:44085", "right": "pull_request:44129", "accept": false, - "reason": "GPT-J and SpeechT5 are different model code paths." + "reason": "GPT-J and SpeechT5 are unrelated model-specific refactors." }, { "left": "pull_request:43996", "right": "pull_request:44007", "accept": false, - "reason": "CVT/FNet and RegNet/ResNet/RT-DETR are separate model families." + "reason": "FNet/CVT and RegNet/ResNet are separate model implementations; not a duplicate." }, { "left": "pull_request:43996", "right": "pull_request:44129", "accept": false, - "reason": "CVT/FNet and SpeechT5 are unrelated changes." + "reason": "FNet/CVT and SpeechT5 are unrelated code paths." }, { "left": "pull_request:44044", "right": "pull_request:44129", "accept": false, - "reason": "DeBERTa-v2 and SpeechT5 are different implementations." + "reason": "DeBERTa and SpeechT5 are different model refactors." }, { "left": "pull_request:44044", "right": "pull_request:44154", "accept": false, - "reason": "DeBERTa-v2 and ViT-S are separate model families and code paths." + "reason": "DeBERTa and Vits are separate model-specific changes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen and EfficientNet are different model code paths." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet and GPT-J/CodeGen are unrelated refactors." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is conservative and grounded in the packet. The PR representative choice is defensible, and the accepted soft-edge pairs all look like same-file or same-model variants of the same output-tracing refactor, so they appear mergeable into a single bugfix PR. The false verdicts are also appropriately cautious across unrelated model families." + "feedback": "The summary is well grounded in the packet and stays conservative about unrelated model-specific refactors. The duplicate/overlapping PR calls are also supported: GPT-J/CodeGen (44066/44085/44722) and GPT-Neo (44018/44068) share the same code paths, while the false pairings are appropriately rejected." }, "evaluator_used": true, "retried": false, @@ -19412,7 +19507,7 @@ { "label": "pull_request", "component_index": 14, - "component_count": 27, + "component_count": 28, "review_unit_index": 2, "review_unit_count": 2, "status": "reviewed", @@ -19439,70 +19534,70 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "fb9235df1ab10975f38deb20ad582ba858fe527b4c5c81737d2700b8ef1d9b9e", + "prepared_review_unit_hash": "bc8106fc812e44d16604a9b15a8cf14b907b9a07bbcfc880496b1d620358f821", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84890, - "estimated_input_tokens": 21223, - "estimated_eval_tokens": 42702 + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 }, "budget": { "node_count": 8, "item_count": 8, "soft_pair_count": 6, - "serialized_chars": 17124, - "estimated_input_tokens": 4281, - "estimated_eval_tokens": 8818 + "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 related model-by-model refactors to standardized output tracing/capture_outputs, but they target different architectures and code paths. None of the soft pairs looks like a true duplicate PR.", - "confidence": 0.93, + "summary": "This cluster is a set of separate model-specific PRs all migrating different Transformers models to standardized output tracing/capture helpers. They share a pattern, but they are not duplicates of the same underlying change or code path.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44056 is the best single representative of the cluster because it shows the common refactor pattern clearly: adding capture_outputs plumbing plus an explicit _can_record_outputs mapping in the model base class.", + "canonical_pr_reason": "44073 is the clearest representative of the shared refactor pattern: it introduces `can_return_tuple`/`capture_outputs` and model-specific output hook registration in a self-contained, well-scoped VisualBert change.", "best_issue_reason": null, - "best_pr_reason": "PR 44056 is a strong cluster representative: it is self-contained, clearly about standardized output tracing, and is broad enough to stand in for the shared refactor theme without being tied to a one-off model edge case.", + "best_pr_reason": "44073 is the best overall representative PR because its title and diff most clearly express the standardized output-tracing migration without being overly model-specific or incomplete.", "soft_edge_verdicts": [ { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "Same broad refactor theme, but different model families (FocalNet vs ImageGPT) and different concrete code paths; too model-specific to be one merged PR." + "reason": "Both are output-tracing refactors, but for different models (FocalNet vs ImageGPT). Different classes and code paths; not one mergeable fix." }, { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Both modernize output tracing, but VisualBert and TextNet are separate implementations with different forward paths and tests, not the same underlying change." + "reason": "VisualBert and TextNet are separate model implementations. Same refactor theme, but not the same underlying bug/change." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Related decorator/capture_outputs cleanup, but MPNet and VisualBert are distinct model-specific refactors rather than one shared bug fix or change." + "reason": "Both standardize output tracing, but MPNet and VisualBert have distinct encoder/attention code paths and separate model-specific hook wiring." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "Same general migration pattern, but MPNet and TextNet touch different model internals and are not plausibly one PR for the same code-path problem." + "reason": "Different models and different forward paths. This is a shared migration pattern, not a duplicate PR." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model implementations; the similarity is only in the standardized output refactor pattern, not a shared concrete fix." + "reason": "RegNet and SqueezeBert are unrelated model implementations; these are parallel refactors, not the same fix." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and SpeechT5 both adjust output tracing, but they fix separate model-specific forward/output handling and should not be merged as duplicates." + "reason": "SqueezeBert and SpeechT5 touch different model stacks, and SpeechT5 also targets a specific issue. Not the same concrete change." } ] }, @@ -19516,7 +19611,7 @@ { "label": "pull_request", "component_index": 15, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19587,304 +19682,298 @@ "pull_request:44002|pull_request:46106", "pull_request:43998|pull_request:46106" ], - "prepared_review_unit_hash": "b0cc381fef03f1a371f3a2bd88d8e4761d0ef8a68a3171332af35b8bc7065513", + "prepared_review_unit_hash": "849968201de66f33101dc0f547ed0df52a49fb235df4d21cf38574b9620342bf", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 44, - "serialized_chars": 43392, - "estimated_input_tokens": 10848, - "estimated_eval_tokens": 21952 + "serialized_chars": 43396, + "estimated_input_tokens": 10849, + "estimated_eval_tokens": 21954 }, "budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 44, - "serialized_chars": 43392, - "estimated_input_tokens": 10848, - "estimated_eval_tokens": 21952 + "serialized_chars": 43396, + "estimated_input_tokens": 10849, + "estimated_eval_tokens": 21954 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are mostly independent model-specific output-tracing refactors, not true duplicates. I would reject all soft-edge merges and treat this as a thematically similar but heterogeneous cluster.", - "confidence": 0.93, + "summary": "This cluster is a batch of model-specific output-tracing refactors plus one unrelated docs toctree edit. The shared wording/pattern makes them look similar, but they are not the same underlying bug or change, so none of the soft edges should be merged.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44003 is the strongest representative of the output-tracing refactor pattern, but the surrounding PRs are separate model-specific changes rather than duplicates.", + "canonical_pr_reason": "No single canonical PR: these are separate per-model refactors with only template-level similarity, not one duplicated fix.", "best_issue_reason": null, - "best_pr_reason": "PR 44003 has the broadest and most core tracing changes, spanning two related model files and the new output-capturing helpers, so it is the best representative PR in this cluster.", + "best_pr_reason": "44025 is the closest representative because it covers two related depth_anything variants and shows the refactor pattern cleanly, but it still isn't merge-equivalent to the rest of the batch.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Same tracing-refactor theme, but different model implementations and code paths." + "reason": "Different models and code paths (timm_backbone vs upernet); same refactor style only." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Different models; only the refactor pattern is similar." + "reason": "Different model families and forward paths; similarity is only the shared output-tracing refactor pattern." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Different model families; not the same underlying change." + "reason": "Separate model implementations (timm_backbone vs univnet), not the same fix." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Both are tracing refactors, but they target different architecture code paths." + "reason": "Unrelated architectures and files; no shared underlying bug/change." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "Separate model codepaths; similarity is only in output-tracing cleanup." + "reason": "Different model code paths (timm_backbone vs rwkv); only the refactor template matches." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "Different models; not a shared bug or mergeable single fix." + "reason": "MobileNetV1 and UnivNet are unrelated; same tracing refactor idea, not one duplicate change." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Same broad refactor style, but unrelated model implementations." + "reason": "Different models and outputs; no common concrete bug to merge." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "Different model families; this is not one concrete fix." + "reason": "Different model families and forward implementations; not the same underlying change." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "Both are refactors, but they affect different model wrappers." + "reason": "Separate encoder/decoder vs MobileNetV1 refactors; template similarity only." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Separate model-specific changes, not the same underlying change." + "reason": "Different subsystems and code paths; not merge-equivalent." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "Distinct model codepaths; similar refactor theme only." + "reason": "Vision-text dual encoder and UnivNet are unrelated; same refactor motif only." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Different model implementations, so not duplicates." + "reason": "Different models and forward paths; no shared underlying fix." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Output-tracing refactor in different model stacks." + "reason": "Distinct model families (CLIP-style dual encoder vs mamba variants); not duplicates." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Separate model families; no single shared bug." + "reason": "Different target models and code paths; similar output-tracing cleanup only." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "Both refactor output handling, but in unrelated models." + "reason": "UnivNet and UperNet are unrelated; same refactor pattern, different change." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "Different codepaths and model architectures." + "reason": "Different architectures and implementation details; not the same bug or fix." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "Different model families; not mergeable as one PR." + "reason": "Different model families and forward signatures; only the shared tracing refactor matches." }, { "left": "pull_request:44001", - "right": "pull_request:46106", + "right": "pull_request:44004", "accept": false, - "reason": "Docs toctree cleanup is unrelated to model output tracing." + "reason": "UnivNet vs CodeGen are unrelated code paths; not a duplicate PR pair." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "Same refactor motif, but different concrete model implementations." + "reason": "UperNet and Mamba/Falcon-Mamba are unrelated; no single concrete change in common." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Different models and codepaths; not the same fix." + "reason": "Different model families and fixes; same refactor theme only." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "Separate model codepaths; only the high-level theme matches." + "reason": "Falcon/Mamba output tracing and CodeGen output tracing are separate model-specific changes." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "Different models; not one underlying bug or change." + "reason": "TimmBackbone and CodeGen are unrelated; the similarity is only cosmetic." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "Different model-specific output behavior, not a duplicate." + "reason": "Different model paths and behavior; no mergeable duplicate fix here." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Different model implementations; only loosely similar." - }, - { - "left": "pull_request:44000", - "right": "pull_request:44004", - "accept": false, - "reason": "Both are tracing-related, but they fix different model codepaths." + "reason": "UperNet and CodeGen touch different subsystems, so this is not the same change." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Different model families; same refactor style is not enough." + "reason": "Depth Anything vs Vision Encoder-Decoder are separate model implementations; only the refactor pattern is shared." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Separate wrappers in different modalities; not duplicates." + "reason": "Different models and output schemas; not the same underlying bug or change." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "Different model implementations and call paths." + "reason": "Depth-anything refactor and SuperPoint refactor are unrelated code paths." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "Different models; not a single concrete fix." + "reason": "Different model families (depth estimation vs RWKV); no duplicate fix." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Same output-tracing pattern, but unrelated models." + "reason": "Depth-anything and DPR are unrelated; same refactor style only." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "Different model wrappers and codepaths." + "reason": "Vision Encoder-Decoder and RWKV are different code paths and fixes." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "Different encoder-decoder wrappers; not mergeable into one PR." + "reason": "Different model stacks (vision-text vs speech encoder-decoder); not merge-equivalent." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "Different models; only the refactor theme overlaps." + "reason": "Encoder-decoder refactor vs SuperPoint refactor are unrelated changes." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "Different model implementations; not the same underlying change." + "reason": "Different model families and output behavior; only the template matches." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "Separate model codepaths with similar tracing cleanup." + "reason": "Speech Encoder-Decoder and SuperPoint are unrelated models; not the same fix." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "Different models; no shared concrete bug." + "reason": "Different model code paths; no shared underlying bug/change." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "Different model-specific refactors, not duplicates." + "reason": "Speech Encoder-Decoder vs DPR are separate changes, not duplicates." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "Unrelated model implementations; same pattern only." + "reason": "SuperPoint and RWKV are unrelated; similarity is only the refactor style." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "Different model families and codepaths." + "reason": "Different model implementations and behavior; not the same underlying change." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "Different models; not the same fix." + "reason": "RWKV and DPR are unrelated code paths; no common duplicate fix." }, { "left": "pull_request:44027", "right": "pull_request:46106", "accept": false, - "reason": "Docs-only toctree change is unrelated to tracing refactors." + "reason": "A docs toctree edit is not the same kind of change as a Speech Encoder-Decoder code refactor." }, { "left": "pull_request:44026", "right": "pull_request:46106", "accept": false, - "reason": "Documentation navigation change is unrelated to model code." + "reason": "Documentation navigation change is unrelated to the Vision Encoder-Decoder refactor." }, { "left": "pull_request:44001", "right": "pull_request:46106", "accept": false, - "reason": "Docs toctree cleanup does not overlap with model output tracing." + "reason": "UnivNet code refactor and docs toctree update are different kinds of changes." }, { "left": "pull_request:44002", "right": "pull_request:46106", "accept": false, - "reason": "Unrelated docs-only change versus model refactor." + "reason": "UperNet model refactor is unrelated to the docs-only toctree change." }, { "left": "pull_request:43998", "right": "pull_request:46106", "accept": false, - "reason": "Docs-only navigation fix is unrelated to the model refactor cluster." + "reason": "TimmBackbone code refactor and docs navigation change do not fix the same thing." } ] }, @@ -19898,7 +19987,7 @@ { "label": "pull_request", "component_index": 16, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19917,52 +20006,52 @@ "pull_request:44662|pull_request:45075", "pull_request:45075|pull_request:45401" ], - "prepared_review_unit_hash": "c9e9ff1c40e8350e918d6a26e0cdb75b371974d6cd30e194e872ce27dccb3ea5", + "prepared_review_unit_hash": "2d9bc19f34434295ee25bca505623ecfd244f2955d09508d60f9537adeae9132", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 10912, - "estimated_input_tokens": 2728, - "estimated_eval_tokens": 5712 + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 10912, - "estimated_input_tokens": 2728, - "estimated_eval_tokens": 5712 + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-addition PRs for different architectures (PenguinVL, DeepSeek-OCR-2, and Voxtral TTS). They share only common registration/docs plumbing, not the same underlying bug or change, so none should be clustered as duplicates.", - "confidence": 0.97, + "summary": "These are three distinct model-addition PRs (PenguinVL, DeepSeek-OCR-2, Voxtral-TTS). They overlap only on boilerplate registration/docs files and do not appear to fix the same underlying code-path or model implementation, so none should be clustered as duplicates.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45075 is the most complete standalone model addition in the set, with model, image-processing, conversion, docs, and tests all present.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #45075 is the strongest representative of this cluster because it contains the most fully fleshed-out implementation and validation surface.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different model families and tasks: PenguinVL vs Voxtral TTS. Shared files are only generic auto-registration/docs, not the same code-path fix." + "reason": "Different model integrations (PenguinVL vs Voxtral-TTS). Shared files are only auto-registry/docs scaffolding, not the same implementation or bug fix." }, { "left": "pull_request:44662", "right": "pull_request:45075", "accept": false, - "reason": "Different model integrations: PenguinVL and DeepSeek-OCR-2. Overlap is limited to boilerplate model registry changes and repo checks." + "reason": "Both add separate models and touch common Transformers registration files, but they are unrelated code changes with different architectures, processors, and tests." }, { "left": "pull_request:45075", "right": "pull_request:45401", "accept": false, - "reason": "DeepSeek-OCR-2 and Voxtral TTS are unrelated model additions. Similarities are broad Transformers scaffolding, not a single mergeable change." + "reason": "DeepSeek-OCR-2 and Voxtral-TTS are unrelated model additions; overlap is limited to shared framework plumbing, which is insufficient to merge as one PR." } ] }, @@ -19976,7 +20065,7 @@ { "label": "pull_request", "component_index": 17, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -19992,12 +20081,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "8d708633c890c9291223e1f16189d0647f677acf4de391980a14401db18dbb04", + "prepared_review_unit_hash": "a464cfe6140a5fff55d89efcf10ae43c109abbc2508cdce4f3f204381015e527", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -20005,7 +20094,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -20014,18 +20103,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two PRs about fusion mechanisms, but they target different levels and concrete changes: one introduces a generic module-fusion API, the other wires n-to-1 kernel fusion through KernelConfig and model loading integration. Related topic, not the same underlying change.", - "confidence": 0.92, + "summary": "These are related kernel/module-fusion feature PRs, but they target different abstractions and code paths: one introduces a new Module Fusion API, while the other wires n-to-1 kernel fusion through KernelConfig and modeling/integration changes. They should not be treated as duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45363 is the better canonical representative because it makes a concrete end-to-end feature change across core loading/integration files, whereas PR 44979 is a lower-level API scaffold for module fusion. They overlap in theme but not in exact code-path change.", + "canonical_pr_reason": "PR 45363 is the better canonical representative because it implements the more concrete end-to-end feature path (KernelConfig-driven fusion) across the main integration points and includes a usage example.", "best_issue_reason": null, - "best_pr_reason": "PR 45363 is the strongest single representative of the cluster: it is more feature-complete, user-facing, and spans the actual integration path for kernel fusion. PR 44979 is adjacent infrastructure rather than the same change.", + "best_pr_reason": "PR 45363 is the stronger overall representative for this cluster since it is the more complete user-facing fusion change and is easier to map to a specific merged capability.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Both concern fusion, but not the same concrete bug/change. 44979 adds a new generic Module Fusion API and tests, while 45363 implements n-to-1 kernel fusion via KernelConfig and model/integration updates. Different code paths and not plausibly one merged PR." + "reason": "44979 adds a standalone Module Fusion API, while 45363 adds n-to-1 kernel fusion via KernelConfig and touches different integration/code paths. They share theme, not the same underlying bug/change." } ] }, @@ -20039,7 +20128,7 @@ { "label": "pull_request", "component_index": 18, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20055,7 +20144,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "6dd846faa10f4363fb5c1252eca9220c7e8335472cab11ffaa90b32ffa5dbf61", + "prepared_review_unit_hash": "06fc293f5a2c69356a0796b5e3e23800e0ce1617c5c60c3fbcccbf3edef44ef5", "component_budget": { "node_count": 2, "item_count": 2, @@ -20077,18 +20166,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs fix the same AttributeError in `_patch_mistral_regex` by switching from `tokenizer.backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer`; 45317 is the more complete version because it adds tests.", - "confidence": 0.98, + "summary": "Two PRs implement the same Mistral regex patch for the same issue (#45081); the later PR adds tests and is the more complete representative.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45317 is the better canonical PR: it addresses the same concrete bug as 45086 and includes test coverage, making it the more complete fix.", + "canonical_pr_reason": "PR #45317 is the more complete version of the same fix: it makes the same AttributeError patch and also adds test coverage, so it is the best canonical PR for this cluster.", "best_issue_reason": null, - "best_pr_reason": "PR 45317 is the best overall PR candidate because it fixes the exact same code path and adds tests, so it is the strongest representative of this bugfix.", + "best_pr_reason": "PR #45317 is the strongest standalone candidate because it contains the same code-path fix as #45086 plus tests, making it the more merge-ready implementation.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same underlying bug and same code change in `_patch_mistral_regex`; 45317 is essentially a superset with tests, so these are duplicates." + "reason": "Both PRs fix the same concrete AttributeError in `_patch_mistral_regex` by changing `tokenizer.backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer` in the same file; #45317 is simply a more complete follow-up with tests." } ] }, @@ -20102,7 +20191,7 @@ { "label": "pull_request", "component_index": 19, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20118,7 +20207,7 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "d085fa4062656c444b428394995c56fd54f6b291aae166932c92a42334c3e792", + "prepared_review_unit_hash": "8f3329e455b7557b961e710950a02e55601f9b272a50e8681b15887ded426701", "component_budget": { "node_count": 2, "item_count": 2, @@ -20140,18 +20229,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs touch Nemotron H model code, but they target different changes: one is a conversion/refactor inheritance update, the other is a broader model split and docs/registration update. They are related, but not duplicates.", + "summary": "These are related NemotronH model refactors, but they do not look like the same concrete change. PR 45115 focuses on inheritance/conversion mapping for NemotronH, while PR 45569 is a broader model split introducing NemotronHDense/NemotronHSparse, docs, auto-mapping, and tests.", "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45569 is the broader, more central Nemotron H change set: it introduces the dense/sparse model split, updates auto-registration, docs, and tests, making it the better representative for the cluster.", + "canonical_pr_reason": "PR 45569 is the broader, more complete architecture change and likely the best representative for the NemotronH family restructuring; PR 45115 is a narrower preparatory refactor around conversion/mapping.", "best_issue_reason": null, - "best_pr_reason": "PR 45569 is the best overall representative because it covers the larger end-to-end model packaging change rather than only the narrower conversion-mapping/refactor work.", + "best_pr_reason": "PR 45569 most directly captures the end-state change: dispatcher over dense/sparse variants plus model, auto, docs, and test updates. It is the best single PR to anchor this cluster.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "They overlap in Nemotron H files, but the concrete changes differ: 45115 focuses on inheritance/conversion-key renaming, while 45569 introduces the dense/sparse model split and associated docs/auto-mapping updates. Not the same fix and not something that should be merged as one PR." + "reason": "They share NemotronH-related files, but 45115 is a narrow inheritance/conversion-mapping refactor, while 45569 is a broader split into dense/sparse model variants with docs and tests. They are related work, not the same mergeable fix." } ] }, @@ -20165,7 +20254,7 @@ { "label": "pull_request", "component_index": 20, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20181,7 +20270,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "f49968ea68817946484073e3bcd757bfec9dd1071f00e85857982199e3669acf", + "prepared_review_unit_hash": "26027937e4d88c502677f1ac5f43cecb8f68be4d301d116fe7755bb320166e98", "component_budget": { "node_count": 2, "item_count": 2, @@ -20203,18 +20292,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related feature PRs in the same heterogeneous-config effort, but they are not duplicates: one adds per-layer configuration support, the other adds broader heterogeneous model/runtime support that builds on it.", - "confidence": 0.85, + "summary": "These two PRs are related feature work in the heterogeneity area, but they are not the same concrete change: one adds per-layer config plumbing, while the other adds broader heterogeneous model/runtime support that builds on it.", + "confidence": 0.81, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45333 is the cleaner foundational change: it introduces heterogeneous per-layer configuration support with a narrower, self-contained scope.", + "canonical_pr_reason": "PR #45332 is the broader umbrella PR: it covers heterogeneous model behavior across cache/model/masking paths and appears to incorporate the config groundwork rather than being the same standalone change as #45333.", "best_issue_reason": null, - "best_pr_reason": "PR #45333 is the best representative for this cluster because it defines the core config-layer abstraction that the larger modeling PR depends on.", + "best_pr_reason": "PR #45332 is the best representative of the cluster because it spans the wider end-to-end heterogeneous support; #45333 is narrower config-only infrastructure.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Related but not the same change: #45333 is config-only support, while #45332 extends into heterogeneous modeling/cache/runtime paths. They address adjacent layers of the feature rather than one identical fix." + "reason": "Reject: #45333 is per-layer configuration plumbing, while #45332 adds heterogeneous model/runtime support and only shares the same subsystem and helper files. They are related, but not the same underlying fix/change and would not plausibly be merged as one PR." } ] }, @@ -20228,7 +20317,7 @@ { "label": "pull_request", "component_index": 21, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20244,7 +20333,7 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "7bd98fe65dfd75002a914ff0dc1e627458997bc6a74f7802bc05ca47b8817b0f", + "prepared_review_unit_hash": "d2acc15ead8a138a54b8e333731bd239e0f2360ed4b232afea6aea7d3119acfe", "component_budget": { "node_count": 2, "item_count": 2, @@ -20266,18 +20355,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related only by broad typing cleanup, but they target different changes: one is a wide type-checking pass across many `src/transformers/*` files, while the other adds shared typing aliases/protocols and specific typing fixes around `modeling_utils` integration. They do not look like the same underlying bug or mergeable duplicate PR.", - "confidence": 0.88, + "summary": "Both pull requests are typing-related, but they address different changes: one is a broad type-checking cleanup across many files, while the other adds reusable typing protocols/aliases for modeling and PEFT-related code. They are related in theme, not duplicates.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45425 is the better canonical representative because it has a tighter, more coherent scope: shared typing helpers in `_typing.py` plus corresponding fixes in a small set of integration points. PR 45415 is a much broader repository-wide type-checking sweep.", + "canonical_pr_reason": "PR 45415 is the broader umbrella change in the cluster: it applies type-checking cleanup across many `src/transformers/*` files and updates type-ignore policy, so it is the most representative PR if one needs a single canonical item.", "best_issue_reason": null, - "best_pr_reason": "PR 45425 is the best fit for the cluster because it is the more concrete code change and centers on a specific typing gap, whereas PR 45415 is a large mechanical type-checking pass spanning many unrelated files.", + "best_pr_reason": "PR 45415 is the best overall representative because it covers the widest surface area and is the clearest cluster anchor for the typing cleanup theme.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Both are typing-related, but they fix different things: 45415 is a broad type-checking cleanup, while 45425 introduces typing protocol/alias support and targeted integration fixes. They are not the same underlying change and would not plausibly merge into one PR." + "reason": "Both are about typing, but they fix different concrete things: 45415 is a broad typing-pass / ignore-policy sweep across many files, while 45425 introduces new typing aliases and protocols for `modeling_utils`/PEFT integration. The overlap is thematic and partial, not the same underlying change, so they should remain separate." } ] }, @@ -20291,7 +20380,7 @@ { "label": "pull_request", "component_index": 22, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20309,7 +20398,7 @@ "pull_request:45513|pull_request:46084", "pull_request:46084|pull_request:46098" ], - "prepared_review_unit_hash": "c1e2adcd89419d0b011eea4a84cfef0867374d0048b03c78c307f60caae3cb82", + "prepared_review_unit_hash": "584a6308539d9c38dfbe37b975bd38830bd8f4ed81958258cd6837ec0491bb81", "component_budget": { "node_count": 3, "item_count": 3, @@ -20331,30 +20420,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two Mamba2 PRs are duplicates of the same cache/seq_len>1 bug; the Qwen3.5 PR is a separate model-specific fix and should not be merged with them.", - "confidence": 0.95, + "summary": "Two PRs are about the same Mamba2 cached-forward bug with `use_cache` and `seq_len > 1`; the Qwen3.5 PR is unrelated and only similar at a high level.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46098 is the most complete representation of the Mamba2 fix: it targets the same issue as #46084, touches the same model/tests, and its title/description clearly state the incorrect-use_cache-with-seq_len>1 bug across CPU and GPU paths, plus docs/tests updates.", + "canonical_pr_reason": "PR #46098 is the best canonical PR because it targets the same Mamba2 `use_cache` seq_len>1 bug as #46084, but appears broader and more complete (explicit issue target, docs, tests, and both CPU/GPU path coverage).", "best_issue_reason": null, - "best_pr_reason": "PR #46098 is the best cluster representative because it is explicit about the concrete bug, covers the same code path as #46084, and adds the user-facing docs/tests that make the change easier to understand and validate.", + "best_pr_reason": "PR #46098 is the strongest representative of the cluster: it directly references the tracking issue, covers the core Mamba2 cached-forward incorrect-results bug, and includes the most complete fix surface.", "soft_edge_verdicts": [ { "left": "pull_request:45513", "right": "pull_request:46084", "accept": false, - "reason": "Both concern cached multi-token forward behavior, but they are for different model families and different implementations (Qwen3.5 GDN linear attention vs. Mamba2). Similar symptom, not the same underlying fix." + "reason": "Different model families and bugs: Qwen3.5 GDN linear attention vs Mamba2 cached forward. The similarity is only the general pattern of multi-token cached decoding." }, { "left": "pull_request:46084", "right": "pull_request:46098", "accept": true, - "reason": "Same issue target, same Mamba2 files, and both fix use_cache with seq_len>1 in the same code path. #46098 is a fuller version of the #46084 fix rather than a distinct change." + "reason": "Same underlying Mamba2 `use_cache` with `seq_len > 1` bug, same target issue, same code area and tests; #46098 looks like a broader version of the same fix." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is conservative and grounded in the packet. The Qwen3.5 PR is clearly a different model family and should not be merged with the Mamba2 fix. The 46084/46098 pair shares the same issue target and files, and 46098 looks like a fuller, mergeable superset of the same bugfix rather than a distinct change." + "feedback": "The cluster summary is grounded in the packet. The 45513 vs 46084 soft rejection is conservative and supported by the clearly different model families and bugs. The 46084 vs 46098 soft acceptance is also supported: both target issue 46032, touch the same Mamba2 files, and 46098 appears to be a broader version of the same fix. No overstatement detected." }, "evaluator_used": true, "retried": false, @@ -20365,7 +20454,7 @@ { "label": "pull_request", "component_index": 23, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20381,7 +20470,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "20c31e8882a3371f947e8e707764908921428f847bea42905a5dbb3136eb53b0", + "prepared_review_unit_hash": "0cf0adb3ed36830afaf9bb140ae2de108bfb4ba8cf53bb2ded5581b0c9038e6c", "component_budget": { "node_count": 2, "item_count": 2, @@ -20403,18 +20492,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are related by expert-parallelism support, but they fix different bugs in different code paths: 45548 is about preserving `distributed_config`/`has_ep` for EP + DeepSpeed ZeRO-3 loading under `accelerate launch`, while 45662 prevents expert parameters from being overwritten after distributed sharding in FSDP2. They are adjacent, not duplicates.", - "confidence": 0.97, + "summary": "Both pull requests are EP-related, but they fix different concrete bugs: one is about preserving distributed_config / EP state during loading, the other about expert tensors being overwritten by rank-0 broadcast in FSDP2. They are similar in subsystem, not in underlying change.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45662 is the better canonical representative for this cluster because it fixes a concrete runtime loading overwrite bug in the sharded-parameter path; PR 45548 is a separate EP/ZeRO-3 serialization and capability-flag change.", + "canonical_pr_reason": "PR #45662 is the more concrete, narrowly described bug fix: expert parameters are being overwritten by rank-0 broadcast during EP/FSDP2 loading. It reads as a specific code-path correction rather than a broader loading/serialization cleanup.", "best_issue_reason": null, - "best_pr_reason": "PR 45662 is the strongest overall fit for the cluster theme: it has a focused fix in the model-loading path and addresses a specific failure mode, whereas 45548 is about config persistence and EP detection rather than the same bug.", + "best_pr_reason": "PR #45662 is the strongest representative of this cluster because it targets a clearly defined failure mode in expert-parallel loading and has the more actionable fix surface.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Different underlying failures: 45548 preserves `distributed_config` and adds `has_ep` for EP + ZeRO-3/accelerate loading, while 45662 wraps post-shard parameters to stop rank-0 broadcast from overwriting experts under FSDP2. Shared subsystem only, not the same change." + "reason": "Do not merge as duplicates: #45548 fixes EP + DeepSpeed ZeRO-3 loading/serialization via accelerate launch, while #45662 fixes EP + FSDP2 rank-0 broadcast overwriting expert parameters. Same subsystem, different underlying bugs and different fix paths." } ] }, @@ -20428,7 +20517,7 @@ { "label": "pull_request", "component_index": 24, - "component_count": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20444,7 +20533,7 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "6df6b01e5822cc0664cbd75e6fddd10cb688e66ecf863ff4aa1c1c262de5c085", + "prepared_review_unit_hash": "7819ca7930a0dd5a671d4b4c7711d1a5ca46efa0371ba4287d9b8264d6123a9c", "component_budget": { "node_count": 2, "item_count": 2, @@ -20466,18 +20555,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same xdist-related `captured_info` collision problem in testing debug logs. PR #45645 appears to be the more complete implementation because it includes the core `testing_utils.py` fix plus the CI/workflow changes needed to consume multiple `captured_info*.txt` files.", - "confidence": 0.9, + "summary": "Both PRs address the same xdist collision around captured_info artifacts in testing_utils; 45645 is the broader superset with CI log handling, while 45639 is the narrower core fix.", + "confidence": 0.84, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45645 is the stronger canonical choice because it covers the core bug fix and the downstream CI/log-collection changes, making it the more complete solution to the shared xdist artifact issue.", + "canonical_pr_reason": "PR 45645 is the more complete change set: it includes the xdist-safe testing_utils fix plus workflow updates to consume multiple captured_info files and related cleanup, making it the better representative of the underlying bug and its CI impact.", "best_issue_reason": null, - "best_pr_reason": "PR #45645 is best suited overall since it preserves debug logs end-to-end: it fixes the writer-side collision and updates workflow/log handling to read all generated files.", + "best_pr_reason": "PR 45645 is the strongest overall candidate because it fixes the core artifact-collision problem and also preserves CI debug log visibility; 45639 is a subset of that change.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Same underlying xdist collision problem for `captured_info` artifacts; #45645 expands on #45639 with the same core `testing_utils.py` fix and compatible workflow support, so they could plausibly be merged into one PR." + "reason": "Both PRs fix the same concrete captured_info xdist collision in testing_utils, and 45645 plausibly subsumes 45639 by adding the CI-side handling needed for the same fix." } ] }, @@ -20491,7 +20580,70 @@ { "label": "pull_request", "component_index": 25, - "component_count": 27, + "component_count": 28, + "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-45670-2", + "nodes": [ + "pull_request:45670", + "pull_request:46229" + ], + "soft_pairs": [ + "pull_request:45670|pull_request:46229" + ], + "prepared_review_unit_hash": "cf892b87826a363e78dd6f369c8da6048bba076ec8677d5691e61610810e2980", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4185, + "estimated_input_tokens": 1047, + "estimated_eval_tokens": 2350 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4185, + "estimated_input_tokens": 1047, + "estimated_eval_tokens": 2350 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster contains two unrelated pull requests: one is a small AutoModel registry fix for GLM ASR, and the other is a new multi-token prediction generation implementation. They do not look like duplicates or the same underlying change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46229 is the better canonical representative because it is the larger, more substantive change and centers on a single new generation feature/module.", + "best_issue_reason": null, + "best_pr_reason": "PR #46229 is the strongest overall representative of the cluster: it is self-contained, technically substantial, and clearly defines a distinct code-path change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45670", + "right": "pull_request:46229", + "accept": false, + "reason": "These touch different code paths and solve different problems: #45670 only updates AutoModel registration for 'glmasr', while #46229 adds a new MTP generation module. They are not the same bug/change and would not plausibly merge into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 26, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20556,8 +20708,8 @@ }, { "label": "pull_request", - "component_index": 26, - "component_count": 27, + "component_index": 27, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20580,12 +20732,12 @@ "pull_request:46073|pull_request:46196", "pull_request:46154|pull_request:46196" ], - "prepared_review_unit_hash": "4588ad3f4409ffb1a887f5753bbdfbe4c65de90c06f0677dc41dc82483a38b01", + "prepared_review_unit_hash": "ecd9e8f8373e98832100fb9c8289d013ccb6b919120639273c7ac548cdaa7704", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 10495, + "serialized_chars": 10496, "estimated_input_tokens": 2624, "estimated_eval_tokens": 5504 }, @@ -20593,7 +20745,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 10495, + "serialized_chars": 10496, "estimated_input_tokens": 2624, "estimated_eval_tokens": 5504 }, @@ -20602,48 +20754,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All four PRs are model-specific additions for `attn_implementation`/SDPA dispatch, but they target different model implementations and code paths. They are similar in feature theme, not duplicates of the same underlying fix.", - "confidence": 0.92, + "summary": "These PRs share a common theme (attention-dispatch support), but each targets a different model and code path, so they are not duplicates.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46196 is the broadest and most representative of the cluster: it adds SDPA dispatch for two closely related model variants (SeamlessM4T and SeamlessM4T v2) and includes corresponding tests, making it the strongest canonical example of this feature pattern.", + "canonical_pr_reason": "No canonical PR: CTRL, T5, ShieldGemma2, and SeamlessM4T/v2 each have separate model-specific implementations and test/docs changes.", "best_issue_reason": null, - "best_pr_reason": "PR 46196 is the best representative PR because it is the most complete model-specific SDPA dispatch change in the group and covers multiple related variants with tests, while the others are narrower single-model adaptations.", + "best_pr_reason": "No single best PR for deduping; they are distinct per-model changes rather than one underlying fix.", "soft_edge_verdicts": [ { "left": "pull_request:46069", "right": "pull_request:46073", "accept": false, - "reason": "Same general SDPA-dispatch theme, but ShieldGemma2 and CTRL are different model families with different concrete code paths; they should not be merged as duplicates." + "reason": "Different models and code paths: ShieldGemma2 image classification vs CTRL causal LM; same general feature, not the same change." }, { "left": "pull_request:46069", "right": "pull_request:46154", "accept": false, - "reason": "These are separate model implementations (ShieldGemma2 vision classifier vs T5 seq2seq) adding the same feature type, not the same bug/change." + "reason": "Different model families and attention implementations; one is ShieldGemma2, the other is T5, so they are not the same underlying bug/change." }, { "left": "pull_request:46069", "right": "pull_request:46196", "accept": false, - "reason": "Different model families and files; this is shared feature work, not one underlying code-path fix." + "reason": "Distinct multimodal model codebases (ShieldGemma2 vs SeamlessM4T/v2); too different to merge as duplicates." }, { "left": "pull_request:46073", "right": "pull_request:46154", "accept": false, - "reason": "CTRL and T5 are distinct architectures and implementations; similar SDPA support does not make them duplicates." + "reason": "Both add SDPA dispatch, but for different architectures (CTRL vs T5) and different attention paths, so not duplicates." }, { "left": "pull_request:46073", "right": "pull_request:46196", "accept": false, - "reason": "Different concrete model code paths (CTRL vs SeamlessM4T/v2), so they are not mergeable as one PR." + "reason": "CTRL and SeamlessM4T/v2 are unrelated model implementations with separate fixes; same feature area only." }, { "left": "pull_request:46154", "right": "pull_request:46196", "accept": false, - "reason": "T5 and SeamlessM4T/v2 are different model stacks; these are separate SDPA enablement changes, not the same fix." + "reason": "T5 and SeamlessM4T/v2 have different attention code paths and model behavior; not the same concrete bug or change." } ] }, @@ -20656,8 +20808,8 @@ }, { "label": "pull_request", - "component_index": 27, - "component_count": 27, + "component_index": 28, + "component_count": 28, "review_unit_index": 1, "review_unit_count": 1, "status": "reviewed", @@ -20673,7 +20825,7 @@ "soft_pairs": [ "pull_request:46152|pull_request:46207" ], - "prepared_review_unit_hash": "0c13bcb38527c26b261cbe65ffd1c1e3003fcf35034e87e26c3eff972d4ed301", + "prepared_review_unit_hash": "09bbdf87d91f5f2453ced59a2e7f15296ef4c53c18c9e5713a53051ab1d9b8dc", "component_budget": { "node_count": 2, "item_count": 2, @@ -20695,18 +20847,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs remove file-based source inspection, but they target different helpers in different modules (`import_utils` vs `modeling_utils`) and fix different behavior, so they are not duplicates.", - "confidence": 0.92, + "summary": "These PRs are related by theme (removing source-file reads/refactoring introspection), but they address different code paths in different modules, so they are not duplicates.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 46207 is the better representative because it targets a specific code path in `modeling_utils` and includes tests; PR 46152 is a separate `import_utils` refactor.", + "canonical_pr_reason": "PR 46207 is the more concrete and self-contained change: it targets a specific modeling-utils helper, adds tests, and fixes a narrowly scoped behavior. PR 46152 is a broader utility refactor in a different module.", "best_issue_reason": null, - "best_pr_reason": "PR 46207 is the strongest standalone change in this cluster: it replaces source-reading with introspection for `_can_set_*_implementation` and has accompanying tests, while 46152 is unrelated beyond the broad theme of avoiding file reads.", + "best_pr_reason": "PR 46207 is the best representative of the cluster because it has a clearly defined functional target and test coverage, making it easier to evaluate than the more infrastructure-oriented PR 46152.", "soft_edge_verdicts": [ { "left": "pull_request:46152", "right": "pull_request:46207", "accept": false, - "reason": "Same high-level theme, but different modules, different concrete helpers, and different behavior changes; they do not look like the same underlying fix and would not plausibly merge into one PR." + "reason": "Both PRs avoid reading source files, but they fix different problems: 46152 changes import-utils path resolution, while 46207 replaces source-reading with class introspection for attention-implementation detection in modeling_utils. Different modules, different code paths, not one mergeable fix." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 6220d1e8280a452fc27756b41c420744ba06aa7c..e8a21cc61d50484320fc23859def2a5875b77911 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260527t000053z", + "analysis_id": "hybrid-model-20260527t060031z", "archived_artifacts": { - "hybrid": "snapshots/20260527T000053Z/analysis-runs/hybrid-model-20260527t000053z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260527T000053Z/analysis-runs/hybrid-model-20260527t000053z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-05-27T00:05:40Z", + "published_at": "2026-05-27T06:09:40Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260527T000053Z", + "snapshot_id": "20260527T060031Z", "variant": "hybrid" } diff --git a/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.json b/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..ffbf388b06b25fdb42fb08cbccbc98ced9b2832c --- /dev/null +++ b/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.json @@ -0,0 +1,2194 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260527T060031Z", + "generated_at": "2026-05-27T06:09:38Z", + "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-46032-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46032.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 46032, + "canonical_pr_number": 46084, + "issue_numbers": [ + 46032 + ], + "pr_numbers": [ + 46084, + 46098 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46084, + "right_pr_number": 46098, + "code_similarity": 0.43, + "size_similarity": 0.5, + "file_overlap": 0.667, + "area_overlap": 0.226, + "patch_similarity": 0.339, + "shared_filenames": [ + "src/transformers/models/mamba2/modeling_mamba2.py", + "tests/models/mamba2/test_modeling_mamba2.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/mamba2/modeling_mamba2.py", + "left_ranges": [ + [ + 256, + 262 + ], + [ + 303, + 353 + ], + [ + 461, + 467 + ], + [ + 471, + 491 + ], + [ + 496, + 502 + ], + [ + 599, + 609 + ] + ], + "right_ranges": [ + [ + 255, + 373 + ], + [ + 481, + 495 + ] + ] + }, + { + "filename": "tests/models/mamba2/test_modeling_mamba2.py", + "left_ranges": [ + [ + 206, + 236 + ], + [ + 285, + 322 + ] + ], + "right_ranges": [ + [ + 476, + 542 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43240-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43240.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43240, + "canonical_pr_number": 43251, + "issue_numbers": [ + 43240 + ], + "pr_numbers": [ + 43251, + 43254 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43251, + "right_pr_number": 43254, + "code_similarity": 0.794, + "size_similarity": 0.64, + "file_overlap": 1.0, + "area_overlap": 0.667, + "patch_similarity": 0.882, + "shared_filenames": [ + "src/transformers/loss/loss_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/loss/loss_utils.py", + "left_ranges": [ + [ + 30, + 50 + ] + ], + "right_ranges": [ + [ + 30, + 43 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-41115-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41115.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41115, + "canonical_pr_number": 45613, + "issue_numbers": [ + 41115 + ], + "pr_numbers": [ + 41116, + 45613 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41116, + "right_pr_number": 45613, + "code_similarity": 0.805, + "size_similarity": 0.885, + "file_overlap": 0.75, + "area_overlap": 0.854, + "patch_similarity": 0.693, + "shared_filenames": [ + "docs/source/en/_toctree.yml", + "docs/source/en/model_doc/minicpm3.md", + "src/transformers/models/auto/modeling_auto.py", + "src/transformers/models/minicpm3/__init__.py", + "src/transformers/models/minicpm3/configuration_minicpm3.py", + "src/transformers/models/minicpm3/modeling_minicpm3.py", + "src/transformers/models/minicpm3/modular_minicpm3.py", + "tests/models/minicpm3/__init__.py", + "tests/models/minicpm3/test_modeling_minicpm3.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/_toctree.yml", + "left_ranges": [ + [ + 713, + 720 + ] + ], + "right_ranges": [ + [ + 713, + 720 + ] + ] + }, + { + "filename": "docs/source/en/model_doc/minicpm3.md", + "left_ranges": [ + [ + 1, + 67 + ] + ], + "right_ranges": [ + [ + 1, + 45 + ] + ] + }, + { + "filename": "src/transformers/models/auto/modeling_auto.py", + "left_ranges": [ + [ + 285, + 291 + ], + [ + 700, + 706 + ], + [ + 1298, + 1304 + ] + ], + "right_ranges": [ + [ + 699, + 705 + ], + [ + 1301, + 1307 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/__init__.py", + "left_ranges": [ + [ + 1, + 29 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/configuration_minicpm3.py", + "left_ranges": [ + [ + 1, + 141 + ] + ], + "right_ranges": [ + [ + 1, + 126 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 544 + ] + ], + "right_ranges": [ + [ + 1, + 522 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modular_minicpm3.py", + "left_ranges": [ + [ + 1, + 444 + ] + ], + "right_ranges": [ + [ + 1, + 342 + ] + ] + }, + { + "filename": "tests/models/minicpm3/test_modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 109 + ] + ], + "right_ranges": [ + [ + 1, + 136 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45647-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45647.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45647, + "canonical_pr_number": 45738, + "issue_numbers": [ + 45647 + ], + "pr_numbers": [ + 45737, + 45738 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 45737, + "right_pr_number": 45738, + "code_similarity": 0.526, + "size_similarity": 0.11, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.024, + "shared_filenames": [ + "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py", + "left_ranges": [ + [ + 491, + 497 + ] + ], + "right_ranges": [ + [ + 491, + 497 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45561-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45561, + "canonical_pr_number": 45645, + "issue_numbers": [ + 45561 + ], + "pr_numbers": [ + 45639, + 45645 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45639, + "right_pr_number": 45645, + "code_similarity": 0.532, + "size_similarity": 0.619, + "file_overlap": 0.5, + "area_overlap": 0.543, + "patch_similarity": 0.451, + "shared_filenames": [ + "src/transformers/testing_utils.py", + "tests/utils/test_testing_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/testing_utils.py", + "left_ranges": [ + [ + 3529, + 3536 + ], + [ + 3753, + 3787 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 86 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46100-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 46100, + "issue_numbers": [], + "pr_numbers": [ + 46100, + 46114 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46100, + "right_pr_number": 46114, + "code_similarity": 1.0, + "size_similarity": 1.0, + "file_overlap": 1.0, + "area_overlap": 1.0, + "patch_similarity": 1.0, + "shared_filenames": [ + "tests/models/vit/test_modeling_vit.py" + ], + "shared_file_areas": [ + { + "filename": "tests/models/vit/test_modeling_vit.py", + "left_ranges": [ + [ + 216, + 226 + ] + ], + "right_ranges": [ + [ + 216, + 226 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44002-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44002, + "issue_numbers": [], + "pr_numbers": [ + 44002, + 44029 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44002, + "right_pr_number": 44029, + "code_similarity": 0.135, + "size_similarity": 0.43, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.329, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, + { + "cluster_id": "cluster-44018-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44068, + "issue_numbers": [], + "pr_numbers": [ + 44018, + 44068 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44018, + "right_pr_number": 44068, + "code_similarity": 0.766, + "size_similarity": 0.939, + "file_overlap": 1.0, + "area_overlap": 0.425, + "patch_similarity": 0.866, + "shared_filenames": [ + "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "tests/models/gpt_neo/test_modeling_gpt_neo.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "left_ranges": [ + [ + 26, + 46 + ], + [ + 138, + 143 + ], + [ + 182, + 187 + ], + [ + 283, + 295 + ], + [ + 327, + 341 + ], + [ + 348, + 354 + ], + [ + 360, + 369 + ], + [ + 399, + 406 + ], + [ + 411, + 419 + ], + [ + 428, + 433 + ], + [ + 472, + 492 + ], + [ + 507, + 513 + ], + [ + 519, + 528 + ], + [ + 541, + 559 + ], + [ + 562, + 567 + ], + [ + 595, + 601 + ], + [ + 607, + 614 + ], + [ + 627, + 643 + ], + [ + 685, + 690 + ], + [ + 708, + 714 + ], + [ + 720, + 727 + ], + [ + 740, + 757 + ], + [ + 761, + 766 + ], + [ + 780, + 786 + ], + [ + 791, + 798 + ], + [ + 807, + 822 + ], + [ + 840, + 845 + ] + ], + "right_ranges": [ + [ + 26, + 31 + ], + [ + 34, + 44 + ], + [ + 136, + 141 + ], + [ + 180, + 185 + ], + [ + 281, + 293 + ], + [ + 325, + 339 + ], + [ + 346, + 352 + ], + [ + 358, + 367 + ], + [ + 397, + 404 + ], + [ + 409, + 417 + ], + [ + 426, + 434 + ], + [ + 463, + 483 + ], + [ + 498, + 504 + ], + [ + 510, + 519 + ], + [ + 532, + 550 + ], + [ + 553, + 564 + ], + [ + 586, + 592 + ], + [ + 598, + 605 + ], + [ + 618, + 634 + ], + [ + 676, + 688 + ], + [ + 699, + 705 + ], + [ + 711, + 718 + ], + [ + 731, + 748 + ], + [ + 752, + 762 + ], + [ + 771, + 777 + ], + [ + 782, + 789 + ], + [ + 798, + 813 + ], + [ + 831, + 836 + ] + ] + }, + { + "filename": "tests/models/gpt_neo/test_modeling_gpt_neo.py", + "left_ranges": [ + [ + 458, + 464 + ] + ], + "right_ranges": [ + [ + 458, + 464 + ] + ] + } + ] + } + ] + } + ], + "duplicate_issues": [], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 45613, + "duplicate_pr_numbers": [ + 41116 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-44002-2", + "canonical_pr_number": 44002, + "duplicate_pr_numbers": [ + 44029 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44002-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-44018-2", + "canonical_pr_number": 44068, + "duplicate_pr_numbers": [ + 44018 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45561-3", + "canonical_pr_number": 45645, + "duplicate_pr_numbers": [ + 45639 + ], + "target_issue_number": 45561, + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45647-3", + "canonical_pr_number": 45738, + "duplicate_pr_numbers": [ + 45737 + ], + "target_issue_number": 45647, + "reason": "PRs in cluster-45647-3 are treated as duplicates because they converge on issue #45647 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46032-3", + "canonical_pr_number": 46084, + "duplicate_pr_numbers": [ + 46098 + ], + "target_issue_number": 46032, + "reason": "PRs in cluster-46032-3 are treated as duplicates because they converge on issue #46032 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46100-2", + "canonical_pr_number": 46100, + "duplicate_pr_numbers": [ + 46114 + ], + "target_issue_number": null, + "reason": "PRs in cluster-46100-2 are treated as duplicates because they share soft_similarity evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-45647-3", + "issue_number": 45647, + "reason": "Issue #45647 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 13 discussion comments plus 8 inbound references.", + "score": 164.0 + }, + "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/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..2aa5d0b504b739b10fafe99cd26231b157e0cf23 --- /dev/null +++ b/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,20873 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260527T060031Z", + "generated_at": "2026-05-27T06:09:38Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43425", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43827", + "issue:43828", + "issue:43866", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45081", + "issue:45161", + "issue:45237", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:45874", + "issue:45901", + "issue:45902", + "issue:45920", + "issue:46015", + "issue:46165" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:45507", + "issue:44292|issue:45237", + "issue:30277|issue:35545", + "issue:35545|issue:42915", + "issue:35545|issue:41720", + "issue:45901|issue:46015", + "issue:44509|issue:44512", + "issue:35545|issue:36010", + "issue:44805|issue:45823", + "issue:44918|issue:46165", + "issue:43638|issue:44805", + "issue:43638|issue:46165", + "issue:43596|issue:44805", + "issue:44805|issue:45161", + "issue:44560|issue:46165", + "issue:43596|issue:46165", + "issue:44805|issue:44918", + "issue:43572|issue:46165", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43039|issue:45161", + "issue:43975|issue:44457", + "issue:44292|issue:44457", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:45859|issue:45902", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:39290|issue:45081", + "issue:45823|issue:45874", + "issue:43425|issue:44292", + "issue:34928|issue:41720", + "issue:43828|issue:45161", + "issue:44918|issue:45161", + "issue:43828|issue:44292", + "issue:43716|issue:45685", + "issue:43541|issue:43716", + "issue:44292|issue:44918", + "issue:43716|issue:44292", + "issue:45507|issue:45685", + "issue:43039|issue:44918", + "issue:41762|issue:43638", + "issue:43827|issue:44512", + "issue:43716|issue:44560", + "issue:43828|issue:45685", + "issue:44918|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:34928|issue:36010", + "issue:45237|issue:46015", + "issue:43039|issue:44805", + "issue:41762|issue:43596", + "issue:39290|issue:41720", + "issue:45237|issue:45901", + "issue:43901|issue:44512", + "issue:44841|issue:45823", + "issue:43039|issue:44560", + "issue:43975|issue:46015", + "issue:45812|issue:45920", + "issue:44863|issue:46015", + "issue:43866|issue:44863", + "issue:44292|issue:46015" + ], + "prepared_review_unit_hash": "f760006c58d5943870603243aef41d69b3184c6414fb83f73f29977276588052", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 69, + "serialized_chars": 48405, + "estimated_input_tokens": 12102, + "estimated_eval_tokens": 24460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of separate bug families rather than one single duplicate set. The strongest duplicate family is the AutoTokenizer v5 regression affecting multiple model families (45812/45920), with a smaller duplicate pair around ZeRO-3 Bert loading (43596/43638) and one around Qwen3_5MoeTextRotaryEmbedding offload/materialization behavior (45859/45902).", + "confidence": 0.83, + "canonical_issue_reason": "issue:45920 is the best umbrella issue: it describes the broader AutoTokenizer token-id regression across multiple model families, while issue:45812 is the narrower Granite-only report.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the strongest global representative because it is the broadest, most up-to-date report of the tokenizer regression and already consolidates several affected models.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different failures: missing config attribute vs streaming chat/completions inference error." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Unrelated backend/hardware errors: AMD GPU failure vs MPS histogram not implemented for Int." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different model/code paths: image-preprocessor dtype mismatch vs GPT-OSS AMD GPU failure." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Both are runtime mismatches, but they affect different models and code paths." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different problem surfaces: AMD GPU execution vs GraniteMoEHybrid calling an invalid method." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Different model-specific runtime errors; no shared concrete code path." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch and ONNX export failure are unrelated." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ONNX export error vs FineGrainedFP8Config failure; different model features and code paths." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "Export-time failure vs device-mapping cuda assert; not the same bug." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "Task crash vs docs/example update; no shared underlying bug." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions in v5, but they concern different removed features and likely separate edits." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "Model export issue and GenerationMixin import error are unrelated." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Mask/index shape mismatch and Gemma4 device mismatch are different failures." + }, + { + "left": "issue:44918", + "right": "issue:46165", + "accept": false, + "reason": "TRL SFT embedding unpacking and resume/trust_remote_code are unrelated." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both are IndexError-ish, but the models and execution paths differ substantially." + }, + { + "left": "issue:43638", + "right": "issue:46165", + "accept": false, + "reason": "ZeRO-3 Bert loading and resume-training/trust_remote_code are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Same broad IndexError symptom, but not the same model/path or root cause." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Mask/index tensor mismatch is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Video StopIteration in Qwen3-vl embedding is unrelated to resume-training behavior." + }, + { + "left": "issue:43596", + "right": "issue:46165", + "accept": false, + "reason": "Different code paths and symptoms." + }, + { + "left": "issue:43572", + "right": "issue:46165", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig is unrelated to model-resume issues." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "Processor voice_preset bug is unrelated to ModernBERT ONNX export." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but they are different failures on different models and code paths." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same ZeRO-3 Bert loading failure class: both report index-0-out-of-bounds during loading/init of Bert models, with only wording/model specificity differing." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross-entropy invocation and GPT-OSS tensor-parallel failure are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Detokenization regression and LoRA merge/save mismatch are different bugs." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error does not match LoRA weight merge/save inconsistency." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "Activation-checkpointing/FSDP recomputation mismatch is not the same as the mask-shape IndexError." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Different DeepSpeed problems: weight-dimension error vs collective mismatch." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": true, + "reason": "Same Qwen3_5MoeTextRotaryEmbedding offload/materialization bug family; both point to broken handling of rotary-embedding state across device moves/offload." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "Bert ZeRO-3 loading and GPT-OSS TP failure are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "Activation checkpointing/FSDP issue is unrelated to the custom_trainer compute_loss signature change." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "Different models and failure modes." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Multimodal dtype mismatch and invalid-method call are unrelated." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3 sliding_window_pattern regression and Mistral tokenizer regex patch crash are unrelated." + }, + { + "left": "issue:45823", + "right": "issue:45874", + "accept": false, + "reason": "Auto device-map forward mismatch and expensive reverse embedding lookup are different Gemma4 issues." + }, + { + "left": "issue:43425", + "right": "issue:44292", + "accept": false, + "reason": "Torch version incompatibility and Qwen-3-8B-NVFP4 runtime error are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:41720", + "accept": false, + "reason": "FSDP activation-checkpointing bug and Qwen3 auto device-map cuda assert are different issues." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Phi autocast dtype mismatch and GPT-OSS tensor-parallel failure are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "Qwen3.5 input-embedding unpacking with TRL is not the same bug as GPT-OSS TP failure." + }, + { + "left": "issue:43828", + "right": "issue:44292", + "accept": false, + "reason": "Different models and failure modes." + }, + { + "left": "issue:43716", + "right": "issue:45685", + "accept": false, + "reason": "Image-preprocessor dtype mismatch and MPS histogram int limitation are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43716", + "accept": false, + "reason": "Torch Dynamo grouped_mm tracing failure and Mistral image-preprocessor dtype mismatch are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:44918", + "accept": false, + "reason": "Different Qwen-related errors with different subsystems and root causes." + }, + { + "left": "issue:45507", + "right": "issue:45685", + "accept": false, + "reason": "GraniteMoEHybrid invalid method and MPS histogram int error are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44918", + "accept": false, + "reason": "Liger Kernel cross-entropy call and Qwen3.5 embedding unpacking are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": false, + "reason": "Both are ZeRO-3 load/index errors, but Gemma3 loading and Bert loading are different code paths and likely separate bugs." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions, but they refer to different removed v5 features and separate documentation fixes." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Mistral image dtype mismatch and Qwen3-vl video StopIteration are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch and MPS histogram limitation are unrelated." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 embedding unpacking and MPS histogram limitation are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Different runtime issues on different model/code paths." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and custom_trainer signature mismatch are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "Activation-checkpointing/FSDP issue is unrelated to GenerationMixin import failure." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "GPT-OSS AMD GPU failure and outdated RAG docs are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross-entropy issue and mask/tensor shape IndexError are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Similar ZeRO-3 symptom, but Gemma3 loading and Bert loading are different model families and likely separate bugs." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Gemma3 config attribute regression with vLLM and Qwen3 auto device mapping cuda assert are unrelated." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "AMD GPU support issue and table-question-answering crash are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "return_all_scores docs mismatch and transformers run-command docs removal are separate documentation issues." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Voxtral processor failure and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint loading and RAG docs are unrelated." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Checkpoint corruption in Ovis2 and NemotronH checkpoint loading are different problems." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "Qwen NVFP4 runtime error and outdated RAG docs are unrelated." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same AutoTokenizer v5 regression; 45920 is the broader multi-model report and 45812 is the narrower Granite-only instance." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays appropriately conservative. The accepted soft-edge pairs are plausible same-bug families based on the issue titles, and the rejected pairs mostly avoid over-claiming across clearly different model/code paths. The umbrella issue selection around 45920 vs 45812 is also well supported by the titles." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 2, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36010", + "issue:39186", + "issue:39290", + "issue:40990", + "issue:41093", + "issue:41762", + "issue:43054", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43645", + "issue:43716", + "issue:43792", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44589", + "issue:44841", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45161", + "issue:45362", + "issue:45399", + "issue:45464", + "issue:45542", + "issue:45632", + "issue:45685", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45901", + "issue:45902" + ], + "soft_pairs": [ + "issue:39290|issue:43531", + "issue:43792|issue:45901", + "issue:43854|issue:43866", + "issue:41093|issue:43638", + "issue:43425|issue:45070", + "issue:43596|issue:45800", + "issue:43975|issue:45901", + "issue:43792|issue:45070", + "issue:45362|issue:45464", + "issue:44292|issue:45901", + "issue:39290|issue:45823", + "issue:45464|issue:45685", + "issue:43404|issue:43866", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:44841|issue:45084", + "issue:36010|issue:39290", + "issue:45800|issue:45823", + "issue:41093|issue:41762", + "issue:43572|issue:45070", + "issue:43296|issue:43366", + "issue:43541|issue:45161", + "issue:44387|issue:45005", + "issue:43828|issue:45464", + "issue:44291|issue:45070", + "issue:43366|issue:45464", + "issue:45020|issue:45800", + "issue:43531|issue:45070", + "issue:43541|issue:44589", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45759|issue:45902", + "issue:45542|issue:45800", + "issue:39186|issue:44918", + "issue:43716|issue:44918", + "issue:40990|issue:44960", + "issue:43645|issue:45399", + "issue:45632|issue:45698", + "issue:43404|issue:45005" + ], + "prepared_review_unit_hash": "4c6c5128649708929ffa9530bf0fcbd3547fe8a3b804df720a66e12fc16eb647", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 39, + "serialized_chars": 39129, + "estimated_input_tokens": 9783, + "estimated_eval_tokens": 19822 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a heterogeneous issue cluster with mostly false-positive similarities. The only strong duplicate-looking pair is the remote-code custom-module loading/cache collision reports.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45698 is the best canonical issue because it states the broad user-facing failure (\u201cfrom_pretrained loads wrong custom module after save_pretrained\u201d) and appears to be the general symptom behind the more specific cache-path-collision report in issue:45632.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is the best overall issue to anchor the cluster: it is open, broadly phrased, and covers the central custom-module loading problem more directly than the narrower cache-path collision variant.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Different models and different failures; one is a missing config attribute, the other is a sliding-window model issue." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Unrelated models and subsystems; no shared bug path." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Both are model-loading problems, but the models and failure modes are different." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both mention zero-3/indexing symptoms, but the specific errors and model paths differ too much to be the same bug." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "One is torch-version compatibility, the other is a PretrainedConfig/pydantic regression." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "A DeepSpeed ZeRO-3 indexing crash is not the same as a torch/transformers version incompatibility." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Different features and errors; detokenization vs table QA crash." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "No shared code path; Whisper runtime failure and PretrainedConfig serialization break are unrelated." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Both involve Qwen3.5 and chat, but one is a transformers chat crash and the other is a streaming API failure." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "Completely different model/task and failure mode." + }, + { + "left": "issue:39290", + "right": "issue:45823", + "accept": false, + "reason": "Different models and distinct runtime issues." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Chat streaming failure and MPS integer histogram error are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Unrelated models; tied-weights bug vs corrupted checkpoint report." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Different hardware/code paths and error types." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Both are numeric/runtime errors, but the model, backend, and exact failure are different." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Processor failure for one model is unrelated to a template-node compile error." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "ImportError for GenerationMixin and Gemma3 config attribute errors are unrelated." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Torch version incompatibility and Gemma4 device mismatch are separate problems." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are ZeRO-3 related index crashes, but the symptoms and triggering paths differ." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Both are v5 regressions, but one is a missing config field and the other is a pydantic model break." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "Different model families and different loading/support problems." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Grouped_mm tracing failure is not the same as GPT-OSS tensor-parallel support." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Int4 OOM regression and tied-weights issue are different bug classes." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Autocast dtype mismatch and streaming chat API failure are unrelated." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Unexpected init_empty_weights argument and PretrainedConfig field regression are distinct." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "GGUF GPT-OSS support and Qwen3.5 chat streaming are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Remote-code breakage and torch/transformers version incompatibility are different failure modes." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Sliding-window model bug and config field regression do not match." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Grouped_mm tracing error and missing Float8 storage object are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Same broad ZeRO-3 area, but the concrete bugs and triggers differ." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Both mention siglip2, but one is about embedding quality while the other is a loading/pipeline correctness issue." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "One is config unwrapping for multimodal CausalLM instantiation; the other is a Qwen3.5 MoE embedding initialization memory bug." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "TensorBoard/TensorFlow backend handling and torch version incompatibility are unrelated." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP weight-shape error and embedding unpacking failure in SFT trainer are different bugs." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Mistral image-preprocessor dtype mismatch is unrelated to Qwen3.5 embedding unpacking." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "Perplexity quality issue and GLM5 issue are not the same underlying bug." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Custom-model notebook initialization and flash-attn fallback gating are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": true, + "reason": "Both point to the same remote-code/custom-module loading path, with a cache-path collision plausibly causing the wrong module to be loaded after save/load." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Both mention tied weights, but they target different models and different failure contexts." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative. The main linkage between issue:45632 and issue:45698 is well supported by the titles, and the other soft-edge rejections are appropriately cautious rather than overstated." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 3, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:39692", + "issue:41553", + "issue:41628", + "issue:41762", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43493", + "issue:43643", + "issue:43644", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:45005", + "issue:45020", + "issue:45237", + "issue:45399", + "issue:45507", + "issue:45663", + "issue:45685", + "issue:45701", + "issue:45800", + "issue:45823", + "issue:45859", + "issue:45902" + ], + "soft_pairs": [ + "issue:42915|issue:45399", + "issue:44530|issue:45005", + "issue:43873|issue:45005", + "issue:44863|issue:45005", + "issue:44360|issue:44512", + "issue:44918|issue:45507", + "issue:43975|issue:44292", + "issue:44492|issue:44512", + "issue:44292|issue:45507", + "issue:43950|issue:44534", + "issue:43644|issue:43950", + "issue:39692|issue:43054", + "issue:43828|issue:44928", + "issue:36331|issue:41762", + "issue:41628|issue:45800", + "issue:43950|issue:45902", + "issue:43872|issue:44589", + "issue:43866|issue:44403", + "issue:43054|issue:45701", + "issue:41553|issue:43883", + "issue:38175|issue:43054", + "issue:45663|issue:45902", + "issue:44910|issue:45859", + "issue:41553|issue:43881", + "issue:42947|issue:44387", + "issue:43054|issue:43493", + "issue:44403|issue:44451", + "issue:43975|issue:45237", + "issue:44403|issue:44488", + "issue:43643|issue:45020", + "issue:42947|issue:43856", + "issue:41762|issue:42915", + "issue:43881|issue:43883", + "issue:45685|issue:45823" + ], + "prepared_review_unit_hash": "273c8b3e27c67ba23d635700ec96046b0b8a4b1bfebecda49cf6039548b81820", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37579, + "estimated_input_tokens": 9395, + "estimated_eval_tokens": 19046 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly heterogeneous, with one clear duplicate-looking pair around the Transformers v5 non-persistent buffer corruption regression. I accepted that edge; the rest look like related-but-distinct bugs or entirely different issues.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 43950 is the clearest canonical issue: it names the regression, the code path (`from_pretrained()`), and the underlying non-persistent buffer corruption. Its twin 44534 is a shorter rephrasing of the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "43950 is the best issue to keep as the representative bug report because it is the most specific and actionable description of the duplicate pair, with the regression and mechanism stated explicitly.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Both are loading/compatibility failures, but one is FineGrainedFP8/Qwen3MoE and the other is flash-attn fallback gating; different code paths and symptoms." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "Qwen3.5 PagedAttentionCache crash vs tied-weights regression on translation models; not the same bug." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior is a different failure mode from tied-weights handling." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading failure is unrelated to the translation tied-weights issue." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "DSA indexer implementation concern vs a docs typo; unrelated." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Different models and different failures: embedding unpacking with TRL SFT vs invalid method call in GraniteMoEHybrid." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Detokenization bug in DeepSeek Coder is not the same as Qwen-3 NVFP4 runtime error." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "Cache-strategy typo in docs is unrelated to removed run-command docs drift." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Different model families and distinct runtime failures." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Near-identical regression reports about Transformers v5 corrupting/filling non-persistent buffers with junk; same underlying bug." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": false, + "reason": "Both mention v5 regressions, but one is about `AutoConfig` missing fields and the other about buffer corruption; different mechanisms." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "SigLIP2 docs/example errors are different from a broad text-embedding quality discrepancy." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Both involve Qwen3.5/typing issues, but one is autocast dtype mismatch and the other is RLHF gradient explosion from a different position_ids/SDPA path." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "CustomTrainer kwarg mismatch and Gemma3 ZeRO-3 loading IndexError are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45800", + "accept": false, + "reason": "AutoImageProcessor import failure is not the same as torch/transformers version incompatibility." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "Related buffer/materialization theme, but 45902 is a model-specific uninitialized-memory bug rather than the same general from_pretrained corruption report." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "bitsandbytes Int8Params kwarg incompatibility is unrelated to missing Float8 storage." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Corrupted checkpoint vs loading noise; no shared bug." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "Different reports about embedding quality vs tokenization changes; not the same concrete defect." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad AutoTokenizer error message for Voxtral vs missing `all_tied_weights_keys` on Molmo; different issues." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "SigLIP2 zero-probabilities vs text-embedding quality regression are separate concerns." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states bug is unrelated to Qwen3_5MoE rotary embedding memory initialization." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Both touch Qwen3.5, but one is Flash Attention illegal access from 3D position_ids and the other is CPU offload incompatibility." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Different models and failures: Voxtral tokenizer loading vs glm-4v loading failure." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Gradient checkpointing/LoRA behavior is unrelated to int4 quantization memory growth." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "Both concern SigLIP2, but one is quality regression and the other is HF-vs-JAX implementation discrepancy; not the same exact bug." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Loading noise is not the same as a model failing to load ScandiBERT." + }, + { + "left": "issue:43975", + "right": "issue:45237", + "accept": false, + "reason": "Detokenization bug is unrelated to AMD GPU loading failure." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Two separate loading-related issues with different models and symptoms." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Remote-code loading problems are broadly related, but one is missing fields from `AutoConfig` and the other is a wider compatibility complaint." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "LoRA gradient checkpointing vs MoE training memory inefficiency are not the same underlying defect." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading IndexError and Qwen3MoE FP8 failure are distinct model/runtime issues." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Two separate model-loading failures with different root causes." + }, + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "MPS histogram kernel limitation is unrelated to Gemma4 device-mismatch during forward." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the lone accepted edge (43950/44534) is well supported by nearly identical titles and the shared non-persistent buffer regression theme. The rejected edges are stated conservatively and mostly distinguish different models, mechanisms, or symptom classes without overstating equivalence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 4, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43404", + "issue:43638", + "issue:43742", + "issue:43782", + "issue:43792", + "issue:43824", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43940", + "issue:43950", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44560", + "issue:44661", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45356", + "issue:45399", + "issue:45538", + "issue:45588", + "issue:45823", + "issue:45902", + "issue:45923" + ], + "soft_pairs": [ + "issue:43782|issue:44960", + "issue:44991|issue:45399", + "issue:43824|issue:43872", + "issue:42491|issue:44368", + "issue:45020|issue:45823", + "issue:44661|issue:45020", + "issue:43828|issue:45823", + "issue:45081|issue:45399", + "issue:43742|issue:45399", + "issue:45020|issue:45356", + "issue:43404|issue:43950", + "issue:43638|issue:45823", + "issue:42915|issue:45020", + "issue:43827|issue:44509", + "issue:43872|issue:44291", + "issue:43792|issue:44220", + "issue:43257|issue:43866", + "issue:42915|issue:43872", + "issue:41553|issue:44991", + "issue:43257|issue:45305", + "issue:36331|issue:41093", + "issue:42947|issue:45923", + "issue:43940|issue:45313", + "issue:43854|issue:44863", + "issue:44928|issue:45588", + "issue:43872|issue:45538", + "issue:41093|issue:44805", + "issue:41553|issue:44488", + "issue:44479|issue:44560", + "issue:44530|issue:44863", + "issue:24643|issue:30064", + "issue:45325|issue:45902" + ], + "prepared_review_unit_hash": "8893ceb9917a0398648acd1d8d88ce6d23ad87cc1a3c57624fb0abecb44a0ad3", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37733, + "estimated_input_tokens": 9434, + "estimated_eval_tokens": 19124 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are distinct Transformers regressions around model loading, tokenizers, DeepSpeed, PEFT, quantization, or docs. None of the soft pairs look like the exact same underlying bug/change, so all should stay separate.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45020 is the broadest and most representative umbrella: it covers recent-version loading regressions in a generic way and can anchor nearby model/tokenizer breakages without being too specific.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best global issue representative because it is broad, user-facing, and closest to the common theme of version-induced loading failures across models.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Different problems: Qwen3VL weight_only loading error vs GLM5 model loading; no evidence they share the same failing code path." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading failure for a specific model vs flash-attn fallback gating; unrelated subsystems and fixes." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Import error for a missing class vs bitsandbytes Int8Params constructor incompatibility; different failure modes." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Both mention LoRA/Qwen3.5 warnings, but one is a hf4.x vs hf5.x model incompatibility and the other is a tie_word_embeddings warning; not the same bug." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "Broad remote_code breakage vs a Gemma4 device_map forward mismatch; too different in trigger and code path." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Tokenizer mapping-name edge case vs generic remote_code regression; related area, but not the same defect." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Autocast dtype mismatch in Phi-tiny-MoE vs Gemma4 device mismatch; distinct runtime issues." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Mistral regex patch crashing in tokenizer loading vs flash-attn2 fallback selection; unrelated." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "Key error loading a specific model vs flash-attention fallback logic; no shared underlying bug." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "General remote_code loading regressions vs Kimi-K2.5 tokenizer codec/fix_mistral_regex warning regression; different components." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Untied lm_head weights in Mistral3 vs non-persistent buffer corruption in from_pretrained; separate state-management bugs." + }, + { + "left": "issue:43638", + "right": "issue:45823", + "accept": false, + "reason": "DeepSpeed zero3 index error for non-pretrained Bert vs Gemma4 PLE device mismatch; unrelated training/forward paths." + }, + { + "left": "issue:42915", + "right": "issue:45020", + "accept": false, + "reason": "FineGrainedFP8Config failure for Qwen3 MoE vs broad remote_code regressions; different precision/configuration issue." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": false, + "reason": "These are the same documentation topic, but the titles indicate separate reports; without more evidence, don't merge." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": false, + "reason": "bitsandbytes Int8Params error vs init_empty_weights/_is_hf_initialized error; similar weight-init area but not the same bug." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Whisper model runtime failure vs _torch_extract_fbank_features issue; likely different audio pipeline failures." + }, + { + "left": "issue:43257", + "right": "issue:43866", + "accept": false, + "reason": "Qwen3 MOE weights not converted under accelerate+deepspeed vs corrupted Ovis2 checkpoint; different models and causes." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Qwen3 MoE FP8 config failure vs bitsandbytes constructor incompatibility; different layers of the stack." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Bad AutoTokenizer error message for Voxtral vs tokenizer loading regression for EMBEDDIA/est-roberta; same broad area, not same bug." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed vs gradient averaging with GAS and DeepSpeed; training/loading issues are different." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "CustomTrainer signature mismatch vs mask/tensor shape IndexError; unrelated trainer API vs tensor indexing bug." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Gradient checkpointing ineffective with PEFT LoRA vs missing supports_gradient_checkpointing flag on a remote-code model; not the same change." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Qwen3-Next ZeRO-3 missing weights vs Qwen3.5 ZeRO-3 missing language_model weights; similar symptom but different models and likely different loader logic." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Unable to load GLM-4.7-Flash in tests vs NemotronH checkpoints not loading; distinct model implementations." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "RLHF gradient explosion from 3D position_ids and SDPA fallback vs flash_attention crash on s_aux=None; unrelated attention-path bugs." + }, + { + "left": "issue:43872", + "right": "issue:45538", + "accept": false, + "reason": "bitsandbytes incompatibility vs CLIPTokenizer model_max_length sentinel value; different subsystems." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": false, + "reason": "Both are mask/tensor shape errors, but they are separate reports with no evidence of the same root cause." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Bad error message for Voxtral vs inability to load cjvt/sleng-bert; different tokenizer/model-loading problems." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "Both are Qwen3.5/VL regressions, but one is a video-input regression and the other is a StopIteration in embedding video handling; not clearly the same bug." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "PagedAttentionCache linear_attention failure in Qwen3.5 vs NemotronH checkpoint loading; different model families and code paths." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "DeepSpeed training weight-dimension error vs image processor void segmentation-map failure; unrelated." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Qwen2.5-VL rope index temporal scaling bug vs uninitialized inv_freq after meta->to_empty(cuda); both model internals, but clearly different defects." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 5, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43720", + "issue:43824", + "issue:43828", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44155", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:45030", + "issue:45276", + "issue:45313", + "issue:45357", + "issue:45584", + "issue:45663", + "issue:45685", + "issue:45823", + "issue:45834" + ], + "soft_pairs": [ + "issue:43366|issue:43828", + "issue:45663|issue:45823", + "issue:43366|issue:43531", + "issue:44155|issue:45584", + "issue:43950|issue:45834", + "issue:44560|issue:45685", + "issue:43824|issue:43957", + "issue:43645|issue:43646", + "issue:43550|issue:44910", + "issue:42915|issue:43957", + "issue:43366|issue:44589", + "issue:44560|issue:44918", + "issue:42617|issue:43366", + "issue:36010|issue:42915", + "issue:44661|issue:45823", + "issue:44560|issue:44805", + "issue:43632|issue:45834", + "issue:38175|issue:43994", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:44661|issue:44805", + "issue:44877|issue:45030", + "issue:43317|issue:44756", + "issue:43644|issue:44534", + "issue:38175|issue:39692", + "issue:42491|issue:43720", + "issue:41720|issue:42915", + "issue:44960|issue:44977", + "issue:44451|issue:44488", + "issue:42915|issue:44756", + "issue:45313|issue:45357", + "issue:42491|issue:45276", + "issue:43708|issue:44075", + "issue:43644|issue:43645" + ], + "prepared_review_unit_hash": "8dae01133f0021020bcf2e8e32b4ac3d5db916940aed7f14a0f17c9fb6e11ce2", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37794, + "estimated_input_tokens": 9449, + "estimated_eval_tokens": 19154 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous, with a few small duplicate mini-clusters around the same regression or model family. I accepted only pairs that look like the same underlying bug/change; most others are just related-by-topic at best.", + "confidence": 0.79, + "canonical_issue_reason": "Issue 44534 is the cleanest anchor for one of the true duplicate groups: it states the non-persistent-buffer regression broadly and matches an exact duplicate title.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44534 is the best overall issue to keep because it is the most general, clearly scoped regression report among the accepted duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "Different problems: GGUF/gpt-oss support vs a Phi-tiny-MoE dtype mismatch." + }, + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Same model family, but different failure modes: FSDP2 KeyError vs device_map auto device mismatch." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Unrelated Qwen3-MoE sliding_window bug vs gpt-oss GGUF support request." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different audio/Whisper code paths and symptoms; not the same bug." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Non-persistent-buffer corruption vs Kosmos2.5 long-ocr index error are unrelated." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Different architectures and failures: Qwen3-VL video StopIteration vs MPS int histogram support." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "ImportError for Qwen2.5-VL is unrelated to meta-device model loading breakage." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Both describe the same Transformers v5 custom model initialization regression; one just adds the Jupyter-notebook context." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different bugs: Bamba torch.compile/SDPA crash vs Qwen3.5 flash-attention memory access." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Qwen3-MoE FP8 failure and meta-device loading are different code paths." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "GGUF support request vs Float8 storage lookup error; no shared root cause." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py runtime failure is unrelated to GGUF model support." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "GenerationMixin import error is unrelated to Qwen3-MoE FP8 handling." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "Tokenizer-mapping/new-model helper bug vs Gemma4 PLE device mismatch; different issues." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Both are Qwen-related, but one is a video StopIteration bug and the other is a mask shape mismatch." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "_is_hf_initialized regression is unrelated to Kosmos2.5 long-input indexing." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Both report the same SigLIP2-base-patch16-224 bad-output regression, just with different symptom descriptions." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Zero3/Bert index error and tokenizer-mapping helper failure are different bugs." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Zero3/Bert empty-shard IndexError is unrelated to add-new-model-like tokenizer mapping." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping bug and tensor mask shape mismatch are unrelated." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": true, + "reason": "Both look like the same stricter config-validation regression preventing model configs from loading." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "GPU+CPU offload dequantized loading failure is unrelated to Strix Halo mmap OOM." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Exact same non-persistent-buffer junk regression in Transformers v5." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "SigLIP2 output corruption is not the same as the documentation example's processor/quantization errors." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Qwen3-MoE LoRA compatibility and BitNet packed-weight loading are different issues." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Different failure modes: auto device mapping cudaErrorAssert vs FineGrainedFP8Config." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 is unrelated to Qwen3.5 flash-attention generation failures." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": true, + "reason": "Both are the same latest-version model-loading failure for specific BERT models." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "ZeRO-3 language_model weight loading and save_pretrained visual-key corruption are different bugs." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Trainer resume/max_steps math bug is unrelated to SGD optimizer args being ignored." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Non-persistent-buffer corruption and custom model initialization breakage are different regressions." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The accepted duplicate pairs are supported by strong title/body similarity, especially the non-persistent-buffer regression around 43644/44534, and the rest of the pair judgments read as cautious rather than overstated. No PR mergeability issue applies here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 6, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "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:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44479", + "issue:44488", + "issue:44756", + "issue:44877", + "issue:44910", + "issue:44912", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:46050" + ], + "soft_pairs": [ + "issue:44488|issue:44991", + "issue:43854|issue:43931", + "issue:45488|issue:45812", + "issue:39692|issue:43994", + "issue:43257|issue:43854", + "issue:42491|issue:43931", + "issue:43856|issue:43873", + "issue:43856|issue:44387", + "issue:45362|issue:45406", + "issue:44960|issue:46050", + "issue:42491|issue:45292", + "issue:44164|issue:44291", + "issue:44910|issue:45588", + "issue:45276|issue:45335", + "issue:36010|issue:39404", + "issue:43404|issue:43454", + "issue:43425|issue:43957", + "issue:41720|issue:44155", + "issue:39692|issue:44960", + "issue:44402|issue:44488", + "issue:42947|issue:45704", + "issue:45663|issue:45759", + "issue:43873|issue:44387", + "issue:44291|issue:44756", + "issue:45663|issue:45859", + "issue:44155|issue:45823", + "issue:43335|issue:45663", + "issue:44292|issue:44912", + "issue:44387|issue:44756", + "issue:43957|issue:44756", + "issue:42947|issue:43381", + "issue:45276|issue:45292", + "issue:44164|issue:44756", + "issue:43854|issue:45216", + "issue:43381|issue:44928", + "issue:44479|issue:44877", + "issue:45588|issue:45663" + ], + "prepared_review_unit_hash": "18c3af5791d812a9e1a7cdaec9f3769649932a3d8fa014fcd540028296bb07b6", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 38940, + "estimated_input_tokens": 9735, + "estimated_eval_tokens": 19726 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a false-positive semantic cluster: the items span unrelated tokenizer loading, quantization, model-loading, flash-attention, and training bugs. The only clear duplicate set is the `resize_token_embeddings` regression affecting `output_embeddings` / `decoder.embed_tokens` across Gemma4 and T5Gemma reports.", + "confidence": 0.72, + "canonical_issue_reason": "issue:45292 is the best canonical issue because it states the broad API-level regression (`resize_token_embeddings` not updating `output_embeddings`) and the other matching reports look like model-specific manifestations of the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45292 is the most general and reusable description of the duplicate set; issue:45276 and issue:45335 are narrower variants of the same embedding-resize failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading failures, but they affect different models and likely different tokenizer paths; too broad to treat as the same bug." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different models and different failure modes (unit test loading vs weight-shape mismatch) indicate separate issues." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but one is a LlamaTokenizer pre-tokenizer override bug and the other is a Granite token-ID regression; not the same concrete fix." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is a documentation/example problem and the other is an actual inference regression; different underlying issues." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is unrelated to GLM-4.7-Flash loading failures." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "Different models and different problems; LoRA compatibility on hf5.x is not the same as a Qwen3-VL shape mismatch." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training memory use and the other is quantization offloading behavior; not the same bug." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both are memory-related, but one is MoE training footprint and the other is int4 quantization reserved-memory OOM; separate regressions." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Different model families and different failure surfaces (chat crash vs processor attribute error)." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "GLM5 is unrelated to the sinq save/load quantization regression." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "LoRA hf4.x/hf5.x compatibility is unrelated to `resize_token_embeddings` not updating embeddings." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Both concern loading/saving internals, but `extra_state` handling and `_is_hf_initialized` argument handling are distinct code paths." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Both are flash-attention crashes, but one is packed-sequence misinterpretation and the other is a missing `s_aux` attribute; different bugs." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": true, + "reason": "Both report the same `resize_token_embeddings` propagation bug, just on different model-specific embedding modules." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "Importing `GenerationMixin` and Whisper `return_language` pipeline behavior are unrelated issues." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both mention lm_head tying, but they are different model implementations and likely separate model-specific fixes." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch-version incompatibility and `torch.device(\"meta\")` loading breakage are different regressions." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Qwen3 device-map CUDA assert and AudioFlamingo3 batch token leak are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "SigLIP2 docs/example problems are unrelated to GLM5 model loading." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "One is tokenizer vocab-size metadata mismatch; the other is a model-specific tokenizer loading failure." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing/PEFT LoRA behavior is unrelated to the apex RMSNorm memory leak." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Different issues: Gemma-4 FSDP2 shared state handling versus Qwen multimodal `text_config` unwrapping." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Quantized offloading behavior and int4 reserved-memory OOM are not the same concrete bug." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "`init_empty_weights` argument handling is unrelated to Strix Halo mmap OOM." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Different code paths: Gemma-4 attention state rebuilding versus Qwen3_5 MoE rotary embedding CPU offload." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "AudioFlamingo3 track leakage and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction and Gemma-4 FSDP2 attention state handling are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Different quantization formats and failures: NVFP4 load error versus MXFP4 fallback to bf16." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Both end in OOM, but one is quantization memory reservation and the other is mmap behavior on specific hardware." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Model loading with meta tensors and disabling mmap for Strix Halo are unrelated fixes." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both mention gradient checkpointing, but one is ineffective under PEFT LoRA and the other is eval-mode incompatibility." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "They describe the same `resize_token_embeddings` regression, with one being the general report and the other a model-specific manifestation." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "Serialization `extra_state` handling is unrelated to mmap/OOM behavior." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "GLM-4.7-Flash loading and Qwen3.5 checkpoint save regression are separate model-specific issues." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Gradient checkpointing eval-mode behavior is unrelated to the Qwen3.5 3D position_ids / SDPA fallback bug." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video-input regression across Qwen VL models is unrelated to strict config blocking granite_speech loading." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Flash attention sink handling and Gemma-4 FSDP2 attention state rebuilding are different problems." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative overall. The only accepted duplicate set is the resize_token_embeddings regression across issue:45276, issue:45292, and issue:45335, which is supported by the titles and soft-pair signal. The other pairwise rejections are appropriately cautious given the limited evidence and model-specific differences." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 7, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36032", + "issue:39404", + "issue:43257", + "issue:43299", + "issue:43425", + "issue:43452", + "issue:43596", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43856", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44265", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44671", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45292", + "issue:45335", + "issue:45684", + "issue:45698", + "issue:45701", + "issue:45759", + "issue:45823", + "issue:45859" + ], + "soft_pairs": [ + "issue:45684|issue:45698", + "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:43257|issue:44863", + "issue:44589|issue:44756", + "issue:39404|issue:45701", + "issue:44683|issue:45823", + "issue:43425|issue:45823", + "issue:43611|issue:43645", + "issue:43425|issue:43716", + "issue:44530|issue:45084", + "issue:39404|issue:44912", + "issue:43906|issue:44530", + "issue:44265|issue:44898", + "issue:44479|issue:44805", + "issue:43994|issue:44079", + "issue:43425|issue:44918", + "issue:44589|issue:45084", + "issue:43901|issue:44509", + "issue:43299|issue:43931", + "issue:44164|issue:45759", + "issue:43596|issue:44683", + "issue:36032|issue:43452", + "issue:44671|issue:44991", + "issue:39404|issue:44671", + "issue:44991|issue:45081", + "issue:43856|issue:44756", + "issue:44164|issue:44479", + "issue:44737|issue:45859" + ], + "prepared_review_unit_hash": "d3a75afa49518dedee36151c0071f45be91042818ecd3c7fa212be4f5f69b751", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37731, + "estimated_input_tokens": 9433, + "estimated_eval_tokens": 19122 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a heterogeneous grab bag of Transformers 5.x regressions and model-specific loading bugs. Only one soft pair looks like the same underlying defect: resize_token_embeddings not updating tied output/decoder embeddings.", + "confidence": 0.81, + "canonical_issue_reason": "Issue 43611 is the broadest representative of the loading-regression subgroup (v5 breaks model loading via `base_model_prefix`), while most other items are narrower variants or unrelated failures.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43611 is the best single anchor for triage because it is concise, broadly scoped, and most representative of the model-loading breakages in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Both involve custom-model save/load, but one is about file permissions during save_pretrained and the other is about loading the wrong custom module." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization is unrelated to Trainer resume_from_checkpoint/max_steps calculation." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Notebook-defined custom model init and extra_state save/from_pretrained handling are different code paths and symptoms." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same bug: resize_token_embeddings is not propagating to tied output/decoder embeddings; 45335 is the t5gemma-specific instance of 45292." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Both are dtype/device related, but one is an autocast mismatch in Phi MoE and the other is a meta-device loading failure for different models." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch version incompatibility and an autocast dtype mismatch are distinct failures." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer vocab metadata mismatch is not the same as a model-specific loading failure for vesteinn/ScandiBERT." + }, + { + "left": "issue:43257", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3 MoE conversion under accelerate+deepspeed and NemotronH checkpoint loading are unrelated model-loading bugs." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Missing Float8 storage support and mmap-related OOM behavior are different problems." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper pipeline return_language regression is unrelated to a general tokenization-version change report." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 PLE device mismatch are different code-path issues." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "Torch 2.10 incompatibility is not the same as a device_map='auto' forward mismatch in Gemma4 PLE." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Both are about v5 breaking custom/model loading, but one is base_model_prefix handling and the other is notebook-defined custom model initialization." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "A version incompatibility report and a Mistral-3 dtype mismatch are not the same bug." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and a template-compilation TypeError are separate failures." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper return_language and MXFP4 quantization fallback are unrelated regressions." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "An isolated reproduction of another issue is not the same as the Qwen3.5 PagedAttentionCache crash." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export with torch_compilable_check and Perceiver non-default-resolution classification are different bugs." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Qwen video-input regression and a mask-shape IndexError are not the same underlying defect." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "Incorrect Siglip outputs and ModelOutput key assignment are distinct behaviors." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch version incompatibility is unrelated to Qwen3.5 embedding unpacking with TRL SFT trainer." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup failure and template-node compilation failure are different issues." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "These are both documentation issues, but they cover different removed/changed pipeline APIs." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3VL MoE conversion/loading and Qwen3-VL weight-shape mismatch are related only broadly, not the same concrete bug." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "extra_state save/load handling and text_config unwrapping for composite configs are separate loader/saver problems." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Zero3 init index error and compiled flex_attention torch>=2.9 failure are unrelated." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5Tokenizer add_special_tokens conflict and gguf_file.from_pretrained breakage are different tokenizer/load paths." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "CamemBERT masked LM prediction errors and est-roberta tokenizer loading failure are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline regression and CamemBERT prediction corruption are unrelated model families and bugs." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer loading failure for est-roberta and a Mistral regex patch crash are separate tokenizer code paths." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3 MoE training memory inefficiency and Strix Halo mmap OOM avoidance are not the same bug." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "extra_state save/load handling is unrelated to the Qwen video-input regression." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "XLNet CPU arange placement and Qwen3_5MoeTextRotaryEmbedding CPU offload compatibility are different device-placement issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The grouping is grounded in the packet and stays mostly conservative. The canonical anchor choice (43611) is defensible as a broad v5 loading-regression representative, and the only accepted soft pair (45292/45335) is strongly supported by the titles. The rejected pairs generally distinguish different code paths or symptom classes without overstating similarity. " + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 8, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:32090", + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42175", + "issue:42915", + "issue:43089", + "issue:43262", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43421", + "issue:43452", + "issue:43482", + "issue:43540", + "issue:43541", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43883", + "issue:44206", + "issue:44220", + "issue:44451", + "issue:44464", + "issue:44488", + "issue:44530", + "issue:44534", + "issue:44661", + "issue:44683", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45507", + "issue:45584", + "issue:45987", + "issue:46178" + ], + "soft_pairs": [ + "issue:43582|issue:43632", + "issue:32090|issue:39404", + "issue:44488|issue:45081", + "issue:44661|issue:44683", + "issue:44488|issue:44987", + "issue:43611|issue:43644", + "issue:40990|issue:43421", + "issue:43716|issue:45084", + "issue:44206|issue:44220", + "issue:36032|issue:42915", + "issue:43883|issue:44488", + "issue:43643|issue:44530", + "issue:34567|issue:40444", + "issue:42175|issue:46178", + "issue:44811|issue:45584", + "issue:43541|issue:45084", + "issue:44451|issue:44987", + "issue:42915|issue:44530", + "issue:43644|issue:43646", + "issue:44464|issue:45987", + "issue:43452|issue:43482", + "issue:36032|issue:43645", + "issue:43317|issue:43482", + "issue:44849|issue:45325", + "issue:44987|issue:44991", + "issue:45084|issue:45507", + "issue:43883|issue:44991", + "issue:43262|issue:43377", + "issue:43645|issue:44534", + "issue:43089|issue:45987", + "issue:43646|issue:44534", + "issue:43322|issue:43540" + ], + "prepared_review_unit_hash": "2cadc8ed113618a03c6c14be2913949a72dc3ac50f1dfe587e37dcc78c604d2f", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36658, + "estimated_input_tokens": 9165, + "estimated_eval_tokens": 18586 + }, + "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, with only a couple of plausible duplicate subgroups: GGUF loading regressions and an audio feature-extraction crash. I\u2019d use the broad GGUF loading report as the best representative issue.", + "confidence": 0.67, + "canonical_issue_reason": "issue:43452 is the broadest, most actionable report here: it covers GGUF loading failing through both AutoTokenizer and AutoModelForCausalLM, making it a better umbrella than the narrower follow-up-style report.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43452 is the strongest global representative because it describes a concrete, cross-API loader regression and plausibly subsumes the more specific Qwen2.5-GGUF failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Different failures: Apple Silicon allocator warmup TypeError vs. `_is_hf_initialized` flag regression." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast error and Whisper `return_language` pipeline bug are unrelated." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-loading problems, but they affect different models and code paths." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "`add-new-model-like`/tokenizer mapping and compiled flex_attention on torch>=2.9 are unrelated." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Separate model-loading regressions with different symptoms and likely different root causes." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are v5 regressions, but one is about `base_model_prefix` loading and the other about buffer initialization." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "Perplexity/tokenization behavior is not the same as runtime post-processor syncing." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Mistral image dtype mismatch and template-node compilation error are unrelated." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": true, + "reason": "Both point to the same audio feature-extraction path; the extractor crash and `_torch_extract_fbank_features()` issue look like the same underlying bug." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "Tokenizer method-name conflict and Qwen3Moe FP8 config failure are unrelated." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Molmo tied-weights attribute error is unrelated to BERT model-loading failures." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "AutoConfig field loss with `trust_remote_code` is unrelated to PagedAttentionCache linear_attention handling." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "Trainer token-count tracking and multimodal IterableDataset loading are different bugs." + }, + { + "left": "issue:42175", + "right": "issue:46178", + "accept": false, + "reason": "Package-extras missing TensorFlow is not the same as an unsupported backend mapping entry for `tensorflow_text`." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both Whisper-related, but one is decode flag handling and the other is empty-transcription generation after alignment." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Grouped_mm/dynamo tracing failure and non-template-node compilation error do not share a code path." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Different model-loading regressions with no clear shared failure mechanism." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Non-persistent buffer junk and custom model initialization breakage are separate regression types." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked generation inconsistency and cache sequence-length return type are related to generation, but not the same bug." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": true, + "reason": "43482 looks like a specific instance of the broader GGUF loading breakage described by 43452." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer special-token conflict and Jupyter custom-model initialization are unrelated." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "Device-map/offload dequantization failure is not the same as GGUF loading in transformers v5." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen variants and different bugs: hidden-states output vs. RoPE position scaling." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are v5 loading regressions, but they affect different artifacts and likely different loaders." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Template compilation error and GraniteMoEHybrid invalid method call are unrelated." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Molmo attribute error is unrelated to tokenizer loading for est-roberta." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Sampling-rate defaulting and MIMI padding-mask behavior are different audio issues." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Notebook custom-model initialization and junk-filled non-persistent buffers are distinct regressions." + }, + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Generation performance overhead and cache length return type are not the same underlying defect." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Custom model initialization and buffer initialization junk are separate bugs." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Llava Next load-time segfault and Qwen3OmniMoe video-processing ValueError are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about duplicates. The GGUF umbrella choice (issue:43452 over issue:43482) is reasonable, and the one accepted soft pair (issue:44206 / issue:44220) is plausibly the same audio-path bug. Rejections are generally cautious 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": 17, + "review_unit_index": 9, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:33666", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:41553", + "issue:41720", + "issue:42915", + "issue:43296", + "issue:43377", + "issue:43404", + "issue:43421", + "issue:43425", + "issue:43531", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43742", + "issue:43792", + "issue:43819", + "issue:43994", + "issue:44062", + "issue:44079", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45084", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45663" + ], + "soft_pairs": [ + "issue:43404|issue:44466", + "issue:44403|issue:45081", + "issue:43994|issue:44610", + "issue:42915|issue:45084", + "issue:43377|issue:43819", + "issue:44265|issue:44610", + "issue:44451|issue:45081", + "issue:44987|issue:45081", + "issue:44220|issue:44683", + "issue:42915|issue:44291", + "issue:43632|issue:44756", + "issue:45216|issue:45357", + "issue:36032|issue:43742", + "issue:44062|issue:44987", + "issue:42915|issue:44589", + "issue:44479|issue:45325", + "issue:43296|issue:43531", + "issue:43404|issue:45127", + "issue:43531|issue:44877", + "issue:36032|issue:41720", + "issue:40444|issue:45325", + "issue:43611|issue:44534", + "issue:43425|issue:44403", + "issue:43632|issue:44534", + "issue:41553|issue:43582", + "issue:40990|issue:44811", + "issue:43792|issue:44589", + "issue:44079|issue:45663", + "issue:44683|issue:45020", + "issue:33666|issue:40444", + "issue:44403|issue:44991", + "issue:43421|issue:44568" + ], + "prepared_review_unit_hash": "4aa1a8a4d0f46af3b4fe3ef17308df78d019b800fc7c3ffbca62cb972385f004", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36728, + "estimated_input_tokens": 9182, + "estimated_eval_tokens": 18620 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a loose cluster of unrelated transformer issues; only the Qwen3.5 save_pretrained regressions look like true duplicates, while most other pairs are merely same-subsystem or same-model-family similarities.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45216 is the broader original report of the Qwen3.5 save_pretrained regression; issue:45357 reads like a narrower follow-up about the same incorrect checkpoint serialization, specifically the visual encoder keys.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45216 is the best representative issue here because it describes a clear regression with broad user impact and the other close match (45357) is a narrower manifestation of the same save_pretrained problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Both mention lm_head/tied weights, but one is a missing tie in Mistral3ForConditionalGeneration and the other is device-dependent save_pretrained serialization; different bugs." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Unrelated: noisy transformer loading behavior vs a tokenizer crash in Mistral regex patching." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Different vision bugs and code paths: SigLIP2 bad outputs vs OmDet-Turbo processor/model input-size mismatch." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "Unrelated FP8 config failure vs a template-compilation TypeError." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Different models and failure modes: batched-vs-single padding-mask mismatch vs DAC from_latents / STE mismatch." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "Torch export failure with torch_compilable_check is not the same bug as the OmDet-Turbo image-size mismatch." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "ScandiBERT loading failure is unrelated to the Mistral tokenizer regex crash." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Different loading failures on different models; no evidence of the same underlying defect." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio feature extraction bug vs compiled flex_attention failure on newer torch." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "FineGrainedFP8Config failure is distinct from the init_empty_weights/_is_hf_initialized argument error." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "Different regressions: v5 _is_hf_initialized behavior vs mmap OOM on Strix Halo." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Both describe the same Qwen3.5 save_pretrained regression; 45357 is a narrower reproduction focused on incorrect visual encoder keys." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "T5 tokenizer special-token method conflict is unrelated to the MobileLLM key error." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken special-argument error is not the same as loading physical-intelligence/fast." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Different FP8-related failures with different symptoms and code paths." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Related Qwen-VL family, but 44479 is a broad video-input regression across multiple models while 45325 is a specific still-image temporal-position scaling bug; not safe to merge as duplicates." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "PaddleOCR-VL loading failure is unrelated to the Qwen3-MoE sliding_window issue." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both involve tied weights, but one is missing tying in Mistral3 while the other is LoRA-merge corruption with extended vocabulary." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3-MoE sliding_window bug and granite_speech strict-config loading are unrelated." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "T5 tokenizer load error vs Qwen3 auto-device-mapping CUDA assert; different failures." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen2.5-VL-related, but multi-image iterable-dataset finetuning failure is not the same as temporal position_id scaling." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "base_model_prefix loading break and non-persistent buffer junk are separate v5 regressions." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch version incompatibility is unrelated to loading noise." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Different v5 breakages: _is_hf_initialized flag handling vs non-persistent buffer serialization." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad AutoTokenizer error message and AppleSilicon caching_allocator_warmup TypeError are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity/evaluation issue vs Whisper batch_decode skip_special_tokens bug." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper runtime failure vs missing Float8 storage object are unrelated." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment bug and Gemma-4 FSDP2 shared_kv_states KeyError are different code paths." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Compiled flex_attention torch>=2.9 failure is distinct from remote_code loading regressions." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Different Qwen-VL training issues: multi-GPU training vs iterable-dataset multi-image finetuning." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Transformer loading noise is not the same as the est-roberta tokenizer loading failure." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime special-token post-processing sync and a specific add_special_tokens BOS/EOS regression are related in area but not clearly the same defect." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted near-duplicate pair, 45216/45357, is well supported by their titles and timelines as the same Qwen3.5 save_pretrained regression with a narrower follow-up. The remaining pairwise rejections are appropriately cautious and do not overstate duplicate evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 10, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:35707", + "issue:40444", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43381", + "issue:43454", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43701", + "issue:43749", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44462", + "issue:44466", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44829", + "issue:44898", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45542", + "issue:45800", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45923" + ], + "soft_pairs": [ + "issue:43381|issue:45923", + "issue:35707|issue:40444", + "issue:44462|issue:45812", + "issue:44291|issue:44589", + "issue:43482|issue:45538", + "issue:42175|issue:45542", + "issue:44829|issue:45479", + "issue:43482|issue:43525", + "issue:43317|issue:43856", + "issue:44079|issue:44534", + "issue:43818|issue:45357", + "issue:43531|issue:44291", + "issue:41720|issue:45081", + "issue:43377|issue:45491", + "issue:44265|issue:45072", + "issue:40444|issue:45381", + "issue:43644|issue:44079", + "issue:43749|issue:44079", + "issue:42175|issue:43296", + "issue:43632|issue:45072", + "issue:45325|issue:45381", + "issue:43611|issue:43646", + "issue:43792|issue:43957", + "issue:45325|issue:45834", + "issue:43994|issue:44898", + "issue:43454|issue:44466", + "issue:42915|issue:43482", + "issue:43296|issue:45800", + "issue:44479|issue:45823", + "issue:43701|issue:45325" + ], + "prepared_review_unit_hash": "f61456958e42e5421c3b6147c99a38cf4b2d6ab0b1d758593201d1e0c3c4244d", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36891, + "estimated_input_tokens": 9223, + "estimated_eval_tokens": 18702 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of distinct Transformers regressions, not one cohesive duplicate set. The only strong soft-duplicate here is the Qwen2.5-VL position-id regression pair.", + "confidence": 0.67, + "canonical_issue_reason": "issue:45325 is the clearest duplicate anchor: it states the concrete Qwen2.5-VL rope/position-id bug, and issue:45381 appears to be the same regression described from the video-input side.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43611 is the best overall representative of the set because it captures a broad, high-signal Transformers v5 model-loading regression with clear generic scope.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "Different gradient-checkpointing problems: eval-mode misuse vs missing supports_gradient_checkpointing on a trust_remote_code model." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Unrelated failures: generation with inputs_embeds/past_key_values vs Qwen2.5-VL multi-image IterableDataset finetuning." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Both involve tokenization, but one is repo tokenizer loading and the other is wrong Granite token IDs; different bugs." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Different load-path regressions: unexpected _is_hf_initialized arg vs Float8 storage lookup failure." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Different tokenization/model issues: GGUF loading failure vs CLIPTokenizer max_length behavior." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Both mention TensorFlow backend packaging, but the concrete failures differ enough that they do not look like the same bug." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Different training-path bugs: flash_attention_3 degeneracy vs single-label classification with num_labels=1 zero loss." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Different failure modes: Qwen2.5-GGUF loading vs missing pad_token_id on Llama4Config." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Different problems: device_map/offload loading failure vs Qwen3 MoE training memory inefficiency." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "Different internals: ModelOutput key assignment bug vs non-persistent buffer junk fill." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Different model families and symptoms: Video-LLaVA weight/tower issue vs Qwen3.5 visual-key save_pretrained regression." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Sliding-window Qwen3-MoE bug is unrelated to init_empty_weights/_is_hf_initialized loading regression." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Different code paths entirely: auto device-mapping CUDA assert vs tokenizer regex patch crash." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Both are batching/masking issues, but the models and concrete failure modes differ." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export/torch_compilable_check export failure is unrelated to bfloat16 dtype mismatches." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL, but one is multi-image finetuning input handling and the other is video position-id regression." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Different regressions: non-persistent buffer initialization vs ModelOutput key bookkeeping." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading and ModelOutput key assignment are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Packaging/backend install issue vs vLLM model-load failure; not the same bug." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "_is_hf_initialized breakage and bf16 dtype mismatches are separate Transformer v5 regressions." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Same Qwen2.5-VL positional-id/rope-index regression in v5.3.0; one report is image/still-image side, the other is video-input side." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Related v5 model-init/loading regressions, but the titles point to different concrete failure modes." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Different loading/runtime issues: Whisper model failure vs meta-device loading breakage." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Qwen2.5-VL position-id regression is unrelated to Kosmos2.5 OCR index errors." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model failures: SigLIP2 bad outputs vs Perceiver non-default-resolution interpolation failure." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Both mention lm_head weights, but one is untied-weight generation and the other is device-dependent serialization." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different model/config paths: Qwen3MoE FP8 config vs Qwen2.5-GGUF loading." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "PaddleOCR-VL load failure and torch/transformers version incompatibility are not the same bug." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Different Qwen/Gemma regressions: video-input regression vs Gemma4 device-map mismatch." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint-resume key mismatch is unrelated to Qwen2.5-VL rope-index position IDs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays appropriately conservative overall. The only accepted soft-duplicate pair (issue:45325 / issue:45381) is supported by the titles, and the rejected pairs are described cautiously rather than overstated." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 11, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44466", + "issue:44488", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45357", + "issue:45584", + "issue:45701", + "issue:45759", + "issue:45823", + "issue:45920", + "issue:45923", + "issue:46050" + ], + "soft_pairs": [ + "issue:44291|issue:44877", + "issue:43881|issue:44488", + "issue:43606|issue:44265", + "issue:43873|issue:45701", + "issue:43262|issue:43746", + "issue:40990|issue:45584", + "issue:34634|issue:35707", + "issue:44568|issue:45920", + "issue:35707|issue:41720", + "issue:44186|issue:44265", + "issue:43606|issue:43819", + "issue:41720|issue:45823", + "issue:43454|issue:43550", + "issue:39692|issue:40444", + "issue:43454|issue:45072", + "issue:43377|issue:44610", + "issue:43572|issue:45759", + "issue:43257|issue:45923", + "issue:40990|issue:42947", + "issue:43377|issue:44898", + "issue:43317|issue:43873", + "issue:43526|issue:43883", + "issue:43322|issue:46050", + "issue:43881|issue:44991", + "issue:43454|issue:43720", + "issue:41720|issue:43322", + "issue:43296|issue:44991", + "issue:43749|issue:44534", + "issue:44466|issue:45357", + "issue:41720|issue:44977", + "issue:42617|issue:42915", + "issue:40990|issue:45701" + ], + "prepared_review_unit_hash": "cc5859e416b97e2dde55f46e1e569a2d8bc80dc38b708a0914e7f42d38b83f98", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37412, + "estimated_input_tokens": 9353, + "estimated_eval_tokens": 18962 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a collection of unrelated bug reports, with only a loose subtheme around loading/offloading/quantization and device-mapping regressions. None of the soft pairs look like safe duplicates or mergeable PR equivalents.", + "confidence": 0.95, + "canonical_issue_reason": "If forced to pick an anchor, issue 43873 is the broadest representative of the offloading/quantization subtheme, but the set is not a true duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue 43873 is the most central and broadly phrased report in the loading/offloading/quantization area, so it is the best representative issue for the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Different regressions: `_is_hf_initialized`/`init_empty_weights` type error vs strict config loading for `granite_speech`." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "One is a device mismatch on model load, the other is a current-version tokenizer/model loading failure; different code paths and models." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "Bark CPU offload device mismatch is unrelated to `torch.export` failures from `torch_compilable_check`." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Broad loading-related wording, but one is offloading/quantization and the other is tokenization behavior changing by version." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio processor sampling-rate default bug vs GraniteSpeech adapter checkpoint loading; different failure modes." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on one model is not the same as Whisper empty-transcription failure after token alignment." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "Voice preset support in BarkProcessor is unrelated to progressive generation with `inputs_embeds` and `past_key_values`." + }, + { + "left": "issue:44568", + "right": "issue:45920", + "accept": false, + "reason": "Tokenizer special-token regression is unrelated to wrong token IDs across multiple models." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Generic generation bug with cached inputs vs Qwen3 auto device-map CUDA assert; not the same concrete bug." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "NER/batched padding crash in LayoutLMv2Tokenizer is unrelated to export failures caused by `torch_compilable_check`." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Both are model math/runtime issues, but Bark CPU offload mismatch and DAC forward/from_latents mismatch are distinct." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Qwen3 auto device-mapping assert vs Gemma4 PLE device mismatch; similar theme, different model and root cause." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "AyaVision weight-tying bug is unrelated to Bamba torch.compile/SDPA failure." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "SigLIP2 documentation errors are unrelated to Qwen2.5-VL finetuning with iterable multi-image datasets." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Weight tying/garbage text generation is not the same as dtype mismatch in SwitchTransformers and TimmWrapperModel." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "MIMI batching/padding-mask issue vs OmDet-Turbo processor/model input-size mismatch; different components." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "Missing `pad_token_idx` in StableLmConfig is unrelated to failing to unwrap `text_config` for Qwen multimodal configs." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "Qwen3 MOE loading with accelerate+deepspeed is unrelated to a missing gradient-checkpointing capability flag." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity/evaluation anomaly is not the same bug as ineffective gradient checkpointing with PEFT LoRA." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Different model families and problems: MIMI padding-mask mismatch vs Perceiver non-default-resolution interpolation failure." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both mention offloading/loading, but one is dequantized model placement on gpu+cpu offload and the other is quantization/offloading behavior more generally." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "BEiT `reduce_labels` returning one label is unrelated to Molmo missing `all_tied_weights_keys`." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Llava Next load segfault and sinq save/load breakage affect different stages and mechanisms." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "PaddleOCR-VL load failure is not the same as tokenizer loading regression for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "AyaVision weight tying is unrelated to BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3 device-map CUDA assert and Llava Next load segfault are different code paths and failure signatures." + }, + { + "left": "issue:43296", + "right": "issue:44991", + "accept": false, + "reason": "PaddleOCR-VL vLLM loading failure is unrelated to tokenizer loading regression." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and v5 junk-filled non-persistent buffers are both loading-related but not the same bug." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Inconsistent lm_head serialization by device is different from incorrect visual encoder keys in `save_pretrained`." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3 auto device-mapping assert is unrelated to Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "`3d_parallel.py` runtime failure and Qwen3MoE FineGrainedFP8Config failure are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity regression and version-dependent tokenization changes are different symptoms and likely different subsystems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 12, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36010", + "issue:36331", + "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:44387", + "issue:44423", + "issue:44479", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44918", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45704", + "issue:45812", + "issue:45823" + ], + "soft_pairs": [ + "issue:44008|issue:45823", + "issue:44387|issue:45704", + "issue:36331|issue:42617", + "issue:43577|issue:43708", + "issue:44186|issue:45072", + "issue:44877|issue:44991", + "issue:36331|issue:42915", + "issue:43404|issue:45357", + "issue:43450|issue:45701", + "issue:43550|issue:43720", + "issue:43873|issue:44756", + "issue:44423|issue:45406", + "issue:41720|issue:42617", + "issue:44479|issue:44918", + "issue:43550|issue:43606", + "issue:44610|issue:44898", + "issue:45632|issue:45684", + "issue:44186|issue:44610", + "issue:42915|issue:43317", + "issue:43295|issue:45701", + "issue:43526|issue:43881", + "issue:41720|issue:45161", + "issue:36010|issue:42617", + "issue:44671|issue:44743", + "issue:45081|issue:45812", + "issue:43531|issue:43572", + "issue:43316|issue:44913", + "issue:43540|issue:43782", + "issue:42915|issue:43322" + ], + "prepared_review_unit_hash": "f3e7ffb107f64cf1cabc9454ecdb91148fdd8cb0fa86a0b92b58ca99f2a62563", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36546, + "estimated_input_tokens": 9137, + "estimated_eval_tokens": 18530 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All soft edges look like false positives: the paired issues touch different models, different APIs, or different failure modes, so none should be merged as duplicates. The best standalone issue candidate is the broad offloading/quantization report (43873), but it is not a duplicate of the others.", + "confidence": 0.92, + "canonical_issue_reason": "issue:43873 is the strongest canonical issue candidate because it is open, broadly framed, and centrally discussed around a recurring offloading/quantization failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best issue to keep as a tracking reference: it is broad enough to represent a real subsystem pain point and has substantial discussion/inbound attention.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Different models and different bugs: Gemma3n forward variable collision vs Gemma4 device-map forward mismatch." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Both mention memory, but one is int4 reserved-memory/OOM and the other is a T5 apex RMSNorm leak; distinct causes and fixes." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Trainer compute_loss kwarg breakage is unrelated to a failure to run 3d_parallel.py." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "BLIP2 dtype loading bug and Trainer checkpoint max_steps bug are separate code paths." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer padding/truncation crash is unrelated to model dtype mismatches in inference." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Config strictness blocking granite_speech loading is not the same bug as tokenizer loading failure for est-roberta." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "CustomTrainer API mismatch is unrelated to Qwen3MoE FP8 configuration failure." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Mistral3 weight tying bug and Qwen3.5 save_pretrained visual key regression affect different models and save/load paths." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Batched video processor shape bug is unrelated to the tokenization regression." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "torch.compile+SDPA failure is a different problem from unpacking packed weights during accelerate loading." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Both involve offloading/memory, but one is quantization/offload behavior and the other is a Strix Halo mmap OOM workaround." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Serve crashes stem from different processor assumptions: string .to() misuse vs missing _tokenizer attribute." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Qwen3 auto device mapping cuda assert is unrelated to the 3d_parallel.py runtime failure." + }, + { + "left": "issue:44479", + "right": "issue:44918", + "accept": false, + "reason": "Video-input regression in Qwen VL models is not the same as Qwen3.5 embedding unpacking during TRL SFT training." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Different failures: SDPA/torch.compile issue vs CPU offload device mismatch for bark-small." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Processor emitting wrong image size is not the same as a model refusing non-default resolution even with interpolation." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collision for trust_remote_code is unrelated to read-only permission propagation in save_pretrained." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer crash and image processor resolution bug are separate subsystems and symptoms." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Qwen3MoE FP8 config failure and dequantized model offload load failure are different quantization/offload bugs." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor.tokenizer regression is not the same as a generic tokenization-change report." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to glm-4v-9b loading failure." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Auto device-map CUDA assert is not the same as GPT-OSS MoE tensor-parallel support failure." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "GenerationMixin import error has no overlap with 3d_parallel.py execution failure." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT masked-LM prediction regression and Qwen3.5 cache-state reset bug are different model behaviors." + }, + { + "left": "issue:45081", + "right": "issue:45812", + "accept": false, + "reason": "Mistral regex patch crash and Granite token-ID regression are both tokenizer-related but clearly different bugs and model families." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue and StableLmConfig pad_token_idx regression are unrelated." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy is not the same as GPTNeoX rotary_pct reload behavior." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Qwen3OmniMoe video-processing ValueError and Qwen3VL weight_only load error are distinct multimodal load paths." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3MoE FP8 config failure is unrelated to Llava Next segmentation fault on load." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 13, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:29159", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36032", + "issue:36331", + "issue:41720", + "issue:42490", + "issue:43329", + "issue:43388", + "issue:43421", + "issue:43531", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43873", + "issue:43927", + "issue:44077", + "issue:44186", + "issue:44361", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44625", + "issue:44898", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45059", + "issue:45072", + "issue:45381", + "issue:45406", + "issue:45464", + "issue:45479", + "issue:45588", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45859" + ], + "soft_pairs": [ + "issue:44186|issue:44361", + "issue:43421|issue:45812", + "issue:36331|issue:41720", + "issue:29159|issue:36032", + "issue:43388|issue:45059", + "issue:45406|issue:45464", + "issue:30064|issue:34634", + "issue:33357|issue:35707", + "issue:43606|issue:43720", + "issue:43550|issue:43653", + "issue:41720|issue:43746", + "issue:43873|issue:45704", + "issue:43421|issue:45701", + "issue:43531|issue:45759", + "issue:43606|issue:43653", + "issue:34634|issue:41720", + "issue:42490|issue:43329", + "issue:34567|issue:35707", + "issue:44492|issue:44509", + "issue:44514|issue:44545", + "issue:44625|issue:45479", + "issue:44898|issue:45072", + "issue:43421|issue:43927", + "issue:44451|issue:44991", + "issue:45759|issue:45859", + "issue:44077|issue:44625", + "issue:43531|issue:44589", + "issue:43782|issue:45381", + "issue:44987|issue:45042", + "issue:45588|issue:45759" + ], + "prepared_review_unit_hash": "5649b820861e38e3ff7500da17e03dd063f9afa8e67e398e67e04be538c9f7fa", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36953, + "estimated_input_tokens": 9239, + "estimated_eval_tokens": 18734 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a loose tokenizer/model-loading/multimodal-bug bucket rather than a duplicate cluster. None of the soft pairs look like the same underlying defect, so all candidate edges should stay split. If one issue must represent the group, the Granite AutoTokenizer regression (45812) is the strongest concrete tokenizer bug, but it is not a duplicate of the others.", + "confidence": 0.72, + "canonical_issue_reason": "Issue 45812 is the clearest, most concrete tokenizer regression in the set: it describes a silent v4\u2192v5 token-ID break for a broad model family and is more actionable than the vaguer tokenizer reports.", + "canonical_pr_reason": null, + "best_issue_reason": "45812 is the best single issue to anchor this cluster because it is specific, reproducible, and has a clear regression boundary; several others are adjacent tokenizer problems, but not the same bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both are tokenizer failures, but they affect different tokenizer classes and different failure modes (LayoutLMv2 batching/NER vs MLuke task AttributeError)." + }, + { + "left": "issue:43421", + "right": "issue:45812", + "accept": false, + "reason": "Runtime special-token/post-processor refresh is a different bug from Granite models getting wrong token IDs after the v4\u2192v5 tokenizer change." + }, + { + "left": "issue:36331", + "right": "issue:41720", + "accept": false, + "reason": "Trainer API signature mismatch and Qwen3 device-mapping cudaErrorAssert are unrelated code paths and symptoms." + }, + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Generic slow-vs-fast tokenizer inconsistency is not the same as the T5Tokenizer load-time add_special_tokens name conflict." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Different subsystems: metric gathering for tuple labels vs SAM3 text/bbox prompt behavior." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Processor missing _tokenizer in transformers serve is distinct from Qwen3.5 streaming chat/completions failure." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Void segmentation-map processing and Bark voice_preset handling are separate processor bugs." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "MacOS CLIP bus error and inputs_embeds/past_key_values generation issues do not share the same underlying defect." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both mention loading/offload, but one is bark-small CPU offload device mismatch and the other is BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "SDPA/torch.compile failure in Bamba-9B-v2 is unrelated to BigBirdTokenizer mask-token registration and empty decode output." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3 auto device-map assertion and GraniteSpeech PEFT local-checkpoint loading are different model-loading problems." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior and T5 apex.FusedRMSNorm memory leak are different bugs with no shared code path." + }, + { + "left": "issue:43421", + "right": "issue:45701", + "accept": false, + "reason": "A tokenizer backend not refreshing post-processors at runtime is not the same as a broad report that versions change tokenization." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Sliding-window handling in Qwen3-MoE and from_config not unwrapping text_config for Qwen multimodal configs are separate model/config issues." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "These are different CI failures in different models: device mismatch during CPU offload vs mask-token registration in a tokenizer." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping are unrelated bugs in different components." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "A trailing-slash local path import warning and undefined video-branch helpers in multimodal token counting are not the same issue." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token-count tracking and progressive generation with embeds/past KV are unrelated." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache-strategy docs is not the same as stale pipeline-task documentation for removed v5 tasks." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": false, + "reason": "These are effectively the same title, but the pair was not presented with evidence of one underlying report; treat as separate records unless explicit duplication is confirmed." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Propagating num_labels in Qwen3.5 configs is different from the single-label-classification zero-loss bug across sequence-classification models." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Perceiver non-default resolution interpolation and bfloat16 dtype mismatches in unrelated models are not the same defect." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Runtime special-token/post-processor updates differ from DiaConfig losing custom token IDs on save/load." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Loading a specific model failing and tokenizer loading for EMBEDDIA/est-roberta are different model-specific loading bugs." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Qwen multimodal config unwrapping and Qwen3.5-MoE rotary embedding CPU-offload compatibility are separate problems." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "A patchtsmixer post_init policy issue is unrelated to Qwen3.5 num_labels propagation." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Sliding-window handling and missing Float8 storage registration are unrelated." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "weight_only loading for Qwen3VL and video vision_position_ids in Qwen2.5-VL are different multimodal bugs." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Loading a specific model and PIL backend image processors requiring torchvision are different regressions." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention sink-less s_aux handling and Qwen multimodal config unwrapping do not share the same code path." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 14, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:39692", + "issue:40444", + "issue:41720", + "issue:42886", + "issue:43295", + "issue:43381", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43596", + "issue:43618", + "issue:43645", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43819", + "issue:43844", + "issue:43873", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44368", + "issue:44451", + "issue:44610", + "issue:44661", + "issue:44671", + "issue:44843", + "issue:44918", + "issue:44936", + "issue:45072", + "issue:45081", + "issue:45183", + "issue:45356", + "issue:45357", + "issue:45479", + "issue:45663", + "issue:45704", + "issue:45834", + "issue:45859", + "issue:45907", + "issue:45923", + "issue:46050" + ], + "soft_pairs": [ + "issue:44062|issue:44451", + "issue:44368|issue:45357", + "issue:39692|issue:44610", + "issue:43746|issue:46050", + "issue:44918|issue:44936", + "issue:45704|issue:45923", + "issue:43421|issue:43653", + "issue:45663|issue:45923", + "issue:42886|issue:44843", + "issue:43381|issue:43844", + "issue:45704|issue:45859", + "issue:43531|issue:43645", + "issue:41720|issue:45704", + "issue:43421|issue:44186", + "issue:43994|issue:44671", + "issue:43746|issue:43873", + "issue:45081|issue:45356", + "issue:39692|issue:43873", + "issue:43596|issue:44220", + "issue:40444|issue:43746", + "issue:45356|issue:45907", + "issue:30064|issue:36010", + "issue:43720|issue:43819", + "issue:44220|issue:44661", + "issue:43653|issue:43720", + "issue:43540|issue:44008", + "issue:43618|issue:43697", + "issue:43550|issue:45072", + "issue:45183|issue:45479", + "issue:43295|issue:45834", + "issue:43550|issue:43819" + ], + "prepared_review_unit_hash": "74e715da9d84727c01f78c894715af87e7b1123bacae28f47d30a568566bc176", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37462, + "estimated_input_tokens": 9366, + "estimated_eval_tokens": 18988 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a grab bag of unrelated issues across tokenizers, vision-language models, training/inference, and offline loading. None of the soft-edge pairs look like the same underlying bug or change, so I reject all links.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the set is not a duplicate cluster, but a collection of unrelated issues spanning different models and code paths.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44843 is the strongest standalone report: it has a clear failure mode, a specific code path (_patch_mistral_regex), and an actionable offline-mode regression.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "Both involve tokenizers, but one is an AddedToken constructor error and the other is a model loading failure for ScandiBERT; different bugs." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen3.5-related problems: one is a training warning about tie_word_embeddings, the other is a save_pretrained visual-encoder key regression." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Both mention processor/model mismatches, but they affect different models and different failure modes." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "Different persistence/loading bugs: PEFT local checkpoint loading vs quantized model save_pretrained/from_pretrained behavior." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Both mention trainers, but one is about unpacking Qwen3.5 input embeddings and the other about evaluate() after train()." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "Completely different issues: a T5 RMSNorm memory leak versus a missing gradient-checkpointing flag in a trust_remote_code model." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Both touch tokenizer special tokens, but one is runtime post-processor syncing and the other is BigBirdTokenizer mask-token registration." + }, + { + "left": "issue:45663", + "right": "issue:45923", + "accept": false, + "reason": "Different model/training bugs: Gemma-4 FSDP2 shared_kv_states handling vs gradient-checkpointing support flag." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": false, + "reason": "Both involve HF_HUB_OFFLINE, but 42886 is cache loading and 44843 is an unconditional model_info() call in mistral regex patching." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Unrelated training behavior issues: gradient checkpointing in eval mode vs gradient growth under ZeRO-3." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "Different models and code paths: T5 norm memory leak vs Qwen3.5-MoE rotary embedding CPU offload incompatibility." + }, + { + "left": "issue:43531", + "right": "issue:43645", + "accept": false, + "reason": "No shared bug: Qwen3-MoE sliding_window behavior versus Jupyter custom model initialization regression." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "Different subsystems and failure modes: Qwen3 auto device mapping CUDA assert versus T5 norm memory leak." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Both are tokenizer-related, but one is special-token post-processor updates and the other is LayoutLMv2 NER/padding crashes." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different model regressions: SigLIP2 pipeline/AutoModel output issues versus CamemBERT masked-LM prediction changes." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Both mention loading/offloading, but one is PEFT checkpoint loading and the other is quantization/offloading behavior." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both mention fix_mistral_regex, but one is a crash in _patch_mistral_regex and the other is a Kimi-K2.5 codec/warning regression." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "Quantization is only a broad theme here; SigLIP2 docs errors and offloading-with-quantization are different issues." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Different code paths entirely: DeepSpeed zero3/BertModel versus audio feature extraction." + }, + { + "left": "issue:40444", + "right": "issue:43746", + "accept": false, + "reason": "Different domains and bugs: Qwen2.5-VL multi-image IterableDataset training versus GraniteSpeech checkpoint loading." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "No relation: tokenizer codec/fix_mistral_regex regression versus list_repo_files missing revision forwarding." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Completely unrelated: image processor segmentation-map handling versus an ImportError for GenerationMixin." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "Different quantization-related model bugs with distinct mechanisms: packed-weight loading versus missing STE in DAC.from_latents." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio feature extraction and tokenizer-mapping model registration are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "Tokenizer special-token decoding bug versus BitNet packed-weight loading bug; no shared underlying defect." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Both are multimodal, but one is Qwen3OmniMoe video input handling and the other is Gemma 3n variable-name collision with audio masks." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "Both are output regressions in v5, but they affect different models and different tensors/paths." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Different inference/compiler issues in different models: SDPA/torch.compile failure versus bfloat16 dtype mismatches." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "One is validation/error-message logic for input_ids vs inputs_embeds; the other is a zero-loss classification bug." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Both involve processor/tokenizer-style workflows, but the first is a regression around processor.tokenizer/images and the second is a long-OCR index error." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Completely different bugs: Bamba torch.compile/SDPA versus DAC latent reconstruction with missing STE." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 15, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43322", + "issue:43335", + "issue:43388", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43577", + "issue:43606", + "issue:43643", + "issue:43644", + "issue:43653", + "issue:43716", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43818", + "issue:43819", + "issue:43950", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44442", + "issue:44448", + "issue:44479", + "issue:44488", + "issue:44671", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45381", + "issue:45584", + "issue:45685", + "issue:45834", + "issue:45987", + "issue:46032" + ], + "soft_pairs": [ + "issue:43388|issue:46032", + "issue:44062|issue:44488", + "issue:43716|issue:44008", + "issue:43335|issue:43756", + "issue:44186|issue:44898", + "issue:43322|issue:45685", + "issue:43653|issue:43819", + "issue:43531|issue:43643", + "issue:43950|issue:45987", + "issue:43643|issue:44877", + "issue:44843|issue:45081", + "issue:30064|issue:36331", + "issue:43540|issue:45381", + "issue:44448|issue:44671", + "issue:43421|issue:44811", + "issue:41720|issue:44991", + "issue:43577|issue:43818", + "issue:39692|issue:45834", + "issue:43782|issue:44977", + "issue:44821|issue:45084", + "issue:45584|issue:45834", + "issue:44186|issue:44442", + "issue:43644|issue:43749", + "issue:43606|issue:45072", + "issue:44479|issue:45834", + "issue:34634|issue:36010" + ], + "prepared_review_unit_hash": "c7bdcca96b9abf9391bef14ad15a75609003f3565b7355facf4b62ee253e5aad", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35744, + "estimated_input_tokens": 8936, + "estimated_eval_tokens": 18128 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous issue reports with no true duplicate group. The only borderline pair is the two Mistral regex/tokenizer-loading reports, but they describe different failure modes and should stay separate.", + "confidence": 0.94, + "canonical_issue_reason": "issue:44843 is the closest thing to a cluster anchor: it is the broadest and most central tokenizer-loading regression here, and it has the only plausible near-match with another item in the set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44843 is the best representative issue overall because it targets a concrete, recurring loading-path regression and is broader than the more specific follow-up crash in issue:45081.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Different subsystems: evaluation gathering vs Mamba2 cache behavior." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "One is an AddedToken API error; the other is a tokenizer loading failure for a specific model." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Different code paths and failures: image preprocessor dtype mismatch vs Gemma3n variable collision." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "Unrelated model/config bug vs RoPE-layer counting issue." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer crash for NER/padding is unrelated to Perceiver image-size handling." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Segfault in Llava loading is unrelated to an MPS histogram/int dtype error." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Tokenizer special-token registration bug is unrelated to DAC latent/STE mismatch." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to AutoConfig trust_remote_code field loss." + }, + { + "left": "issue:43950", + "right": "issue:45987", + "accept": false, + "reason": "Non-persistent buffer corruption is unrelated to StaticCache return-type issues." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-loading issues, but they affect different mechanisms and models." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Same broad file/function area, but different bugs: offline model_info call vs missing backend_tokenizer attribute." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Different training/image-processing failures with no shared code path." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both mention Qwen video, but the concrete failures and likely fixes are different models/code paths." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Different model regressions; same version does not make them duplicates." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Runtime special-token/post-processor update bug is not the same as batch_decode skip_special_tokens behavior." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "Unrelated issues: device-mapping crash vs tokenizer loading failure for a different model." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Blip2 dtype initialization and Video-LLaVA weight/attention structure are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "Doc-example/model mismatch is unrelated to long OCR index errors in Kosmos2.5." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Different Qwen generation regressions with different triggers and code paths." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading is unrelated to a template compilation TypeError." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Whisper empty-transcription handling is unrelated to Kosmos2.5 long-input indexing." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer crash in LayoutLMv2 is unrelated to FastSpeech2Conformer tokenizer loading." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Both involve loading/state, but not the same bug or fix." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Device offload mismatch and bf16 dtype mismatch are different issues." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Video regression in Qwen models is unrelated to Kosmos2.5 OCR indexing." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "Bark voice_preset behavior and GenerationMixin import failure are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 16, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36331", + "issue:42175", + "issue:43262", + "issue:43329", + "issue:43526", + "issue:43577", + "issue:43582", + "issue:43650", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:43994", + "issue:44060", + "issue:44291", + "issue:44368", + "issue:44485", + "issue:44509", + "issue:44554", + "issue:44561", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45406", + "issue:45538", + "issue:45600", + "issue:45701", + "issue:45987", + "issue:46192", + "issue:46209" + ], + "soft_pairs": [ + "issue:43577|issue:43582", + "issue:44561|issue:45020", + "issue:43825|issue:43994", + "issue:44805|issue:44936", + "issue:45987|issue:46192", + "issue:44671|issue:45701", + "issue:44610|issue:45072", + "issue:28218|issue:29159", + "issue:44485|issue:44509", + "issue:43720|issue:45072", + "issue:43782|issue:45325", + "issue:43906|issue:46209", + "issue:43825|issue:43827", + "issue:42175|issue:45081", + "issue:43582|issue:44857", + "issue:43749|issue:44291", + "issue:45081|issue:45406", + "issue:43262|issue:43819", + "issue:34634|issue:36331", + "issue:43526|issue:43697", + "issue:43650|issue:43906", + "issue:44060|issue:44368", + "issue:43653|issue:44912", + "issue:43742|issue:45020", + "issue:43526|issue:44554", + "issue:45084|issue:45538", + "issue:43329|issue:45600" + ], + "prepared_review_unit_hash": "84197617e6f879eb57bc66e0ab4a700f941c00a39ae78d1677c075a3cedccd60", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35476, + "estimated_input_tokens": 8869, + "estimated_eval_tokens": 17994 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated Transformers bug reports. The only strong duplicate-looking pair is the remote-code regression pair (44561 / 45020); the rest are too model-, feature-, or subsystem-specific to merge as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 45020 is the best umbrella issue: it is broad, current, and matches the clearest duplicate candidate about remote-code breakage in recent Transformers versions.", + "canonical_pr_reason": null, + "best_issue_reason": "45020 has the broadest scope and strongest overlap with a likely duplicate, while the other issues are more specific or clearly unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "Different failure modes: one is a model dtype-loading mismatch, the other is an Apple Silicon allocator warmup TypeError." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": true, + "reason": "Both describe the same underlying v5/regression class: recent Transformers changes breaking trust_remote_code/remote_code models." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "One is an error-message/docs issue around translation pipelines; the other is a model output quality problem with AutoModel/pipeline." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "An IndexError in masking/indexing is unrelated to trainer.evaluate() failing after train()." + }, + { + "left": "issue:45987", + "right": "issue:46192", + "accept": false, + "reason": "Different processor/cache contract bugs: seq-length return type vs return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "A CamemBERT masked-LM regression is not the same underlying bug as a broad tokenization-change report." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "These are unrelated model/processor bugs: input-size mismatch vs dtype mismatches in inference." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": false, + "reason": "Both are tokenizer-related, but the titles do not clearly point to the same concrete bug; one is about added-token spacing, the other a broad slow/fast inconsistency report." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE discussion is unrelated to docs still referencing removed pipeline tasks." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight loading in BitNet is a different code path from dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3VL weight_only load failures and Qwen2.5-VL RoPE index scaling are distinct issues." + }, + { + "left": "issue:43906", + "right": "issue:46209", + "accept": false, + "reason": "An isolated reproduction of an older issue is unrelated to FP8 inference failures on B200." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Both mention v5 pipeline removals, but one is runtime messaging and the other is documentation drift; not the same bug." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Backend dependency installation and Mistral regex patching are unrelated." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Apple Silicon allocator warmup and CUDA AMP loss crashes are different failures." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "FSDP CPU RAM-efficient loading and init_empty_weights argument errors are separate loading regressions." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Tokenizer regex patching crash and Gemma4Processor missing _tokenizer are different processor/tokenizer bugs." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio processor sampling-rate defaults and DAC from_latents/STE mismatch are unrelated." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "BarkProcessor voice_preset and CustomTrainer compute_loss signature changes do not share an underlying code path." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "BEiT reduce_labels behavior and RTDetrV2 output drift are different model/processor regressions." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "A vague request for data and an isolated reproduction of another issue are not duplicates." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both mention tied-weights warnings, but the model-specific warning behavior and the alleged cross-layer tying bug are not clearly the same issue." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "A tokenizer special-token registration bug is unrelated to MXFP4 quantization loading fallback." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "A key error while loading a specific model is not the same as a broad remote_code regression report." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative. The only duplicate-like pair it elevates is 44561/45020, which is supported by the titles, and the rest are appropriately treated as unrelated or too specific. No evidence appears overstated." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 17, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:29159", + "issue:34567", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:42490", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43452", + "issue:43454", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43606", + "issue:43638", + "issue:43643", + "issue:43645", + "issue:43696", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43854", + "issue:43927", + "issue:43931", + "issue:44387", + "issue:44479", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44617", + "issue:44756", + "issue:44811", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45042", + "issue:45356", + "issue:45459", + "issue:45759", + "issue:45803", + "issue:45859" + ], + "soft_pairs": [ + "issue:44589|issue:45042", + "issue:43638|issue:44936", + "issue:43746|issue:44811", + "issue:43452|issue:44991", + "issue:44387|issue:44617", + "issue:43550|issue:44912", + "issue:43819|issue:43927", + "issue:36010|issue:43645", + "issue:43257|issue:43329", + "issue:43854|issue:44977", + "issue:39692|issue:44912", + "issue:43720|issue:44610", + "issue:45459|issue:45803", + "issue:38175|issue:44610", + "issue:36010|issue:36331", + "issue:43454|issue:43577", + "issue:43531|issue:45042", + "issue:43262|issue:44811", + "issue:43329|issue:44479", + "issue:43257|issue:43696", + "issue:42490|issue:43257", + "issue:43643|issue:45759", + "issue:43606|issue:45859", + "issue:43329|issue:43450", + "issue:43749|issue:44756", + "issue:34567|issue:40990", + "issue:44568|issue:45356", + "issue:43257|issue:43931", + "issue:29159|issue:34567" + ], + "prepared_review_unit_hash": "f5382abc69d6b255bb3663a1bdacca79b85f15e93238d5ea9f60c74e5a37e8f0", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36089, + "estimated_input_tokens": 9023, + "estimated_eval_tokens": 18302 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: most items are unrelated tokenizer, trainer, loading, or multimodal bugs. No soft pair is a safe duplicate merge under a strict same-root-cause standard.", + "confidence": 0.78, + "canonical_issue_reason": "No single canonical issue stands out because the set spans many unrelated bugs; the overall cluster is too mixed to collapse into one duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 is the closest thing to a representative issue for the only semi-coherent subgroup (video-input regressions across Qwen/VLM models), but it still does not subsume the rest of the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Different failures in different subsystems: float8 storage loading vs PIL image processor torchvision dependency." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Both are training/eval failures, but one is a deepspeed/BERT zero3 index error and the other is a trainer lifecycle regression." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Local checkpoint loading with PEFT adapters is unrelated to Whisper batch_decode ignoring skip_special_tokens." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "One is gguf_file breaking from_pretrained loading; the other is a tokenizer regression for a specific model family." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but the underlying causes and code paths are different model/runtime issues." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "Torch.compile SDPA failure in Bamba is not the same bug as MXFP4 loading fallback in git-oss." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "DAC latent conversion/STE mismatch is unrelated to DiaConfig losing custom token IDs on save/load." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "GenerationMixin import failure is a different problem from Jupyter custom-model initialization in Transformers 5.0." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3-MoE accelerate/deepspeed weight conversion and multimodal video token counting are different bugs." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "GLM-4.7-Flash load/test issues do not match the Qwen3.5 flash-attention generation bug." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "SigLIP2 docs example errors are unrelated to git-oss quantization loading behavior." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "BitNet packed-weight unpacking during accelerate loading is unrelated to OmDet-Turbo processor image-size mismatch." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Both hide errors, but they are different code paths and different failure modes, so not the same duplicate." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 is not the same as a processor/model input-size mismatch." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "GenerationMixin import and custom compute_loss kwargs are separate Trainer/AutoModel regressions." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "AyaVision weight tying and BLIP2 dtype loading are different model-loading defects." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior has no concrete overlap with PIL backend torchvision requirements." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio apply_chat_template sample-rate defaulting and Whisper batch_decode token skipping are different processor behaviors." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "43329 is a specific undefined-helper bug in multimodal token counting; 44479 is a broader video-input regression across models, not clearly the same root cause." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Qwen3-MoE weight conversion and GPT-oss OOM are unrelated loading/performance problems." + }, + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "Trailing-slash local path warning is unrelated to accelerate/deepspeed weight conversion for Qwen3-MoE." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Both involve config/model construction, but one is trust_remote_code missing fields and the other is text_config unwrapping for composite multimodal configs." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "CPU offload incompatibility appears in both, but the affected models and failure mechanisms are different." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Undefined helpers in video token counting are not the same as batched video processor output shape errors." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and Strix Halo mmap OOM are separate memory-loading issues." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token counting not updating is unrelated to GPT-oss perplexity results." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer special-token insertion regression and Kimi codec/regex warning issues are different tokenizer bugs." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3-MoE conversion/loading and Qwen3-VL weight-shape mismatch are different model-loading failures." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer slow/fast inconsistency is unrelated to TrainerState token accounting." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 18, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43452", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43824", + "issue:43927", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44560", + "issue:44568", + "issue:44737", + "issue:44821", + "issue:44849", + "issue:44898", + "issue:44912", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45137", + "issue:45357", + "issue:45381", + "issue:45405", + "issue:45488", + "issue:45538", + "issue:45588", + "issue:45701", + "issue:45907" + ], + "soft_pairs": [ + "issue:42915|issue:43742", + "issue:44291|issue:45042", + "issue:42915|issue:44821", + "issue:44849|issue:45701", + "issue:44568|issue:45488", + "issue:44987|issue:45538", + "issue:43452|issue:43582", + "issue:41720|issue:43540", + "issue:43697|issue:45357", + "issue:43927|issue:44912", + "issue:42915|issue:45538", + "issue:44479|issue:45381", + "issue:33357|issue:34567", + "issue:43540|issue:44479", + "issue:42175|issue:43824", + "issue:43540|issue:44560", + "issue:43317|issue:45538", + "issue:44062|issue:44991", + "issue:44987|issue:45907", + "issue:24643|issue:29159", + "issue:45137|issue:45405", + "issue:43577|issue:45072", + "issue:28282|issue:34689", + "issue:43742|issue:44291", + "issue:43257|issue:43756", + "issue:42175|issue:44488", + "issue:28282|issue:29127", + "issue:42915|issue:43749", + "issue:43582|issue:43782", + "issue:44737|issue:45588", + "issue:38175|issue:44898" + ], + "prepared_review_unit_hash": "aa70bce33535ce5ac6c117817aef01d82c3be02338b196b0b4ee0a4d80915a0f", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37051, + "estimated_input_tokens": 9263, + "estimated_eval_tokens": 18782 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly heterogeneous issue cluster with one credible duplicate pair around Qwen VL video-input regression. The rest are separate bugs across tokenizer, loading, quantization, device-mapping, and backend behavior.", + "confidence": 0.87, + "canonical_issue_reason": "Issue 44479 is the best canonical issue for the only real overlap: it describes a broad Qwen video-input regression affecting multiple related models, and 45381 looks like a narrower report of the same problem.", + "canonical_pr_reason": null, + "best_issue_reason": "44479 is the strongest representative issue in this set because it is broad, concrete, and has the clearest duplicate-like companion report; most other items are too unrelated to serve as a cluster anchor.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different failures: Qwen3Moe FineGrainedFP8 loading vs MobileLLM key error." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated code paths: init_empty_weights/_is_hf_initialized vs PIL image processor torchvision dependency." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "Different subsystems: FP8 Qwen3Moe loading vs AutoImageProcessor URL loading." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Different bugs: Qwen3.5 hidden-states behavior vs general tokenization changes across versions." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Both tokenizer-related, but different model families and different failure mechanisms; not the same bug." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Unrelated: model loading failure for one repo vs CLIPTokenizer max_length bug." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "Completely different areas: gguf/AutoModel loading vs Apple Silicon allocator warmup TypeError." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Different Qwen issues and different symptoms: auto device-map CUDA assert vs video-input ValueError." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "Different regressions: RTDetrV2 output drift vs save_pretrained key serialization." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Different model-loading failures with different causes: config token ID loss vs MXFP4 fallback." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "No overlap beyond being Transformers issues; separate topics." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Both report the same Qwen VL video-input regression; 45381 is a narrower manifestation of the broader v5.3.0 breakage in 44479." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "Unrelated: macOS CLIP bus error vs TrainerState token-count tracking." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Both concern video input, but different model families and symptoms; not enough evidence of one underlying bug." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Different problems: packaging extras missing TensorFlow vs missing Qwen2.5-VL symbol import." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Different issues: Qwen3 MoE weight conversion vs Smollm3 RoPE-layer mismatch." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Similar area, but separate Qwen video-input bugs with different failures and model families." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Unrelated: device_map/offload loading bug vs CLIPTokenizer max_length." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "Both tokenizer loading failures, but the reports do not clearly point to the same underlying AddedToken bug." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Different API regressions: model loading for a repo vs list_repo_files not forwarding revision." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "Distinct bugs: DeepSpeed training RuntimeError vs tokenizer slow/fast inconsistency." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "Unrelated: ZeRO3 deque error vs PEFT version bump." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both dtype-related, but different models and different failure modes; not the same code-path problem." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "Different issues: missing PyTorch import vs a specific model-loading regression." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Distinct loading bugs with different stack traces and causes." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Separate model-specific bugs; no shared code-path evidence." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Packaging/backend dependency issue vs model loading failure; not duplicates." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "Different errors and domains: missing torch import vs LayoutLMv3 box validation message." + }, + { + "left": "issue:43749", + "right": "issue:45405", + "accept": false, + "reason": "Different issues: FSDP CPU RAM efficient loading regression vs unreleased PEFT version bump." + }, + { + "left": "issue:43582", + "right": "issue:43782", + "accept": false, + "reason": "Different failures: Apple Silicon allocator TypeError vs Qwen3VL weight_only loading error." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "Unrelated: XLNet device placement bug vs flash-attention integration crash." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different model behaviors and failure modes; not the same underlying bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: it identifies one plausible duplicate-like pair (44479/45381) and treats the rest as separate issues. The soft-edge verdicts are generally conservative and do not overclaim duplication from the available titles/evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 19, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:24643", + "issue:29127", + "issue:34567", + "issue:36331", + "issue:41762", + "issue:43404", + "issue:43452", + "issue:43493", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43643", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43906", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44488", + "issue:44514", + "issue:44589", + "issue:44625", + "issue:44792", + "issue:44821", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44936", + "issue:44964", + "issue:45042", + "issue:45081", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45600", + "issue:45823", + "issue:45920", + "issue:45987", + "issue:46121", + "issue:46202" + ], + "soft_pairs": [ + "issue:43540|issue:45823", + "issue:43404|issue:45987", + "issue:43526|issue:45479", + "issue:43906|issue:45381", + "issue:43493|issue:43994", + "issue:44291|issue:44821", + "issue:34567|issue:36331", + "issue:43756|issue:45325", + "issue:44913|issue:46121", + "issue:44062|issue:45081", + "issue:44792|issue:44964", + "issue:45042|issue:45081", + "issue:43742|issue:44589", + "issue:45357|issue:45920", + "issue:43452|issue:44488", + "issue:43531|issue:44821", + "issue:43643|issue:45375", + "issue:45405|issue:45600", + "issue:44625|issue:44877", + "issue:24643|issue:29127", + "issue:43720|issue:44898", + "issue:43697|issue:45216", + "issue:43906|issue:46202", + "issue:41762|issue:44936", + "issue:44514|issue:45381", + "issue:43688|issue:43749" + ], + "prepared_review_unit_hash": "84a95db6ec4e5576626fa6937da946ec8e85b4afcf78f66024583e17f04b359a", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35835, + "estimated_input_tokens": 8959, + "estimated_eval_tokens": 18174 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a set of loosely related bug reports connected only by superficial model/subsystem similarity. I found two likely true duplicate pairs: SigLIP2 discrepancy reports, and Qwen2.5-VL video/position-id reports. The rest are distinct bugs and should not be merged.", + "confidence": 0.7, + "canonical_issue_reason": "issue:43493 is the strongest root-cause style report among the clear duplicate pairs: it describes the underlying SigLIP2 HF-vs-JAX implementation discrepancy, while the paired issue is just a concrete symptom on a specific checkpoint.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 is the best overall issue for the cluster as a whole because it is a broad, current regression report with multiple affected models and clear user impact, making it the most useful general anchor if one issue had to represent the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43540", + "right": "issue:45823", + "accept": false, + "reason": "Different models and failures: Qwen3OmniMoe video processing vs Gemma4 device placement under auto device mapping." + }, + { + "left": "issue:43404", + "right": "issue:45987", + "accept": false, + "reason": "Unrelated bugs: missing lm_head tying in Mistral3 vs StaticCache sequence-length return type." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "Different layer of the stack: BeitImageProcessorFast label reduction bug vs sequence-classification zero-loss bug across models." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": true, + "reason": "Both point to the same Qwen2.5-VL video/position-id code path; 43906 is an isolated reproduction of the same underlying bug." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 implementation problem: one report is the general HF-vs-JAX discrepancy, the other is a specific model/pipeline symptom." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Different failure modes: init_empty_weights/_is_hf_initialized argument handling vs loading AutoImageProcessor from a URL." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "Both trainer-related, but they describe different bugs in different trainer code paths." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both involve RoPE, but the concrete bugs are different models and different position/rope calculations." + }, + { + "left": "issue:44913", + "right": "issue:46121", + "accept": false, + "reason": "Both touch rope config handling, but one is silent config reload fallback and the other is a TypeError during JSON-loaded validation." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Distinct tokenizer bugs: AddedToken special-arg duplication vs Mistral regex patching against a tokenizer without backend_tokenizer." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Different models and symptoms: janus test failure vs Phi-4 multimodal model loading failure." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "PIL/image-processor torchvision dependency regression is unrelated to Mistral tokenizer regex patching." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "Different loading failures: MobileLLM key error vs Float8 storage lookup failure." + }, + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Both are save/load regressions, but they affect different APIs and different model families with no shared code path evidence." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "Both are loading issues, but one is gguf_file/from_pretrained handling and the other is a specific model load failure without clear shared cause." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Unrelated subsystems: Qwen3-MoE sliding-window behavior vs image-processor loading from URL." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both are config/strictness issues, but one is generic trust_remote_code field loss and the other is a specific Qwen vision config field being dropped." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "Different maintenance issues: PEFT version pin vs stale auto_mappings references breaking CI." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Different config bugs: num_labels propagation vs strict config rejecting granite_speech." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "Different problems and layers: DeepSpeed training RuntimeError vs LayoutLMv3 error-message clarity." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "Different model/loading paths: BitNet packed-weight unpacking vs Perceiver non-default-resolution handling." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Different regressions: RTDetrV2 output drift vs Qwen3.5 save_pretrained checkpoint corruption." + }, + { + "left": "issue:43906", + "right": "issue:46202", + "accept": false, + "reason": "Both mention model/tokenization behavior, but one is a Qwen2.5-VL reproduction and the other is a Gemma pipeline roundtrip issue." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "Different bugs: Gemma3 ZeRO-3 load failure vs trainer.evaluate() after trainer.train()." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL issues, but one is batched chat-template padding behavior and the other is video vision_position_ids; different code paths." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Different training/runtime issues: auxiliary-loss normalization vs FSDP CPU RAM efficient loading." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The broad anchor choice for issue:45920 is defensible from the title, and the two accepted duplicate pairs are plausible given the shared model/topic signals. The rejected soft-edge pairs are also reasoned at a cautious level. No obvious overstatement that would justify rejecting." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 20, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:28282", + "issue:29159", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:35141", + "issue:38175", + "issue:41720", + "issue:43295", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43819", + "issue:43906", + "issue:43927", + "issue:44361", + "issue:44442", + "issue:44464", + "issue:44466", + "issue:44484", + "issue:44493", + "issue:44514", + "issue:44568", + "issue:44704", + "issue:44743", + "issue:44811", + "issue:44843", + "issue:44936", + "issue:44991", + "issue:45005", + "issue:45290", + "issue:45356", + "issue:45372", + "issue:45440", + "issue:45542", + "issue:45701", + "issue:45704", + "issue:45823", + "issue:46050", + "issue:46170", + "issue:46178" + ], + "soft_pairs": [ + "issue:45372|issue:46170", + "issue:43452|issue:43577", + "issue:44466|issue:45005", + "issue:44493|issue:46050", + "issue:43819|issue:44811", + "issue:30064|issue:33666", + "issue:44936|issue:45823", + "issue:43452|issue:44843", + "issue:45542|issue:46178", + "issue:44484|issue:44704", + "issue:43295|issue:45356", + "issue:28282|issue:32090", + "issue:44361|issue:44442", + "issue:43653|issue:44361", + "issue:28282|issue:33357", + "issue:44514|issue:45290", + "issue:44464|issue:44743", + "issue:38175|issue:43295", + "issue:28282|issue:35141", + "issue:41720|issue:43906", + "issue:43653|issue:44568", + "issue:43577|issue:43697", + "issue:43688|issue:45704", + "issue:44843|issue:44991", + "issue:43531|issue:43742", + "issue:43646|issue:43749", + "issue:29159|issue:33357", + "issue:43688|issue:45440", + "issue:43653|issue:43927", + "issue:43688|issue:45701" + ], + "prepared_review_unit_hash": "0467617e4ab7314dc8a06a8b4d17c820d42d89c309421fbc0c4e6489e2b1ae14", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36842, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is broadly heterogeneous: it mixes unrelated tokenizer, model-loading, training, generation, backend, and platform-specific regressions. None of the soft-similarity pairs look like the same concrete bug/change, so no duplicates should be merged here.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the items span many unrelated failure modes, so there is no representative issue that safely subsumes the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue either; the cluster is not one duplicate family, so any single issue would be misleading as a representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Different import/dependency failures: Gemma processor loading vs Windows AMD ROCm torch.distributed import/type errors." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "Different bugs: gguf_file breaks tokenizer/model loading, while this is a BLIP2 dtype preservation issue." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both mention tied weights, but one is serialization/device-dependent lm_head.weight behavior and the other is a broader translation-model tied-weights issue." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Different persistence problems: unexpected position-id keys vs quantized model save/load breakage." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "Different code paths: DAC latent reconstruction mismatch vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Unrelated areas: void segmentation-map processing vs Qwen2-VL multi-GPU training." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Different failures: trainer lifecycle regression vs Gemma4 device-map mismatch during forward." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "Both involve tokenizer loading, but one is gguf_file handling and the other is an offline model_info / regex patch regression." + }, + { + "left": "issue:45542", + "right": "issue:46178", + "accept": false, + "reason": "Both are backend dependency issues, but they involve different backends and different failure modes (tensorboard/TF vs tensorflow_text)." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "Different APIs: save_pretrained shard-size behavior vs AutoProcessor kwargs propagation to cached_file." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer/processor regressions, but they affect different models and different root causes." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Missing PyTorch ImportError vs a Trainer distributed broadcast TypeError; not the same underlying bug." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer failures: MLuke task AttributeError vs AutoTokenizer load failure for FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizers and symptoms: BigBird special-token registration vs MLuke tokenizer AttributeError." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Different environments and failure types: missing PyTorch vs MacOS bus error on a CLIP model." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Same general API family, but different crashes and inputs in apply_chat_template; not one concrete bug." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Both touch generation/cache behavior, but on different models and distinct code-path bugs." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Unrelated: zero probabilities in SigLIP2 vs processor.tokenizer regression." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "Missing-library ImportError vs token-embedding resize/post_init reinitialization bug." + }, + { + "left": "issue:41720", + "right": "issue:43906", + "accept": false, + "reason": "The second issue is only an isolated reproduction of another report; there is no evidence it is the same Qwen3 device-map cudaErrorAssert bug." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both involve special tokens, but they are different tokenizers and distinct behaviors." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Different problems: BLIP2 dtype casting vs RTDetrV2 output divergence in v5." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Different model bugs: auxiliary-loss normalization vs an apex RMSNorm memory leak." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but one is offline/model_info behavior and the other is a specific model tokenizer load failure." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Different model-loading/runtime bugs: Qwen3-MoE sliding_window vs MobileLLM key error." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Different regression classes: custom model initialization breakage vs FSDP CPU RAM-efficient loading." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer inconsistency vs MacOS bus error; unrelated symptoms and subsystems." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Both involve MoE models, but one is auxiliary-loss normalization and the other is a native-vs-remote implementation divergence." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "Different tokenizer/config bugs: BigBird special-token registration vs DiaConfig losing custom token IDs." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Unrelated: auxiliary-loss normalization vs a broad tokenization-version 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": 17, + "review_unit_index": 21, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:28218", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:34567", + "issue:36010", + "issue:36331", + "issue:41720", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43643", + "issue:43688", + "issue:43720", + "issue:43742", + "issue:43756", + "issue:43825", + "issue:43927", + "issue:43994", + "issue:44361", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44554", + "issue:44561", + "issue:44610", + "issue:44625", + "issue:44821", + "issue:44849", + "issue:44913", + "issue:44977", + "issue:45357", + "issue:45491", + "issue:45588", + "issue:45823", + "issue:46002", + "issue:46050", + "issue:46192", + "issue:46209" + ], + "soft_pairs": [ + "issue:45491|issue:46002", + "issue:43550|issue:43927", + "issue:44464|issue:44977", + "issue:43643|issue:44561", + "issue:43550|issue:44361", + "issue:44479|issue:44625", + "issue:41720|issue:44464", + "issue:44977|issue:46209", + "issue:43577|issue:45588", + "issue:32090|issue:34567", + "issue:43577|issue:44554", + "issue:43504|issue:43550", + "issue:43504|issue:43720", + "issue:43504|issue:44610", + "issue:32090|issue:33666", + "issue:33666|issue:36010", + "issue:44466|issue:46050", + "issue:43742|issue:44821", + "issue:43720|issue:43927", + "issue:45823|issue:46050", + "issue:43606|issue:43927", + "issue:43452|issue:44451", + "issue:43262|issue:43295", + "issue:44849|issue:46209", + "issue:43278|issue:43688", + "issue:43994|issue:46192", + "issue:43525|issue:43572", + "issue:33666|issue:36331", + "issue:43720|issue:44361", + "issue:43756|issue:44913", + "issue:44625|issue:46002", + "issue:43493|issue:43825", + "issue:28218|issue:33357", + "issue:44849|issue:45357" + ], + "prepared_review_unit_hash": "2370325747955e6cb4f1eedc96ba6451d3cb3a7f270be2d83392001cbd7c1f02", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37932, + "estimated_input_tokens": 9483, + "estimated_eval_tokens": 19222 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the soft pairs share generic Transformers terms (config, loading, training, device/dtype, tokenizer) but describe different concrete bugs and code paths. I would not merge any of the suggested pairs as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Different failures: Gemma3 mixed-length sliding-window attention NaNs vs classification-head NaNs on checkout/init." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "One is Bamba-9B-v2 torch.compile SDPA, the other DiaConfig save/load custom token IDs and generation IndexError." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Both involve generation, but one is chunked generation with compiled forward; the other is Qwen3.5 flash-attention inference failure." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both are trust_remote_code-adjacent, but one is missing fields from AutoConfig.from_pretrained and the other is removal of is_torch_fx_available breaking remote-code models." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Bamba torch.compile/SDPA bug is unrelated to MLukeTokenizer AttributeError on tasks." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression for several Qwen VL models is not the same as Qwen3.5 num_labels propagation." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert is a different path from chunked generation inconsistency under compiled forward." + }, + { + "left": "issue:44977", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug is unrelated to FP8 inference failure on B200." + }, + { + "left": "issue:43577", + "right": "issue:45588", + "accept": false, + "reason": "Blip2 dtype persistence and flash_attention.py s_aux=None crash are unrelated model/runtime bugs." + }, + { + "left": "issue:32090", + "right": "issue:34567", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error is different from TrainerState.num_input_tokens_seen not updating." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Blip2 dtype mismatch is unrelated to the MPS attention value/query head-dim correctness issue." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Legacy field blocking BEiT semantic segmentation loading is not the same as Bamba SDPA compile failure." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy preset loading failure is unrelated to BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "BEiT legacy-field load bug is different from OmDet-Turbo processor/model input-size mismatch." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error and Qwen2-VL multi-GPU training are different training issues." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Qwen2-VL multi-GPU training is unrelated to GenerationMixin import failure." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "lm_head.weight serialization inconsistency is different from sinq quantized model save/load breakage." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "MobileLLM loading key error is unrelated to AutoImageProcessor loading from URL." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "BitNet packed-weight handling during loading is unrelated to DiaConfig custom token ID persistence." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Gemma4 device mismatch with auto device map is unrelated to sinq quantization save/load regression." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch for bark-small is unrelated to DiaConfig save/load token IDs." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "gguf_file loader breakage for AutoTokenizer/AutoModel is unrelated to ScandiBERT load failure." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio processor chat-template sample-rate default bug is unrelated to processor.tokenizer/image-passing regression." + }, + { + "left": "issue:44849", + "right": "issue:46209", + "accept": false, + "reason": "Qwen3.5 output_hidden_states bug is unrelated to FP8 inference failure on B200." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Embedding dtype changing between train/eval is unrelated to auxiliary-loss normalization in OLMoE/GPT Oss." + }, + { + "left": "issue:43994", + "right": "issue:46192", + "accept": false, + "reason": "SigLIP2 bad outputs from AutoModel/pipeline are unrelated to return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Llama4Config missing pad_token_id and StableLmConfig missing pad_token_idx are similar in shape but distinct model config fields and fixes." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Qwen2-VL multi-GPU training is unrelated to CustomTrainer.compute_loss unexpected num_items_in_batch." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet loading/unpacking issue is unrelated to MLukeTokenizer task AttributeError." + }, + { + "left": "issue:43756", + "right": "issue:44913", + "accept": false, + "reason": "Smollm3 RoPE-layer count bug is unrelated to GPTNeoXConfig rotary_pct reload defaulting." + }, + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "Qwen3.5 num_labels propagation is unrelated to NaNs in classification heads after checkout/init." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy is unrelated to pipeline translation-task error messaging in v5." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer extra-space-after-added-token bug is unrelated to CLIP MacOS bus error." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3.5 hidden-states bug is unrelated to incorrect visual encoder keys saved by Qwen3.5." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 22, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:33357", + "issue:33666", + "issue:39692", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43493", + "issue:43504", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44112", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44493", + "issue:44568", + "issue:44589", + "issue:44734", + "issue:44792", + "issue:44843", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45405", + "issue:45520", + "issue:45584", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:46050" + ], + "soft_pairs": [ + "issue:44964|issue:45405", + "issue:39692|issue:43493", + "issue:44568|issue:45812", + "issue:43540|issue:44977", + "issue:43334|issue:43572", + "issue:45200|issue:45229", + "issue:44488|issue:44843", + "issue:43493|issue:43844", + "issue:43278|issue:43844", + "issue:43334|issue:45759", + "issue:39692|issue:43295", + "issue:43756|issue:45083", + "issue:43257|issue:43299", + "issue:43646|issue:44466", + "issue:43824|issue:44589", + "issue:33357|issue:33666", + "issue:45229|issue:45823", + "issue:43504|issue:43606", + "issue:43927|issue:44361", + "issue:44589|issue:45520", + "issue:43572|issue:44291", + "issue:44464|issue:45584", + "issue:43278|issue:43701", + "issue:43504|issue:43653", + "issue:43606|issue:44112", + "issue:43575|issue:45405", + "issue:44792|issue:45405", + "issue:43994|issue:44493", + "issue:44423|issue:44734", + "issue:43873|issue:46050" + ], + "prepared_review_unit_hash": "0832b321236872c2c0a1bd3480dc2e4810bf0e820a33d3a211c79d6840fa5b71", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36884, + "estimated_input_tokens": 9221, + "estimated_eval_tokens": 18698 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most pairs are related only at a broad subsystem level (tokenizers, multimodal models, quantization, device/offload, or v5 regressions) but point to different concrete failures, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "No single issue cleanly represents the rest as a duplicate cluster; if one anchor is needed, issue 43873 is the broadest and most actively discussed, but it is still a separate bug from the others.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the best triage anchor because it is current, active, and broadly framed, but it should not be used to collapse the other issues into one duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Different problems: model-loading failure for Phi-4 multimodal vs an unreleased PEFT version bump." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Both involve SigLIP2, but one is a docs/example error and the other is an implementation-vs-JAX discrepancy." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "Tokenizer regressions, but for different models and different failure modes (BOS/EOS omission vs wrong Granite token IDs)." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Distinct Qwen multimodal bugs: video-input processing error vs flash-attention generation issue." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Both are missing pad-token fields, but they affect different configs/models and different attribute names." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Gemma4 text-only token-type handling is unrelated to the Gemma4 multi-GPU CUDA OOM report." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "One is a model-specific load failure; the other is offline-mode breakage from an unconditional hub call." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation mismatch is unrelated to the ZeRO-3 gradient amplification bug." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype drift in eval is a different issue from the HfDeepSpeedConfig/ZeRO-3 gradient problem." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Different multimodal-config regressions: missing pad_token_id vs not unwrapping text_config in AutoModelForCausalLM.from_config." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "Both are regressions around model usage, but one is a SigLIP2 docs example and the other is a processor/tokenizer API break." + }, + { + "left": "issue:43756", + "right": "issue:45083", + "accept": false, + "reason": "Separate model-specific bugs: Smollm3 RoPE layer handling vs qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both mention Qwen3 MoE, but one is accelerate+deepspeed conversion and the other is a dev-version load regression for Qwen3VL MoE." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Different v5 regressions: custom model initialization vs lm_head.weight serialization depending on device." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Import error for a missing class is unrelated to the Float8 storage-type lookup failure." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "Completely different models and failures: MacOS bus error on CLIP vs Qwen2-VL multi-GPU training." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Both are Gemma4, but one is inference OOM and the other is a forward device mismatch with device_map=\"auto\"." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy-field load failure for BEiT is unrelated to the CPU offload device mismatch in bark-small." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizer/config bugs: DiaConfig save-load token IDs vs MLukeTokenizer task-time AttributeError." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Float8 storage lookup failure is not the same as a flash_attn import-utils KeyError on Python 3.13." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Missing pad_token_idx after v5 update is separate from init_empty_weights passing an unexpected _is_hf_initialized argument." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Different generation bugs: compiled-forward chunked generation mismatch vs empty-transcription failure after align_special_tokens." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype drift is unrelated to a resume_from_checkpoint key mismatch." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are distinct bugs." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "A real device mismatch bug and a stale GraniteSpeech CI test are not the same underlying issue." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "TP OOM for Qwen2-57B is unrelated to the PEFT minimum-version bump." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to the PEFT versioning issue." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 nonsensical outputs and unexpected position-id keys are different symptoms with different likely causes." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Serve crashes differ: one is a multimodal '.to' AttributeError, the other is KV-cache continuation indexing." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Both concern quantization/offloading, but one is an offloading behavior bug and the other is save/load corruption after sinq quantization." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 23, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29159", + "issue:29942", + "issue:30333", + "issue:32090", + "issue:34567", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:42490", + "issue:43232", + "issue:43295", + "issue:43450", + "issue:43493", + "issue:43540", + "issue:43572", + "issue:43582", + "issue:43696", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43957", + "issue:44186", + "issue:44521", + "issue:44554", + "issue:44610", + "issue:44849", + "issue:44964", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45084", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45325", + "issue:45356", + "issue:45381", + "issue:45406", + "issue:45446", + "issue:45520", + "issue:45588", + "issue:45823", + "issue:46050", + "issue:46121", + "issue:46192", + "issue:46209" + ], + "soft_pairs": [ + "issue:29942|issue:30333", + "issue:45406|issue:46121", + "issue:45588|issue:46050", + "issue:42490|issue:43696", + "issue:43844|issue:44964", + "issue:28218|issue:34567", + "issue:29159|issue:32090", + "issue:44521|issue:45290", + "issue:43582|issue:44977", + "issue:44554|issue:45446", + "issue:45084|issue:45520", + "issue:43540|issue:45200", + "issue:15354|issue:38175", + "issue:45200|issue:45823", + "issue:45072|issue:45198", + "issue:37428|issue:43957", + "issue:43232|issue:44849", + "issue:44186|issue:46192", + "issue:43493|issue:43701", + "issue:45381|issue:46209", + "issue:45823|issue:46192", + "issue:43295|issue:43450", + "issue:43295|issue:44610", + "issue:43572|issue:44987", + "issue:43572|issue:43957", + "issue:44554|issue:45356", + "issue:39692|issue:46192", + "issue:37428|issue:43824", + "issue:45200|issue:45325" + ], + "prepared_review_unit_hash": "d5607f5e065fb60fabd435c78876293c9501d54ac277218ed5019e9f7a316022", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36487, + "estimated_input_tokens": 9122, + "estimated_eval_tokens": 18500 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated individual Transformers issues rather than a duplicate cluster. None of the soft edges look like the same underlying bug or change, so all are rejected.", + "confidence": 0.97, + "canonical_issue_reason": "If one issue must be chosen as a representative, issue 44521 is the clearest, concrete bug report with active discussion and inbound references; it is the most broadly representative item in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the strongest standalone issue for triage: it has a precise repro, clear user impact, and external references, making it the best representative issue here.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention 2 test failures and MLflow reporting on job failure are unrelated subsystems and failure modes." + }, + { + "left": "issue:45406", + "right": "issue:46121", + "accept": false, + "reason": "Gemma4 processor _tokenizer access and rope-parameter JSON list handling are different code paths with different errors." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "A flash_attention aux-argument crash is unrelated to sinq quantized model save/load corruption." + }, + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Local-path import warning and GPT-oss CUDA OOM are clearly different issues." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "ZeRO-3 gradient behavior and Phi-4 multimodal loading are unrelated bugs." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer spacing after added tokens does not match TrainerState token-count tracking." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Slow-vs-fast tokenizer inconsistency is unrelated to NoneType broadcast errors in Trainer." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both involve apply_chat_template, but one is zeroed assistant masks for multimodal inputs while the other is a crash on tool-call messages without content; not the same concrete bug." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation problems are unrelated." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "An MPS attention correctness issue and a PyTorch version check bug in flex_attention are different problems." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Template-node compilation failure and flash_attn import KeyError on Python 3.13 are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video-input processing failure in Qwen3OmniMoe is unrelated to Gemma4 mm_token_type_ids defaults." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "torch.jit.script GeneratorExp export failure and unexpected zero probabilities in SigLIP2 are unrelated." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Missing mm_token_type_ids defaults and Gemma4 device-map mismatch are different forward-path bugs." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "bfloat16 dtype mismatches in inference and Wav2Vec2 save/tokenization failures do not match." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "A missing flash attention helper import and torch.device(\"meta\") model-loading breakage are different compatibility issues." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs sync after sync_gpus and Qwen3.5 output_hidden_states behavior are unrelated generation bugs." + }, + { + "left": "issue:44186", + "right": "issue:46192", + "accept": false, + "reason": "LayoutLMv2 tokenizer crashes on NER/padding are not the same as return_tensors being ignored when text_kwargs is set." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:45381", + "right": "issue:46209", + "accept": false, + "reason": "Qwen2.5-VL video position-id bug and FP8 inference failure on B200 are different hardware/model issues." + }, + { + "left": "issue:45823", + "right": "issue:46192", + "accept": false, + "reason": "Gemma4 device-map placement bug and return_tensors precedence bug are not the same failure." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor.tokenizer regression with image inputs is unrelated to video processor output shape for batched inputs." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Custom processor.tokenizer regression and OmDet-Turbo image-size mismatch are different model-specific issues." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression and loading physical-intelligence/fast on transformers>=5.1.0 are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Missing StableLmConfig pad_token_idx and meta-device loading failures affect different mechanisms." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness with mismatched head dims and Kimi-K2.5 tokenizer codec regression are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:46192", + "accept": false, + "reason": "SigLIP2 documentation errors and return_tensors precedence are not the same bug." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Missing flash-support import and missing Qwen2.5-VL symbol export are different import errors." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Gemma4 multimodal token-type defaults and Qwen2.5-VL rope-index scaling are unrelated model bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 24, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:28218", + "issue:35141", + "issue:38175", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43334", + "issue:43388", + "issue:43450", + "issue:43493", + "issue:43525", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43653", + "issue:43819", + "issue:43824", + "issue:43827", + "issue:43844", + "issue:43856", + "issue:43873", + "issue:43901", + "issue:43975", + "issue:44112", + "issue:44206", + "issue:44442", + "issue:44466", + "issue:44514", + "issue:44521", + "issue:44610", + "issue:44792", + "issue:44857", + "issue:44898", + "issue:45083", + "issue:45127", + "issue:45198", + "issue:45412", + "issue:45520", + "issue:45563", + "issue:45584", + "issue:46170", + "issue:46178", + "issue:46192", + "issue:46209" + ], + "soft_pairs": [ + "issue:43295|issue:43873", + "issue:45520|issue:46178", + "issue:44898|issue:45412", + "issue:43550|issue:43844", + "issue:44466|issue:45127", + "issue:43334|issue:43525", + "issue:43450|issue:43575", + "issue:44792|issue:45563", + "issue:43824|issue:43975", + "issue:41720|issue:46209", + "issue:42915|issue:43824", + "issue:43388|issue:43526", + "issue:42490|issue:43493", + "issue:43278|issue:43856", + "issue:28218|issue:35141", + "issue:43653|issue:44112", + "issue:43526|issue:46192", + "issue:38175|issue:43493", + "issue:44442|issue:45198", + "issue:43975|issue:45520", + "issue:38175|issue:45412", + "issue:44514|issue:44521", + "issue:43262|issue:43450", + "issue:45520|issue:46170", + "issue:44206|issue:45584", + "issue:43819|issue:44112", + "issue:44610|issue:44857", + "issue:43827|issue:43901", + "issue:43819|issue:45083" + ], + "prepared_review_unit_hash": "81b1cf32396df91c8872ddf470374cac6fc897b59501e9223e17815a733918a6", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36003, + "estimated_input_tokens": 9001, + "estimated_eval_tokens": 18258 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All soft pairs are false positives: they share superficial themes (tokenizers, vision models, imports, docs, etc.) but describe different bugs, models, or code paths. No pull requests are present.", + "confidence": 0.92, + "canonical_issue_reason": "No pair looks like a true duplicate cluster. If one issue must serve as a representative, issue:45520 is the broadest core-runtime import bug and is the least model-specific of the set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue:45520 is the best standalone representative because it concerns shared import infrastructure and has broader applicability than the model-specific or docs-only reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Different failures: processor.tokenizer / batched image handling regression vs quantization offloading behavior." + }, + { + "left": "issue:45520", + "right": "issue:46178", + "accept": false, + "reason": "Both are import-related, but one is a flash_attn KeyError on Python 3.13 and the other is a BACKENDS_MAPPING error for tensorflow_text; different code paths and causes." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver resolution/interpolation failure is unrelated to RT-DETR memory not being released after deletion." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "torch.compile + SDPA crash in Bamba is not the same as gradient explosion with DeepSpeed ZeRO-3." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Both mention tied weights, but one is device-dependent serialization of lm_head.weight and the other is LoRA merge collapse with extended vocabulary." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Same missing-attribute symptom, but different configs/models (Qwen3-VL vs Llama4) and likely separate fixes." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video batched-shape bug and Qwen2-57B tensor-parallel OOM are unrelated." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "A failing janus generate_images test is not the same as a stale warning about num_return_sequences in paged generate()." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "ImportError for Qwen2_5_VLForConditionalGeneration and DeepSeek-Coder detokenization regression are different bugs." + }, + { + "left": "issue:41720", + "right": "issue:46209", + "accept": false, + "reason": "Different models, hardware, and symptoms: Qwen3 auto device mapping cudaErrorAssert vs FP8 inference failure on B200." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to a Qwen2.5-VL import error." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Both involve labels, but one drops elements in gather_for_metrics and the other reduces Beit image labels incorrectly; different layers and behavior." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "Trailing-slash local model loading warning is unrelated to the SigLIP2 HF-vs-JAX discrepancy." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Embedding dtype drift during evaluate and Qwen3 MoE memory usage are unrelated issues." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization when resizing embeddings with tie_word_embedding=False." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug and GraniteSpeech CI test flake are different problems." + }, + { + "left": "issue:43526", + "right": "issue:46192", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to return_tensors being ignored when text_kwargs is passed." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": false, + "reason": "Both mention SigLIP2, but one reports zero probabilities while the other is a broader HF-vs-JAX implementation mismatch; not enough to treat as the same concrete bug." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "FastSpeech2Conformer tokenizer loading failure and Wav2Vec2 save_pretrained/tokenization failure are separate audio-tokenizer bugs." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Detokenization regression in DeepSeek-Coder v5 is unrelated to a flash_attn import KeyError." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 zero probabilities and RT-DETR memory leak are entirely different bugs." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both are Qwen2.5-VL chat-template issues, but one crashes on batched padding=False input while the other returns all-zero assistant masks; different symptoms and likely different fixes." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate default bug is unrelated to batched video processor shape output." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Both are import failures, but one is flash_attn on Python 3.13 and the other is a Windows/ROCm torch.distributed dependency issue." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Unsupported center arg in LasrFeatureExtractor is unrelated to Whisper failing on empty transcription after align_special_tokens." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC.from_latents mismatch and a GraniteSpeech stale CI test have no shared underlying code path." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Processor output size mismatch for OmDet-Turbo is unrelated to an AMP/CUDA crash in LwDetrImageLoss." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs issues, but they concern different pipeline deprecations and different pages; not the same bug/change." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "DAC STE mismatch and qwen3_omni_moe feature-extract length helper behavior are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 25, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:39692", + "issue:42490", + "issue:43278", + "issue:43335", + "issue:43425", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43720", + "issue:43819", + "issue:43825", + "issue:43854", + "issue:43931", + "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:45478", + "issue:45588", + "issue:45593", + "issue:45823", + "issue:45859", + "issue:45920", + "issue:46192" + ], + "soft_pairs": [ + "issue:37428|issue:43425", + "issue:43335|issue:43550", + "issue:43606|issue:45083", + "issue:44857|issue:45593", + "issue:29942|issue:32090", + "issue:44448|issue:44743", + "issue:45127|issue:45245", + "issue:45290|issue:45381", + "issue:45200|issue:45588", + "issue:37428|issue:44351", + "issue:43335|issue:43653", + "issue:43611|issue:43931", + "issue:44206|issue:45823", + "issue:43550|issue:44112", + "issue:43819|issue:44493", + "issue:44855|issue:45083", + "issue:43335|issue:43504", + "issue:45588|issue:45859", + "issue:29127|issue:32090", + "issue:43278|issue:43825", + "issue:43335|issue:45083", + "issue:39692|issue:43526", + "issue:39692|issue:44493", + "issue:45310|issue:45478", + "issue:43720|issue:44112", + "issue:43335|issue:44112", + "issue:43335|issue:45072", + "issue:44448|issue:44779", + "issue:42490|issue:43854", + "issue:44360|issue:44561", + "issue:45920|issue:46192", + "issue:43606|issue:44857", + "issue:34689|issue:35141" + ], + "prepared_review_unit_hash": "7ab05a267208bec123d92f9f7ff4e48545c6428d178a6d738d406c63274802ad", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37806, + "estimated_input_tokens": 9452, + "estimated_eval_tokens": 19160 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are unrelated regression reports across different models/subsystems. The only clear duplicate-like pair is the two near-identical Qwen3.5 MoE from_pretrained failures.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45310 is the clearest canonical issue: it is the older, more specific report for the Qwen3.5 MoE from_pretrained regression, and issue:45478 reads like a follow-up/duplicate of the same failure.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the best representative issue overall because it captures a concrete, reproducible model-loading regression and has a strong duplicate counterpart, making it the most suitable anchor for triage.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Different failures: flash-attention helper import error vs Torch 2.10 compatibility break." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Unrelated bugs: sparse-layer config creation vs torch.compile/SDPA failure in Bamba." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "Different code paths: CPU offload device mismatch vs qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "Different model/components: AMP float16 crash in LwDetrImageLoss vs missing auxiliary losses in D-FINE." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 test failures are not the same as the Trainer _gpu_broadcast_one NoneType error." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Different regressions: Pegasus output drift between v4/v5 vs Qwen recurrent-state reset with cache." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA/tied-embedding collapse is unrelated to the category-count limit RuntimeError." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Chat template tool-call crash is unrelated to Qwen2.5-VL video vision_position_ids behavior." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 text-only token-type handling is unrelated to the flash_attention s_aux=None crash." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Different import failures: flash-attention utility symbol missing vs HybridCache import error." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Sparse-layer config bug and BigBirdTokenizer special-token registration are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "base_model_prefix loading breakage is not the same as Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "Unsupported feature-extractor arg and Gemma4 PLE device mismatch are different problems." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Torch.compile SDPA bug in Bamba is unrelated to a stale CI device-override test." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC from_latents/STE mismatch is unrelated to unexpected position-id keys across models." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Python 3.13 IndentationError on import is unrelated to qwen3_omni_moe output-length behavior." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "Sparse-layer creation bug and Beit legacy-field loading failure do not share the same code-path." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Both involve Qwen3.5/MoE-ish flash/offload areas, but one is an AttributeError on sink-less flash attention while the other is CPU-offload incompatibility in rotary embedding." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMv3 error-message clarity issue is unrelated to the Trainer broadcast NoneType crash." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype drift during evaluate is unrelated to the pipeline translation-task error message regression." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug is unrelated to qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "SigLIP2 docs example errors and BeitImageProcessorFast reduce_labels bug are different issues." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "Documentation example failures are unrelated to position-id key regressions." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Near-identical Qwen3.5 MoE from_pretrained failures with the same regression surface; these look like the same underlying bug." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is unrelated to the stale GraniteSpeech CI device override test." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "Sparse-layer config creation and GraniteSpeech CI device-override flakiness are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "Sparse-layer creation bug is unrelated to bfloat16 dtype mismatches in inference." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus v4/v5 output drift is unrelated to DeepSeek tokenizer regression." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path trailing-slash import warning is unrelated to GLM-4.7-Flash model loading in tests." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer missing ReLU is unrelated to removal of is_torch_fx_available affecting trust_remote_code models." + }, + { + "left": "issue:45920", + "right": "issue:46192", + "accept": false, + "reason": "AutoTokenizer wrong token IDs regression is not the same as silently ignoring return_tensors when text_kwargs is passed." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch in Bark is unrelated to LwDetrImageLoss AMP crashes." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Llama 3.2 Vision model-loading regression is unrelated to embedding reinitialization after resizing token embeddings." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative overall. The duplicate-like call on issue:45310 / issue:45478 is supported by the identical titles and high similarity, and the rejected soft pairs are consistent with the packet. No PR-mergeability concerns apply here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 26, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:28218", + "issue:29942", + "issue:32090", + "issue:33357", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:40444", + "issue:41720", + "issue:42490", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43412", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43701", + "issue:43723", + "issue:43746", + "issue:43873", + "issue:43927", + "issue:44448", + "issue:44462", + "issue:44493", + "issue:44554", + "issue:44561", + "issue:44568", + "issue:44779", + "issue:44857", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45083", + "issue:45198", + "issue:45229", + "issue:45372", + "issue:45406", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45698", + "issue:45803", + "issue:45823", + "issue:46209" + ], + "soft_pairs": [ + "issue:38175|issue:44493", + "issue:44977|issue:45229", + "issue:44857|issue:45072", + "issue:43746|issue:46209", + "issue:44448|issue:44977", + "issue:40444|issue:43526", + "issue:28218|issue:32090", + "issue:43701|issue:45083", + "issue:29942|issue:35141", + "issue:32090|issue:33357", + "issue:43550|issue:45083", + "issue:43723|issue:44554", + "issue:37428|issue:44987", + "issue:44561|issue:44779", + "issue:43873|issue:45229", + "issue:43927|issue:45698", + "issue:44462|issue:44561", + "issue:43278|issue:43329", + "issue:42490|issue:43575", + "issue:10000|issue:15354", + "issue:41720|issue:45229", + "issue:44977|issue:45563", + "issue:45372|issue:45588", + "issue:44568|issue:45491", + "issue:44857|issue:45588", + "issue:43412|issue:45588", + "issue:45198|issue:45803", + "issue:45372|issue:45406", + "issue:43278|issue:43381", + "issue:45588|issue:45823" + ], + "prepared_review_unit_hash": "e02635811b9a4fc6577f8c5dc5472aeb7d4e0f8eaca48df451d8e8c15697dbd3", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36245, + "estimated_input_tokens": 9062, + "estimated_eval_tokens": 18380 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The soft links are largely topical false positives: these issues span unrelated tokenizer, model-loading, attention, dtype, and runtime bugs. I don't see any pair that is clearly the same underlying defect; the only near-adjacent case is the Gemma4 processor-loading subgroup, but even there the failures differ.", + "confidence": 0.94, + "canonical_issue_reason": "No strong canonical issue exists for this cluster because the items are heterogeneous; issue 45372 is only the closest local representative of the Gemma4 processor-loading subgroup.", + "canonical_pr_reason": null, + "best_issue_reason": "45372 is the best issue to anchor the nearest related subgroup because it has the clearest, most specific failure report, but it is not a good global duplicate hub for the full set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Different symptoms and domains: zero probabilities in SigLIP2 vs unexpected position-id keys; not the same bug." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "One is flash-attention generation behavior, the other is multi-GPU CUDA OOM; different failure modes and code paths." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "Both involve numeric precision, but they hit different models/components and distinct crashes/correctness issues." + }, + { + "left": "issue:43746", + "right": "issue:46209", + "accept": false, + "reason": "PEFT/local-checkpoint loading for GraniteSpeech is unrelated to FP8 inference failure on B200." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Version output differences for Pegasus are not the same as Qwen flash-attention generation failures." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Different subsystems and defects: multimodal iterable-dataset handling vs Beit fast processor label reduction." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to Trainer/_gpu_broadcast_one NoneType errors." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint key mismatch and a helper-function length calculation bug are unrelated." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention test failures and embedding reinitialization on resize are different bugs." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Different runtime failures: NoneType broadcast in Trainer vs MacOS bus error with CLIP." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure is unrelated to qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "Tokenizer loading regression in v5 is not the same as the MPS attention correctness issue." + }, + { + "left": "issue:37428", + "right": "issue:44987", + "accept": false, + "reason": "Missing flash-attention symbol import and loading a specific model in v5.1 are different problems." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 regressions, but one is trust_remote_code FX removal and the other is Deepseek tokenizer correctness." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Quantization/offloading behavior is not the same as Gemma4 multi-GPU OOM." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "Save/load config token IDs and loading the wrong custom module after save_pretrained are distinct bugs." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to removal of is_torch_fx_available." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype drift and an undefined-variable bug in multimodal token counting are different issues." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "Local-path import warning is unrelated to tensor-parallel OOM while loading Qwen2-57B-A14B-Instruct." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel issue and a torch.jit GeneratorExp export problem are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Different models and failures: auto device mapping cudaErrorAssert vs multi-GPU CUDA OOM." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Flash-attention generation failure and a stale warning about num_return_sequences are not the same bug." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Related Gemma4/flash-attention area, but one is an import error for ReasoningEffort and the other is an AttributeError on s_aux=None; different root causes." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer add_special_tokens behavior and Gemma3 NaN embeddings are unrelated defects." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "AMP/CUDA loss crash is not the same as flash_attention.py crashing on sink-less models." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export failures and flash_attention.py AttributeError are different code paths and bugs." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save/tokenization failures are unrelated to FuyuBatchFeature's bare-except problem." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both involve Gemma4 processor loading, but one is an import dependency failure and the other is a serve-time missing-_tokenizer AttributeError; not the same concrete bug." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Embedding dtype drift and gradient checkpointing being unusable in eval mode are different problems." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "Flash_attention.py crashing on s_aux=None is unrelated to Gemma4 PLE device mismatch under device_map=\"auto\"." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 27, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:25251", + "issue:29127", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43381", + "issue:43421", + "issue:43575", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43873", + "issue:43976", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44442", + "issue:44448", + "issue:44462", + "issue:44466", + "issue:44479", + "issue:44610", + "issue:44743", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45245", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45563", + "issue:45712", + "issue:46121" + ], + "soft_pairs": [ + "issue:43976|issue:45440", + "issue:44448|issue:44466", + "issue:43653|issue:44743", + "issue:25251|issue:32090", + "issue:45459|issue:45712", + "issue:45245|issue:45563", + "issue:44206|issue:44479", + "issue:43065|issue:43994", + "issue:44186|issue:45198", + "issue:43065|issue:45412", + "issue:43611|issue:46121", + "issue:45372|issue:45459", + "issue:44361|issue:45198", + "issue:44871|issue:44977", + "issue:32090|issue:33290", + "issue:42490|issue:43873", + "issue:33357|issue:39692", + "issue:29942|issue:30064", + "issue:29127|issue:34567", + "issue:42915|issue:42994", + "issue:43994|issue:45412", + "issue:44265|issue:45198", + "issue:43066|issue:44462", + "issue:43278|issue:43575", + "issue:43756|issue:45440", + "issue:43421|issue:45491", + "issue:43381|issue:43701", + "issue:43653|issue:44442", + "issue:44610|issue:45412" + ], + "prepared_review_unit_hash": "67263a04103ca719749ee6c04d942947bd4d62028cf7da15e784993e7ecea5f1", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36406, + "estimated_input_tokens": 9102, + "estimated_eval_tokens": 18460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is noisy and mostly spans unrelated Transformers bug reports across tokenizers, model loading, generation, quantization, and processors. I did not find any true duplicate pair among the soft candidates; all should stay separate.", + "confidence": 0.79, + "canonical_issue_reason": "If forced to pick one issue as the cluster anchor, issue 43066 is the broadest user-facing Transformers v5 regression in the set and is closest to the recurring tokenizer/loading theme, though the cluster is not a true duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43066 is the best representative of the cluster because it describes a general v5 tokenizer regression rather than a highly model-specific edge case.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Different problems: a Python version/import compatibility issue versus a model implementation divergence." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "One is a generation output regression for Pegasus; the other is serialization inconsistency for lm_head.weight. Different code paths." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "Tokenizer special-token decode bug versus Qwen recurrent-state/cache reset bug. Not the same underlying issue." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline top_k nesting change and Trainer NoneType broadcast failure are unrelated." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer error masking on missing protobuf versus cleanup of leftover dummy classes. Different fixes." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category cardinality runtime error and a stale warning in paged generate() are distinct bugs." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Unsupported center argument in a feature extractor versus a video-input regression in Qwen models. Different failures." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "SAM3 dummy Conv2d placeholder issue versus SigLIP2 incorrect outputs. Not the same bug." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2Tokenizer crash on NER/padding versus Wav2Vec2 save/tokenization failure; unrelated tokenizer paths." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy layer placeholder cleanup and RT-DETR memory release are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:46121", + "accept": false, + "reason": "base_model_prefix loading regression and rope-parameter validation TypeError are different model/config issues." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Gemma processor import dependency breakage versus hidden tokenizer decode errors; different root causes." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and Wav2Vec2 save/tokenization failure do not share a concrete code path." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Gemma eos_token configuration mismatch versus Qwen flash-attention generation regression; unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Local-path import warning and quantization/offloading behavior are different issues." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "MacOS bus error on a community CLIP model is not the same as documentation example errors for SigLIP2." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention 2 CI failures versus void segmentation map processing failure; unrelated." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity issue and TrainerState token counter not updating are different bugs." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Both are quantization-related, but one fails on load with FineGrainedFP8Config and the other fails while saving quantized models; distinct phases and likely distinct causes." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 nonsensical outputs versus RT-DETR memory leak; unrelated." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export failure on torch_compilable_check and Wav2Vec2 tokenization/save failure are unrelated." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Both involve tokenizers, but one is wrong decoder selection in v5 and the other is AutoTokenizer ignoring tokenizer.json; different underlying failures." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "BF16-to-FP32 dtype change in eval and tensor-parallel OOM are unrelated." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "RoPE layer count bug in Smollm3 and DeepSeekV3 implementation divergence are separate model-specific issues." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime post-processor update behavior and Gemma3 NaN embeddings are different code paths." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Gradient checkpointing in eval mode versus resume_from_checkpoint key mismatch are unrelated training issues." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "BigBirdTokenizer special-token decode issue and AutoTokenizer load failure for FastSpeech2ConformerTokenizer are different tokenizer bugs." + }, + { + "left": "issue:44610", + "right": "issue:45412", + "accept": false, + "reason": "Processor image-size mismatch and RT-DETR memory leak are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 28, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:35141", + "issue:36010", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43334", + "issue:43381", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43646", + "issue:43701", + "issue:43761", + "issue:43824", + "issue:43931", + "issue:44062", + "issue:44077", + "issue:44079", + "issue:44186", + "issue:44206", + "issue:44279", + "issue:44387", + "issue:44442", + "issue:44464", + "issue:44625", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44898", + "issue:44977", + "issue:45072", + "issue:45103", + "issue:45245", + "issue:45372", + "issue:45440", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45834", + "issue:45859", + "issue:45907" + ], + "soft_pairs": [ + "issue:43824|issue:44279", + "issue:29127|issue:33357", + "issue:29942|issue:36010", + "issue:45440|issue:45593", + "issue:43761|issue:45859", + "issue:43526|issue:45834", + "issue:43526|issue:44186", + "issue:45103|issue:45372", + "issue:29127|issue:35141", + "issue:44625|issue:45584", + "issue:43334|issue:43531", + "issue:44442|issue:45072", + "issue:43575|issue:43701", + "issue:43526|issue:44857", + "issue:42915|issue:43931", + "issue:42490|issue:43381", + "issue:44977|issue:45245", + "issue:43575|issue:44387", + "issue:44464|issue:45491", + "issue:43262|issue:44206", + "issue:43646|issue:44077", + "issue:44743|issue:44811", + "issue:43452|issue:45588", + "issue:44062|issue:45907", + "issue:43761|issue:44079", + "issue:33290|issue:34689", + "issue:44442|issue:44898" + ], + "prepared_review_unit_hash": "18cb09e631fb8215bb0725c3b4860ce76bd30a3e031e5ab1dcd6121c05d2a120", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35767, + "estimated_input_tokens": 8942, + "estimated_eval_tokens": 18140 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the proposed edges connect unrelated regressions, model-specific issues, and API/import problems rather than the same underlying bug. I would reject all soft edges and treat the set as noise rather than a duplicate cluster.", + "confidence": 0.96, + "canonical_issue_reason": "No true canonical issue stands out because the items are unrelated; if a single anchor is required, issue 45372 is the least-bad choice since it is open, well-described, and has inbound references, but it is not a duplicate representative.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45372 is the best cluster anchor only in the loose sense that it is open and already referenced; however, it should not be treated as the canonical duplicate because the cluster does not share one underlying bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Different problems: Qwen2.5-VL import failure vs a broad dependency issue; no shared code-path." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 error-message clarity vs a MacOS CLIP bus error; unrelated symptoms and subsystems." + }, + { + "left": "issue:29942", + "right": "issue:36010", + "accept": false, + "reason": "Flash Attention 2 test failures are unrelated to the GenerationMixin import error." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "Different model families and failure modes: DeepseekV3MoE divergence vs D-FINE auxiliary-loss behavior." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "Both involve Qwen-family components, but one is a CLIPVisionModel hidden-state regression and the other is CPU-offload rotary embedding compatibility." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to Kosmos2.5 OCR index errors." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "Both are image/tokenization issues, but they affect different processors and different failure mechanisms." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "auto_docstring annotation handling is unrelated to the mistral_common ReasoningEffort import failure." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 box validation messaging and token-embedding reinitialization are different code paths." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and Whisper empty-transcription generation are unrelated." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Qwen3-VL missing pad_token_id and Qwen3-MoE sliding_window are separate configuration bugs." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "FastSpeech2ConformerTokenizer loading failure is unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Tensor-parallel OOM on Qwen2-57B is not the same as resume_from_checkpoint key mismatch." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "BeitImageProcessorFast label handling and LwDetrImageLoss AMP crashes do not share a concrete implementation bug." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "FineGrainedFP8Config failure for Qwen3Moe and Qwen3-VL weight-shape mismatch are distinct model-loading issues." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "Trailing-slash import warning and gradient-checkpointing eval-mode restriction are unrelated." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug is not the same as the 2^24 category limit error." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory, but one is load-time TP OOM and the other is increased CUDA reserved memory under int4 quantization; not the same bug." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Chunked generation with compiled forward and Gemma3 mixed-length NaN embeddings are different generation/attention failures." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio processor chat-template sampling-rate default and LasrFeatureExtractor unsupported center arg are unrelated audio preprocessing bugs." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Custom model initialization breakage in Transformers 5.0.0 is not the same as patchtsmixer optional post_init handling." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Qwen3.5 recurrent-state reset bug and Whisper batch_decode skip_special_tokens issue affect different logic paths." + }, + { + "left": "issue:43452", + "right": "issue:45588", + "accept": false, + "reason": "gguf_file loading breakage and flash_attention.py s_aux=None crash are separate loader/integration bugs." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken duplicate special-argument error is unrelated to list_repo_files missing revision forwarding." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression and ModelOutput key assignment bug are different abstractions." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "AdaFactor+DeepSpeed OOM and Llama 3.2 Vision model-loading breakage are different failures." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading and Perceiver non-default-resolution image classification failures are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 29, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:33290", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:36010", + "issue:36331", + "issue:42490", + "issue:42907", + "issue:43232", + "issue:43316", + "issue:43381", + "issue:43388", + "issue:43493", + "issue:43525", + "issue:43611", + "issue:43643", + "issue:43650", + "issue:43720", + "issue:43723", + "issue:43749", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44077", + "issue:44188", + "issue:44190", + "issue:44442", + "issue:44479", + "issue:44496", + "issue:44743", + "issue:44855", + "issue:44871", + "issue:44913", + "issue:44987", + "issue:45003", + "issue:45137", + "issue:45216", + "issue:45290", + "issue:45440", + "issue:45479", + "issue:45698", + "issue:45759", + "issue:45803", + "issue:46178", + "issue:46202", + "issue:46209" + ], + "soft_pairs": [ + "issue:33290|issue:34567", + "issue:33290|issue:35141", + "issue:45003|issue:45440", + "issue:43723|issue:45479", + "issue:43381|issue:43873", + "issue:43316|issue:43525", + "issue:43232|issue:43388", + "issue:42490|issue:44743", + "issue:43761|issue:44188", + "issue:44479|issue:45290", + "issue:34689|issue:43611", + "issue:43720|issue:44442", + "issue:44987|issue:46178", + "issue:44442|issue:45803", + "issue:44077|issue:44479", + "issue:43931|issue:45216", + "issue:34689|issue:36010", + "issue:43493|issue:43827", + "issue:33290|issue:36331", + "issue:43316|issue:45759", + "issue:42907|issue:45698", + "issue:45290|issue:46209", + "issue:43873|issue:44871", + "issue:43749|issue:43931", + "issue:43643|issue:44913", + "issue:43650|issue:46202", + "issue:44190|issue:44496", + "issue:44855|issue:45137" + ], + "prepared_review_unit_hash": "563af6ed5ffaec350ac1c20a5e9ff5c8f988e1e8fe1f5d4a13bd710303dd8b4b", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35710, + "estimated_input_tokens": 8928, + "estimated_eval_tokens": 18112 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: most pairs are only weak title-similarity matches and do not look like the same underlying bug or change. No soft edges should be merged.", + "confidence": 0.9, + "canonical_issue_reason": "issue:45698 is the broadest and most representative open bug in the set: a loading/persistence problem around `from_pretrained`/`save_pretrained`, with inbound references and discussion. It is the least specific single issue to anchor the cluster, though the cluster itself is not cohesive.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is the best global representative because it is open, broadly phrased, and centered on a common model-loading workflow that is more likely to recur than the very narrow edge cases elsewhere in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "Different failures: DeepSpeed Adafactor OOM vs TrainerState token counter not updating. Same general training area, but not the same bug or code path." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "Unrelated: optimizer memory blow-up vs embedding resize/post_init reinitialization bug." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Both involve model loading internals, but one is unsafe `sys.modules` access and the other is DeepSeekV3 implementation divergence; not the same underlying issue." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer loading failure vs sequence-classification loss degeneracy; different subsystems and symptoms." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing eval-mode behavior vs quantization/offloading bug. Similar high-level area, but not one concrete defect." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config/API discrepancies, but they affect different models and fields (`Gemma3TextConfig` vs `Llama4Config.pad_token_id`)." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs ordering bug vs metric gathering dropping tuple labels. Different code paths." + }, + { + "left": "issue:42490", + "right": "issue:44743", + "accept": false, + "reason": "Local-path import warning vs recurrent-state reset in Qwen3.5 modular model; not the same bug." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "Hidden states regression in CLIP vision forward vs attention-kernel divergence under `torch.compile`." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Video-input regression for several Qwen VL models vs chat template crash on tool-call assistant messages; unrelated input pipelines." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Model-loading breakage in one release vs base_model_prefix loading regression in another; same broad theme, but not a single concrete fix." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "BitNet packed-weight unpacking during accelerate loading vs AutoTokenizer loading FastSpeech2ConformerTokenizer; different loading layers and model types." + }, + { + "left": "issue:44987", + "right": "issue:46178", + "accept": false, + "reason": "Loading a specific model failing in v5.1.0+ vs backend mapping error for `tensorflow_text`; not the same defect." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer loading failure vs Fuyu tensor conversion exception handling; unrelated code paths." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "Model post-init policy for `patchtsmixer` vs video-input regression in Qwen VL models." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Shape mismatch loading a specific model vs incorrect checkpoint saving for Qwen3.5; both are serialization-related but not the same bug." + }, + { + "left": "issue:34689", + "right": "issue:36010", + "accept": false, + "reason": "Model loading regression vs missing `GenerationMixin` import; different failure modes and fixes." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation mismatch vs docs still referencing removed pipeline APIs; one is model behavior, the other documentation cleanup." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed/Adafactor OOM vs custom trainer loss signature mismatch. No shared underlying bug." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Config API discrepancy for Gemma3TextConfig vs `from_config` text_config unwrapping for Qwen multimodal configs; related to configs but distinct defects." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Both are save/load related, but one is dequantized Ministral saving and the other is wrong custom module reloading after save_pretrained; not the same code-path problem." + }, + { + "left": "issue:45290", + "right": "issue:46209", + "accept": false, + "reason": "Chat template crash on tool-call assistant messages vs FP8 inference failure on B200; unrelated." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading issue vs Gemma eos token config mismatch; both model/runtime issues but not the same bug." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM efficient loading vs Qwen3-VL weight-shape mismatch. Both loading-related, but not the same defect." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Missing fields with `trust_remote_code=True` vs GPTNeoX rotary_pct not persisting on reload; both config serialization issues but distinct." + }, + { + "left": "issue:43650", + "right": "issue:46202", + "accept": false, + "reason": "Generic/low-signal issue vs Gemma pipeline roundtrip failure; no evidence they share the same bug." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in a training script vs unrecognized model/config.json error for a specific repo; different surfaces." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 TorchScript parsing/import error vs DeepSpeed ZeRO3 deque underflow; unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 30, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:25251", + "issue:28282", + "issue:32090", + "issue:35141", + "issue:41950", + "issue:42617", + "issue:43065", + "issue:43295", + "issue:43316", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43452", + "issue:43475", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43673", + "issue:43688", + "issue:43701", + "issue:43716", + "issue:43742", + "issue:43761", + "issue:43825", + "issue:44077", + "issue:44162", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44534", + "issue:44617", + "issue:44743", + "issue:44792", + "issue:44871", + "issue:44877", + "issue:44913", + "issue:45127", + "issue:45200", + "issue:45584", + "issue:45803", + "issue:46050", + "issue:46170", + "issue:46178" + ], + "soft_pairs": [ + "issue:25251|issue:28282", + "issue:43065|issue:44617", + "issue:43295|issue:43825", + "issue:43550|issue:43701", + "issue:43316|issue:44877", + "issue:43606|issue:43701", + "issue:43316|issue:43716", + "issue:43540|issue:44514", + "issue:42617|issue:43575", + "issue:44077|issue:44534", + "issue:44162|issue:44792", + "issue:44871|issue:45584", + "issue:16998|issue:22355", + "issue:43575|issue:43742", + "issue:45803|issue:46170", + "issue:43404|issue:43441", + "issue:43295|issue:44871", + "issue:43673|issue:44743", + "issue:43761|issue:44479", + "issue:41950|issue:43475", + "issue:43388|issue:43688", + "issue:43452|issue:44496", + "issue:43646|issue:44913", + "issue:44871|issue:46050", + "issue:45127|issue:45200", + "issue:43572|issue:46178", + "issue:32090|issue:35141", + "issue:44617|issue:45584" + ], + "prepared_review_unit_hash": "2ecea2b83c198bde38cffbbe3afa47bf843cacafbc7b6321cf1079783a0aa570", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36021, + "estimated_input_tokens": 9006, + "estimated_eval_tokens": 18268 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The items are a heterogeneous mix of unrelated Transformers bugs and regressions. The soft-paired issues share only broad subsystem keywords (e.g. loading, config, video, cache, v5 regressions) but not the same concrete underlying failure, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k output-shape bug and missing `transformers.onnx` import are unrelated failures." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "One is about a dummy Conv2d in Sam3PixelDecoder, the other is CUDA OOM in Sam3Video; different code paths and symptoms." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression and a misleading pipeline error message are separate problems." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile/SDPA failure and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Gemma3TextConfig API mismatch and granite_speech strict-config loading are different config issues." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch has nothing to do with checkpoint resume key mapping." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Config API discrepancy and image preprocessor/model dtype mismatch are distinct bugs." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Both involve Qwen video/text processing, but one is a video-input validation error and the other is batched apply_chat_template padding behavior." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "3d_parallel.py runtime failure and tensor-parallel OOM are different distributed-inference problems." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "Optional post_init enforcement and non-persistent buffer initialization are separate v5 concerns." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 model breakage and a janus test failure do not describe the same bug." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma-3 eos_token_id config inconsistency and Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "model_max_length question for DeBERTa-V3 and missing `transformers.onnx` import are unrelated." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "Tensor-parallel OOM on a large Qwen model and a key error loading MobileLLM are different load-time failures." + }, + { + "left": "issue:45803", + "right": "issue:46170", + "accept": false, + "reason": "Bare except swallowing interrupts in FuyuBatchFeature is unrelated to Windows/ROCm distributed import issues." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Missing lm_head tying in Mistral3ForConditionalGeneration and FlashAttention failure in Ministral-3 are different code-path bugs." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Processor/tokenizer regression and Gemma-3 eos_token_id config mismatch are separate issues." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Missing GenerationMixin cache support during chunked_prefill and recurrent state reset in modular_qwen3_5 are not the same defect." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression and Qwen VL video-input regression affect different models and behaviors." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification pipeline picking image processors and Sam3VisionEncoderOutput missing an attribute are unrelated." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric gathering drops tuple labels, while the other issue is about auxiliary-loss normalization." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "`gguf_file` argument handling and missing `model_type` in config.json are different model-loading failures." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Broad v5 custom model initialization breakage is not the same as GPTNeoX rotary_pct persistence on reload." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "Gemma-3 eos_token_id mismatch and sinq quantized save/load breakage are separate serialization/config bugs." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA merge collapse with tied embeddings and Gemma 4 mm_token_type_ids defaults are unrelated." + }, + { + "left": "issue:43572", + "right": "issue:46178", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig and tensorflow_text backend registration are different issues." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error and output embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Sam3Video CUDA OOM and Whisper empty-transcription failure are not the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 31, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:35141", + "issue:41720", + "issue:42490", + "issue:43065", + "issue:43335", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43644", + "issue:43761", + "issue:43827", + "issue:43931", + "issue:43976", + "issue:43994", + "issue:44060", + "issue:44077", + "issue:44112", + "issue:44190", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44521", + "issue:44534", + "issue:44610", + "issue:44734", + "issue:44737", + "issue:44792", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44933", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45245", + "issue:45357", + "issue:45362", + "issue:45381", + "issue:45406", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:45803", + "issue:45987" + ], + "soft_pairs": [ + "issue:43504|issue:45698", + "issue:44734|issue:45406", + "issue:43525|issue:43531", + "issue:45563|issue:45584", + "issue:43761|issue:44242", + "issue:44792|issue:44977", + "issue:43931|issue:44060", + "issue:44479|issue:44734", + "issue:44534|issue:44913", + "issue:44792|issue:45600", + "issue:44190|issue:44964", + "issue:44464|issue:45563", + "issue:43827|issue:43994", + "issue:44855|issue:45588", + "issue:44521|issue:45381", + "issue:35141|issue:42490", + "issue:41720|issue:45245", + "issue:45803|issue:45987", + "issue:43065|issue:44610", + "issue:44933|issue:45362", + "issue:43335|issue:44855", + "issue:43644|issue:44913", + "issue:44077|issue:44877", + "issue:43976|issue:44855", + "issue:43550|issue:44190", + "issue:43540|issue:44734", + "issue:45127|issue:45357", + "issue:44737|issue:45803", + "issue:44112|issue:45600" + ], + "prepared_review_unit_hash": "0a82fef3c1b2cd654196228609d6206454b1d7a238ca96a8383b18530eb3028b", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36419, + "estimated_input_tokens": 9105, + "estimated_eval_tokens": 18466 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly unrelated one-off bugs across different models and subsystems. I do not see any soft pair that is clearly the same underlying issue, so I reject all duplicate candidates. The best representative issue is the broad serialization/init regression around non-persistent buffers.", + "confidence": 0.92, + "canonical_issue_reason": "issue:44534 is the cleanest broad statement of the non-persistent-buffer junk-initialization regression and is more generic than the version-specific wording in the older report.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44534 is the most actionable and generally applicable issue in the set; it describes a concrete core bug rather than a narrow model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Different failures: Beit semantic-segmentation preset loading with a legacy field vs wrong custom module restored after save_pretrained." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Different code paths: KV-cache continuation indexing in transformers serve vs Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Different bugs: Llama4Config missing pad_token_id vs Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "Different symptoms and APIs: stale num_return_sequences warning vs Whisper empty-transcription failure after align_special_tokens." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "Different model behaviors: CLIPVisionModel hidden_states regression vs load-balancing loss omission when output_router_logits=False." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Different model families and failure modes: janus image-generation test failure vs Qwen3.5 flash-attention generation issue." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "One is a weight-shape mismatch loading Qwen3-VL; the other is an incorrect tied-weights warning in Qwen3-Next." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Both mention multimodal/serve areas, but one is a video-input regression in Qwen VL models and the other is a serve KV-cache indexing crash." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Both are serialization-related, but one is non-persistent buffers being junk-filled and the other is GPTNeoX rotary_pct not surviving reload." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "Unrelated: janus image-generation test failure vs stale auto_mappings references to removed Sam3LiteText configs." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Different scope: local dataset loading in an example script vs a model-loading regression for microsoft/Phi-4-multimodal-instruct." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Different problems: compiled-forward chunked generation inconsistency vs a stale warning about num_return_sequences." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "One is docs still referencing removed pipeline() APIs; the other is a SigLIP2 inference correctness issue." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Different failures: Python 3.13 IndentationError in DebertaV2Model import vs flash_attention crash on s_aux=None." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Both are multimodal-template regressions, but they affect different outputs: assistant_masks vs vision_position_ids." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Completely different issues: token-embedding resize reinit bug vs import warning when local model path ends with a slash." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "No shared underlying bug: Qwen3 auto device mapping cuda assert vs category-count overflow at 2^24." + }, + { + "left": "issue:45803", + "right": "issue:45987", + "accept": false, + "reason": "Different concerns: bare except swallowing errors in FuyuBatchFeature vs StaticCache.get_seq_length returning the wrong type." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Different model components: dummy Conv2d in Sam3PixelDecoder vs OmDet-Turbo processor/model input-size mismatch." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Unrelated: nonexistent image_utils import vs Qwen3.5-35B crash in transformers chat." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "Different failures: SwitchTransformersConfig creates an unwanted sparse layer vs a Python 3.13 parser/import error." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Related only at a high level (config/serialization), but not the same bug: non-persistent buffer junk vs rotary_pct resetting on reload." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "Both involve config handling, but one is optional post_init on patchtsmixer and the other is strict config blocking granite_speech loading." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Both mention Python compatibility, but they are different failures: broad version incompatibility vs a specific IndentationError in one model file." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "Different areas entirely: Bamba torch.compile/SDPA failure vs local dataset loading in an image-classification training script." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Different multimodal bugs: video input ValueError in Qwen3OmniMoe vs wrong tensor indexing during serve KV-cache continuation." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "Different save/load regressions: LoRA merge with extended vocabulary on tied embeddings vs incorrect visual-encoder keys in save_pretrained." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "Different code paths: XLNet CPU-side relative position computation vs a bare except in FuyuBatchFeature.convert_to_tensors()." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "Unrelated CI issues: stale GraniteSpeech device override test vs auto_mappings references to removed Sam3LiteText configs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 32, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36246", + "issue:38175", + "issue:38617", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:42907", + "issue:43065", + "issue:43116", + "issue:43295", + "issue:43335", + "issue:43404", + "issue:43441", + "issue:43479", + "issue:43531", + "issue:43643", + "issue:43650", + "issue:43854", + "issue:43976", + "issue:44016", + "issue:44112", + "issue:44186", + "issue:44190", + "issue:44220", + "issue:44246", + "issue:44279", + "issue:44462", + "issue:44496", + "issue:44561", + "issue:44683", + "issue:44734", + "issue:44938", + "issue:44964", + "issue:44987", + "issue:45003", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45335", + "issue:45375", + "issue:45446", + "issue:45542", + "issue:45584", + "issue:45600", + "issue:45632", + "issue:45907", + "issue:46082", + "issue:46121" + ], + "soft_pairs": [ + "issue:44462|issue:45907", + "issue:43335|issue:44190", + "issue:46082|issue:46121", + "issue:43116|issue:44190", + "issue:45276|issue:45584", + "issue:44938|issue:45542", + "issue:43295|issue:45276", + "issue:43479|issue:45375", + "issue:36246|issue:41628", + "issue:45245|issue:45584", + "issue:43650|issue:44220", + "issue:42175|issue:44279", + "issue:44186|issue:44734", + "issue:44246|issue:45003", + "issue:41950|issue:44016", + "issue:36246|issue:38617", + "issue:44683|issue:45446", + "issue:43404|issue:45216", + "issue:38175|issue:43065", + "issue:44016|issue:44220", + "issue:43643|issue:44496", + "issue:44964|issue:45600", + "issue:44112|issue:44964", + "issue:43854|issue:44964", + "issue:44734|issue:45632", + "issue:44561|issue:45632", + "issue:43976|issue:44987", + "issue:43531|issue:45600", + "issue:43295|issue:45335", + "issue:42907|issue:43441" + ], + "prepared_review_unit_hash": "9a359d109b37b18ad304ec8a362180d7fb9b91d628c8471acd629a52b47974e6", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36651, + "estimated_input_tokens": 9163, + "estimated_eval_tokens": 18582 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is not a tight duplicate cluster; it mixes unrelated import regressions, model/config bugs, pipeline issues, and environment-specific failures. I would not merge any of the soft-edge pairs.", + "confidence": 0.79, + "canonical_issue_reason": "If one issue must serve as canonical, issue 41628 is the broadest user-facing import regression in the set and best represents the general class of API/export breakages.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 41628 is the strongest standalone report: clear failure mode, broad impact, and closest to the import/API regression theme seen in a few neighboring issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "Different bugs: AutoTokenizer ignoring tokenizer.json is a loading precedence issue, while list_repo_files forwarding revision is an Hub API parameter bug." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "Unrelated subsystems: SwitchTransformers config construction vs a local dataset loading failure in an example script." + }, + { + "left": "issue:46082", + "right": "issue:46121", + "accept": false, + "reason": "Different validation paths: LlamaConfig head_dim handling is not the same as rope-parameter JSON list coercion." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Different example-script failures: empty multi-label outputs vs inability to load a local dataset." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Unrelated model behaviors: Gemma4 embedding resizing vs Whisper empty-transcription generation after token alignment." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Separate environment issues: Python 3.14 import/load failure vs TensorFlow backend installation confusion with tensorboard." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Different regression surfaces and code paths: processor/tokenizer image handling vs Gemma4 resize_token_embeddings behavior." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Different multimodal config bugs affecting different models and fields; not the same underlying defect." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but for different symbols and namespaces; they do not point to the same missing export." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Completely unrelated: category count runtime limit vs Whisper transcription edge case." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "No shared bug: an issue about missing data vs a feature-extraction helper crash." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "Broad dependency complaints, but no shared concrete code-path or fix." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Tokenizer/NER padding crash is unrelated to serve-time KV-cache continuation indexing." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import performance and unsafe sys.modules access are different problems, even if both touch startup/import behavior." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Different topics entirely: pipeline processor lookup vs a notebook syntax error." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Both are import errors, but they concern different missing names and different modules." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Same subsystem name, but different failures: compiled flex_attention on new torch versions vs a wrong version gate for AuxRequest import." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Different model-specific regressions: untied lm_head weights vs incorrect checkpoint saving." + }, + { + "left": "issue:38175", + "right": "issue:43065", + "accept": false, + "reason": "Unrelated model bugs in different architectures and code paths." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error and audio feature extraction failure are not the same issue." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "Different config-loading failures: missing fields with trust_remote_code vs an unrecognized-model/config.json problem." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Different breakages: loading a specific multimodal model vs stale autogenerated auto-mappings for removed configs." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "CI-only GraniteSpeech test failure is unrelated to Phi-4 multimodal loading." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Different model families and failure modes; these are not the same bug." + }, + { + "left": "issue:44734", + "right": "issue:45632", + "accept": false, + "reason": "Serve-time tensor indexing bug is unrelated to local trust_remote_code cache-path collisions." + }, + { + "left": "issue:44561", + "right": "issue:45632", + "accept": false, + "reason": "One is a removed helper breaking trust_remote_code models; the other is a cache-path naming collision." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "A Python-version compatibility report is not the same as a specific model-load failure on a version range." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and auto_mappings CI breakage are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Both mention embedding/tokenizer mutation, but they affect different models and different internal components." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Different Ministral issues: saving dequantized models vs FlashAttention compatibility in v5 RC." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 33, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41950", + "issue:42490", + "issue:43295", + "issue:43299", + "issue:43335", + "issue:43381", + "issue:43441", + "issue:43479", + "issue:43525", + "issue:43526", + "issue:43577", + "issue:43650", + "issue:43673", + "issue:43824", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44242", + "issue:44279", + "issue:44336", + "issue:44351", + "issue:44368", + "issue:44496", + "issue:44521", + "issue:44617", + "issue:44779", + "issue:44933", + "issue:45020", + "issue:45042", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45325", + "issue:45335", + "issue:45357", + "issue:45405", + "issue:45584", + "issue:45588", + "issue:45593", + "issue:45632", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:46209" + ], + "soft_pairs": [ + "issue:42490|issue:45632", + "issue:43526|issue:43901", + "issue:43441|issue:45588", + "issue:44496|issue:45042", + "issue:43673|issue:44336", + "issue:43335|issue:45593", + "issue:44016|issue:44062", + "issue:44075|issue:44368", + "issue:43824|issue:44933", + "issue:43299|issue:45216", + "issue:43381|issue:45593", + "issue:43901|issue:44496", + "issue:43976|issue:44279", + "issue:43650|issue:46209", + "issue:44496|issue:45759", + "issue:45335|issue:45584", + "issue:43479|issue:43525", + "issue:44242|issue:45200", + "issue:44521|issue:45325", + "issue:43299|issue:45357", + "issue:38175|issue:44617", + "issue:43577|issue:45216", + "issue:41950|issue:45020", + "issue:44351|issue:45278", + "issue:44279|issue:45405", + "issue:44779|issue:45812", + "issue:41950|issue:45823", + "issue:39692|issue:43901", + "issue:43295|issue:44779" + ], + "prepared_review_unit_hash": "177608cc55e117776954ebf8a20a1e2b4ca8bfc27fec96aa554cf8476e38ed91", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "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": "These items are mostly unrelated issue reports clustered by superficial text similarity rather than true duplicates. The only near-umbrella report is the broad import-regression issue, but even that does not safely subsume the others.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45278 is the broadest representative in the set: it describes a general import-breakage regression after upgrading transformers, which is the closest thing here to an umbrella issue.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best global representative because it is the most general user-facing regression report and could plausibly absorb some specific import-failure manifestations.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42490", + "right": "issue:45632", + "accept": false, + "reason": "Trailing-slash local-path warning is a different bug from trust_remote_code cache collisions on shared leaf names." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels corruption is unrelated to a TextClassificationPipeline docs mismatch." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both touch flash_attention, but one is a Ministral-3 RC failure and the other is an s_aux=None crash in sink-less models." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Unrecognized model/config loading is unrelated to PIL backend processors incorrectly requiring torchvision." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "GenerationMixin cache missing in chunked_prefill has nothing to do with ANSI codes in loading_report." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "Sparse-layer creation in SwitchTransformers and missing auxiliary losses in D-FINE are different model bugs." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error and AddedToken special-argument TypeError are unrelated failures." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "SGD optimizer args not applied is unrelated to the Qwen3.5 tie_word_embeddings warning." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Missing Qwen2_5_VLForConditionalGeneration export is not the same as a nonexistent image_utils import." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3VL loading regression and Qwen3.5 save_pretrained regression affect different code paths." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "A docs/API wording issue is not the same as a model-type loading failure." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python version incompatibility and a generic dependency issue are too broad and not the same bug." + }, + { + "left": "issue:43650", + "right": "issue:46209", + "accept": false, + "reason": "Placeholder text and FP8 inference failure on B200 are unrelated." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Unrecognized model config loading is different from failing to unwrap text_config in composite Qwen multimodal configs." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Decoder embed resize bug and Whisper empty-transcription failure are different generation/model issues." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Default multimodal config initialization and missing pad_token_id on Llama4Config are unrelated attribute/config bugs." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "Router load-balancing loss gating and mm_token_type_ids defaulting are different training-path issues." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "All-zero assistant masks and still-image rope-index scaling are separate multimodal processing bugs." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Model loading failure and incorrect visual-encoder keys on save_pretrained are not the same concrete regression." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Zero probabilities in SigLIP2 is unrelated to Sam3Video CUDA OOM." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Blip2 dtype handling and Qwen3.5 checkpoint saving are different code paths and symptoms." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Pipeline processor lookup bug is unrelated to remote_code model breakage." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "Missing HybridCache import could be one symptom in a broader import regression, but the reports are not the same concrete bug." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "A generic dependency issue is not the same as bumping MIN_PEFT_VERSION to an unreleased release." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Deepseek tokenizer regression and Granite tokenizer regression affect different model families and likely different fixes." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Video-classification processor lookup and Gemma4 PLE device mismatch are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "SigLIP2 doc-example errors are not the same as a TextClassificationPipeline docs mismatch." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Custom processor/tokenizer regression and Deepseek tokenization regression are different failures." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Custom processor/tokenizer regression and Deepseek tokenization regression are different failures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 34, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:33290", + "issue:42175", + "issue:42994", + "issue:43065", + "issue:43262", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43450", + "issue:43650", + "issue:43653", + "issue:43673", + "issue:43723", + "issue:43761", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44291", + "issue:44297", + "issue:44361", + "issue:44462", + "issue:44466", + "issue:44521", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:44898", + "issue:45081", + "issue:45276", + "issue:45397", + "issue:45405", + "issue:45542", + "issue:45593", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45823", + "issue:45834", + "issue:45907", + "issue:46132" + ], + "soft_pairs": [ + "issue:44683|issue:45542", + "issue:43299|issue:43854", + "issue:43976|issue:45405", + "issue:43650|issue:46132", + "issue:44462|issue:45081", + "issue:43450|issue:44279", + "issue:44297|issue:44462", + "issue:45081|issue:45907", + "issue:44016|issue:44291", + "issue:44466|issue:45276", + "issue:43441|issue:45834", + "issue:45600|issue:45712", + "issue:42994|issue:43317", + "issue:43653|issue:45593", + "issue:33290|issue:45656", + "issue:44623|issue:45397", + "issue:44623|issue:44749", + "issue:43065|issue:44898", + "issue:44261|issue:44263", + "issue:44016|issue:45823", + "issue:44361|issue:45593", + "issue:43441|issue:44683", + "issue:44671|issue:44779", + "issue:43723|issue:44162", + "issue:42175|issue:43976", + "issue:43262|issue:44521", + "issue:43901|issue:43994", + "issue:43673|issue:43761" + ], + "prepared_review_unit_hash": "d52ebadb998f684ef4e5b14f0c39484b86734b44b4331f93b14817b0daa69c4b", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36015, + "estimated_input_tokens": 9004, + "estimated_eval_tokens": 18264 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pair looks like the same underlying bug/change; the set is a grab bag of unrelated Transformers regressions, packaging issues, and model-specific failures.", + "confidence": 0.93, + "canonical_issue_reason": "issue:44521 has the strongest signal in this mixed set: it is open, has substantial discussion/inbound references, and describes a concrete regression in a core API (`apply_chat_template`) affecting multimodal workflows.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 is the most actionable and representative issue here because it is well-scoped, actively discussed, and touches a central user-facing API.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Unrelated failures: compiled flex_attention on torch 2.9 vs TensorFlow backend missing from a torch install." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Both are model-loading regressions, but they affect different models and appear to be distinct code-path failures." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Different dependency/version problems: Python support breakage vs an unreleased PEFT minimum version bump." + }, + { + "left": "issue:43650", + "right": "issue:46132", + "accept": false, + "reason": "Completely unrelated: a generic data issue vs an AttentionInterface.register behavior change." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Both involve tokenizers, but one is AutoTokenizer ignoring repo tokenizer.json while the other is a Mistral regex patch AttributeError." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "No substantive overlap; batched video processor shape bug vs generic transformers dependency issue." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "One is save_pretrained metadata mismatch; the other is tokenizer loading ignoring tokenizer.json. Related area, different bug." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Different subsystems entirely: Mistral tokenizer regex patch vs repo file-listing revision forwarding." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Both touch Gemma4 weights, but one is serialization inconsistency and the other is resize_token_embeddings propagation; not the same bug." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "Different model-specific failures with no shared concrete code path." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are CI/cleanup related, but they stem from different stale-reference files and distinct failures." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantization is the broad theme, but saving a quantized model and loading a dequantized model with offload are different problems." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "Tokenizer special-token registration bug vs D-FINE auxiliary-loss logic; unrelated." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Both mention DeepSpeed/optimizer, but one is OOM with Adafactor and the other is double optimizer steps; not the same underlying issue." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Processor save_pretrained missing files vs Gemma-4 zero3 from_pretrained load failure are different save/load bugs." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated: processor serialization bug vs a performance regression in Chinese text filtering." + }, + { + "left": "issue:43065", + "right": "issue:44898", + "accept": false, + "reason": "Both are vision-model issues, but dummy Conv2d in Sam3PixelDecoder and Perceiver interpolation failure are different defects." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Different problems: missing layernorm config precision error vs torch.split return handling in a separate model indexer." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax issue and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "No overlap: MLukeTokenizer AttributeError vs D-FINE auxiliary-loss behavior." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Both are attention-related, but FlashAttention model failure and compiled flex_attention torch-compat failure are distinct code paths." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "Different regression surfaces: masked LM predictions for CamemBERT vs tokenizer output correctness for DeepSeek." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading regression vs a broader ESM2 breakage report; not the same bug." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "Packaging/backend omission and Python-version incompatibility are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Both involve apply_chat_template, but one is audio sampling-rate defaulting and the other is multimodal assistant masks; different bugs." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mismatch for return_all_scores vs an inference-quality bug on a specific model/pipeline." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache missing in chunked_prefill and CLIPVisionModel hidden_states omission are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 35, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36246", + "issue:41628", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:42886", + "issue:43299", + "issue:43352", + "issue:43450", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43596", + "issue:43643", + "issue:43673", + "issue:43784", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:44188", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44351", + "issue:44373", + "issue:44485", + "issue:44496", + "issue:44561", + "issue:44589", + "issue:44617", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:45003", + "issue:45250", + "issue:45292", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45593", + "issue:46132" + ], + "soft_pairs": [ + "issue:43784|issue:45542", + "issue:43450|issue:44617", + "issue:43824|issue:43976", + "issue:44297|issue:45397", + "issue:43525|issue:43881", + "issue:43901|issue:44246", + "issue:43525|issue:43643", + "issue:44188|issue:44336", + "issue:41628|issue:43976", + "issue:44589|issue:45362", + "issue:42491|issue:43299", + "issue:44336|issue:44373", + "issue:45561|issue:45563", + "issue:44295|issue:45397", + "issue:44246|issue:44496", + "issue:44242|issue:45593", + "issue:43531|issue:45362", + "issue:41950|issue:44220", + "issue:43526|issue:45003", + "issue:42886|issue:46132", + "issue:44246|issue:44871", + "issue:44336|issue:44485", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:45003|issue:45561", + "issue:44779|issue:45292", + "issue:43673|issue:44561", + "issue:36246|issue:43596", + "issue:44295|issue:44749", + "issue:44242|issue:44857", + "issue:43352|issue:45250" + ], + "prepared_review_unit_hash": "a611ace7a416df4677abf090768d0ad7d07550cf2f88da354bd3b6054cb7eb31", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36689, + "estimated_input_tokens": 9173, + "estimated_eval_tokens": 18602 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous: many are model-specific import/loading regressions, but the soft pairs do not look like true duplicates of the same concrete bug/change.", + "confidence": 0.31, + "canonical_issue_reason": "Issue 41628 is the broadest user-facing import-regression report in the set and is the closest thing to a central API breakage; the other issues are mostly model-specific failures, config mismatches, docs, or unrelated runtime behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "41628 is the best representative issue for this noisy cluster because it captures a generic transformers import failure, but it is still not a strong duplicate anchor for the rest of the items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Both are import/backend-related, but one is a sentence-transformers import NameError and the other is a TensorBoard-without-TensorFlow backend issue; different code paths." + }, + { + "left": "issue:43450", + "right": "issue:44617", + "accept": false, + "reason": "Both involve video, but one is a batched video-processor shape bug and the other is a CUDA OOM in Sam3Video; not the same defect." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Qwen2.5-VL import failure vs Python version compatibility breakage; unrelated underlying causes." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save metadata mismatch is a serialization bug; gemma-4 zero3 from_pretrained is a model-loading/config issue." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Missing pad_token_id on Llama4Config and GLM-4V loading failure are different model/config problems." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Docs mismatch for return_all_scores vs intermittent slow import are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Both mention missing config fields, but one is a specific Llama4 attribute error and the other is a trust_remote_code AutoConfig fields regression; not 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." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "Generic import-name failure vs Python version support issue; different causes and fixes." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage lookup failure and Qwen3.5 chat crash are unrelated runtime errors." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both are Qwen3 MoE loading regressions, but one is LoRA compatibility across hf4.x/hf5.x and the other is Qwen3VL MoE loading on dev v5; similar area, not clearly the same code-path bug." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "ANSI output in loading_report vs a wrong docstring for position_ids are not duplicates." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "pytest-xdist file race and stale warning in paged generate() touch different subsystems." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "position_ids buffer access bug vs gemma-4 zero3 loading bug are unrelated." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import slowness and eos_token_id configuration inconsistency are different issues." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Missing load-balancing loss when router logits are off vs missing auxiliary losses for D-FINE denoising are distinct training-loss bugs." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and Qwen3.5 chat crash are separate model-specific failures." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline looking for image processors and _torch_extract_fbank_features() are different processing bugs." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "BEiT fast image processor reduce_labels bug vs modeling_utils sys.modules access bug are unrelated." + }, + { + "left": "issue:42886", + "right": "issue:46132", + "accept": false, + "reason": "Offline cache loading failure and AttentionInterface.register behavior change are different concerns." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Repeated pair: import slowness vs Gemma eos_token_id inconsistency are unrelated." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI code emission in a utility vs GLM-5 RoPE implementation discussion are not the same bug." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but one is Qwen2.5-VLImageProcessor and the other is is_offline_mode from huggingface_hub; different symbols and packages." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Different missing exports: Qwen2.5-VLImageProcessor vs HybridCache." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils and xdist captured_info.txt race are unrelated." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "DeepSeek tokenizer regression and resize_token_embeddings not updating output_embeddings are different tokenization/model-head bugs." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Chunked-prefill cache regression and removal of is_torch_fx_available breaking trust_remote_code are different v5 regressions with different failure modes." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "Qwen2.5-VL import error and deepspeed zero3 BertModel IndexError are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "position_ids buffer read issue vs performance slowdown after upgrading transformers are not the same bug." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Load-balancing loss omission and LwDetrImageLoss float16 AMP crash are separate training/runtime issues." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "Nemotron Flash Attention 2 unsupported and a generic Flash Attention 2.0 issue title do not provide enough evidence of the same concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 36, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:42886", + "issue:43065", + "issue:43232", + "issue:43257", + "issue:43262", + "issue:43408", + "issue:43475", + "issue:43540", + "issue:43653", + "issue:43901", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44188", + "issue:44246", + "issue:44336", + "issue:44368", + "issue:44466", + "issue:44479", + "issue:44483", + "issue:44496", + "issue:44877", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45292", + "issue:45335", + "issue:45375", + "issue:45397", + "issue:45412", + "issue:45459", + "issue:45563", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45656", + "issue:45701", + "issue:45759", + "issue:45987", + "issue:46082", + "issue:46123", + "issue:46170", + "issue:46202" + ], + "soft_pairs": [ + "issue:44162|issue:45200", + "issue:43901|issue:45003", + "issue:44496|issue:45003", + "issue:45412|issue:46123", + "issue:44877|issue:45375", + "issue:45292|issue:45584", + "issue:43262|issue:44246", + "issue:45292|issue:45701", + "issue:43257|issue:45137", + "issue:45375|issue:45759", + "issue:45987|issue:46082", + "issue:43232|issue:44336", + "issue:44987|issue:45459", + "issue:44479|issue:46170", + "issue:44995|issue:45200", + "issue:43065|issue:43408", + "issue:43475|issue:43540", + "issue:44466|issue:45335", + "issue:45397|issue:45656", + "issue:42886|issue:45229", + "issue:44060|issue:44077", + "issue:44483|issue:46202", + "issue:44995|issue:45588", + "issue:45137|issue:45600", + "issue:44188|issue:46202", + "issue:44368|issue:44977", + "issue:43653|issue:44162", + "issue:44877|issue:46170", + "issue:44368|issue:45563", + "issue:43931|issue:44368" + ], + "prepared_review_unit_hash": "c4d29dd3877eb727f3c74b2ef2b7feed37a43862779dbf297bb8c5e8b76bf1a9", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36744, + "estimated_input_tokens": 9186, + "estimated_eval_tokens": 18628 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a collection of unrelated model, config, tokenizer, and generation regressions. The soft links are largely broad topical similarity rather than the same concrete bug, so none should be merged.", + "confidence": 0.74, + "canonical_issue_reason": "issue:45003 is the clearest broad core-library bug here: it targets `modeling_utils` and can affect many import/load paths, unlike the more model-specific reports in the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45003 is a strong representative issue because it describes a concrete infrastructure bug in a central code path, making it more reusable as the cluster anchor than the narrower model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "Different failures: ESM2 runtime breakage vs Gemma 4 token-type ID defaults." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Docs mismatch in pipeline behavior vs unsafe `sys.modules` access in core modeling code." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model/config loading error vs unrelated `sys.modules` access bug." + }, + { + "left": "issue:45412", + "right": "issue:46123", + "accept": false, + "reason": "RT-DETR memory leak and MaskGenerationPipeline partial-batch handling are unrelated code paths." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both involve config validation, but they are different models and distinct missing-field/loading failures." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Token-embedding resize behavior vs Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio chat-template sampling-rate bug vs intermittent import slowness." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "Resize-token-embeddings bug vs version-dependent tokenization change." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MOE weight conversion/loading issue vs DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Missing config field dropped by strict validation vs failing to unwrap `text_config` in `from_config`." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "Cache length return-type bug vs LlamaConfig `head_dim` validation error." + }, + { + "left": "issue:43232", + "right": "issue:44336", + "accept": false, + "reason": "Generation kwargs update bug vs ANSI output in loading report." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Model loading failure vs tokenizer error masking when protobuf is missing." + }, + { + "left": "issue:44479", + "right": "issue:46170", + "accept": false, + "reason": "Video-input regression vs Windows/ROCm torch.distributed import dependency problem." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Stale indexer cache crash vs Gemma 4 multimodal token-type defaulting." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Dummy Conv2d placeholder vs model-type mismatch warning." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "SAM 3 missing attribute vs Qwen3OmniMoe video-input ValueError." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Tied-weight serialization inconsistency vs token-embedding resize not propagating to decoder embeddings." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 ZeRO3 loading bug vs optimizer step being called twice under deepspeed." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline tokenizer cache loading bug vs Gemma 4 multi-GPU OOM." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Incorrect tied-weights warning vs patchtsmixer `post_init` policy change." + }, + { + "left": "issue:44483", + "right": "issue:46202", + "accept": false, + "reason": "Chat/completions request rejection vs Gemma pipeline roundtrip token issue." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Stale indexer cache crash vs flash-attention `s_aux=None` AttributeError." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque error vs stale auto-mappings reference to removed configs." + }, + { + "left": "issue:44188", + "right": "issue:46202", + "accept": false, + "reason": "torch.compile attention-kernel divergence vs Gemma roundtrip failure on \u2581." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Tie-word-embeddings warning vs flash-attention generation regression." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token bug vs ESM2 breakage." + }, + { + "left": "issue:44877", + "right": "issue:46170", + "accept": false, + "reason": "Config strictness regression vs Windows ROCm torch.distributed import issue." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "Different stale warnings: tied-weights config warning vs num_return_sequences warning." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch vs tied-word-embeddings warning." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 37, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:41628", + "issue:42371", + "issue:42994", + "issue:43065", + "issue:43232", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43701", + "issue:43827", + "issue:43873", + "issue:44038", + "issue:44162", + "issue:44188", + "issue:44297", + "issue:44361", + "issue:44530", + "issue:44623", + "issue:44661", + "issue:44734", + "issue:44749", + "issue:44869", + "issue:44871", + "issue:44933", + "issue:44945", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45406", + "issue:45459", + "issue:45478", + "issue:45656", + "issue:46077", + "issue:46123", + "issue:46132", + "issue:46170", + "issue:46202" + ], + "soft_pairs": [ + "issue:43673|issue:44734", + "issue:44530|issue:46170", + "issue:43065|issue:46123", + "issue:44297|issue:44623", + "issue:44933|issue:45459", + "issue:44361|issue:44749", + "issue:44869|issue:45356", + "issue:43646|issue:44162", + "issue:43232|issue:43673", + "issue:44623|issue:45656", + "issue:44749|issue:45656", + "issue:43701|issue:44749", + "issue:43673|issue:43827", + "issue:44871|issue:46132", + "issue:45230|issue:45397", + "issue:42994|issue:43482", + "issue:43475|issue:43643", + "issue:43475|issue:45381", + "issue:44038|issue:44623", + "issue:28282|issue:30990", + "issue:46077|issue:46123", + "issue:45406|issue:46202", + "issue:42371|issue:45092", + "issue:41628|issue:44933", + "issue:43475|issue:44661", + "issue:42994|issue:43525", + "issue:45375|issue:45478", + "issue:45081|issue:45103", + "issue:45310|issue:45375", + "issue:43873|issue:44945", + "issue:43232|issue:44188" + ], + "prepared_review_unit_hash": "041bd16f579a6bd46df2fdcf8130b016a7fce8aa61aff62f0f595eb909c5d990", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36874, + "estimated_input_tokens": 9219, + "estimated_eval_tokens": 18694 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a loose collection of mostly unrelated Transformers issues. Several share broad themes like generation, tokenizers, or model-family regressions, but none of the soft pairs look like the same concrete bug/change, so I reject all soft edges.", + "confidence": 0.86, + "canonical_issue_reason": "issue 43873 is the strongest representative issue here: it is detailed, active, has inbound references, and describes a concrete reproducible bug with broader impact.", + "canonical_pr_reason": null, + "best_issue_reason": "issue 43873 is the best standalone triage candidate in this set because it is well-documented, still open, and actionable.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43673", + "right": "issue:44734", + "accept": false, + "reason": "Both involve generation/KV-cache behavior, but they fail in different code paths: chunked-prefill cache handling vs serve continuation tensor indexing." + }, + { + "left": "issue:44530", + "right": "issue:46170", + "accept": false, + "reason": "Completely different problems: Qwen3.5 paged attention group parsing versus Windows/ROCm import and torch.distributed dependency issues." + }, + { + "left": "issue:43065", + "right": "issue:46123", + "accept": false, + "reason": "Different subsystems and symptoms: SAM3 pixel decoder dummy layer versus pipeline batching dropping final results." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Both are save_pretrained-related, but one is tokenizer metadata/class mismatch and the other is processor file emission; not the same bug." + }, + { + "left": "issue:44933", + "right": "issue:45459", + "accept": false, + "reason": "Different issues: a missing/nonexistent image_utils import versus tokenizer error handling that hides protobuf-related exceptions." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated failures: MLukeTokenizer task AttributeError versus a performance regression after upgrading Transformers." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Both mention token decoding/codec behavior, but they are different model-specific regressions with different failure modes." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Different areas entirely: trust_remote_code config field loss versus an ESM2 model breakage report." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Both are generation-related, but one is after sync_gpus bookkeeping and the other is missing generation cache in chunked prefill." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "No shared bug: missing processor save files versus optimizer stepping twice under deepspeed." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "A data-filtering slowdown regression is unrelated to the deepspeed double-step optimizer bug." + }, + { + "left": "issue:43701", + "right": "issue:44749", + "accept": false, + "reason": "Checkpoint key mismatch and upgrade-related filtering slowdown are different failure modes." + }, + { + "left": "issue:43673", + "right": "issue:43827", + "accept": false, + "reason": "Generation cache bug versus documentation still referencing removed pipeline APIs; one is runtime, the other docs." + }, + { + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "Gemma eos_token_id inconsistency is unrelated to AttentionInterface.register changing callback behavior." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Issue 45230 is too vague to match, and 45397 is a specific gemma-4 ZeRO3 from_pretrained failure." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Related to quantized models in a broad sense, but saving failure and GGUF loading failure are different concrete bugs." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "Different objects and code paths: SAM3 vision encoder output missing an attribute versus AutoConfig trust_remote_code dropping fields." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Both are vision/video-related, but SAM3 output attributes and Qwen2.5-VL vision_position_ids are unrelated bugs." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe loading failure and processor save_pretrained missing files do not describe the same issue." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "An immediate PyTorch ImportError is not the same as Sentence Transformers getting stuck during loading." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Attention mask construction for Gemma4 is unrelated to MaskGenerationPipeline's final-batch handling." + }, + { + "left": "issue:45406", + "right": "issue:46202", + "accept": false, + "reason": "Gemma4 processor initialization failure and Gemma pipeline roundtrip tokenization failure are different bugs." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API guidance is unrelated to remote-code checkpoint incompatibility with v5 meta initialization." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Both are import-related, but they concern different symbols and different failure surfaces; not the same underlying export bug." + }, + { + "left": "issue:43475", + "right": "issue:44661", + "accept": false, + "reason": "A missing SAM3 output attribute is unrelated to add-new-model-like failing on TOKENIZER_MAPPING_NAMES membership." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized saving failure and missing pad_token_id on Llama4Config are different model/config bugs." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but one is a specific missing config field and the other is a generic from_pretrained error; too little evidence they are the same bug." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Both produce AttributeError, but they are in unrelated components: tokenizer regex patching versus auto_docstring processing." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Both involve Qwen3.5-Moe, but one is a generic from_pretrained failure while the other is a specific vision-config field omission." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization/offloading behavior and pipeline-parallel output correctness are different code paths and symptoms." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Generation kwargs bookkeeping and attention kernel divergence under torch.compile are not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 38, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42673", + "issue:42831", + "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43504", + "issue:43526", + "issue:43673", + "issue:43704", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43976", + "issue:44038", + "issue:44079", + "issue:44188", + "issue:44222", + "issue:44242", + "issue:44351", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44792", + "issue:44869", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45103", + "issue:45137", + "issue:45200", + "issue:45310", + "issue:45356", + "issue:45561", + "issue:45712", + "issue:45753", + "issue:45812" + ], + "soft_pairs": [ + "issue:41628|issue:44351", + "issue:42673|issue:43704", + "issue:44964|issue:45137", + "issue:42994|issue:43824", + "issue:42831|issue:44222", + "issue:44945|issue:45753", + "issue:36010|issue:41628", + "issue:44222|issue:45310", + "issue:43232|issue:44792", + "issue:44038|issue:45310", + "issue:43673|issue:44188", + "issue:42175|issue:45712", + "issue:43526|issue:45561", + "issue:44496|issue:45561", + "issue:43504|issue:45561", + "issue:42175|issue:44933", + "issue:43377|issue:44242", + "issue:42898|issue:44568", + "issue:42222|issue:45200", + "issue:43976|issue:44933", + "issue:30990|issue:35141", + "issue:43278|issue:45137", + "issue:43479|issue:44869", + "issue:43479|issue:44079", + "issue:44623|issue:45310", + "issue:30333|issue:33290", + "issue:42898|issue:45812", + "issue:43784|issue:43824", + "issue:44734|issue:45103", + "issue:43723|issue:44568", + "issue:42898|issue:45356" + ], + "prepared_review_unit_hash": "bb784333617437e71ac4740af354522633a7e49c3c399bedd0e2a0d9dbfc7f2f", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36897, + "estimated_input_tokens": 9225, + "estimated_eval_tokens": 18706 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of unrelated Transformers bugs. The only clear duplicate pair is the repeated Qwen3ForCausalLM VRAM leak report; the rest are similar in subsystem at best, but describe different concrete failures.", + "confidence": 0.87, + "canonical_issue_reason": "issue:42673 is the earliest and fuller report of the exact same Qwen3ForCausalLM VRAM leak problem; issue:43704 repeats the same title later with less context.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the strongest representative because it is the original, detailed report for the only clear duplicate issue in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they reference different missing symbols (`AutoImageProcessor` vs `HybridCache`) and likely different export gaps." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same title and same underlying bug: Qwen3ForCausalLM leaks VRAM when used across multiple dataloader threads." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Model loading failure vs DeepSpeed ZeRO3 deque error; these are unrelated code paths." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Quantized model saving failure and a Qwen2.5-VL import error are different problems." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "Both mention FP8, but one is an accuracy regression and the other is a save_pretrained/MoE serialization issue." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline parallelism output corruption and a Qwen3.5 image infinite loop are different bugs." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Both are import failures, but for different missing APIs and likely different root causes." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "Different symptoms and model families: FP8 MoE save_pretrained vs Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation kwargs/cache handling and a janus image-generation test failure do not describe the same bug." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Related model family naming, but the reported failures are different and not the same concrete issue." + }, + { + "left": "issue:43673", + "right": "issue:44188", + "accept": false, + "reason": "Chunked prefill cache behavior and torch.compile attention-kernel branching are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Packaging/backend dependency issues and leftover dummy classes are different maintenance problems." + }, + { + "left": "issue:43526", + "right": "issue:45561", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and a pytest-xdist file race are unrelated." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Model config loading failure vs xdist race; no shared underlying bug." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Legacy BEiT preset loading and xdist capture-file race are different issues." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Torch extra/backend packaging and a missing image_utils import are not the same defect." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Missing padding-mask support in MIMI encoder and missing load-balancing loss when router logits are off are different code paths." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer cleanup behavior change and BOS/EOS insertion regression are distinct tokenizer bugs." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "VitPose breakage and Gemma 4 multimodal token type defaults are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python version compatibility and missing image_utils import are unrelated." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Sentence-Transformers loading hang and embedding resize reinitialization are different issues." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift in evaluation and a ZeRO3 deque IndexError are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Phi4 multimodal config defaults and Whisper timestamp decode crash are different bugs." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Config initialization bug and ModelOutput key assignment bug are separate problems." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Processor save_pretrained missing files and Qwen3.5 MoE from_pretrained error are different failures." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLFlow reporting after training failure and Adafactor OOM in DeepSpeed are unrelated." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different behaviors and different model families." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Sentence-Transformers import NameError and Qwen2.5-VL import error are unrelated import issues." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "KV-cache continuation indexing in serve and auto_docstring annotation handling are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer loading in v5 and BOS/EOS token insertion regression are distinct tokenizer issues." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "v5 cleanup behavior and Kimi-K2.5 codec handling are different tokenizer regressions." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: the only accepted duplicate pair is indeed issue:42673 / issue:43704, with 42673 being earlier and more detailed. The other soft-edge rejections are conservative and consistent with the titles and descriptions, and there are no PR mergeability concerns here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 39, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30333", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42898", + "issue:43122", + "issue:43232", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43761", + "issue:43874", + "issue:44060", + "issue:44190", + "issue:44206", + "issue:44222", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44373", + "issue:44448", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44623", + "issue:44779", + "issue:44933", + "issue:44977", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45245", + "issue:45310", + "issue:45405", + "issue:45406", + "issue:45656", + "issue:45712", + "issue:45834", + "issue:46082", + "issue:46132", + "issue:46209" + ], + "soft_pairs": [ + "issue:42898|issue:44779", + "issue:43874|issue:45092", + "issue:44060|issue:45125", + "issue:42898|issue:44448", + "issue:45405|issue:45712", + "issue:44295|issue:45656", + "issue:44448|issue:46132", + "issue:44933|issue:45712", + "issue:44222|issue:44297", + "issue:44297|issue:45656", + "issue:43479|issue:43644", + "issue:44190|issue:46082", + "issue:44222|issue:44623", + "issue:30333|issue:34689", + "issue:42831|issue:44623", + "issue:43122|issue:46209", + "issue:30333|issue:36331", + "issue:22355|issue:28282", + "issue:28282|issue:30333", + "issue:45245|issue:46132", + "issue:30333|issue:35141", + "issue:45071|issue:46082", + "issue:41628|issue:42757", + "issue:44206|issue:45103", + "issue:44297|issue:45310", + "issue:43723|issue:44977", + "issue:43232|issue:45834", + "issue:44263|issue:45092", + "issue:43761|issue:46082", + "issue:44222|issue:44484", + "issue:44295|issue:44623", + "issue:44373|issue:44485", + "issue:44483|issue:45406", + "issue:44514|issue:45103" + ], + "prepared_review_unit_hash": "6483c58bd454d63b558fbe46a04d83e6b4295b01ff7ee44825e202d348e814f2", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37646, + "estimated_input_tokens": 9412, + "estimated_eval_tokens": 19080 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are not a duplicate set; they span unrelated bugs and regressions across tokenization, model loading, multimodal processors, config validation, FP8, and tooling. The soft pairs mostly share only broad subsystem similarity, not the same underlying defect.", + "confidence": 0.17, + "canonical_issue_reason": "issue:42898 is the closest broad v5 tokenizer-regression candidate, but it still only represents one narrow symptom and does not unify the rest of the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42898 is the best single representative by breadth of nearby v5/tokenization reports, though the cluster is too heterogeneous to treat it as a real canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 tokenizer regressions, but one is clean_up_tokenization_spaces behavior and the other is Deepseek output drift; different code paths." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Missing image-patch method in an image processor is unrelated to remote-code checkpoint meta-initialization incompatibility." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Tied-weights warning in Qwen3-Next is a different defect from missing _tp_plan in Qwen3_5Moe." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Tokenization cleanup behavior change is not the same as Pegasus output differences across v4/v5." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "A version pin bump for PEFT is unrelated to leftover dummy class cleanup and dir() leakage." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "Reading a registered position_ids buffer is unrelated to deepspeed causing optimizer.step to run twice." + }, + { + "left": "issue:44448", + "right": "issue:46132", + "accept": false, + "reason": "Pegasus output drift and AttentionInterface.register behavior changes are separate code paths." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Nonexistent image_utils import and dummy_pt_objects cleanup are different import-layer issues." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "FP8 save_pretrained for MoE is not the same as tokenizer_class mismatch in tokenizer_config.json." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer save metadata mismatch and deepspeed double-step are unrelated failures." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Defaulting multimodal configs when None is unrelated to non-persistent buffers being filled with junk." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Local dataset loading in an example script is unrelated to LlamaConfig head_dim validation." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 save_pretrained on MoE and processor.save_pretrained missing files are different serialization bugs." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "MLFlow job status reporting and Llama 3.2 Vision model-loading breakage do not share the same bug." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues are unrelated to missing processor save_pretrained artifacts." + }, + { + "left": "issue:43122", + "right": "issue:46209", + "accept": false, + "reason": "Tokenizer behavior drift from 4.57.3 to 5.0 is unrelated to FP8 inference failure on B200." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "MLFlow reporting after failed training is not the same as compute_loss signature breakage." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Missing transformers.onnx and missing torch for AutoModel are both import issues, but they affect different modules and causes." + }, + { + "left": "issue:28282", + "right": "issue:30333", + "accept": false, + "reason": "PyTorch-not-found ImportError and MLFlow job reporting are unrelated." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "Category-count overflow is unrelated to AttentionInterface.register changing registered function behavior." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "MLFlow failure reporting and embedding reinitialization in post_init are different training/model-init bugs." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "PretrainedConfig type-checking regression is unrelated to LlamaConfig head_dim rejection." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "AutoImageProcessor import failure and is_offline_mode import failure come from different packages and symbols." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Unsupported center arg in a feature extractor is unrelated to auto_docstring crashing on future annotations." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer save metadata mismatch and Qwen3.5 MoE from_pretrained error are separate serialization/load failures." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading regression and Qwen3.5 flash-attention generation failure are different runtime paths." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to Kosmos2.5 OCR long-input indexing." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "torch.split indexing semantics in GlmMoeDsaIndexer do not match remote-code meta-init incompatibility." + }, + { + "left": "issue:43761", + "right": "issue:46082", + "accept": false, + "reason": "CLIPVision hidden_states regression and LlamaConfig head_dim validation are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 MoE save_pretrained and max_shard_size default policy are different concerns." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "position_ids buffer read errors are unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A docstring mistake for position_ids is not the same as a GLM-5 RoPE implementation discussion." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "chat/completions endpoint rejection is unrelated to Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:44514", + "right": "issue:45103", + "accept": false, + "reason": "Batched apply_chat_template crash and auto_docstring future-annotations crash are different processor/documentation code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 40, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:22355", + "issue:28218", + "issue:30990", + "issue:33357", + "issue:38175", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43122", + "issue:43232", + "issue:43479", + "issue:43576", + "issue:43646", + "issue:43723", + "issue:43756", + "issue:43784", + "issue:43824", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44360", + "issue:44373", + "issue:44442", + "issue:44521", + "issue:44568", + "issue:44671", + "issue:44704", + "issue:44829", + "issue:44871", + "issue:44913", + "issue:44991", + "issue:45092", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45593", + "issue:45753", + "issue:46002", + "issue:46050", + "issue:46132" + ], + "soft_pairs": [ + "issue:44263|issue:44360", + "issue:45092|issue:45357", + "issue:43723|issue:44871", + "issue:30990|issue:38175", + "issue:42907|issue:46050", + "issue:44261|issue:45092", + "issue:43723|issue:44991", + "issue:44829|issue:46002", + "issue:43723|issue:44442", + "issue:43479|issue:44913", + "issue:42831|issue:45310", + "issue:41628|issue:43824", + "issue:44263|issue:44373", + "issue:42831|issue:44038", + "issue:39401|issue:44568", + "issue:42757|issue:45278", + "issue:43232|issue:45593", + "issue:42886|issue:42994", + "issue:42886|issue:43066", + "issue:22355|issue:28218", + "issue:43756|issue:44060", + "issue:42831|issue:45753", + "issue:44117|issue:44704", + "issue:44671|issue:46132", + "issue:43576|issue:43784", + "issue:43646|issue:46002", + "issue:43122|issue:44521", + "issue:42371|issue:44263", + "issue:30990|issue:33357" + ], + "prepared_review_unit_hash": "c898a99c8bc5cf522786cc95a8ea7193c11748fb8e4c6e65cd90486292feb846", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35928, + "estimated_input_tokens": 8982, + "estimated_eval_tokens": 18220 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a mix of unrelated Transformers v5 regressions; none of the soft pairs look like true duplicates of the same underlying bug/change, so all soft edges are rejected. The most representative issue is the broad tokenizer-loading regression in issue 43723.", + "confidence": 0.62, + "canonical_issue_reason": "issue:43723 is the broadest, most central tokenizer-loading regression in this set and overlaps conceptually with several nearby tokenizer-related reports, making it the best canonical issue candidate.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43723 best captures the shared theme of v5 tokenizer/from_pretrained breakage without being too model-specific.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Same DSA/GLM-MoE area, but the titles indicate different defects: split return handling vs missing ReLU." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Both are multimodal checkpoint/config issues, but one is meta-init incompatibility and the other is incorrect visual-encoder save keys." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Both involve tokenizer/model loading in v5, but one is an AutoTokenizer load failure and the other is an eos_token_id config mismatch." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Unrelated model-loading stall vs zero-probability inference bug." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Both touch quantized save/load, but the concrete failure modes differ: dequantized Ministrals/Devstrals saving vs sinq round-trip corruption." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "Different problems: RMS norm epsilon precision mismatch vs remote-code checkpoint meta-init incompatibility." + }, + { + "left": "issue:43723", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-load regressions, but the affected code paths and model-specific failures are not shown to be the same." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Different training pathologies: flash_attention_3 degeneration vs NaNs in classification heads after checkout/init." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "Both are AutoTokenizer load issues, but they appear model-specific and there is no evidence they share the same root cause." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Different configuration bugs: multimodal sub-config initialization vs rotary_pct not persisting on reload." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "Accuracy problem in FineGrainedFP8 vs a Qwen3.5 MoE from_pretrained error; unrelated failures." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and likely different export paths." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "Different issue types: runtime tensor indexing behavior vs a docstring correction." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "Separate bugs in different areas: FP8 accuracy vs Qwen3-VL-Moe loading." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Different tokenizer regressions: offset_mapping vs add_special_tokens/BOS-EOS handling." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "Related only at a very broad 'import error' level; the specific missing symbol and scope are different." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation-state update bug vs loss-computation bug in D-FINE; unrelated." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Cache/offline tokenizer loading bug vs quantized model saving failure." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Both tokenizer-related, but one is offline cache loading and the other is a decoder-type mismatch in v5." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "Completely different areas: missing onnx module import vs tokenizer spacing behavior." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Different model families and different logic errors: RoPE layer dropout vs tied-weights warning behavior." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "Unrelated: FP8 accuracy issue vs Qwen3_5 infinite-loop generation on a specific image." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "Both are from_pretrained plumbing issues, but one is tokenizer mapping resolution and the other is cached_file kwargs forwarding." + }, + { + "left": "issue:44671", + "right": "issue:46132", + "accept": false, + "reason": "CamemBERT MLM regression and AttentionInterface.register behavior change are different bugs." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Different failures: env CLI breakage vs import-time NameError in sentence-transformers integration." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Custom model initialization breakage is not the same as NaNs in classification heads." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer output drift vs multimodal assistant-mask failure; not the same code path." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 settings warning/behavior issue vs a tensor split/indexer bug." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Different model/runtime failures: sentence-transformers load hang vs MacOS bus error with a CLIP model." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 41, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:22355", + "issue:28218", + "issue:29159", + "issue:30990", + "issue:36296", + "issue:39401", + "issue:41628", + "issue:42757", + "issue:42907", + "issue:43116", + "issue:43122", + "issue:43232", + "issue:43278", + "issue:43408", + "issue:43452", + "issue:43475", + "issue:43784", + "issue:43824", + "issue:43906", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44188", + "issue:44222", + "issue:44261", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44360", + "issue:44373", + "issue:44485", + "issue:44488", + "issue:44514", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44743", + "issue:44779", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45290", + "issue:45446", + "issue:45478", + "issue:45600" + ], + "soft_pairs": [ + "issue:44295|issue:44297", + "issue:44704|issue:45290", + "issue:44188|issue:44373", + "issue:42907|issue:44945", + "issue:43232|issue:44743", + "issue:43784|issue:45600", + "issue:44261|issue:44373", + "issue:39401|issue:44488", + "issue:41628|issue:45278", + "issue:43976|issue:44945", + "issue:43122|issue:44514", + "issue:43408|issue:43475", + "issue:42907|issue:43937", + "issue:39401|issue:44521", + "issue:43116|issue:45200", + "issue:44625|issue:44829", + "issue:42907|issue:44964", + "issue:44360|issue:44485", + "issue:44038|issue:44297", + "issue:44038|issue:44295", + "issue:28218|issue:30990", + "issue:43122|issue:44779", + "issue:36296|issue:43116", + "issue:42907|issue:43452", + "issue:44038|issue:44222", + "issue:43824|issue:44351", + "issue:43122|issue:43906", + "issue:45216|issue:45478", + "issue:43784|issue:45446", + "issue:42757|issue:44351", + "issue:22355|issue:29159", + "issue:43278|issue:44945" + ], + "prepared_review_unit_hash": "1fba7439c5fa4cd99654846ff334152de6aee76fb043526134c75173af5fb137", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36902, + "estimated_input_tokens": 9226, + "estimated_eval_tokens": 18708 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No duplicate cluster emerges here: the items span unrelated tokenizer, import, model-loading, and training bugs. I rejected every soft edge as a true duplicate match; the only strong standalone issue is the open multimodal assistant_mask bug.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 44521 is the clearest current bug report: open, detailed, reproducible, and already attracting references, so it is the best anchor issue among this otherwise unrelated set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the most actionable and representative standalone issue in the group; the others are either narrower, already closed, or clearly on different code paths.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Different bugs: one is about reading a bufferized position_ids tensor, the other about tokenizer_config metadata written by save_pretrained." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "Different code paths: cached_file kwargs passthrough vs apply_chat_template crashing on tool-call assistant messages." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "One is an attention-kernel branching issue under torch.compile; the other is just a docstring problem for position_ids." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Unrelated failures: dequantized model saving vs incorrect output under pipeline parallelism." + }, + { + "left": "issue:43232", + "right": "issue:44743", + "accept": false, + "reason": "Both are generation-state bugs, but the symptoms and affected logic differ: sync_gpus kwargs handling vs recurrent state reset in modular_qwen3_5." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Different regressions: a missing nn import in sentence-transformers integration vs stale auto_mappings references to removed configs." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "Model numerical precision/config issue vs a docstring mismatch; not the same underlying bug." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Both touch tokenization, but they concern different models and different failures (offset_mapping vs load failure)." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "The second is a broad upgrade regression with many import errors; the first is a specific missing export. Too broad to treat as the same bug." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Different problem domains: Python-version compatibility vs incorrect outputs from pipeline parallelism." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Both involve tokenization/chat handling, but one is a version-to-version tokenization regression and the other is an apply_chat_template crash on batched multimodal input." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Same model family, but different concrete bugs: model-type warning vs missing fpn_position_embeddings attribute." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Different Qwen/Mistral-related failures: saving dequantized models vs invalid GenerationConfig for GLM-5." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Both are tokenizer/template issues, but one is Qwen3 offset_mapping and the other is assistant_masks being all-zero for multimodal inputs." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Different tasks and symptoms: run_classification.py multi-label inference vs Gemma 4 fine-tuning token type defaults." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "Different bugs: num_labels propagation in Qwen3.5 config vs degenerate training with flash_attention_3." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Both are model-loading failures, but for different models and different causes; not the same code-path problem." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "Completely different topics: DSA indexer activation vs GLM-5 RoPE implementation." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "One is a Qwen3-VL-Moe bug; the other is tokenizer_config metadata mismatch. Different root causes." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Different failures: Qwen3-VL-Moe bug vs reading position_ids after buffer registration." + }, + { + "left": "issue:28218", + "right": "issue:30990", + "accept": false, + "reason": "Tokenizer spacing issue vs Sentence Transformers loading hang; no shared underlying bug." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different behaviors (general tokenization drift vs DeepSeek incorrect results)." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Different training bugs in unrelated workflows: tensor parallel training vs example-script multi-label classification output." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Different loading/saving paths and different model APIs; not the same concrete bug." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Different MoE-related issues: Qwen3-VL-Moe behavior vs FP8 save_pretrained handling." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols from different packages and likely different fixes." + }, + { + "left": "issue:43122", + "right": "issue:43906", + "accept": false, + "reason": "The second is just an isolated reproduction of another issue; the titles do not indicate the same concrete failure as the tokenization regression." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Related Qwen3.5 family, but one is save_pretrained checkpoint correctness and the other is from_pretrained loading error; too different to merge as one bug." + }, + { + "left": "issue:43784", + "right": "issue:45446", + "accept": false, + "reason": "Sentence-transformers import failure vs a PyTorch version-check bug in flex_attention; unrelated code paths." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but one is from huggingface_hub and the other is a transformers symbol export; different dependencies and fixes." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "Module import failure for transformers.onnx vs slow/fast tokenizer inconsistency; unrelated issues." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Different runtime problems: dtype drift between train/eval and incorrect outputs under pipeline parallelism." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 42, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:33453", + "issue:38617", + "issue:39401", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:43064", + "issue:43066", + "issue:43116", + "issue:43441", + "issue:43450", + "issue:43493", + "issue:43611", + "issue:43673", + "issue:43825", + "issue:43827", + "issue:43874", + "issue:43906", + "issue:43937", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44291", + "issue:44380", + "issue:44485", + "issue:44704", + "issue:44779", + "issue:44936", + "issue:44938", + "issue:44977", + "issue:45071", + "issue:45072", + "issue:45083", + "issue:45278", + "issue:45357", + "issue:45446", + "issue:45478", + "issue:45563", + "issue:45588", + "issue:45753", + "issue:46002", + "issue:46077" + ], + "soft_pairs": [ + "issue:42175|issue:44938", + "issue:42673|issue:45072", + "issue:43992|issue:44704", + "issue:43493|issue:45753", + "issue:43066|issue:44779", + "issue:22355|issue:29127", + "issue:38617|issue:45278", + "issue:43611|issue:46002", + "issue:43937|issue:44485", + "issue:42222|issue:43441", + "issue:43976|issue:45446", + "issue:43116|issue:43493", + "issue:42371|issue:43874", + "issue:44075|issue:44936", + "issue:44038|issue:45478", + "issue:45357|issue:45478", + "issue:43064|issue:43673", + "issue:43116|issue:43825", + "issue:45083|issue:45446", + "issue:44380|issue:46077", + "issue:43937|issue:45563", + "issue:44117|issue:45588", + "issue:33453|issue:39401", + "issue:39401|issue:44779", + "issue:43827|issue:45446", + "issue:42673|issue:43906", + "issue:44291|issue:45071", + "issue:43450|issue:43986", + "issue:43937|issue:44977" + ], + "prepared_review_unit_hash": "3f164a69579ba898b28382b9a8803a66a051bc6ffe672328c413546302636b0b", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36040, + "estimated_input_tokens": 9010, + "estimated_eval_tokens": 18276 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a grab-bag of Transformers v5 regression reports across imports, tokenizers, model-specific bugs, docs, and training/runtime issues. The soft pairs mostly share broad vocabulary but not the same concrete code path, so I would reject all as duplicates; the broadest umbrella issue is the v5 upgrade/import-regression report.", + "confidence": 0.24, + "canonical_issue_reason": "issue:45278 is the broadest umbrella in the set: it describes multiple import failures after upgrading to Transformers 5.5.0, which is the closest match to the recurring v5 breakage pattern across the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best representative because it is high-level and can subsume several v5 import/load regressions, unlike the narrower model- or tokenizer-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Different failures: missing TensorFlow backend in an install extra vs Python 3.14 load failure. Same release context, not the same bug." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM multi-threaded dataloading is unrelated to bfloat16 dtype mismatches in CI." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Missing embed_tokens.weight during UMT5 loading is a model-specific checkpoint issue, not the same as AutoProcessor kwargs forwarding." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "Both concern Qwen-related behavior, but one is a SigLIP2 implementation discrepancy and the other is a Qwen3_5 generation loop on a specific image." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Wrong tokenizer decoder type in v5 is not the same as Deepseek tokenizer output regressions; different tokenizer code paths." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "Import error for transformers.onnx is unrelated to a LayoutLMv3 box-validation error message." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "The former is a specific missing symbol import; the latter is a broad multi-import regression report. Not the same concrete failure." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "Base_model_prefix loading breakage is a model-loading compatibility bug, while NaNs in classification heads are an initialization/numerics issue." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "GenerationConfig validation for GLM-5 is not the same as a RoPE implementation discussion." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are distinct model-specific bugs." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python version compatibility regression is unrelated to a flex_attention version-check bug." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Classification script returning empty multi-label results is unrelated to SigLIP2 HF/JAX discrepancy." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API warning changes and missing get_number_of_image_patches are different issues in different subsystems." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "SGD optimizer arguments not used is not the same as trainer.evaluate() failing after train()." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL-Moe compatibility and Qwen3.5 Moe from_pretrained error are related only by model family, not the same code-path bug." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "Incorrect visual encoder keys on save_pretrained is a save bug; from_pretrained loading failure is a separate load bug." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "Wrong optimizer state/params in FSDP2+PEFT training is unrelated to missing GenerationMixin cache in chunked prefill." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Empty multi-label classification results are not the same as the pipeline() translation-support error message problem." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Feature-extract output length behavior in qwen3_omni_moe is unrelated to the flex_attention PyTorch version check." + }, + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "GPT2 attention scaling under SDPA/FlashAttention and Gemma4 bidirectional attention masking are different attention bugs." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "GLM-5 generation config invalidity is not the same as a paged generate warning for num_return_sequences." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None is unrelated to flash_attention.py crashing on s_aux=None." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Tokenizer loading regression and Qwen3 offset_mapping bug are different tokenizer failures with different symptoms." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer bugs, but offset_mapping corruption and Deepseek incorrect results are not the same underlying defect." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs still referencing pipeline() and a flex_attention version check are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "Multi-thread VRAM leak and an isolated reproduction of another issue are not the same bug." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument during init_empty_weights is unrelated to PretrainedConfig type-checking breakage." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video processor shape bug and missing torchvision crash on AutoProcessor loading are different failures." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "GenerationConfig validation for GLM-5 is unrelated to Qwen3.5 flash-attention generation issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 43, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42490", + "issue:42617", + "issue:42673", + "issue:42831", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43408", + "issue:43421", + "issue:43577", + "issue:43582", + "issue:43673", + "issue:43742", + "issue:43937", + "issue:44075", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44410", + "issue:44466", + "issue:44483", + "issue:44484", + "issue:44493", + "issue:44496", + "issue:44829", + "issue:44933", + "issue:45071", + "issue:45081", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45405", + "issue:45446", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45561", + "issue:45656", + "issue:46002", + "issue:46178" + ], + "soft_pairs": [ + "issue:44466|issue:46002", + "issue:45405|issue:45446", + "issue:44117|issue:45081", + "issue:45341|issue:45561", + "issue:42222|issue:46178", + "issue:44410|issue:45310", + "issue:38617|issue:41628", + "issue:44829|issue:45216", + "issue:42617|issue:45341", + "issue:42673|issue:43582", + "issue:39401|issue:43421", + "issue:42490|issue:42673", + "issue:43408|issue:43742", + "issue:42371|issue:44261", + "issue:38617|issue:44220", + "issue:44484|issue:45310", + "issue:44933|issue:45278", + "issue:43937|issue:44496", + "issue:42222|issue:45542", + "issue:43066|issue:43122", + "issue:42371|issue:42831", + "issue:44483|issue:45464", + "issue:42175|issue:42222", + "issue:44075|issue:45656", + "issue:43012|issue:43577", + "issue:44410|issue:45478", + "issue:43673|issue:45341", + "issue:43097|issue:44493", + "issue:41628|issue:45071" + ], + "prepared_review_unit_hash": "a8b87e2b8d342371f640ea493fa1750468c62184483b4456d5789d26a8c231d2", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36144, + "estimated_input_tokens": 9036, + "estimated_eval_tokens": 18328 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose collection of Transformers v5 regressions and API/import/model-loading complaints, but most items are not duplicates of one another. The soft edges are mostly false positives from broad lexical similarity rather than the same concrete bug.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster\u2019s recurring API/import regression theme, making it the best single issue anchor among otherwise heterogeneous reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 best fits as a cluster representative because it captures the widest user-facing regression pattern (post-upgrade import failures) seen across several nearby issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "Different bugs: lm_head serialization inconsistency vs NaNs in classification heads after checkout/init." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "Unrelated: a PEFT version bump issue vs a PyTorch version check bug in flex_attention." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Both tokenizer-related, but one is a TOKENIZER_MAPPING_NAMES None-handling bug and the other is a Mistral regex patch crash." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "Both mention testing utils, but one is a vague general bug and the other is a specific pytest-xdist race on captured_info.txt." + }, + { + "left": "issue:42222", + "right": "issue:46178", + "accept": false, + "reason": "Different model/backend failures: vitpose breakage vs missing tensorflow_text in BACKENDS_MAPPING." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen variants and different failure modes: missing projections vs from_pretrained error." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols and different code paths." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "Training degeneration under flash_attention_3 is not the same as incorrect save_pretrained checkpoint serialization." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "Unrelated: a runtime script failure vs a testing_utils bug." + }, + { + "left": "issue:42673", + "right": "issue:43582", + "accept": false, + "reason": "Different issues entirely: VRAM leak in dataloader threads vs Apple Silicon TypeError in allocator warmup." + }, + { + "left": "issue:39401", + "right": "issue:43421", + "accept": false, + "reason": "Both tokenizer-related, but offset_mapping is a different problem from post-processor updates after runtime special-token changes." + }, + { + "left": "issue:42490", + "right": "issue:42673", + "accept": false, + "reason": "Local-path import warning is unrelated to the Qwen3ForCausalLM VRAM leak." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "Different model-loading problems: wrong model type warning vs a key error for MobileLLM." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 behavior settings and RMS norm epsilon precision are unrelated issues." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "Missing configuration_utils import and _torch_extract_fbank_features are different code-path failures." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "A shard-size question is unrelated to the Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "Both concern imports, but one is a specific missing image_utils export and the other is a broad upgrade-wide import regression; not enough to treat as the same bug." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Different model/config failures: invalid GenerationConfig vs missing model_type in config.json." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "Vitpose breakage is unrelated to the missing TensorFlow backend when only tensorboard is installed." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Related tokenization regression area, but one is decoder type selection and the other is broader tokenization differences across versions." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 API usage and FineGrainedFP8 accuracy are distinct issues." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Different API failures: generic /v1/chat/completions rejection vs a Qwen3.5 streaming inference failure." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "Backend installation packaging issue vs vitpose model breakage are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "Optimizer args being ignored is different from optimizer.step being called twice with deepspeed." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both dtype-related, but one is a compile-time warning and the other is incorrect dtype propagation on load." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen-related, but they describe different model families and different from_pretrained failures." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Chunked_prefill GenerationMixin cache regression is unrelated to the testing_utils bug." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removed tie_embeddings_and_encoder_decoder and unexpected position-id keys are different regressions." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "Different regressions: missing AutoImageProcessor export vs PretrainedConfig type-checking breakage." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 44, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:38617", + "issue:42757", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43335", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43502", + "issue:43531", + "issue:43673", + "issue:43824", + "issue:43873", + "issue:43874", + "issue:43906", + "issue:43931", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44368", + "issue:44410", + "issue:44484", + "issue:44485", + "issue:44704", + "issue:44829", + "issue:44987", + "issue:45003", + "issue:45071", + "issue:45103", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45542", + "issue:45588", + "issue:46050", + "issue:46123", + "issue:46172" + ], + "soft_pairs": [ + "issue:43874|issue:44261", + "issue:44263|issue:44485", + "issue:45071|issue:45542", + "issue:44263|issue:46172", + "issue:44360|issue:46172", + "issue:43012|issue:43408", + "issue:43441|issue:44829", + "issue:44484|issue:45478", + "issue:43931|issue:44410", + "issue:43335|issue:45071", + "issue:43931|issue:44038", + "issue:43673|issue:46123", + "issue:43097|issue:43452", + "issue:45003|issue:45341", + "issue:43986|issue:45341", + "issue:44117|issue:44987", + "issue:43986|issue:44484", + "issue:43874|issue:44263", + "issue:43408|issue:44368", + "issue:43986|issue:45372", + "issue:43531|issue:45341", + "issue:45440|issue:46172", + "issue:44261|issue:44485", + "issue:43097|issue:46050", + "issue:43064|issue:44315", + "issue:43097|issue:45588", + "issue:38617|issue:42757", + "issue:44062|issue:44117", + "issue:44410|issue:45375", + "issue:44038|issue:44410", + "issue:38617|issue:44351", + "issue:43502|issue:43992", + "issue:43097|issue:43873", + "issue:44704|issue:45103", + "issue:43986|issue:44117", + "issue:43824|issue:43986", + "issue:43066|issue:43906" + ], + "prepared_review_unit_hash": "23b0b8c37900f10c7d24cca0aef44a14d3b005b23c4a9753ec0a824f772aa960", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 38784, + "estimated_input_tokens": 9696, + "estimated_eval_tokens": 19648 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated transformer issues with superficial overlap in model names, loading paths, or dependency/import errors. None of the soft-edge pairs look like the same underlying bug or fix, so I would not merge any of them; issue 45071 is the best representative because it is a broad core-regression report.", + "confidence": 0.95, + "canonical_issue_reason": "Broad, cross-cutting regression in core config/type checking; more central than the model-specific and dependency-specific reports in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "Impacts a core code path and is likely to be the most generally useful anchor for triage among these mostly narrow issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Different models and failures: missing image-patch method vs LayerNorm epsilon precision issue." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "Unrelated GLM-MoE indexing behavior vs GLM-5 RoPE implementation." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Core config type-checking regression vs TensorFlow backend detection error." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "Different bugs in different model families and tensor math paths." + }, + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "DSA ReLU complaint is unrelated to deepseek-v4 transpose behavior." + }, + { + "left": "issue:43012", + "right": "issue:43408", + "accept": false, + "reason": "Precision warning during bfloat16 compilation vs tokenizer/model type mismatch warning." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "FlashAttention inference failure for Ministral-3 vs flash_attention_3 training degeneration." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "save_pretrained shard-size question is unrelated to Qwen3.5-Moe loading failure." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch vs Qwen3next missing projections." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "SwitchTransformers config edge case vs global PreTrainedConfig type-check regression." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Different Qwen multimodal loading problems with different symptoms and code paths." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Generation cache/chunked prefill issue vs pipeline partial-batch result dropping." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "Removed tie_embeddings API warning is unrelated to gguf_file loading breakage." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "sys.modules access bug in modeling_utils vs a testing_utils bug." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "Missing torchvision crash vs an unrelated testing utility bug." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling vs loading failure for a specific model repo." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Video-model processor dependency crash is unrelated to save_pretrained shard sizing." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Different GLM-related failures: image processor method missing vs tensor split return handling." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "Model-type warning mismatch vs tie_word_embeddings warning during Qwen3.5 LoRA." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Different missing-dependency/import issues affecting different processor paths." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Qwen3-MoE sliding window behavior is unrelated to a testing utility bug." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeekV3 MoE divergence vs deepseek-v4 comb transpose bug." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "LayerNorm epsilon precision issue is unrelated to RoPE implementation." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Removed API warning is unrelated to quantized model save/load corruption." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Distributed optimizer-state bug vs Liger Kernel not applied in model_init." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Tie-embeddings deprecation/removal warning is unrelated to flash_attention.py crash on sink-less models." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different missing symbols from different modules." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "AddedToken keyword collision is unrelated to tokenizer mapping None handling." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Different Qwen-family issues: missing projections vs missing config field being dropped." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL-Moe loading bug is unrelated to qwen3next missing attention projections." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Distinct import errors from different modules and symbols." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "local_files_only still making requests vs missing embed_tokens weight on UMT5Encoder." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Embedding-tying removal warning is unrelated to quantization offloading behavior." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "AutoProcessor kwargs passthrough bug vs auto_docstring annotation crash." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Missing torchvision processor crash vs tokenizer mapping assumption bug." + }, + { + "left": "issue:43824", + "right": "issue:43986", + "accept": false, + "reason": "Qwen2.5-VL import error is unrelated to video-model loading without torchvision." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Tokenizer decoder type issue is unrelated to the reproduced issue referenced by 38071." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 45, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42490", + "issue:42757", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43502", + "issue:43526", + "issue:43550", + "issue:43582", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43792", + "issue:43824", + "issue:43825", + "issue:43867", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44162", + "issue:44222", + "issue:44230", + "issue:44368", + "issue:44380", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44655", + "issue:44683", + "issue:44938", + "issue:45290", + "issue:45478", + "issue:45479", + "issue:45698", + "issue:45753", + "issue:46123", + "issue:46132", + "issue:46172", + "issue:46178", + "issue:46209" + ], + "soft_pairs": [ + "issue:44938|issue:46178", + "issue:43208|issue:43550", + "issue:43992|issue:44484", + "issue:43825|issue:44655", + "issue:43012|issue:43550", + "issue:43012|issue:44368", + "issue:42907|issue:44655", + "issue:43352|issue:44380", + "issue:42907|issue:45753", + "issue:44162|issue:46172", + "issue:43352|issue:43976", + "issue:44380|issue:46132", + "issue:43792|issue:46123", + "issue:44380|issue:45479", + "issue:43064|issue:44380", + "issue:44483|issue:44514", + "issue:43208|issue:43526", + "issue:44222|issue:45478", + "issue:36683|issue:43582", + "issue:43012|issue:44683", + "issue:43618|issue:43761", + "issue:42490|issue:43012", + "issue:42757|issue:43824", + "issue:43618|issue:44380", + "issue:45698|issue:46172", + "issue:38617|issue:43502", + "issue:44038|issue:44230", + "issue:44483|issue:45290", + "issue:43756|issue:44485", + "issue:43066|issue:46209", + "issue:43299|issue:43867", + "issue:36683|issue:43475" + ], + "prepared_review_unit_hash": "629f81c964ea2c00b47db2b4ed31927c8aec16bfeedb159c16d59a7dad6f42f8", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36944, + "estimated_input_tokens": 9236, + "estimated_eval_tokens": 18728 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of unrelated Transformers bug reports around v5 regressions, loading/saving, imports, and model-specific failures. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates.", + "confidence": 0.56, + "canonical_issue_reason": "issue:45698 is the broadest and most active open report here, centered on from_pretrained/save_pretrained custom-module loading behavior, which is the closest thing to a representative theme across the mostly serialization/loading-adjacent items in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is the best global issue candidate because it is open, has substantial discussion and inbound references, and covers a concrete user-facing loading/saving regression rather than a model-specific edge case.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44938", + "right": "issue:46178", + "accept": false, + "reason": "Python 3.14 load failure and invalid backend-name error are unrelated import/dispatch problems." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training bugs and Bamba compile/SDPA failure affect different models and code paths." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing embed_tokens.weight on from_pretrained is unrelated to a max_shard_size question." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Incorrect pipeline task error messaging is different from saving Pipeline objects." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "A bf16 compile warning is not the same as a model-specific torch.compile SDPA failure." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "General bf16 compile warning and tie_word_embeddings warning are different symptoms with different causes." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Both are save-related, but dequantized model saving and Pipeline.save_pretrained are distinct code paths." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Nemotron Flash Attention 2 support and GPT2 attention scaling under SDPA/FA are different bugs." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "A dequantized-save failure is unrelated to a Qwen3.5 image-induced infinite loop." + }, + { + "left": "issue:44162", + "right": "issue:46172", + "accept": false, + "reason": "ESM2 loading breakage and a DeepSeek tensor-op bug are different model-specific issues." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Flash Attention support warning and Python version incompatibility are unrelated." + }, + { + "left": "issue:44380", + "right": "issue:46132", + "accept": false, + "reason": "Attention scaling being ignored is not the same as AttentionInterface.register changing registered function behavior." + }, + { + "left": "issue:43792", + "right": "issue:46123", + "accept": false, + "reason": "Whisper runtime failure and partial-batch pipeline result dropping are different pipeline bugs." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention backend scaling and degenerate zero loss classification are unrelated regressions." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "FSDP2 optimizer-state corruption and GPT2 attention-scaling bugs do not share a concrete code path." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Chat-completions request rejection and chat-template batching crash are both chat-related but distinct failures." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "xLSTM training issues and BeitImageProcessorFast reduce_labels behavior are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "FP8 save_pretrained failure and Qwen3.5 MoE from_pretrained failure are different operations and symptoms." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "Missing Gemma3Config vocab_size and Apple Silicon TypeError are unrelated." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "A precision warning is not the same as compiled flex_attention failing on torch >= 2.9." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Both affect CLIP, but one drops attentions while the other drops hidden_states; not the same bug." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "Local-model-path slash warning and bf16 compile warning are unrelated." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different failure modes." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIP attentions regression and GPT2 attention-scaling regression are separate model-specific bugs." + }, + { + "left": "issue:45698", + "right": "issue:46172", + "accept": false, + "reason": "Custom module loading after save_pretrained and a DeepSeek tensor transform error are unrelated." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "ImportError for layer_type_validation and API calls despite local_files_only=True are different problems." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Qwen3-VL-Moe loading breakage and FP8 inference support are related only by model family, not by the same bug." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Chat API request handling and apply_chat_template tool-call crash are different layers of the stack." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Smollm3 RoPE-layer omission and GLM-5 RoPE implementation discussion are not the same underlying issue." + }, + { + "left": "issue:43066", + "right": "issue:46209", + "accept": false, + "reason": "Wrong tokenizer decoder type and FP8 inference on B200 are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Qwen3VL-Moe loading regression and sorted-state_dict load error are distinct loading failures." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Gemma3Config vocab_size missing and SAM3 output missing fpn_position_embeddings are unrelated attribute errors." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 46, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:36010", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42490", + "issue:42757", + "issue:42890", + "issue:42994", + "issue:43010", + "issue:43064", + "issue:43299", + "issue:43352", + "issue:43525", + "issue:43575", + "issue:43611", + "issue:43618", + "issue:43867", + "issue:43901", + "issue:43931", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44075", + "issue:44077", + "issue:44242", + "issue:44246", + "issue:44355", + "issue:44380", + "issue:44492", + "issue:44625", + "issue:44655", + "issue:44855", + "issue:44938", + "issue:45071", + "issue:45216", + "issue:45276", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45478", + "issue:45561", + "issue:45593", + "issue:45753", + "issue:45850", + "issue:46077", + "issue:46082", + "issue:46202" + ], + "soft_pairs": [ + "issue:45276|issue:46077", + "issue:42994|issue:44655", + "issue:43901|issue:44655", + "issue:43299|issue:45753", + "issue:44077|issue:44380", + "issue:36010|issue:42757", + "issue:36683|issue:44077", + "issue:44625|issue:46077", + "issue:43611|issue:44655", + "issue:43352|issue:43937", + "issue:43867|issue:45216", + "issue:46077|issue:46082", + "issue:43976|issue:44938", + "issue:36683|issue:43525", + "issue:43352|issue:45593", + "issue:43618|issue:45216", + "issue:36683|issue:44242", + "issue:43352|issue:43931", + "issue:43064|issue:44492", + "issue:43010|issue:45278", + "issue:43992|issue:45310", + "issue:43867|issue:46202", + "issue:44355|issue:44855", + "issue:43867|issue:45375", + "issue:41669|issue:44246", + "issue:36683|issue:43064", + "issue:44938|issue:45071", + "issue:42890|issue:45561", + "issue:43010|issue:45850", + "issue:43867|issue:45357", + "issue:42490|issue:44075", + "issue:43992|issue:45478", + "issue:43618|issue:44077", + "issue:43010|issue:43976", + "issue:36296|issue:43575" + ], + "prepared_review_unit_hash": "c6c7548030a3345045d0e95195321b04f619e71b7fc26e4592b9c6e0c4a20acf", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37687, + "estimated_input_tokens": 9422, + "estimated_eval_tokens": 19100 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose collection of transformer v5 regression reports across import, config, loading, and save/pretrained paths. The only plausible duplicate edge is the import-time slowdown pair; the rest are distinct model- or bug-specific issues and should stay separate.", + "confidence": 0.64, + "canonical_issue_reason": "issue:45278 is the broadest and most representative issue in the cluster: it captures the general post-upgrade import-regression theme ('many import errors after update from 4.57.0 to 5.5.0') that several other items orbit around.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best single issue to represent the cluster because it is broad, actionable, and closer to the shared version-regression theme than the many model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Different Gemma4 bugs: token embedding resize propagation vs causal attention-mask construction." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "One is quantized-model saving failure; the other is pipeline save_pretrained serialization." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Docs mismatch about return_all_scores is not the same bug as pipeline save_pretrained failing." + }, + { + "left": "issue:43299", + "right": "issue:45753", + "accept": false, + "reason": "Different Qwen3VL issues: MoE model loading regression vs an image-triggered infinite loop." + }, + { + "left": "issue:44077", + "right": "issue:44380", + "accept": false, + "reason": "Unrelated: patchtsmixer post_init validation versus GPT2 attention scaling in SDPA/FlashAttention." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols from different packages and different fixes." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "Gemma3 missing vocab_size and patchtsmixer post_init are unrelated config issues." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and Gemma4 attention-mask behavior are different model bugs." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Base_model_prefix loading regression and pipeline save_pretrained are different code paths." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Flash Attention 2 support gate and invalid GenerationConfig are unrelated failures." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "State_dict ordering load error versus incorrect Qwen3.5 checkpoint saving are different serialization bugs." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Gemma4 attention-mask bug and LlamaConfig head_dim validation are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both mention Python compatibility, but they are different version-specific failures with different causes." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different missing config attributes in different model families (Gemma3 vs Llama4)." + }, + { + "left": "issue:43352", + "right": "issue:45593", + "accept": false, + "reason": "Flash Attention support for Nemotron and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIPOutput attentions assignment and Qwen3.5 save_pretrained corruption are separate bugs." + }, + { + "left": "issue:36683", + "right": "issue:44242", + "accept": false, + "reason": "Missing Gemma3Config vocab_size is unrelated to MoE load-balancing loss gating." + }, + { + "left": "issue:43352", + "right": "issue:43931", + "accept": false, + "reason": "Nemotron Flash Attention support and Qwen3-VL weight-shape mismatch are different model-loading problems." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Distributed optimizer-state corruption and a cache-strategy typo are not the same issue." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Cache update no_grad behavior and import errors after upgrade are different parts of the stack." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "UMT5 missing embed_tokens.weight is unrelated to Qwen3.5 MoE from_pretrained failures." + }, + { + "left": "issue:43867", + "right": "issue:46202", + "accept": false, + "reason": "State_dict sorting load error and Gemma pipeline roundtrip on \u2581 are different serialization/pathology reports." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Compiled Python file execution errors are distinct from the Python 3.13 decorator/comment parsing regression." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "State_dict load failure and missing Qwen3.5 vision config field are different issues." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "These look like the same import-time slowdown problem: 41669 identifies the import-star root cause and 44246 reports the same slow-import symptom." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python 3.14 load failure and PretrainedConfig type-checking breakage are different regressions." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "SAM HQ test flakiness from missing set_seed and an xdist file race are different test issues." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "A cache update behavior change and a vague 'breaks too many models' complaint are not the same concrete bug." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "State_dict sorting load error and Qwen3.5 visual-encoder save key corruption are different save/load regressions." + }, + { + "left": "issue:42490", + "right": "issue:44075", + "accept": false, + "reason": "Local-path slash import warning and SGD args not being used are unrelated." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "UMT5Encoder weight-missing and Qwen3.5 MoE from_pretrained errors are different model-specific failures." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIPOutput attentions regression and patchtsmixer post_init gating are unrelated." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache no_grad behavior and Python-version compatibility failure are unrelated." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Both mention tensor parallelism, but one is a generic training bug and the other is a specific TP load OOM; not the same concrete issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays within the packet evidence: it frames the cluster conservatively as a mixed set of v5 regression reports and picks 45278 as the broadest representative issue without overclaiming. The only accepted soft edge, 41669\u219444246, is well-grounded by the shared import-time slowdown symptom. The rest of the pairwise rejections are cautious and supported by the titles, with no obvious overstatement." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 47, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:28282", + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:41628", + "issue:42548", + "issue:42757", + "issue:42831", + "issue:42890", + "issue:43010", + "issue:43065", + "issue:43232", + "issue:43299", + "issue:43502", + "issue:43576", + "issue:43618", + "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:44857", + "issue:44961", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45125", + "issue:45341", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45600", + "issue:46132" + ], + "soft_pairs": [ + "issue:28282|issue:33453", + "issue:43618|issue:44857", + "issue:43618|issue:44977", + "issue:44360|issue:45440", + "issue:42890|issue:44964", + "issue:43502|issue:44062", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:43723|issue:44556", + "issue:43576|issue:46132", + "issue:44387|issue:44556", + "issue:44222|issue:45125", + "issue:43010|issue:43065", + "issue:42890|issue:45341", + "issue:42548|issue:44351", + "issue:43502|issue:44485", + "issue:44355|issue:44749", + "issue:45125|issue:45478", + "issue:30333|issue:33453", + "issue:43010|issue:43232", + "issue:45125|issue:45375", + "issue:30990|issue:31795", + "issue:43502|issue:44222", + "issue:43299|issue:44556", + "issue:42831|issue:43874", + "issue:43502|issue:44961", + "issue:43827|issue:44961", + "issue:30333|issue:36296", + "issue:44355|issue:45071", + "issue:44315|issue:44360", + "issue:44355|issue:44655", + "issue:33453|issue:34689", + "issue:42757|issue:43576", + "issue:43576|issue:45071", + "issue:33453|issue:36296", + "issue:44222|issue:44230", + "issue:43576|issue:43824", + "issue:41628|issue:42548" + ], + "prepared_review_unit_hash": "85e6b0ca235e677902de5001c0407305fe8c2d5f8e550c6447cf77e076b12955", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 38, + "serialized_chars": 38478, + "estimated_input_tokens": 9620, + "estimated_eval_tokens": 19496 + }, + "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 share broad v5-era breakage themes (imports, loading, CLI, quantization, tensor parallelism). I found one plausible duplicate edge in the Qwen3.5 MoE from_pretrained/tensor-parallel-plan reports; the rest are too coarse or describe different code paths.", + "confidence": 0.62, + "canonical_issue_reason": "issue:44556 is the clearest representative of the recurring v5 loading/regression theme: a concrete upgrade-related failure to reload checkpoints after moving from v4.57 to v5.2/5.3.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44556 is the best single issue to stand in for the cluster because it is concrete, user-facing, and directly captures the v4\u2192v5 model-loading breakage pattern seen across several nearby reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "Different failures: missing PyTorch dependency vs tokenizer-loading regression." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "Unrelated model/output bug vs FP16 AMP crash." + }, + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "Different code paths; CLIPOutput attention assignment is not the same as Qwen3.5 flash-attention generation." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "Different models and failures; DSA ReLU discussion is not the same as DeepSeekV3MoE divergence." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Different test stability issue vs Phi-4 multimodal load failure." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "Different bugs: offline/local-files behavior vs AddedToken constructor argument collision." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "CI flakiness in a test file is unrelated to GraniteSpeech device override behavior." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Different scope: seed-dependent test flakiness vs auto_mappings referencing removed configs." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Both are loading-related, but tokenizer.from_pretrained breakage is a different symptom and likely a different fix than checkpoint reload compatibility." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "CLI/environment command breakage is unrelated to AttentionInterface.register behavior change." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Memory/OOM under int4 quantization is a different failure than checkpoint reload compatibility." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained support and missing tensor-parallel plan are different problems." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Cache/layer no_grad decoration and Sam3PixelDecoder dummy Conv2d are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Different test bug entirely." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols and do not look like the same underlying bug." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "API/local_files_only bug is unrelated to GLM-5 RoPE implementation." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled-Python-file errors are unrelated to the slowdown/filtering regression." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": true, + "reason": "Same underlying Qwen3.5 MoE from_pretrained failure in v5.4.0; the missing _tp_plan tensor-parallel fix matches the concrete error report." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLflow reporting after failed training and tokenizer loading are unrelated." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "No_grad decoration for cache/layer updates is unrelated to the generation kwargs sync issue." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Tensor-parallel loading failure is not the same as missing deepstack_visual_indexes in config." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Sentence-transformers stuck loading and documentation confusion are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Local-files-only network calls and FP8 save_pretrained bugs are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Qwen3VL MoE loading and generic checkpoint reload breakage are different model-specific failures." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FineGrainedFP8 accuracy and GLM46V image-patch counting are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "API-request leakage is unrelated to the placeholder issue titled racoon." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs still referencing pipeline() is unrelated to racoon." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLflow failure reporting and tensor-parallel training are different bugs." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled file execution errors are unrelated to PretrainedConfig type-checking regressions." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel application with model_init and DSA ReLU discussion are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled Python file errors and Pipeline.save_pretrained failures are different issues." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer-loading regression and Llama 3.2 vision model loading regression are different model/code paths." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Missing is_offline_mode import and broken transformers env command are different failures." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "CLI command breakage and PretrainedConfig type checking are unrelated." + }, + { + "left": "issue:33453", + "right": "issue:36296", + "accept": false, + "reason": "Tokenizer-loading regression and tensor-parallel training bug are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "Both mention FP8/MoE, but one is save_pretrained and the other is inference support; not the same bug." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "CLI/env command breakage and Qwen2_5_VL import failure are different issues." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Different missing imports: AutoImageProcessor vs PreTrainedModel." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall: the summary reflects the packet\u2019s mix of mostly unrelated v5-era issues, and the only accepted soft edge (issue:45125 \u2194 issue:45478) is cautiously stated and plausible from the titles. The other verdicts stay conservative rather than overreaching." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 48, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:31795", + "issue:36296", + "issue:38617", + "issue:41669", + "issue:42371", + "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:44297", + "issue:44315", + "issue:44355", + "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:45362", + "issue:45468", + "issue:45542", + "issue:45753", + "issue:45850", + "issue:46095", + "issue:46143", + "issue:46144", + "issue:46178", + "issue:46209", + "issue:46218" + ], + "soft_pairs": [ + "issue:44961|issue:45468", + "issue:31795|issue:38617", + "issue:44355|issue:46178", + "issue:43576|issue:44938", + "issue:44315|issue:46209", + "issue:46143|issue:46144", + "issue:44908|issue:44961", + "issue:44485|issue:44961", + "issue:36296|issue:43901", + "issue:43673|issue:44492", + "issue:36296|issue:43526", + "issue:45362|issue:45850", + "issue:44623|issue:45230", + "issue:44273|issue:45542", + "issue:44038|issue:45125", + "issue:44230|issue:46095", + "issue:44273|issue:45003", + "issue:44393|issue:45071", + "issue:43576|issue:44556", + "issue:43525|issue:45850", + "issue:44393|issue:45753", + "issue:41669|issue:43673", + "issue:44556|issue:45092", + "issue:43572|issue:45850", + "issue:42371|issue:44492", + "issue:44492|issue:45850", + "issue:45278|issue:45850", + "issue:41669|issue:46218", + "issue:44246|issue:44273", + "issue:42890|issue:44908", + "issue:43408|issue:44315", + "issue:44704|issue:44908", + "issue:44297|issue:45230", + "issue:43976|issue:44393", + "issue:43352|issue:44393" + ], + "prepared_review_unit_hash": "cd3e3c3de119bbcb89a4d53cff76182f0866b3cb0de7cb460ea8e672b816384b", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37488, + "estimated_input_tokens": 9372, + "estimated_eval_tokens": 19000 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are largely separate bug reports/regressions across different models, APIs, and environments. None of the soft pairs look like the same underlying defect, so there is no safe duplicate merge among them.", + "confidence": 0.88, + "canonical_issue_reason": "issue:45850 is the broadest umbrella report ('Since 5.0 version it breaks too many models') and best matches the cluster\u2019s general v5-regression theme, even though the individual reports are not duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45850 is the best global issue anchor because it is the most general, open-ended regression report covering many downstream breakages.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "Unrelated issue content: a one-word/non-descriptive report vs a Gemma-4 positional-encoding bug." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation confusion about forward inputs vs an import error for layer_type_validation." + }, + { + "left": "issue:44355", + "right": "issue:46178", + "accept": false, + "reason": "Compiled Python-file loading errors vs a backend mapping error for tensorflow_text; different failure modes." + }, + { + "left": "issue:43576", + "right": "issue:44938", + "accept": false, + "reason": "Broken `transformers env` command vs Python 3.14 import/load failure; not the same code path." + }, + { + "left": "issue:44315", + "right": "issue:46209", + "accept": false, + "reason": "Model-init/Liger-kernel behavior vs FP8 inference failure on B200; different problems." + }, + { + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "Passing `**kwargs` through RoFormer methods vs attention-interface usage; related subsystem but distinct bug." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler keyword handling vs an unrelated issue titled 'racoon'." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "RoPE implementation discussion vs an unrelated issue; no shared bug." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training bug vs docs mentioning `return_all_scores`; completely different concerns." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Chunked-prefill cache behavior vs a cache-strategies typo; not the same defect." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor-parallel training issue vs BeitImageProcessorFast label reduction bug; different code paths." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Qwen3.5 chat crash is a specific model failure, while 45850 is a broad v5 regression umbrella." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "`processor.save_pretrained` missing files vs a generic 'Bug report' with no concrete shared symptom." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading malfunction vs undefined TensorFlow backend from tensorboard-only install; unrelated." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Qwen3-VL-Moe bug vs missing `_tp_plan` for tensor parallelism; same family, but not the same concrete issue." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 inference support request vs a weight-conversion-script question; not duplicates." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Lazy loading problems vs unsafe `sys.modules` access in modeling_utils; different underlying bug." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Qwen3-VL bounding-box hallucination/error vs PretrainedConfig type-checking breakage; unrelated." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "`transformers env` command breakage vs checkpoint reload incompatibility after upgrade; different symptoms and code paths." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "Missing `pad_token_id` on Llama4Config is a specific config bug, not the broad v5 regression umbrella." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "Missing `pad_token_idx` in StableLmConfig is a specific model config issue, not the same as the generic v5-breakage report." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 behavior settings vs a typo in cache strategies; no common underlying bug." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "A typo issue is not the same as a general claim that v5 breaks many models." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "Both are broad v5-regression complaints, but one is about import errors after upgrade and the other about many models; too general to treat as the same bug." + }, + { + "left": "issue:41669", + "right": "issue:46218", + "accept": false, + "reason": "Import-star usage/performance issue vs a specific missing-module import error for Qwen3_5ForCausalLM." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Slow import timing vs lazy-loading malfunction; related theme, different defect." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Missing `set_seed()` in tests vs scheduler kwarg handling; unrelated." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "sam3_video/sam3_tracker warning vs Liger kernel application with model_init; different bugs." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "`AutoProcessor.from_pretrained` kwargs forwarding vs inverse-sqrt scheduler kwargs; not the same component or bug." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer class mismatch on save vs a vague bug report with no matching failure mode." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python-version compatibility problem vs Qwen3-VL bounding-box output bug; unrelated." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention 2 support gap vs Qwen3-VL output hallucination/error; different issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 49, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:41669", + "issue:42371", + "issue:42757", + "issue:42913", + "issue:43066", + "issue:43298", + "issue:43502", + "issue:43519", + "issue:43576", + "issue:43606", + "issue:43618", + "issue:43704", + "issue:43827", + "issue:43874", + "issue:43976", + "issue:44077", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44295", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44655", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44945", + "issue:44998", + "issue:45071", + "issue:45230", + "issue:45310", + "issue:45341", + "issue:45468", + "issue:45478", + "issue:45850", + "issue:46018", + "issue:46095", + "issue:46172", + "issue:46218" + ], + "soft_pairs": [ + "issue:44295|issue:45230", + "issue:43576|issue:44908", + "issue:41669|issue:44351", + "issue:44908|issue:45071", + "issue:41669|issue:45341", + "issue:43874|issue:45850", + "issue:44492|issue:44945", + "issue:43976|issue:46218", + "issue:44077|issue:44861", + "issue:44393|issue:45478", + "issue:43827|issue:44908", + "issue:31795|issue:33453", + "issue:44315|issue:44829", + "issue:43576|issue:44273", + "issue:44655|issue:44861", + "issue:44360|issue:45468", + "issue:45230|issue:45310", + "issue:44273|issue:44908", + "issue:46018|issue:46172", + "issue:43576|issue:46218", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:41669|issue:43704", + "issue:31795|issue:34689", + "issue:31515|issue:31795", + "issue:44485|issue:45468", + "issue:43298|issue:46095", + "issue:43704|issue:44485", + "issue:45850|issue:46218", + "issue:43502|issue:43519", + "issue:42371|issue:43704", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:43606|issue:45468", + "issue:44263|issue:45468", + "issue:44261|issue:45468", + "issue:44351|issue:46218", + "issue:42757|issue:46218", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44492|issue:44998", + "issue:43618|issue:44998" + ], + "prepared_review_unit_hash": "9b1cb3bbf8a47bd46822211466a77931fcb573c72f797ca28cb41edb897a9787", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 42, + "serialized_chars": 39418, + "estimated_input_tokens": 9855, + "estimated_eval_tokens": 19966 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous, with many unrelated loading, tokenizer, docs, import, and model-specific issues. The clearest canonical thread is the Qwen3.5 MoE from_pretrained regression, while the soft-edge pairs mostly share only broad Transformers-v5 or loading-related wording and should stay separate.", + "confidence": 0.56, + "canonical_issue_reason": "issue:45310 is the most concrete, reproducible bug report in the cluster: it names a specific failing path (Qwen3.5 MoE + transformers>=5.4.0 + from_pretrained) and is representative of the loading-regression theme.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the best overall representative because it is specific enough to anchor the cluster, describes an actionable failure mode, and is more precise than the surrounding broad/vague loading complaints.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "One is a position_ids buffer access bug; the other is only a generic bug report with no matching failure mode." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "Broken env command and scheduler kwargs bug are unrelated subsystems and concrete failures." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star performance issue and missing HybridCache export are different problems." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler keyword handling and PretrainedConfig type-checking are distinct code paths." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Model import cleanup/performance and a testing_utils typo are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:45850", + "accept": false, + "reason": "A missing image-processor method in one model is not the same as a broad complaint that many models break in v5." + }, + { + "left": "issue:44492", + "right": "issue:44945", + "accept": false, + "reason": "Cache-strategy typo and pipeline-parallelism output corruption do not describe the same bug." + }, + { + "left": "issue:43976", + "right": "issue:46218", + "accept": false, + "reason": "Python version compatibility and a missing Qwen3_5 module import are separate import/runtime issues." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Optional post_init enforcement and tied-weights list handling are different model-init bugs." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL bounding-box hallucination is unrelated to a Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing pipeline() and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion for model.forward and tokenizer-loading regression are not the same underlying problem." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel application in model_init and flash_attention_3 degenerate training are different training-path issues." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Broken env command and lazy loading malfunction are not the same bug." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Saving pipeline objects and crashing on _tied_weights_keys list are unrelated save/load behaviors." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer ReLU discussion and Gemma-4 audio positional encoding are different model internals." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "A generic bug report is not a duplicate of a specific from_pretrained error report." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading failure and scheduler kwargs ignoring are unrelated." + }, + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "DeepSeek-V4 gating behavior and a comb transpose bug are different DeepSeek issues." + }, + { + "left": "issue:43576", + "right": "issue:46218", + "accept": false, + "reason": "Broken env command and missing Qwen3_5 import are unrelated commands/imports." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are different vision/text model bugs." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Timestamp math in a processor and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star performance regression and VRAM leak in multi-threaded use are different issues." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Forward-doc confusion and Llama 3.2 model-loading breakage are not the same bug." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Slow checkpoint loading and documentation confusion about forward inputs are unrelated." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "GLM-5 RoPE implementation and Gemma-4 audio positional encoding are different architecture bugs." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Continuous batching for audio models and DeepSeek-V4 weight-conversion question are not the same underlying issue." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leak in threaded inference and GLM-5 RoPE implementation do not match." + }, + { + "left": "issue:45850", + "right": "issue:46218", + "accept": false, + "reason": "Broad complaint that v5 breaks many models and a specific missing module import are not duplicates." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "Offline file-loading network requests and Qwen3VL timestamp calculation are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 config migration and VRAM leak in threaded inference are separate concerns." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Timestamp calculation in a processor and missing processor save files are different processor bugs." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Processor timestamp math and multi-thread VRAM leak are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch in bark-small is unrelated to Gemma-4 audio positional encoding." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "GlmMoeDsaIndexer torch.split behavior and Gemma-4 audio positional encoding are different internals." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "MLA q_a_layernorm epsilon precision issue and Gemma-4 audio positional encoding are not the same bug." + }, + { + "left": "issue:44351", + "right": "issue:46218", + "accept": false, + "reason": "Missing HybridCache export and missing Qwen3_5ForCausalLM import are separate import/export failures." + }, + { + "left": "issue:42757", + "right": "issue:46218", + "accept": false, + "reason": "huggingface_hub is_offline_mode import breakage is unrelated to Qwen3_5 module import failure." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Both mention tokenizer v5 behavior, but they describe different concrete tokenizer regressions." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs handling and an unrelated placeholder issue title are not the same bug." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "Cache-strategy typo and unrelated placeholder issue are not duplicates." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions regression is unrelated to the placeholder issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 50, + "review_unit_count": 50, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-431", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:33453", + "issue:36296", + "issue:37934", + "issue:38617", + "issue:42936", + "issue:43240", + "issue:43298", + "issue:43525", + "issue:43572", + "issue:43668", + "issue:43717", + "issue:43813", + "issue:44031", + "issue:44230", + "issue:44556", + "issue:44623", + "issue:44861", + "issue:44998", + "issue:45290", + "issue:45310", + "issue:45850", + "issue:46018", + "issue:46144" + ], + "soft_pairs": [ + "issue:44861|issue:45290", + "issue:44998|issue:45850", + "issue:44556|issue:44998", + "issue:43525|issue:44861", + "issue:43572|issue:44861", + "issue:31515|issue:38617", + "issue:44230|issue:44623", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:43298|issue:44230", + "issue:43717|issue:46144", + "issue:42936|issue:46018", + "issue:43668|issue:44031", + "issue:42936|issue:43240", + "issue:43240|issue:46144", + "issue:42936|issue:44031", + "issue:42936|issue:46144", + "issue:42936|issue:43813", + "issue:43240|issue:43668", + "issue:43668|issue:43813", + "issue:37934|issue:43668", + "issue:42936|issue:43668" + ], + "prepared_review_unit_hash": "420eecf687ba8d7b1338d36000ace2d9034f9b4b1a15dbc9e64c5a98ea8c03bb", + "component_budget": { + "node_count": 431, + "item_count": 431, + "soft_pair_count": 1595, + "serialized_chars": 825278, + "estimated_input_tokens": 206320, + "estimated_eval_tokens": 412896 + }, + "budget": { + "node_count": 25, + "item_count": 25, + "soft_pair_count": 24, + "serialized_chars": 21130, + "estimated_input_tokens": 5283, + "estimated_eval_tokens": 10822 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are unrelated issues spanning loading regressions, config/API breakages, model-specific bugs, and a few generic feature requests. I do not see any true duplicate pair here; the soft-similarity matches are mostly surface-level topic overlap.", + "confidence": 0.91, + "canonical_issue_reason": "issue:44556 is the best representative because it is a concrete, user-facing version-regression report about checkpoint reloads, which is the closest substantive theme shared by several other items in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44556 is the most actionable and broadly representative issue in the set: it names the failure mode, the affected upgrade path, and a clear regression scenario.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Different bugs in different code paths: tied-weight key extraction crash vs chat template crashes on tool-call assistant messages." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "Unemployment is not a technical bug report; the other is a broad model-regression complaint, so they are unrelated." + }, + { + "left": "issue:44556", + "right": "issue:44998", + "accept": false, + "reason": "A checkpoint reload regression is unrelated to the non-technical \"Unemployment\" issue." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id on Llama4Config and an AttributeError in _get_tied_weight_keys are different failures in different subsystems." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "StableLmConfig padding attribute regression is not the same bug as tied-weight key handling." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow from_pretrained checkpoint loading and an ImportError for layer_type_validation are distinct problems." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 inference support for Qwen models is unrelated to missing files in processor.save_pretrained()." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLflow reporting after failed training and slow checkpoint loading are different code paths and outcomes." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "from_pretrained performance and tensor-parallel training bugs are not the same underlying defect." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 support request and Qwen3.5 MoE from_pretrained error are different model-specific issues." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Checkpoint load slowness and tokenizer loading regression are both loading-related but not the same concrete bug." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Continuous batching audio-model support is a different feature gap than FP8 inference support." + }, + { + "left": "issue:43717", + "right": "issue:46144", + "accept": false, + "reason": "Mamba init_weights behavior and RoFormer attention-interface usage are unrelated model implementation issues." + }, + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2Former ignore_value handling and DeepSeek-V4 shared-expert gating are unrelated model bugs." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "A type-hint bug in ModernBERTConfig and a tokenizer regex warning regression are different issues." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Ignored loss parameter handling and kwargs not forwarded to loss calculation are separate bugs, though both touch loss code." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss kwargs forwarding and attention-interface usage are different subsystems with no shared concrete failure." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2Former ignore_value handling does not match the tokenizer regex warning issue." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "Different model code paths: Mask2Former loss handling vs RoFormer attention implementation." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Mask2Former ignore_value handling and a typo in quantization_operation are unrelated." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Loss kwargs forwarding bug and ModernBERTConfig type-hint bug are unrelated." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "A config type-hint issue and a typo fix are not the same underlying bug." + }, + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "A question about Llama4TextL2Norm being RMSNorm is unrelated to ModernBERTConfig norm_eps typing." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Mask2Former loss logic and ModernBERTConfig typing are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10519-3", + "nodes": [ + "issue:10519", + "issue:11005", + "issue:11307" + ], + "soft_pairs": [ + "issue:10519|issue:11307", + "issue:10519|issue:11005", + "issue:11005|issue:11307" + ], + "prepared_review_unit_hash": "79bcc05d6892c55ae0f82f5045e5df4de37c28db7f56536151c44d798f1bac3d", + "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 unrelated feature requests/questions: truncation behavior, ReduceLROnPlateau-like scheduling, and Wav2Vec2 word time offsets. They do not describe the same underlying bug or change, so this cluster should not be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a good canonical representative here because the topics are unrelated; if forced to pick a single issue, #10519 is the most broadly phrased feature request, but it is not a duplicate of the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:10519", + "right": "issue:11307", + "accept": false, + "reason": "Different domains and requested behaviors: truncation from beginning vs Wav2Vec2 word boundary timestamps." + }, + { + "left": "issue:10519", + "right": "issue:11005", + "accept": false, + "reason": "One is about token truncation strategy, the other about learning-rate scheduling functionality; no shared underlying bug." + }, + { + "left": "issue:11005", + "right": "issue:11307", + "accept": false, + "reason": "Unrelated requests in different parts of the library: scheduler behavior versus Wav2Vec2 time 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": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-13244-3", + "nodes": [ + "issue:13244", + "issue:15980", + "issue:16618" + ], + "soft_pairs": [ + "issue:13244|issue:16618", + "issue:15980|issue:16618" + ], + "prepared_review_unit_hash": "7ad21377992da39338aa6ae60ff20a09973bd02e4a2cf2ce374067a67088fcc8", + "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 look like the same bug: one is about Tapas tokenization differing from TensorFlow code, one is about an unhelpful auth error when downloading a private model, and one is a generic model-loading failure for bert-base-uncased. The only loose commonality is model loading, but the concrete failure modes differ.", + "confidence": 0.88, + "canonical_issue_reason": "Issue 16618 is the broadest and most representative of the cluster because it concerns a core model-loading failure, whereas 15980 is specifically an authentication/error-message problem and 13244 is about tokenizer behavior.", + "canonical_pr_reason": null, + "best_issue_reason": "16618 is the best issue to keep as the cluster representative since it is the most general user-facing model-load problem; however, it should not be treated as a duplicate of the other two issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Tapas tokenization mismatch vs. bert-base-uncased loading failure are different subsystems and different failure modes." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Private-model login/download error message is an auth/UX issue, while 16618 is a model-load failure for a public model; not the same underlying bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-26350-2", + "nodes": [ + "issue:26350", + "issue:33260" + ], + "soft_pairs": [ + "issue:26350|issue:33260" + ], + "prepared_review_unit_hash": "942127d98fb66aec749c5cdf0b97b35ed2e84f2a4a283b6f10921fd2ac58f293", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two different community-contribution feature requests: one for Flash Attention 2 support on more architectures, the other for GGUF support on more architectures. They share the same broad pattern but not the same underlying change or bug.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different requested features and code paths: Flash Attention 2 architecture support vs GGUF architecture support. Similar wording and contribution framing are too broad to treat as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 5, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-27088-8", + "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:46222", + "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:46222", + "issue:38435|issue:44365", + "issue:38435|issue:43834", + "issue:38435|issue:46226", + "issue:43834|issue:44370", + "issue:43834|issue:46222", + "issue:43834|issue:44365", + "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": "39ffaa3671b3302b6ae30b5b6f477b0dfefee359a99ddec76c620cb8c377b66c", + "component_budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 25, + "serialized_chars": 13678, + "estimated_input_tokens": 3420, + "estimated_eval_tokens": 7096 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 25, + "serialized_chars": 13678, + "estimated_input_tokens": 3420, + "estimated_eval_tokens": 7096 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a set of i18n translation request issues for different languages. The titles/body templates are highly similar, but they represent separate translation tasks rather than the same underlying bug or change, so none should be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "No canonical issue: the items are separate translation requests for different languages, and the similarity is mostly from a shared issue template.", + "canonical_pr_reason": null, + "best_issue_reason": "No single issue is a good global duplicate target; the closest matches are generic template issues, but they still do not represent the same underlying translation task.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Different language-request issues; same template, but no evidence they track the same translation task." + }, + { + "left": "issue:27088", + "right": "issue:46222", + "accept": false, + "reason": "Different language-request issues; superficial template similarity only." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "Turkish vs another translation request; not the same underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Different language request; similar template does not make them duplicates." + }, + { + "left": "issue:27088", + "right": "issue:38435", + "accept": false, + "reason": "Turkish vs Romanian translation requests are distinct tasks." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "Turkish vs Spanish translation requests are separate issues." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "Spanish translation request vs generic template issue; no same-language/task evidence." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "Same translation template, but likely different language-specific task." + }, + { + "left": "issue:28936", + "right": "issue:38435", + "accept": false, + "reason": "Spanish and Romanian translation requests are distinct." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Shared template only; no sign these are the same translation work item." + }, + { + "left": "issue:38435", + "right": "issue:44370", + "accept": false, + "reason": "Different translation requests; similarity is only from the i18n template." + }, + { + "left": "issue:38435", + "right": "issue:46222", + "accept": false, + "reason": "Different language-specific translation tasks." + }, + { + "left": "issue:38435", + "right": "issue:44365", + "accept": false, + "reason": "Romanian translation request vs another language request; not duplicates." + }, + { + "left": "issue:38435", + "right": "issue:43834", + "accept": false, + "reason": "Same i18n template, but not the same underlying request." + }, + { + "left": "issue:38435", + "right": "issue:46226", + "accept": false, + "reason": "Romanian translation request vs unrelated Thai text; clearly not duplicates." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": false, + "reason": "Both are generic translation templates, but there is no evidence they refer to the same language or task." + }, + { + "left": "issue:43834", + "right": "issue:46222", + "accept": false, + "reason": "Template similarity only; no explicit target-language match." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": false, + "reason": "Separate translation issues despite near-identical wording." + }, + { + "left": "issue:43834", + "right": "issue:46226", + "accept": false, + "reason": "Generic i18n template vs unrelated Thai title; not the same issue." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": false, + "reason": "Near-identical placeholder titles, but no evidence they are the same language request." + }, + { + "left": "issue:44365", + "right": "issue:46222", + "accept": false, + "reason": "Same template, but distinct translation tasks are plausible and should not be merged." + }, + { + "left": "issue:44365", + "right": "issue:46226", + "accept": false, + "reason": "Generic translation issue vs unrelated Thai text; not duplicates." + }, + { + "left": "issue:44370", + "right": "issue:46222", + "accept": false, + "reason": "Exact template match is not enough; could still be separate language-specific issues." + }, + { + "left": "issue:44370", + "right": "issue:46226", + "accept": false, + "reason": "Generic template issue vs unrelated Thai title; no duplicate signal." + }, + { + "left": "issue:46222", + "right": "issue:46226", + "accept": false, + "reason": "Generic i18n translation request vs unrelated Thai text; 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": 6, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-36817-16", + "nodes": [ + "issue:36817", + "issue:36879", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322", + "issue:46133" + ], + "soft_pairs": [ + "issue:36817|issue:41211", + "issue:36817|issue:43671", + "issue:36817|issue:36879", + "issue:36879|issue:41211", + "issue:36879|issue:43909", + "issue:36879|issue:44121", + "issue:36879|issue:43837", + "issue:36879|issue:43671", + "issue:37015|issue:43837", + "issue:37015|issue:44121", + "issue:37015|issue:43909", + "issue:36879|issue:37015", + "issue:37015|issue:42503", + "issue:41115|issue:43837", + "issue:41115|issue:44121", + "issue:41115|issue:42503", + "issue:41115|issue:45027", + "issue:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42738|issue:45027", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43671|issue:43909", + "issue:43837|issue:43909", + "issue:43837|issue:45322", + "issue:43837|issue:44121", + "issue:43909|issue:44205", + "issue:44121|issue:45322", + "issue:43837|issue:45027", + "issue:37015|issue:45027", + "issue:43837|issue:45177", + "issue:45322|issue:46133" + ], + "prepared_review_unit_hash": "960c4a8c45a236dfad14c057b8df5f2898ebe5d7fda74da47fc03e15dd5c465d", + "component_budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21657, + "estimated_input_tokens": 5415, + "estimated_eval_tokens": 11086 + }, + "budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21657, + "estimated_input_tokens": 5415, + "estimated_eval_tokens": 11086 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a loose collection of model-support request issues, but the items are for different models and often different modalities/tasks. They are not duplicates of one another; at most they share the generic pattern of requesting new model additions.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: these are distinct model-request issues, not the same underlying bug/change.", + "canonical_pr_reason": null, + "best_issue_reason": "No single best issue stands out as a representative duplicate target; the requests span unrelated models, so a canonical merge point would be misleading.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "EuroBert config vs DEIMv2 request; different models and code paths." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "BERT config request vs Qwen3-TTS support; different model families and tasks." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert vs RF-DETR; unrelated architectures and modalities." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR object detection vs DEIMv2; different concrete models." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR vs LFM2.5 Audio; detection and audio are unrelated." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "RF-DETR vs weight-sparse transformer; no same underlying change." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR vs Qwen3-ASR; different models and tasks." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "RF-DETR vs Qwen3-TTS; different model families." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT config request vs Qwen3-ASR support; unrelated models." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT vs sparse transformer; different architecture requests." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT vs LFM2.5 Audio; unrelated model types." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are different models with different integration work." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT vs ModernVBERT; similar naming only, not the same model." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture request vs Qwen3-ASR support; different targets." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 vs weight-sparse transformer; distinct model additions." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 vs ModernVBERT; both model requests, but not duplicates." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 vs Voxtral TTS; different model families and modalities." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 vs Qwen3-TTS; unrelated model support requests." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT vs Qwen3-ASR; different models/tasks." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT vs Qwen3-TTS; no shared code-path problem." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "Specific model request vs generic BERT-like/RoPE discussion; not duplicates." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "Generic BERT-like model request vs audio TTS model; unrelated." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Qwen3-TTS and Qwen3-ASR are different tasks and likely separate implementations." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Both are TTS-related, but they request different models and code paths." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS vs DeepSeek-OCR2; different model domains." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Audio model support vs sparse transformer request; not the same change." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS vs LFM2.5 Audio; both audio-related but different models." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR vs LFM2.5 Audio; different audio models and tasks." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "ASR model request vs EUPE vision-language encoder; unrelated." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR support vs sparse transformer request; different change types." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "Audio model request vs SAM3-LiteText; different modalities and code paths." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer request vs EUPE encoder; not the same model or bug." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR vs Voxtral TTS; both audio, but distinct tasks and models." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT config vs Voxtral TTS support; unrelated requests." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "ASR support vs OCR2 model request; different model families." + }, + { + "left": "issue:45322", + "right": "issue:46133", + "accept": false, + "reason": "EUPE and TIPSv2 are different vision-language models; 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": "issue", + "component_index": 7, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41084-2", + "nodes": [ + "issue:41084", + "issue:43037" + ], + "soft_pairs": [ + "issue:41084|issue:43037" + ], + "prepared_review_unit_hash": "821b46ba1d6c91d6ac20e6b66383e8caaf38ae153b2457351055f8ccee9b5fba", + "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": "The two issues are not duplicates: one is about enabling block decoding, the other is a request for DeepSeek v3.2 model support. They describe different features/problems with no shared targets or code path evidence.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #41084 is the more technical, implementation-oriented request and is the better representative of the cluster if one must be chosen, though it is not actually the same underlying problem as #43037.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #41084 is the best single issue candidate because it is more concrete and likely to map to an actionable code change; #43037 is a separate model-support request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "Different topics and likely different implementation paths: block decoding vs DeepSeek v3.2 support. No explicit links, shared targets, filenames, or bug signature indicate 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": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43598-2", + "nodes": [ + "issue:43598", + "issue:43600" + ], + "soft_pairs": [ + "issue:43598|issue:43600" + ], + "prepared_review_unit_hash": "c0e5508c1f4822971e9ede5aa5046d65b0dc84485d6215f7ccd51dfffbe3bf1f", + "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 high level (custom trainers in Trainer), but they describe different problems: one is about `remove_unused_column` customizability, the other about FLOPS estimation handling custom trainers. Not duplicates.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43598 is the more central and broadly framed Trainer customization concern; issue 43600 is a narrower follow-on about FLOPS estimation rather than the core customization behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "They share the theme of custom trainers, but the concrete issues differ: one targets `remove_unused_column` behavior, the other targets FLOPS estimation logic. Not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 9, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43684-2", + "nodes": [ + "issue:43684", + "issue:44405" + ], + "soft_pairs": [ + "issue:43684|issue:44405" + ], + "prepared_review_unit_hash": "a0c78dc5e184c9f67fc294b2025bd85402428f14de21f09c6380b00864748cb7", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are not duplicates: one requests auto-class registration for Qwen3-Omni, while the other asks for sequence-classification support for Qwen3.5. They touch different model families and different AutoModel paths.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 43684 is the broader and more central request around auto-model registration, so it is the better representative issue if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43684 is the cleaner cluster representative because it concerns generic AutoModel/AutoModelForConditionalGeneration registration rather than a single task-specific head.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different underlying changes: Qwen3-Omni auto registration vs Qwen3.5 sequence-classification support. They are not the same bug or fix 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": "issue", + "component_index": 10, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43835-4", + "nodes": [ + "issue:43835", + "issue:43845", + "issue:43846", + "issue:43859" + ], + "soft_pairs": [ + "issue:43835|issue:43845", + "issue:43835|issue:43846", + "issue:43835|issue:43859", + "issue:43845|issue:43846", + "issue:43845|issue:43859", + "issue:43846|issue:43859" + ], + "prepared_review_unit_hash": "249b3f80038f0322f7a1a52c9700b2d05986153c6b7f79c1e8a2cc11e8a20352", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "One issue (\u201cFett\u201d) appears unrelated. The three \u201chuggingface\u201d issues look like a small duplicate cluster based on identical titles and very similar creation timing/body size.", + "confidence": 0.83, + "canonical_issue_reason": "issue:43845 is the earliest of the near-identical \u201chuggingface\u201d issues, so it is the best representative for the duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43845 is the clearest cluster anchor: same title as the other likely duplicates, earliest timestamp, and no evidence that it differs materially from them.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Different title; no evidence these describe the same underlying issue." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Different title; no indication of the same bug or request." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "Different title and no shared concrete issue details." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": true, + "reason": "Identical title, nearly identical creation time, and matching body size strongly suggest the same underlying issue." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": true, + "reason": "Same title and clearly overlapping metadata; likely duplicate reports of the same issue." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": true, + "reason": "Identical title and strong similarity in metadata indicate the same underlying issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary and pairwise verdicts are grounded in the packet-level metadata: issue:43835 has a different title, while issue:43845/43846/43859 share the same title and close timestamps, with body-length similarity supporting a conservative duplicate-cluster call. The canonical choice of issue:43845 as the earliest representative is also supported." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "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": "2383e2cb1af7c8fa6c8282c53b4f7d46792ded1accacdb6b7632a0d62a6d42ef", + "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": "The cluster is heterogeneous: multiple spam placeholders, one generic documentation suggestion, and one unrelated company-name request. There is no shared concrete bug or change, so no duplicate canonicalization is justified.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44371 is only marginally the best representative because it is an earliest, minimal placeholder, but the cluster is too mixed to treat any item as a true canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are spam placeholders, but there is no concrete underlying bug or change to unify." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Same as above: identical placeholder titles do not establish the same issue." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "Spam placeholder versus a documentation enhancement request; different topics." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder versus a request about 'Google LLC'; no shared problem statement." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "Generic docs enhancement and spam placeholder are unrelated." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Generic docs enhancement and spam placeholder are unrelated." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Both are placeholders, but that is not enough to conclude the same underlying issue." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder versus unrelated 'Google LLC' request." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder versus unrelated 'Google LLC' request." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45145-2", + "nodes": [ + "issue:45145", + "issue:45151" + ], + "soft_pairs": [ + "issue:45145|issue:45151" + ], + "prepared_review_unit_hash": "b444870143764218620fedd0c6e661e8977edec4b9595175239ec07659d06f52", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1644, + "estimated_input_tokens": 411, + "estimated_eval_tokens": 1078 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1644, + "estimated_input_tokens": 411, + "estimated_eval_tokens": 1078 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two closed issues look like likely duplicates: they have the same title and a strong text-similarity signal, with only limited metadata differences. The evidence is enough to cluster them together, but not enough to claim they are proven identical without seeing the full bodies.", + "confidence": 0.87, + "canonical_issue_reason": "Issue #45145 is the better canonical representative because it was created earlier and matches #45151 on title plus a strong similarity signal, making it the safer anchor for this likely duplicate pair.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best issue for the cluster because it is the earliest report and appears to be the same likely duplicate topic as #45151, based on the matching title and similarity signal.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Accept as a likely duplicate: the titles are identical and the packet shows a solid soft-similarity signal, but the evidence is limited because the bodies are not shown and there is some distinguishing metadata." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The clustering is plausible, but the writeup overstates the evidence. The packet only shows a soft_similarity signal (Jaccard ~0.63) plus matching titles and similar body lengths; it does not show the body contents, and there is distinguishing metadata such as different creation times and inbound references. Please soften claims like \"same underlying issue,\" \"fully equivalent,\" and \"no distinguishing metadata\" to reflect that this is only a likely duplicate, not a proven match." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45295-2", + "nodes": [ + "issue:45295", + "issue:45373" + ], + "soft_pairs": [], + "prepared_review_unit_hash": "f5fd95fada2f530ab0999e343c17c16266cde0c7182bb22509972ca44827cd49", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues describe the same missing Gemma 4 sequence-classification support; the later, more specific open issue is the better cluster representative.", + "confidence": 0.94, + "canonical_issue_reason": "issue #45373 is the clearest canonical tracker: it is open, explicitly names the missing class `Gemma4ForSequenceClassification`, and states the module-level gap relative to Gemma 2/3.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #45373 best captures the concrete request and likely remaining work, while #45295 is a shorter earlier duplicate with less detail and is already closed.", + "best_pr_reason": null, + "soft_edge_verdicts": [] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45672-2", + "nodes": [ + "issue:45672", + "issue:45674" + ], + "soft_pairs": [ + "issue:45672|issue:45674" + ], + "prepared_review_unit_hash": "808fd1470b07eea820021d7b2bef0383affb059ecbd801d0ed9d1aae5d492310", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1668, + "estimated_input_tokens": 417, + "estimated_eval_tokens": 1090 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1668, + "estimated_input_tokens": 417, + "estimated_eval_tokens": 1090 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related only at a very broad level (4-bit inference / quantization), but they describe different bugs: one is a Gemma4 audio-module uint8/torch.finfo TypeError, the other is a BitsAndBytesConfig skip-module regression that drops the default lm_head exclusion and triggers an AssertionError.", + "confidence": 0.97, + "canonical_issue_reason": "issue:45674 is the better anchor for the cluster because it is the more general configuration regression and has higher discussion/activity; issue:45672 is a model-specific Gemma4 failure with a different root cause.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45674 has the clearer, broader repro and is more likely to serve as the representative report for quantized-inference breakage, while issue:45672 is narrowly scoped to Gemma4 audio modules.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Reject: both concern 4-bit inference, but the failure modes and code paths differ substantially (Gemma4 audio-module torch.finfo on uint8 vs. BitsAndBytesConfig llm_int8_skip_modules clearing lm_head exclusion)." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45762-2", + "nodes": [ + "issue:45762", + "issue:45763" + ], + "soft_pairs": [ + "issue:45762|issue:45763" + ], + "prepared_review_unit_hash": "9234982dd41d7ff6c7ad6313a07048f34b5f8f4e751d2b39f309b9d6356d90b8", + "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": "These two closed issues do not appear to be duplicates: they have different names and dates, with no shared problem description or technical context.", + "confidence": 0.98, + "canonical_issue_reason": "Neither issue is a clear canonical duplicate target; the titles refer to different people and contain no evidence of the same underlying issue.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is suitable as a global best issue for this cluster because the items look unrelated rather than duplicate reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "Different names and dates, no common bug report or matching context; this looks unrelated rather than 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": 16, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45764-2", + "nodes": [ + "issue:45764", + "issue:45765" + ], + "soft_pairs": [ + "issue:45764|issue:45765" + ], + "prepared_review_unit_hash": "d4ba2403b946c1deec5746aa5a57f1cddd53625dc0301bc611c8cdaaa53e8788", + "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 just separate name/date titles and do not describe the same bug, change, or request. They should not be clustered together.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "Different personal-name/date titles with no shared problem statement, target, or context; not the same underlying issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 17, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45869-5", + "nodes": [ + "issue:45869", + "issue:45870", + "issue:45871", + "issue:45872", + "issue:45873" + ], + "soft_pairs": [ + "issue:45869|issue:45870", + "issue:45869|issue:45871", + "issue:45870|issue:45871", + "issue:45870|issue:45872", + "issue:45870|issue:45873", + "issue:45871|issue:45872", + "issue:45871|issue:45873", + "issue:45872|issue:45873" + ], + "prepared_review_unit_hash": "85f5d96ed41f20dcaef927d62d21a86f9a18172a4cf55d5fe1efc19c8fbb8b0f", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5727, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5727, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster contains two separate exact-duplicate pairs: 45870/45871 and 45872/45873. Issue 45869 is a different financial dataset promotion/sample request and does not match the DataForge example/service issues.", + "confidence": 0.98, + "canonical_issue_reason": "45870 is the earliest and cleanest representative of the exact-duplicate pair titled 'Example: Financial time series prediction with DataForge'.", + "canonical_pr_reason": null, + "best_issue_reason": "45870 is the strongest single issue to keep as a canonical representative because it is an exact duplicate target, earliest created, and has a fully generic title for the repeated example issue.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Different topics: a dataset bundle promo/sample vs a DataForge example issue." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Different topics: a dataset bundle promo/sample vs a DataForge example issue." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Same title and same underlying example issue; exact duplicate." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Related financial/DataForge theme, but this is a different requested feature/topic: example prediction vs scraping service." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Same as 45872: different underlying issue, not the same concrete request." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Different underlying requests; exact duplicate only with 45870, not with the scraping-service issue." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Different underlying requests; exact duplicate only with 45870, not with the scraping-service issue." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Same title and same underlying DataForge scraping-service request; exact duplicate." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 45870/45871 and 45872/45873 are well-supported exact-duplicate pairs by identical titles and body similarity, while 45869 is clearly a different dataset-promo request. The soft-edge rejections between the two topic clusters are conservative and consistent with the available evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 28, + "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-39895-13", + "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" + ], + "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: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": "45115f25b79dd681b8bc046b985fe23add2af7e5d3a7df5f868842205351f9e1", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 18, + "serialized_chars": 34774, + "estimated_input_tokens": 8694, + "estimated_eval_tokens": 17644 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 18, + "serialized_chars": 34774, + "estimated_input_tokens": 8694, + "estimated_eval_tokens": 17644 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Only one pair appears to be a true duplicate/fork: the two fixed_cross_entropy PRs for issue 43240. The rest are unrelated PRs across different models or features and should not be merged.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43251 is the best canonical representative: it explicitly targets issue 43240 and contains the clearer direct fix to fixed_cross_entropy.", + "best_issue_reason": null, + "best_pr_reason": "PR 43251 is the strongest standalone PR in this set for the same reason: explicit issue linkage and a concrete, minimal change to the shared loss path.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "Different models/features (VideoPrism docs/model add vs ExecuTorch dynamic-shape test); not the same underlying change." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer API addition vs ExecuTorch export test; unrelated code paths." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Mistral tokenizer helper methods vs adding MobileLLM model files; different subsystems and goals." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor robustness changes vs new MobileLLM model implementation; not the same bug or change." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Cross-entropy kwargs fix vs adding a new model; unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Loss utility fix vs Mistral4 test/model work; different concrete problems." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "ExecuTorch dynamic-shape test vs Mistral4 test/model work; only broad testing theme overlaps." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "ExecuTorch export test vs MobileLLM model addition; unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "New MobileLLM model addition vs config-resolution fix in modeling_utils; no shared code path." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "New MobileLLM model addition vs Falcon output-interface refactor; unrelated model implementations." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactor vs model-creation skill/docs; no common underlying bug or change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon output-collection refactor vs Mistral4 test fixes; different code paths." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test fixes vs layrnorm->layernorm renaming across CLIP-like models; unrelated fixes." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 test fixes vs TDT loss kernel integration; different functionality." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Mistral4 test fixes vs config resolution ordering fix; not the same issue." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 test/model work vs model-creation skill repository changes; unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "New MobileLLM model addition vs model-creation skill; the latter is tooling/docs, not the model change itself." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs fix the same concrete loss-path issue in src/transformers/loss/loss_utils.py by extending fixed_cross_entropy to forward supported cross-entropy kwargs for issue 43240; they are plausible duplicates." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The verdicts are grounded in the packet: the only mergeable same-bug pair is PR 43251/43254 on loss_utils.py for issue 43240, while the other pairs involve distinct model additions, tests, or tooling changes and are conservatively rejected." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 28, + "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": "b83a5fb80fef75147d1766147c0d5381ca4f18bb0027bff2c8fbae2039cc5baf", + "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 PRs, but they are not duplicates: one is the base model implementation and the other is a later realtime/tokenizer extension.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 40546 is the foundational VibeVoice implementation and the broader, earlier change; PR 42781 is a later realtime-focused extension built on the same model family.", + "best_issue_reason": null, + "best_pr_reason": "PR 40546 is the best representative for the cluster because it is the main initial integration, has substantial review activity, and is not a draft.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Do not merge: 40546 adds VibeVoice support, while 42781 adds VibeVoice Realtime and the acoustic tokenizer. They share model-family files but target different functionality and would not plausibly be one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 3, + "component_count": 28, + "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": "9e53778aa5b28f61297df49e9f8de9f36fa44e8b73db64ca624425119445c80d", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are unrelated: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. They do not fix the same bug or implement the same change, so this is not a duplicate cluster.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": "If a representative PR is needed, #41224 is the more focused and better-scoped change; #40755 is a separate TimesFM feature in a different code path and model family.", + "best_issue_reason": null, + "best_pr_reason": "#41224 is the stronger standalone PR to keep as the cluster representative because it is narrower and has substantial review activity, but it is not a duplicate of #40755.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different models, different functionality, and different code paths: TimesFM covariate forecasting vs DINOv3 image classification support. They are 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": 4, + "component_count": 28, + "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": 28, + "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": "cbf5c2858422f83689c1238420600c8a5592a66987670a0f7fa7499375570d39", + "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": "These are three separate model-addition PRs (DeepSeek OCR, Isaac, Kimi2-6). They share Transformers boilerplate and auto-registration changes, but they do not fix the same underlying bug or implement the same concrete model path, so they should not be clustered as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different models and codepaths: DeepSeek OCR vs Isaac. Shared docs/auto-mapping scaffolding is generic model-onboarding work, not the same change." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Different model families and implementations: Isaac vs Kimi2-6. Overlap is limited to standard integration files, which is insufficient for duplicate PRs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 6, + "component_count": 28, + "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": "d5520462a5e32a6c2b5270c31c7aba0fa12ea20d9c78fb2f10f3a62c814c8279", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related workflow-maintenance PRs, but they are not duplicates: one is a broad GitHub Actions upgrade for Node 24 compatibility, while the other is a narrower refresh of docker/build-push-action versions in docker image workflows.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #42975 is the better canonical representative because it is the broader, more clearly scoped workflow-action compatibility upgrade across many files, centered on Node 24 support.", + "best_issue_reason": null, + "best_pr_reason": "PR #42975 is the best cluster anchor: it captures the main infra-wide dependency modernization theme and covers the widest shared workflow surface, whereas #42976 is a separate docker-action version bump.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "They touch overlapping GitHub Actions workflows, but the concrete changes differ: #42975 upgrades many actions for Node 24 compatibility, while #42976 mainly bumps docker/build-push-action in docker image workflows. They are adjacent maintenance PRs, not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 28, + "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-42978-11", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45487", + "pull_request:45490", + "pull_request:46109", + "pull_request:46135", + "pull_request:46151", + "pull_request:46162", + "pull_request:46214", + "pull_request:46225" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:42978|pull_request:46225", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43448|pull_request:46225", + "pull_request:43451|pull_request:45490", + "pull_request:43451|pull_request:46225", + "pull_request:45487|pull_request:46214", + "pull_request:45487|pull_request:46151", + "pull_request:45490|pull_request:46225", + "pull_request:45490|pull_request:46162", + "pull_request:46109|pull_request:46214", + "pull_request:46109|pull_request:46151", + "pull_request:46135|pull_request:46214", + "pull_request:46135|pull_request:46151", + "pull_request:46151|pull_request:46214", + "pull_request:46162|pull_request:46214", + "pull_request:46162|pull_request:46225" + ], + "prepared_review_unit_hash": "6d3711fd08e70be02f7d202ca77275518449b905332eaafcbf11a1ed7f71f383", + "component_budget": { + "node_count": 11, + "item_count": 11, + "soft_pair_count": 20, + "serialized_chars": 34733, + "estimated_input_tokens": 8684, + "estimated_eval_tokens": 17624 + }, + "budget": { + "node_count": 11, + "item_count": 11, + "soft_pair_count": 20, + "serialized_chars": 34733, + "estimated_input_tokens": 8684, + "estimated_eval_tokens": 17624 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is not a true duplicate set: it mixes several unrelated model-addition PRs and several separate test/expectation tweaks. The only thematic overlap is the small XPU/distributed-test expectation subgroup, but those still target different test suites and code paths.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single PR is a real duplicate canonical, but PR 46109 is the closest representative of the only tightly related subset because it is the broadest of the XPU/distributed-test expectation updates and connects to the other expectation PRs more than the model-addition PRs do.", + "best_issue_reason": null, + "best_pr_reason": "PR 46109 is the best global representative of the cluster\u2019s only coherent slice: generic distributed trainer test expectations. It is still not a duplicate of the others, but it is the most central among the related test-only changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different model additions (ViT NEPA vs Molmo); shared docs/auto wiring is incidental, not the same change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "Different model additions (ViT NEPA vs Molmo2); same subsystem, but distinct models and implementations." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Different model additions (ViT NEPA vs CTSM) with unrelated architectures and files." + }, + { + "left": "pull_request:42978", + "right": "pull_request:46225", + "accept": false, + "reason": "Different model additions (ViT NEPA vs Rio2); not the same underlying feature." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are separate model families/PRs, not one bug or one code change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo vs CTSM are unrelated model additions despite shared auto/doc plumbing." + }, + { + "left": "pull_request:43448", + "right": "pull_request:46225", + "accept": false, + "reason": "Molmo and Rio2 are distinct model integrations, not a duplicate fix." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are different model additions with no shared underlying change." + }, + { + "left": "pull_request:43451", + "right": "pull_request:46225", + "accept": false, + "reason": "Molmo2 and Rio2 are unrelated model additions." + }, + { + "left": "pull_request:45487", + "right": "pull_request:46151", + "accept": false, + "reason": "AltCLIP/ChineseCLIP model-parallel fix is unrelated to pipeline token-classification test expectations." + }, + { + "left": "pull_request:45487", + "right": "pull_request:46214", + "accept": false, + "reason": "AltCLIP/ChineseCLIP parallelism fix is a different code path from HRM text XPU expectations." + }, + { + "left": "pull_request:45490", + "right": "pull_request:46225", + "accept": false, + "reason": "CTSM and Rio2 are separate model-addition PRs; same auto-registration patterns do not make them duplicates." + }, + { + "left": "pull_request:45490", + "right": "pull_request:46162", + "accept": false, + "reason": "Model addition for CTSM is unrelated to Florence2/LFM2-VL XPU test expectations." + }, + { + "left": "pull_request:46109", + "right": "pull_request:46151", + "accept": false, + "reason": "Distributed trainer parity tests and token-classification pipeline expectations are different test suites and failure modes." + }, + { + "left": "pull_request:46109", + "right": "pull_request:46214", + "accept": false, + "reason": "Both are test-related, but one is distributed trainer parity and the other is HRM text expectations; not the same change." + }, + { + "left": "pull_request:46135", + "right": "pull_request:46214", + "accept": false, + "reason": "Kosmos2.5 integration expectations and HRM text expectations are distinct model-specific test adjustments." + }, + { + "left": "pull_request:46135", + "right": "pull_request:46151", + "accept": false, + "reason": "Kosmos2.5 integration test updates and pipeline token-classification expectations are unrelated." + }, + { + "left": "pull_request:46151", + "right": "pull_request:46214", + "accept": false, + "reason": "Both add expectations, but for different components and failure signatures; not one underlying bug." + }, + { + "left": "pull_request:46162", + "right": "pull_request:46214", + "accept": false, + "reason": "Florence2/LFM2-VL XPU expectations and HRM text XPU expectations are separate model-specific test changes." + }, + { + "left": "pull_request:46162", + "right": "pull_request:46225", + "accept": false, + "reason": "Florence2/LFM2-VL XPU expectations are unrelated to adding the Rio2 model." + } + ] + }, + "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": 28, + "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-43291-3", + "nodes": [ + "pull_request:43291", + "pull_request:43665", + "pull_request:44815" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43665|pull_request:44815" + ], + "prepared_review_unit_hash": "a5a469b0458fe17c41e90ea5bac0d127df59cd6c2b0fc0209d30c8562032fb3d", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three PRs are unrelated: one changes Whisper decoding logic, one is a test-only patch for CLIP/ViT, and one fixes FP8/dequant loading paths across several components. None should be clustered as the same underlying change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:44815 is the strongest standalone change: it has a concrete bug-fix scope across the FP8 loading/dequant path and includes the most substantive implementation work.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:44815 is the best representative PR in this set because it addresses a specific runtime correctness issue with real code-path changes, unlike the test-only or narrow Whisper patch.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Different models and different code paths: Whisper tokenizer decoding vs CLIP/ViT test edits. No shared underlying bug or mergeable common fix." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Both touch tokenization-related files, but they fix unrelated model-specific behaviors in different areas (Whisper decoding vs Mistral tokenization/FP8 loading). Not the same bug." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "Test-only CLIP/ViT assertions are unrelated to the FP8/dequant loading fix. They do not target the same concrete code-path problem." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 9, + "component_count": 28, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43488-5", + "nodes": [ + "pull_request:43488", + "pull_request:43532", + "pull_request:45932", + "pull_request:46100", + "pull_request:46114" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:46114", + "pull_request:43488|pull_request:43532", + "pull_request:43532|pull_request:46114", + "pull_request:45932|pull_request:46100", + "pull_request:46100|pull_request:46114" + ], + "prepared_review_unit_hash": "c300c73a714a7274812a73fa76814a5de4f927f60250162a43b7aace2af6d53f", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11339, + "estimated_input_tokens": 2835, + "estimated_eval_tokens": 5926 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The strongest duplicate signal is the ViT test-migration pair 46100/46114; the other PRs are distinct. The packet\u2019s deterministic_accept flag on 45932/46100 is not supported by the visible diff, so it should not be treated as the same change.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46100 is the cleanest representative of the ViT test-migration duplicate pair: it is the minimal diff showing the inserted failing tests, while 46114 is a forked copy with the same change but a more specialized title.", + "best_issue_reason": null, + "best_pr_reason": "PR 46100 is the best global PR representative because it most clearly captures the underlying ViT test-migration change. 46114 is a near-copy of that same change, while 45932 and the other PRs touch different code paths or broader CI/modeling plumbing.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:46114", + "accept": false, + "reason": "Both mention ViT tests, but 43488 also changes setup/versioning and dependency plumbing; 46114 is only a tiny test-migration PR, so they are not the same underlying change." + }, + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "These are different dummy PRs: 43488 is packaging/version/dependency setup plus a test file, while 43532 rewrites workflow/comment-CI and related utilities." + }, + { + "left": "pull_request:43532", + "right": "pull_request:46114", + "accept": false, + "reason": "Shared ViT test filename is not enough; 43532 is CI/workflow plumbing with broader unrelated edits, while 46114 is a minimal failing-test migration change." + }, + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "The deterministic_accept flag is inconsistent with the visible diffs: 45932 changes model code and CI secrets across several files, while 46100 only adds deliberately failing ViT tests, so they are not the same change." + }, + { + "left": "pull_request:46100", + "right": "pull_request:46114", + "accept": true, + "reason": "These are effectively the same ViT test-migration change: same file, same inserted failing tests, and the same narrow purpose, so they look like duplicate copies of one PR." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary is not fully grounded in the packet. The packet marks pull_request:45932 | pull_request:46100 as deterministic_accept=true, but the analyst says the only true duplicate pair is 46100/46114. That overstates the evidence and ignores the strongest signal in the packet. Because of that, the global/canonical PR reasoning is not conservative enough. The 46100/46114 duplicate claim may be plausible from the diffs, but the writeup cannot assert exclusivity here." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 10, + "component_count": 28, + "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": "fc011bde4e24b815cfb08205bf207e3adb9e9aef912cd2dda1bcf2e9761266b8", + "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": "These PRs are closely related in scope and infrastructure, but they add different pipelines for different tasks: promptable concept segmentation vs promptable visual segmentation. They should not be treated as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: the two pull requests implement distinct pipeline features for different segmentation tasks, not the same underlying change.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a better duplicate canonical than the other; both are standalone feature additions, and the visual-segmentation PR is not a substitute for the concept-segmentation PR.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Different pipelines and different target tasks/code paths; they share framework/docs scaffolding but not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 11, + "component_count": 28, + "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": "5590ee7fff176c2be8c71ba3686a579581d26c7559945374b9e69b6390053049", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs both touch `trainer.py`, but they address different extension points in `Trainer`: one adds `_loss_is_scaled_for_ga` for gradient-accumulation loss scaling, the other adds `_get_signature_columns` for column filtering. They are related only by file and broad subsystem, not by the same underlying bug or change.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "No clear canonical PR: the patches solve different problems in `Trainer` and should not be merged as duplicates.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a duplicate of the other; if forced to pick a representative, `#43651` is slightly more self-contained and narrowly scoped, but `#43663` is equally valid for its separate feature.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Reject: both modify `trainer.py`, but one changes gradient-accumulation loss scaling and the other refactors signature-column inspection for custom trainers. They do not fix the same concrete code-path problem and would not plausibly be merged into one PR 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": 12, + "component_count": 28, + "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": "8f0762a004e04e6f03f10485ebb193df9fbacc08950a8041d221f4964fc4e5c9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3664, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3664, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs touch `src/transformers/cli/serve.py`, but they address different runtime/type-hint problems. The stronger canonical candidate is the narrower `pydantic`-related NameError fix; the other is a separate typing/annotation cleanup. Soft match rejected.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43842 is the more specific and self-contained fix: it guards the `TypeAdapter` import behind `TYPE_CHECKING` and quotes the annotation so the CLI does not require `pydantic` at runtime.", + "best_issue_reason": null, + "best_pr_reason": "PR #43842 is the best representative of a concrete bug fix because it directly targets the runtime failure mode described in its title and diff, with minimal unrelated changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Both touch the same file, but they fix different underlying issues: #43656 is a broader annotations/`__future__` typing change, while #43842 specifically fixes `TypeAdapter` being unavailable when `pydantic` is not installed. They are not the same code-path bug and would not be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 13, + "component_count": 28, + "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": 28, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44066|pull_request:44071", + "pull_request:44018|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44066|pull_request:44068", + "pull_request:44019|pull_request:44722", + "pull_request:44086|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44116", + "pull_request:44018|pull_request:44116", + "pull_request:44068|pull_request:44722", + "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:44722", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:44071|pull_request:44086", + "pull_request:43997|pull_request:44073", + "pull_request:44017|pull_request:44071", + "pull_request:44073|pull_request:44101", + "pull_request:44059|pull_request:44071", + "pull_request:44018|pull_request:44059", + "pull_request:44073|pull_request:44161", + "pull_request:44056|pull_request:44101", + "pull_request:44059|pull_request:44068", + "pull_request:44101|pull_request:44161", + "pull_request:44017|pull_request:44018", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:43997|pull_request:44018", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:44018|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44018|pull_request:44129", + "pull_request:44010|pull_request:44073", + "pull_request:43997|pull_request:44074", + "pull_request:44010|pull_request:44018", + "pull_request:44071|pull_request:44129", + "pull_request:43997|pull_request:44129", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154" + ], + "prepared_review_unit_hash": "2f74dcab86a770e0a07b14c38c49704cc87a4a01b7b56b113443136807633a5a", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 + }, + "budget": { + "node_count": 27, + "item_count": 27, + "soft_pair_count": 72, + "serialized_chars": 75091, + "estimated_input_tokens": 18773, + "estimated_eval_tokens": 37802 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of independent model-by-model refactors to the new output-tracing/capture_outputs interface. The only clear duplicate/overlapping PRs are the GPT-J/CodeGen changes; the DeBERTa, SpeechT5, ViLT, Vits, RegNet/ResNet, and similar PRs are related by theme but not the same concrete code-path change.", + "confidence": 0.86, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44722 is the best canonical representative for the duplicate GPT-J/CodeGen refactor: it targets the same concrete output-tracing change as 44066/44085, explicitly ties to #43979, and appears to cover the shared code path most completely.", + "best_issue_reason": null, + "best_pr_reason": "44722 is the strongest global PR candidate because it is the clearest concrete implementation of the shared GPT-J/CodeGen output-tracing refactor and the most plausible merge target among the near-duplicates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different models and different code paths (FNet/CVT vs GPT-J); same theme only, not the same underlying change." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "FNet/CVT output-tracing refactor is unrelated to DeBERTa's attention/output changes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Both are GPT-J output-tracing refactors and touch the same underlying model code path; they look like overlapping implementations of the same change." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "RegNet/ResNet output-tracing migrations are separate model implementations, not one duplicate bug or change." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet output tracing and GPT-J/CodeGen output tracing are different model-specific refactors." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen refactor vs EfficientNet refactor; broad standardization theme only." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa are unrelated code paths, so this is not a duplicate." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen refactor does not match MGP-STR's model-specific output handling." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen and MPT are separate model refactors, not the same concrete bug/change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Both target GPT-Neo output-tracing standardization in the same file/code path and are plausibly duplicate implementations of the same refactor." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are different model code paths, so this is only a thematic match." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen vs GPT-Neo are distinct model refactors." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet output-tracing refactor and GPT-J/CodeGen refactor are unrelated model-specific changes." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR output tracing is a different model and code path from GPT-J/CodeGen." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are different implementations; same refactor theme is not enough." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are unrelated model changes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are separate model refactors, not duplicates." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are unrelated code paths." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are different model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are distinct model-specific changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo refactor is separate from GPT-J/CodeGen refactor." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated changes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR do not share the same concrete code path." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44085", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are different model-specific refactors." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT2 output tracing and MGP-STR output tracing are separate model implementations." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J and MGP-STR are different code paths; only the general output-tracing theme overlaps." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and XLM/Flaubert are distinct model refactors, not the same underlying change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT2 and MPT are different model code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "SegFormer output-tracing refactor is unrelated to GPT-J/CodeGen." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "SegFormer and LongT5 are separate model-specific refactors." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "SegFormer and VisualBert are unrelated model changes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44086", + "accept": false, + "reason": "MPT and MGP-STR are unrelated model-specific refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet output tracing and VisualBert output tracing are different code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44071", + "accept": false, + "reason": "SegFormer and MPT are separate model refactors." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44101", + "accept": false, + "reason": "VisualBert and XLM/Flaubert are not the same implementation change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT2 and GPT-Neo are different model code paths." + }, + { + "left": "pull_request:44101", + "right": "pull_request:44161", + "accept": false, + "reason": "Flaubert/XLM and LongT5 are unrelated refactors." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44018", + "accept": false, + "reason": "SegFormer and GPT-Neo are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are separate model refactors." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT and ImageGPT are different model code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert/XLM are unrelated model-specific changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet/ResNet and GPT-Neo are not the same underlying change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are different model refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet/ResNet and LongT5 are unrelated code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet and LongT5 are separate model changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44161", + "accept": false, + "reason": "GPT-Neo and LongT5 are different model code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44071", + "accept": false, + "reason": "RegNet/ResNet and MPT are unrelated refactors." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are different model-specific changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are distinct model code paths." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert and VisualBert are unrelated model refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet/ResNet and TextNet are separate model-specific changes." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44044", + "accept": false, + "reason": "GPT-Neo and DeBERTa are different model code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa and TextNet are unrelated changes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "Same GPT-J/CodeGen output-tracing refactor with overlapping file changes; these are the closest duplicate pair in the cluster." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Both are GPT-J output-tracing refactors that touch the same underlying implementation and look mergeable into one change." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-J and SpeechT5 are unrelated model-specific refactors." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": false, + "reason": "FNet/CVT and RegNet/ResNet are separate model implementations; not a duplicate." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": false, + "reason": "FNet/CVT and SpeechT5 are unrelated code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": false, + "reason": "DeBERTa and SpeechT5 are different model refactors." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44154", + "accept": false, + "reason": "DeBERTa and Vits are separate model-specific changes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen and EfficientNet are different model code paths." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet and GPT-J/CodeGen are unrelated refactors." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet and stays conservative about unrelated model-specific refactors. The duplicate/overlapping PR calls are also supported: GPT-J/CodeGen (44066/44085/44722) and GPT-Neo (44018/44068) share the same code paths, while the false pairings are appropriately rejected." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 28, + "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": "bc8106fc812e44d16604a9b15a8cf14b907b9a07bbcfc880496b1d620358f821", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84896, + "estimated_input_tokens": 21224, + "estimated_eval_tokens": 42704 + }, + "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": "This cluster is a set of separate model-specific PRs all migrating different Transformers models to standardized output tracing/capture helpers. They share a pattern, but they are not duplicates of the same underlying change or code path.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "44073 is the clearest representative of the shared refactor pattern: it introduces `can_return_tuple`/`capture_outputs` and model-specific output hook registration in a self-contained, well-scoped VisualBert change.", + "best_issue_reason": null, + "best_pr_reason": "44073 is the best overall representative PR because its title and diff most clearly express the standardized output-tracing migration without being overly model-specific or incomplete.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "Both are output-tracing refactors, but for different models (FocalNet vs ImageGPT). Different classes and code paths; not one mergeable fix." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "VisualBert and TextNet are separate model implementations. Same refactor theme, but not the same underlying bug/change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "Both standardize output tracing, but MPNet and VisualBert have distinct encoder/attention code paths and separate model-specific hook wiring." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "Different models and different forward paths. This is a shared migration pattern, not a duplicate PR." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated model implementations; these are parallel refactors, not the same fix." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert and SpeechT5 touch different model stacks, and SpeechT5 also targets a specific issue. 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": 15, + "component_count": 28, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43998-14", + "nodes": [ + "pull_request:43998", + "pull_request:43999", + "pull_request:44000", + "pull_request:44001", + "pull_request:44002", + "pull_request:44003", + "pull_request:44004", + "pull_request:44025", + "pull_request:44026", + "pull_request:44027", + "pull_request:44028", + "pull_request:44029", + "pull_request:44030", + "pull_request:46106" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030", + "pull_request:44027|pull_request:46106", + "pull_request:44026|pull_request:46106", + "pull_request:44001|pull_request:46106", + "pull_request:44002|pull_request:46106", + "pull_request:43998|pull_request:46106" + ], + "prepared_review_unit_hash": "849968201de66f33101dc0f547ed0df52a49fb235df4d21cf38574b9620342bf", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 44, + "serialized_chars": 43396, + "estimated_input_tokens": 10849, + "estimated_eval_tokens": 21954 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 44, + "serialized_chars": 43396, + "estimated_input_tokens": 10849, + "estimated_eval_tokens": 21954 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a batch of model-specific output-tracing refactors plus one unrelated docs toctree edit. The shared wording/pattern makes them look similar, but they are not the same underlying bug or change, so none of the soft edges should be merged.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: these are separate per-model refactors with only template-level similarity, not one duplicated fix.", + "best_issue_reason": null, + "best_pr_reason": "44025 is the closest representative because it covers two related depth_anything variants and shows the refactor pattern cleanly, but it still isn't merge-equivalent to the rest of the batch.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models and code paths (timm_backbone vs upernet); same refactor style only." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different model families and forward paths; similarity is only the shared output-tracing refactor pattern." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Separate model implementations (timm_backbone vs univnet), not the same fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Unrelated architectures and files; no shared underlying bug/change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model code paths (timm_backbone vs rwkv); only the refactor template matches." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "MobileNetV1 and UnivNet are unrelated; same tracing refactor idea, not one duplicate change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models and outputs; no common concrete bug to merge." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families and forward implementations; not the same underlying change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Separate encoder/decoder vs MobileNetV1 refactors; template similarity only." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "Different subsystems and code paths; not merge-equivalent." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Vision-text dual encoder and UnivNet are unrelated; same refactor motif only." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models and forward paths; no shared underlying fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Distinct model families (CLIP-style dual encoder vs mamba variants); not duplicates." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Different target models and code paths; similar output-tracing cleanup only." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "UnivNet and UperNet are unrelated; same refactor pattern, different change." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and implementation details; not the same bug or fix." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families and forward signatures; only the shared tracing refactor matches." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "UnivNet vs CodeGen are unrelated code paths; not a duplicate PR pair." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "UperNet and Mamba/Falcon-Mamba are unrelated; no single concrete change in common." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "Different model families and fixes; same refactor theme only." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Falcon/Mamba output tracing and CodeGen output tracing are separate model-specific changes." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "TimmBackbone and CodeGen are unrelated; the similarity is only cosmetic." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "Different model paths and behavior; no mergeable duplicate fix here." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "UperNet and CodeGen touch different subsystems, so this is not the same change." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Depth Anything vs Vision Encoder-Decoder are separate model implementations; only the refactor pattern is shared." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Different models and output schemas; not the same underlying bug or change." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "Depth-anything refactor and SuperPoint refactor are unrelated code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families (depth estimation vs RWKV); no duplicate fix." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "Depth-anything and DPR are unrelated; same refactor style only." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "Vision Encoder-Decoder and RWKV are different code paths and fixes." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "Different model stacks (vision-text vs speech encoder-decoder); not merge-equivalent." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "Encoder-decoder refactor vs SuperPoint refactor are unrelated changes." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "Different model families and output behavior; only the template matches." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "Speech Encoder-Decoder and SuperPoint are unrelated models; not the same fix." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model code paths; no shared underlying bug/change." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "Speech Encoder-Decoder vs DPR are separate changes, not duplicates." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "SuperPoint and RWKV are unrelated; similarity is only the refactor style." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "Different model implementations and behavior; not the same underlying change." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "RWKV and DPR are unrelated code paths; no common duplicate fix." + }, + { + "left": "pull_request:44027", + "right": "pull_request:46106", + "accept": false, + "reason": "A docs toctree edit is not the same kind of change as a Speech Encoder-Decoder code refactor." + }, + { + "left": "pull_request:44026", + "right": "pull_request:46106", + "accept": false, + "reason": "Documentation navigation change is unrelated to the Vision Encoder-Decoder refactor." + }, + { + "left": "pull_request:44001", + "right": "pull_request:46106", + "accept": false, + "reason": "UnivNet code refactor and docs toctree update are different kinds of changes." + }, + { + "left": "pull_request:44002", + "right": "pull_request:46106", + "accept": false, + "reason": "UperNet model refactor is unrelated to the docs-only toctree change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:46106", + "accept": false, + "reason": "TimmBackbone code refactor and docs navigation change do not fix the same thing." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 16, + "component_count": 28, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44662-3", + "nodes": [ + "pull_request:44662", + "pull_request:45075", + "pull_request:45401" + ], + "soft_pairs": [ + "pull_request:44662|pull_request:45401", + "pull_request:44662|pull_request:45075", + "pull_request:45075|pull_request:45401" + ], + "prepared_review_unit_hash": "2d9bc19f34434295ee25bca505623ecfd244f2955d09508d60f9537adeae9132", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10913, + "estimated_input_tokens": 2729, + "estimated_eval_tokens": 5714 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three distinct model-addition PRs (PenguinVL, DeepSeek-OCR-2, Voxtral-TTS). They overlap only on boilerplate registration/docs files and do not appear to fix the same underlying code-path or model implementation, so none should be clustered as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different model integrations (PenguinVL vs Voxtral-TTS). Shared files are only auto-registry/docs scaffolding, not the same implementation or bug fix." + }, + { + "left": "pull_request:44662", + "right": "pull_request:45075", + "accept": false, + "reason": "Both add separate models and touch common Transformers registration files, but they are unrelated code changes with different architectures, processors, and tests." + }, + { + "left": "pull_request:45075", + "right": "pull_request:45401", + "accept": false, + "reason": "DeepSeek-OCR-2 and Voxtral-TTS are unrelated model additions; overlap is limited to shared framework plumbing, which is insufficient to merge as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 28, + "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": "a464cfe6140a5fff55d89efcf10ae43c109abbc2508cdce4f3f204381015e527", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related kernel/module-fusion feature PRs, but they target different abstractions and code paths: one introduces a new Module Fusion API, while the other wires n-to-1 kernel fusion through KernelConfig and modeling/integration changes. They should not be treated as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45363 is the better canonical representative because it implements the more concrete end-to-end feature path (KernelConfig-driven fusion) across the main integration points and includes a usage example.", + "best_issue_reason": null, + "best_pr_reason": "PR 45363 is the stronger overall representative for this cluster since it is the more complete user-facing fusion change and is easier to map to a specific merged capability.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "44979 adds a standalone Module Fusion API, while 45363 adds n-to-1 kernel fusion via KernelConfig and touches different integration/code paths. They share theme, not the same underlying bug/change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 28, + "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": "06fc293f5a2c69356a0796b5e3e23800e0ce1617c5c60c3fbcccbf3edef44ef5", + "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": "Two PRs implement the same Mistral regex patch for the same issue (#45081); the later PR adds tests and is the more complete representative.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45317 is the more complete version of the same fix: it makes the same AttributeError patch and also adds test coverage, so it is the best canonical PR for this cluster.", + "best_issue_reason": null, + "best_pr_reason": "PR #45317 is the strongest standalone candidate because it contains the same code-path fix as #45086 plus tests, making it the more merge-ready implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Both PRs fix the same concrete AttributeError in `_patch_mistral_regex` by changing `tokenizer.backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer` in the same file; #45317 is simply a more complete follow-up 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": 28, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45115-2", + "nodes": [ + "pull_request:45115", + "pull_request:45569" + ], + "soft_pairs": [ + "pull_request:45115|pull_request:45569" + ], + "prepared_review_unit_hash": "8f3329e455b7557b961e710950a02e55601f9b272a50e8681b15887ded426701", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related NemotronH model refactors, but they do not look like the same concrete change. PR 45115 focuses on inheritance/conversion mapping for NemotronH, while PR 45569 is a broader model split introducing NemotronHDense/NemotronHSparse, docs, auto-mapping, and tests.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45569 is the broader, more complete architecture change and likely the best representative for the NemotronH family restructuring; PR 45115 is a narrower preparatory refactor around conversion/mapping.", + "best_issue_reason": null, + "best_pr_reason": "PR 45569 most directly captures the end-state change: dispatcher over dense/sparse variants plus model, auto, docs, and test updates. It is the best single PR to anchor this cluster.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "They share NemotronH-related files, but 45115 is a narrow inheritance/conversion-mapping refactor, while 45569 is a broader split into dense/sparse model variants with docs and tests. They are related work, not the same mergeable 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": 20, + "component_count": 28, + "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": "26027937e4d88c502677f1ac5f43cecb8f68be4d301d116fe7755bb320166e98", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5398, + "estimated_input_tokens": 1350, + "estimated_eval_tokens": 2956 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5398, + "estimated_input_tokens": 1350, + "estimated_eval_tokens": 2956 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are related feature work in the heterogeneity area, but they are not the same concrete change: one adds per-layer config plumbing, while the other adds broader heterogeneous model/runtime support that builds on it.", + "confidence": 0.81, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45332 is the broader umbrella PR: it covers heterogeneous model behavior across cache/model/masking paths and appears to incorporate the config groundwork rather than being the same standalone change as #45333.", + "best_issue_reason": null, + "best_pr_reason": "PR #45332 is the best representative of the cluster because it spans the wider end-to-end heterogeneous support; #45333 is narrower config-only infrastructure.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Reject: #45333 is per-layer configuration plumbing, while #45332 adds heterogeneous model/runtime support and only shares the same subsystem and helper files. They are related, but not the same underlying fix/change and would not plausibly be merged as one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 21, + "component_count": 28, + "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": "d2acc15ead8a138a54b8e333731bd239e0f2360ed4b232afea6aea7d3119acfe", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5514, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5514, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests are typing-related, but they address different changes: one is a broad type-checking cleanup across many files, while the other adds reusable typing protocols/aliases for modeling and PEFT-related code. They are related in theme, not duplicates.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45415 is the broader umbrella change in the cluster: it applies type-checking cleanup across many `src/transformers/*` files and updates type-ignore policy, so it is the most representative PR if one needs a single canonical item.", + "best_issue_reason": null, + "best_pr_reason": "PR 45415 is the best overall representative because it covers the widest surface area and is the clearest cluster anchor for the typing cleanup theme.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Both are about typing, but they fix different concrete things: 45415 is a broad typing-pass / ignore-policy sweep across many files, while 45425 introduces new typing aliases and protocols for `modeling_utils`/PEFT integration. The overlap is thematic and partial, not the same underlying change, so they should remain separate." + } + ] + }, + "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": 28, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45513-3", + "nodes": [ + "pull_request:45513", + "pull_request:46084", + "pull_request:46098" + ], + "soft_pairs": [ + "pull_request:45513|pull_request:46084", + "pull_request:46084|pull_request:46098" + ], + "prepared_review_unit_hash": "584a6308539d9c38dfbe37b975bd38830bd8f4ed81958258cd6837ec0491bb81", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7428, + "estimated_input_tokens": 1857, + "estimated_eval_tokens": 3970 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7428, + "estimated_input_tokens": 1857, + "estimated_eval_tokens": 3970 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs are about the same Mamba2 cached-forward bug with `use_cache` and `seq_len > 1`; the Qwen3.5 PR is unrelated and only similar at a high level.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46098 is the best canonical PR because it targets the same Mamba2 `use_cache` seq_len>1 bug as #46084, but appears broader and more complete (explicit issue target, docs, tests, and both CPU/GPU path coverage).", + "best_issue_reason": null, + "best_pr_reason": "PR #46098 is the strongest representative of the cluster: it directly references the tracking issue, covers the core Mamba2 cached-forward incorrect-results bug, and includes the most complete fix surface.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45513", + "right": "pull_request:46084", + "accept": false, + "reason": "Different model families and bugs: Qwen3.5 GDN linear attention vs Mamba2 cached forward. The similarity is only the general pattern of multi-token cached decoding." + }, + { + "left": "pull_request:46084", + "right": "pull_request:46098", + "accept": true, + "reason": "Same underlying Mamba2 `use_cache` with `seq_len > 1` bug, same target issue, same code area and tests; #46098 looks like a broader version of the same fix." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet. The 45513 vs 46084 soft rejection is conservative and supported by the clearly different model families and bugs. The 46084 vs 46098 soft acceptance is also supported: both target issue 46032, touch the same Mamba2 files, and 46098 appears to be a broader version of the same fix. No overstatement detected." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 23, + "component_count": 28, + "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": "0cf0adb3ed36830afaf9bb140ae2de108bfb4ba8cf53bb2ded5581b0c9038e6c", + "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 pull requests are EP-related, but they fix different concrete bugs: one is about preserving distributed_config / EP state during loading, the other about expert tensors being overwritten by rank-0 broadcast in FSDP2. They are similar in subsystem, not in underlying change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45662 is the more concrete, narrowly described bug fix: expert parameters are being overwritten by rank-0 broadcast during EP/FSDP2 loading. It reads as a specific code-path correction rather than a broader loading/serialization cleanup.", + "best_issue_reason": null, + "best_pr_reason": "PR #45662 is the strongest representative of this cluster because it targets a clearly defined failure mode in expert-parallel loading and has the more actionable fix surface.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Do not merge as duplicates: #45548 fixes EP + DeepSpeed ZeRO-3 loading/serialization via accelerate launch, while #45662 fixes EP + FSDP2 rank-0 broadcast overwriting expert parameters. Same subsystem, different underlying bugs and different fix paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 28, + "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": "7819ca7930a0dd5a671d4b4c7711d1a5ca46efa0371ba4287d9b8264d6123a9c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same xdist collision around captured_info artifacts in testing_utils; 45645 is the broader superset with CI log handling, while 45639 is the narrower core fix.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45645 is the more complete change set: it includes the xdist-safe testing_utils fix plus workflow updates to consume multiple captured_info files and related cleanup, making it the better representative of the underlying bug and its CI impact.", + "best_issue_reason": null, + "best_pr_reason": "PR 45645 is the strongest overall candidate because it fixes the core artifact-collision problem and also preserves CI debug log visibility; 45639 is a subset of that 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 in testing_utils, and 45645 plausibly subsumes 45639 by adding the CI-side handling needed for the same fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 25, + "component_count": 28, + "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-45670-2", + "nodes": [ + "pull_request:45670", + "pull_request:46229" + ], + "soft_pairs": [ + "pull_request:45670|pull_request:46229" + ], + "prepared_review_unit_hash": "cf892b87826a363e78dd6f369c8da6048bba076ec8677d5691e61610810e2980", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4185, + "estimated_input_tokens": 1047, + "estimated_eval_tokens": 2350 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4185, + "estimated_input_tokens": 1047, + "estimated_eval_tokens": 2350 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster contains two unrelated pull requests: one is a small AutoModel registry fix for GLM ASR, and the other is a new multi-token prediction generation implementation. They do not look like duplicates or the same underlying change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46229 is the better canonical representative because it is the larger, more substantive change and centers on a single new generation feature/module.", + "best_issue_reason": null, + "best_pr_reason": "PR #46229 is the strongest overall representative of the cluster: it is self-contained, technically substantial, and clearly defines a distinct code-path change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45670", + "right": "pull_request:46229", + "accept": false, + "reason": "These touch different code paths and solve different problems: #45670 only updates AutoModel registration for 'glmasr', while #46229 adds a new MTP generation module. They are not the same bug/change and would not plausibly merge into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 26, + "component_count": 28, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45737-2", + "nodes": [ + "pull_request:45737", + "pull_request:45738" + ], + "soft_pairs": [ + "pull_request:45737|pull_request:45738" + ], + "prepared_review_unit_hash": "cc2ade73a49ec6a8883a7a24c433988645588bcd049e8aad2bce56b1b24daa1c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same MusicgenMelody cache initialization bug in `forward()`: replacing `EncoderDecoderCache(DynamicCache, DynamicCache)` with a single `DynamicCache`. PR #45737 is the more complete duplicate because it bundles the fix with a regression test, while #45738 is the fix-only variant.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45737 is the best canonical PR because it contains the same code fix as #45738 plus a regression test, making it the most complete representation of the bug fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #45737 is the strongest overall candidate: it fixes the concrete cache bug and adds a targeted regression test, whereas #45738 only changes the implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45737", + "right": "pull_request:45738", + "accept": true, + "reason": "Same underlying code-path bug in `musicgen_melody` cache setup; the diffs match on the core fix and the test-only addition in #45737 could plausibly live in the same PR as #45738." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: both PRs make the same `musicgen_melody` cache initialization change, and #45737 only adds a regression test on top of that fix. The soft-edge accept is conservative enough because the diffs are mergeable into one PR for the same bugfix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 27, + "component_count": 28, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-46069-4", + "nodes": [ + "pull_request:46069", + "pull_request:46073", + "pull_request:46154", + "pull_request:46196" + ], + "soft_pairs": [ + "pull_request:46069|pull_request:46073", + "pull_request:46069|pull_request:46154", + "pull_request:46069|pull_request:46196", + "pull_request:46073|pull_request:46154", + "pull_request:46073|pull_request:46196", + "pull_request:46154|pull_request:46196" + ], + "prepared_review_unit_hash": "ecd9e8f8373e98832100fb9c8289d013ccb6b919120639273c7ac548cdaa7704", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 10496, + "estimated_input_tokens": 2624, + "estimated_eval_tokens": 5504 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 10496, + "estimated_input_tokens": 2624, + "estimated_eval_tokens": 5504 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs share a common theme (attention-dispatch support), but each targets a different model and code path, so they are not duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: CTRL, T5, ShieldGemma2, and SeamlessM4T/v2 each have separate model-specific implementations and test/docs changes.", + "best_issue_reason": null, + "best_pr_reason": "No single best PR for deduping; they are distinct per-model changes rather than one underlying fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46069", + "right": "pull_request:46073", + "accept": false, + "reason": "Different models and code paths: ShieldGemma2 image classification vs CTRL causal LM; same general feature, not the same change." + }, + { + "left": "pull_request:46069", + "right": "pull_request:46154", + "accept": false, + "reason": "Different model families and attention implementations; one is ShieldGemma2, the other is T5, so they are not the same underlying bug/change." + }, + { + "left": "pull_request:46069", + "right": "pull_request:46196", + "accept": false, + "reason": "Distinct multimodal model codebases (ShieldGemma2 vs SeamlessM4T/v2); too different to merge as duplicates." + }, + { + "left": "pull_request:46073", + "right": "pull_request:46154", + "accept": false, + "reason": "Both add SDPA dispatch, but for different architectures (CTRL vs T5) and different attention paths, so not duplicates." + }, + { + "left": "pull_request:46073", + "right": "pull_request:46196", + "accept": false, + "reason": "CTRL and SeamlessM4T/v2 are unrelated model implementations with separate fixes; same feature area only." + }, + { + "left": "pull_request:46154", + "right": "pull_request:46196", + "accept": false, + "reason": "T5 and SeamlessM4T/v2 have different attention code paths and model behavior; not the same concrete bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 28, + "component_count": 28, + "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": "09bbdf87d91f5f2453ced59a2e7f15296ef4c53c18c9e5713a53051ab1d9b8dc", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4269, + "estimated_input_tokens": 1068, + "estimated_eval_tokens": 2392 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4269, + "estimated_input_tokens": 1068, + "estimated_eval_tokens": 2392 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are related by theme (removing source-file reads/refactoring introspection), but they address different code paths in different modules, so they are not duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46207 is the more concrete and self-contained change: it targets a specific modeling-utils helper, adds tests, and fixes a narrowly scoped behavior. PR 46152 is a broader utility refactor in a different module.", + "best_issue_reason": null, + "best_pr_reason": "PR 46207 is the best representative of the cluster because it has a clearly defined functional target and test coverage, making it easier to evaluate than the more infrastructure-oriented PR 46152.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46152", + "right": "pull_request:46207", + "accept": false, + "reason": "Both PRs avoid reading source files, but they fix different problems: 46152 changes import-utils path resolution, while 46207 replaces source-reading with class introspection for attention-implementation detection in modeling_utils. Different modules, different code paths, not one mergeable fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/manifest.json b/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..4a31f574442ad76d9e4136d0f5305c164e72adeb --- /dev/null +++ b/snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260527t060031z", + "artifacts": { + "hybrid": "snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-05-27T06:09:40Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260527T060031Z", + "variant": "hybrid" +} diff --git a/snapshots/20260527T060031Z/manifest.json b/snapshots/20260527T060031Z/manifest.json index 2bef7b6a15b7b539cc51224c53f857e3257cf575..4f09290397d4dd1ae3ee4e960a99d39648e8604d 100644 --- a/snapshots/20260527T060031Z/manifest.json +++ b/snapshots/20260527T060031Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 516 }, "extracted_at": "2026-05-27T06:00:31Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260527t060031z", + "runs": { + "hybrid-model-20260527t060031z": { + "analysis_id": "hybrid-model-20260527t060031z", + "artifacts": { + "hybrid": "snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260527T060031Z/analysis-runs/hybrid-model-20260527t060031z/manifest.json", + "model": null, + "published_at": "2026-05-27T06:09:40Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260527T060031Z", "watermark": {